Reservoir Simulation Note

What is Reservoir Simulation ?
• Understand flow and transport in our

reservoirs to make engineering decisions
• Develop mathematical models to

describe pressure, velocity, saturation,
etc.
• Described by coupled, nonlinear partial

differential equations (PDEs)
• Solution to PDEs can’t be found

analytically – we are left with solving
numerically
Reservoir Simulators Help Us Answer Several
Questions
1. How should a field be developed to maximize economic recovery?
2. What is the best enhanced recovery scheme for the reservoir?
3. Why is the reservoir not behaving according to predictions made by

previous engineering studies?
4. What is the ultimate economic recovery of the field?
5. What type of laboratory data is required?
6. Is it necessary to do physical model studies of the reservoir?
7. What is the best completion scheme for wells?
8. From what portion of the reservoir is the production coming?

Modeling and Simulation
Engineering Engineering
Problem Solution
Mathematical Numerical
Problem Solution
All Steps in Simulation are Important
1. Conversion to mathematical model requires understanding physics

– Fluid and rock properties of the reservoir
– Laws that describe flow and transport (conservation of mass, energy, and momentum)
2. Numerical solution to the mathematical problem requires simplifications

and approximations that are still accurate
– Assume that fluid and rock properties are constant over a control volume
– Transform non-linear PDEs into linear system of algebraic equations
– Solve the system of equations
3. Using the math solution to find an engineering solution requires

economics, experience, and good decision-making skills
General Idea
• Break up the reservoir into individual “blocks” or “grids”
• Write algebraic equations for pressure, saturation, etc. each block
• Each block depends on the adjacent block equations so we get a

system of linear equations
• Solve the system of equations. More blocks means more accuracy, but
longer computation time
Simulation Schematic...
Grid block Out Out

or cell
z y
In Out x
In
In
Conservation law...
• {Rate In} - {Rate Out} = {Accumulation}
• For each component (oil, gas, water, energy)
• For each cell
Solving the PDEs that describe flow
• Equations come from a mass balance and implementing Darcy’s law:

∂ ∂po ∂z ∂ φ So
λo −γo =
∂x ∂x ∂x ∂t Bo
∂ ∂pw ∂z ∂ φ Sw
λw −γw =
∂x ∂x ∂x ∂t Bw
• These equations are very complicated.

1. Coupled equations (water and oil)
2. PDEs (time and space – 1, 2 ,or 3D)
3. Variable properties (like permeability and porosity)
• Approximations allow us to get a problem that is easier to solve
T1 −T1
2 P1 b1
−T1 T2 −T3 P2 b
2 2
= 2 ⇒ Ax = b
−T3 T3 −T5/2 P3 b3
2
−T5 T4 P4 b4
2
What tools do I need to create my simulator?
• Write equations for fluid transport in porous media, rock/fluid properties,

and conservation of mass, momentum, and energy
• Use and perform error analysis of approximate numerical techniques to

solve math problems
– Root finding
– Systems of equations
– Interpolation, Integration, etc.
– Solution to differential equations
• Develop computer programs to solve these massive numerical problems

Class Exercise: List up to 3 examples of problems in
the last 2 years you had to:
• Use Matlab and/or program
• Find the root of a nonlinear equation
• Solve a system of equations (Ax=b)
• Curve fit/regression analysis
• Find an integral numerically
• Solve a differential equation numerically

Why can’t I just use a commercial simulator (CMG,
Eclipse) that someone smart already made?
• Great danger in using software if you don’t understand the fundamentals it

was built on
• Simulators have lots of limitations. They can give very misleading results
and only an understanding of the problem and mathematics will help your
recognize when they do.
• Understanding the math and physics separates you from the technician
• Who knows you might be the smart person that creates the next
commercial reservoir simulator
Categories of Reservoir Simulators
• Commercial
– CMG
– Eclipse (Schlumberger)
– Intersect (Schlumberger + Chevron)
– Nexus (Halliburton)
• In House
– Empower (ExxonMobil)
– Cheers (Chevron)
– PSim (ConocoPhillips)
– MoReS – (Shell)
– Powers – (Saudi)
• Academic
– UT-CHEM (UT Chemical Flooding)
– IPARS (UT Multiscale Simulation)
– GPAS (UT compositional)
– TOUGH2 (Lawrence Berkley National Labs)
CHAPTER 1.
SINGLE PHASE FLOW
EQUATIONS
Single Phase Flow Equations
Mass In – Mass Out +/- Generation/Consumption = Accumulation
z Mass in
Control volume can be any
shape
Control
Volume Mass enters and leaves the
Mass out x control volume in any direction
y
Mass Balance in 1D
Mass in Mass out
mx x A∆t − mx x +∆x
A∆t + m∆V ∆t = ρφ∆V t +∆t
− ρφ∆V t
generation
mass in mass out accumulation
mx x A∆t − mx A∆t + m∆V ∆t = ρφ∆V − ρφ∆V
Continuity Equation in 1D mass in
x +∆x
mass out
generation
t +∆t
accumulation
t
Divide the mass balance by A∆x∆t
mx x − mx ρφ t +∆t − ρφ t
x +∆x
+m=
∆x ∆t
Making ∆x and ∆t very small, they become dx and dt
mx x − mx ∂mx ρφ t +∆t − ρφ t ∂ ( ρφ )
lim x +∆x
=− and lim =
∆x → 0 ∆x ∂x ∆t → 0 ∆t ∂t
Expressing mass flux in terms of velocity ( ), and subtracting
we get the continuity equation to be
Continuity Equation
Continuity Equation in Multi-Dimensions
In 3D Cartesian coordinates the equation becomes
∂( ρux ) ∂( ρu y ) ∂( ρuz ) ∂ ( ρφ )
− + + = −m
∂x ∂y ∂z ∂t
More general form of the Continuity Equation
Other names:
- Mass Conservation Equation
- Material Balance Equation
Continuity Equation: What is ?
- accounts for mass entering or leaving a control volume from a

source or sink (e.g. well)
- We model injection/production wells as discrete sources/sinks so they

occupy no volume
- Taking limits without using this variable creates difficulty
mass Mass
m= Time • Unit Volume
time-volume
Units of mass/time per volume
of the control volume
0 Distance
Darcy’s Law
kA∆Φ
Q=−
THE
PUBLIC FOUNTAINS
OF THE CITY OF DIJON
µL
EXPERIENCE AND APPLICATION
PRINCIPLES TO FOLLOW AND FORMULAS TO BE USED
IN THE QUESTION
OF
THE DISTRIBUTION OF WATER

WORK FINISHES WITH
AN APPENDIX RELATING TO THE WATER SUPPLIES OF SEVERAL CITIES
THE FILTERING OF WATER
AND
THE MANUFACTURE OF STRONG PIPES OF LEAD, SHEET METAL AND BITUMEN
BY
HENRY DARCY
INSPECTOR GENERAL OF BRIDGES AND HIGHWAYS
The good quality o f water is one of things which contribute the most to
the health of citizen of a city. There is nothing that the magistrate
should have more of interest in then maintaining the quality of that
which is useful for the drink commune of man and of animal, and
remedies accidents by which these water can be altered, be in the read of
fountain, of river, of brook or they run, be in the place or be conserved
that that one in derive, be finally in the well of or be born of source.
(De JUSSIEU, Hist. de I'Acadimie. royale des sciences, 1733, p. 351.)
1 darcy = 9.869 x 10-9 cm2

PARIS
VICTOR DALMONT, EDITOR,

Successor of Carilian-Gaery and Vor Da lmont,
BOOKSELLER OF THE IMPERIAL CORPS OF BRIDGES, HIGHWAYS AND MINES,
1 darcy = 1.062 x 10-11 ft2 Quay of Augustins, 49.
1856
Darcy’s Experiment
"The Public Fountains of the Town of Dijon”, Dalmont, Paris (1856)
Permeability (k)
Medium is homogeneous if permeability does not vary with space, but

heterogeneous if non-uniform, k=k(x,y,z)
Medium is isotropic if permeability is the same in all directions,

but anisotropic if kx≠ky≠kz. Oftentimes kx ≈ ky but kx /kz≈1-10
kzz
k xx k xy k xz kx 0 0
k = k yx k yy k yz ≈ 0 ky 0
kxx
kyy
k zx k zy k zz 0 0 kz
Off-diagonal terms are only zero if we choose a coordinate system
that makes this true, but generally assume this is the case
Medium could be homogenous and anisotropic or heterogeneous and isotropic

Darcy’s Law into Continuity Equation
In 1D and horizontal flow (gravity neglected)
Replacing fluid velocity in the continuity equation
Assuming homogenous and isotropic permeability and viscosity (NOT always the
case!)
Formation Volume Factor (Bw)
We measure volume at surface but do the mass balance at the reservoir
Surface/Standard Conditions
P = 14.7 psi; T = 60 °F; V = Vsc ρ SC
ρ RC =
depth Bw
Reservoir Conditions
P = PR ; T = TR ; V = VR
Formation volume factor and ρ depend on both pressure and temperature

• Bo = formation volume factor for oil (> 1.0)
• Bg = formation volume factor for gas (<< 1.0)
• Bw = formation volume factor for water (~1.0)
Formation Volume Factor—Continuity Equation
At reservoir conditions, density is:
Replacing density in continuity equation and divide through by ρSC (a constant)
k ∂ 1 ∂p ∂ φ m
= −
µ ∂x Bw ∂x ∂t Bw ρ sc
Using the product rule on the left-hand side of the equation
k 1 ∂ 2 p ∂ 1 ∂p ∂ φ m
+ = −
µ Bw ∂x 2 ∂x Bw ∂x ∂t Bw ρ sc
And chain rule on the left hand side,
2
k 1 ∂2 p ∂ 1 ∂p ∂ φ m
+ = −
µ Bw ∂x 2 ∂p Bw ∂x ∂t Bw ρ sc
Expansion of the Time Derivative
Chain and product rule on time derivative (right hand side)

2
k 1 ∂2 p ∂ 1 ∂p ∂ 1 1 ∂φ ∂p m
+ = φ + −
µ Bw ∂x 2 ∂p Bw ∂x ∂p Bw Bw ∂p ∂t ρ sc
A few definitions:
1 ∂VP 1 ∂φ
cr = = (rock compressibility)
VP ∂p φ ∂p
1 ∂V 1 ∂ρ ∂ 1 −1 ∂Bw
cf = − = = Bw = (fluid compressibility)
V ∂p T
ρ ∂p T
∂p Bw Bw ∂p
ct = cr + c f (total compressibility)
With some manipulation:
2
k 1 ∂2 p 1 ∂ 1 ∂P φ ∂ 1 1 ∂φ ∂p m
+ B = Bw + −
µ Bw ∂x 2 Bw w ∂p Bw ∂x Bw ∂p Bw φ ∂p ∂t ρ sc
cf cf cr
LT 2 LT 2
c f ≡ fluid compressibility cr ≡ rock compressibility
M M
1D Diffusivity Equation
≈0 slightly compressible fluid
2
k 1 ∂ 2 p c f ∂p φ ct ∂p m
+ = −
µ Bw ∂x Bw ∂x
2
Bw ∂t ρ sc
If the fluid is “slightly compressible” (liquid), the compressibility is small (< 10-5)
and constant and terms involving can be ignored.
1D diffusivity (with homogenous fluid and reservoir properties) can be written:

k
Diffusivity constant ≡ α =
µφ ct
k
Mobility ≡ λ =
µ Bw
m
Source ≡ qSC =
ρ SC
If no sources or sinks (wells) are present, we get the “heat equation”
∂ 2 p 1 ∂p
=
∂x 2
α ∂t
Generalized Diffusivity (Heat) Equation
In 2D (x-y plane)
In 3D and potential Φ accounting for gravity

Slightly Compressible Fluids: Liquids
P1 P2
More Pressure
V1 V2
Recall fluid compressibility factor (cf) at constant temperature
1 ∂V 1 ∂ρ ∂ 1 −1 ∂Bw
cf = − = = Bw =
V ∂p T
ρ ∂p T
∂p Bw Bw ∂p
Integrating from a reference point (p0, ρ0), to any other point
p ρ 1
∫ p0
c f dp = ∫
ρ0 ρ
dρ
Taylor Series Expansion
1 1
f ( x + ∆x) = f ( x) + f ′( x)∆x + f ′′( x)∆x 2 + f "'( x)∆x3 + ...
2! 3!
Using Taylor series to expand density around a reference density,
∂ρ 0 1 1
ρ ( p 0 + ∆p) = ρ ( p 0 ) + ( p )∆p + ρ "( p 0 )∆p 2 + ρ "( p 0 )∆p 3 + ...
∂p 2! 3!
∂ρ 0
Differentiate exponential equation for density:
∂p
( p ) = ρ cf e
0 c f ( p0 − p0 )
= ρ 0c f
For slightly compressible (cf < 10-5 psi-1) liquids, higher order terms are small:
Negligible for small cf
1 2 0 2 1 3 0 3
ρ ( p + ∆p) = ρ + c f ρ ∆p +
0 0
c f ρ ∆p + c f ρ ∆p + ...
0
2! 3!
Therefore,
1
ρ ≈ ρ 1+ cf ( p − p )
0 0
≈ 1 + c f ( p − p0 ) Bw0=1 (assume reference is
Bw standard conditions)
Simple 1D Problem: Core Flooding
PB1 PB2
“Heat” Equation
p=pinit
∂ 2 p 1 ∂p
=
∂x 2
α ∂t
x=0 x=L
p ( x, 0 ) = pinit = 0.0 psi

p ( 0, t ) = pB1 = 1000.0 psi
p ( L, t ) = pB 2 = 0.0 psi
Analytical Solution to PDE

α (2 n +1)n π 2t
4 pinit ∞
(−1) n
(2n + 1)π x
p ( x, t ) = pB1 −
π
∑
n = 0 2n + 1
e 4 L2
cos
2L
Analytical Solution
α (2 n +1)n π 2t
4 pinit ∞
(−1)n
(2n + 1)π x
p ( x, t ) = pB1 −
π
∑
n = 0 2n + 1
e 4 L2
cos
2L
Steady state
solution
Time
increasing
Real Reservoirs
That was the easy solution
Real reservoirs have:

• Spatially varying permeability, porosity, etc.
• Time-varying viscosity, formation volume factor

• Geometries that are 2D or 3D and are not on a regular grid
• Sources and sinks (called wells) spaced irregularly throughout the reservoir
• Complex boundary conditions
k φ ct ∂p
∇ ( ∇p + ρ g ∇z ) = − qSC
µ Bw Bw ∂t
(and this is just for single phase flow )

Solving “ugly” PDE
So how do we solve this complex, nonhomogeneous, 3D PDE?
1. Break the reservoir into manageable blocks that have contain reservoir and fluid
properties
2. Write algebraic equations for each block by “discretizing” PDE

3TP1 − TP2 = Q1
−TP1 + 2TP2 − TP3 = Q2
−TP2 + 2TP3 − TP4 = Q3
−TPN −1 + TPN = QN
3. Solve system of linear equations

B B
T+ P n +1 = P n +1 + Q
∆t ∆t
CHAPTER 2.
FINITE DIFFERENCES
Solution to Partial Differential Equations (PDEs)
Analytical solutions to PDEs are rare:
α (2 n +1)n π 2t
∂ p 1 ∂p
2
4 pinit ∞
(−1)n
(2n + 1)π x
∂x 2
=
α ∂t
⇒ p ( ) B1
x , t = p −
π
∑
n = 0 2n + 1
e 4 L2
cos
2L
More generally we seek a numerical approximation to the PDE:
• Usually we transform the PDE into algebraic equations which are easier
to solve
• We “discretize” the derivatives
• Assume constant pressure and properties over an area/block
Solutions for PDEs ∂ 2 p 1 ∂p
=
∂x 2
α ∂t
Analytical Solution Numerical Solution

Need to solve a complex PDE Need to solve algebraic equations
p ( x, t ) time step
Grid block
Continuous Solution P Discrete Solution

P
x
x
Finite Difference Approximations
• Finite differences are used to approximate the numerical value

of a derivative at a point, x.
∂f
f ′( x) =
∂x x
f(x)
• Begin by using a truncated Taylor Series approximation of a

function
1 1
f ( x + ∆x) = f ( x) + f ′( x)∆x + f ′′( x)∆x 2 + f "'( x)∆x3 + ...
2! 3!
• Various finite difference schemes can be employed, with varying

degrees of accuracy. Three common methods employed here:
• Forward difference
• Backward difference
• Centered difference
Forward Difference Approximation
P(x) Tangent line
Secant line
∂P P( x + ∆x) − P( x)
≈
∂x forward ∆x
x x+∆x
x
• Derivative is defined as the “slope of the line tangent to the curve”
• This is difficult because you need two points to make a line
• Secant connects P(x) to P(x+∆x). Slope of the secant line is the “rise”
over the “run”
• Using smaller ∆x should give more accuracy to slope of the tangent

line, in fact they are the same in the limit that ∆x 0
Derivation of Forward Difference
Recall a Taylor Series:

∂P 1 ∂2 P 1 ∂3 P
P( x + ∆x) = P( x) + ∆x + ( ∆x ) 2
+ ( ∆x ) 3
+
∂x x 2! ∂x x
2
3! ∂x x
3
Re-arrange for the first derivative:
∂P P( x + ∆x) − P ( x) 1 ∂2 P 1 ∂3 P
= − ( ∆x ) + ( ∆x ) 2
+
∂x x ∆x 2! ∂x x
2
3! ∂x x
3
forward difference approximation truncation error
Forward approximation is as before; error is proportional to ∆x:
∂P P( x + ∆x) − P( x) P" P "'

≈ R= (∆x) + (∆x) 2 + = O(∆x)
∂x forward ∆x 2! 3!
Backward Difference Approximation for
Secant line
Tangent line
P(x)
∂P P( x) − P( x − ∆x)
≈
∂x backward ∆x
x-∆x x
• Secant now connects P(x) to P(x-∆x). Slope of the secant line is the
“rise” over the “run”

• Backward difference is not necessarily “better” or “worse” than forward

difference
Derivation of Backward Difference
Even terms are now negative in Taylor Series:

∂P 1 ∂2 P 1 ∂3 P
P( x − ∆x) = P( x) − ∆x + ( ∆x ) 2
− ( ∆x ) 3
+
∂x x 2! ∂x x
2
3! ∂x x
3
Re-arrange for the first derivative:
∂P P( x) − P ( x − ∆x) 1 ∂2 P 1 ∂3 P
= + ( ∆x ) − ( ∆x ) 2
+
∂x x ∆x 2! ∂x x
2
3! ∂x x
3
backward difference approximation truncation error
Error in the backward difference approximation is proportional to ∆x, just

like the forward difference:
∂P P( x) − P ( x − ∆x) P" P "'

≈ R= (∆x) − (∆x) 2 + = O(∆x)
∂x backward ∆x 2! 3!
Central Difference Approximation
Tangent
P(x) line
Secant line
∂P P ( x + ∆x) − P( x − ∆x)
≈
∂x centered 2∆x
x-∆x x x+∆x
• Secant now connects P(x+∆x) to P(x-∆x). Slope of the secant line is the
“rise” over the “run”

• Centered difference is potentially more accurate than forward or

backward difference
Derivation of Central Difference Approximation
You can subtract forward and backward Taylor series to get centered
∂P 1 ∂2 P 1 ∂3 P
P( x + ∆x) = P( x) + ∆x + ( ∆x ) 2
+ ( ∆x ) 3
+
∂x x 2! ∂x x
2
3! ∂x x
3
∂P 1 ∂2 P 1 ∂3 P
− P( x − ∆x) = P( x) − ∆x + ( ∆x ) 2
− ( ∆x ) 3
∂x x 2! ∂x 2 x 3! ∂x 3 x
∂P 2 ∂3 P
P( x + ∆x) − P( x − ∆x) = 2 ∆x + ( ∆x ) 3
+ ...
∂x x 3! ∂x x
3
Rearranging gives the centered approximation with only O(∆x2) error!
∂P P( x + ∆x) − P( x − ∆x) P "'

≈ R=− (∆x) 2 + = O(∆x 2 )
∂x centered 2∆x 3!
What about 2nd Derivative?
• Can derive a forward or backward, 2nd derivative
• Here, we’ll derive a centered difference approximation
• Add forward and backward Taylor series together
∂P 1 ∂2 P 1 ∂3 P
P( x + ∆x) = P( x) + ∆x + ( ∆x ) 2
+ ( ∆x ) 3
+
∂x x 2! ∂x 2 x 3! ∂x 3 x
∂P 1 ∂2 P 1 ∂3 P
+ P( x − ∆x) = P ( x) − ∆x + (∆x) −
2
(∆x)3 + ...
∂x x 2! ∂x x
2
3! ∂x x
3
∂2 P
P( x + ∆x) + P( x − ∆x) = 2 P( x) + 2 (∆x) 2 + ...
∂x x
Centered difference approximation is O(∆x2)

P( x + ∆x) − P( x) P( x) − P( x − ∆x)
−
∂2 P P( x + ∆x) − 2 P ( x) + P ( x − ∆x) (∆x) (∆x)
≈ =
∂x 2 centered (∆x) 2 (∆x)
Summary of Important Finite Difference
Approximations
Derivative Finite Difference Approximation Type Error

∂P P ( t + ∆t ) − P ( t ) Forward Difference O ( ∆t )
∂t ∆t
∂P P ( t ) − P ( t − ∆t ) Backward Difference O ( ∆t )
∂t ∆t
∂P P ( t + ∆t ) − P ( t − ∆t ) Centered Difference O ( ∆t 2 )
∂t 2∆t
∂2 P P ( x + ∆x ) − 2 P ( x ) + P ( x − ∆x ) Centered Difference O ( ∆x 2 )
∂x 2 ∆x 2
CHAPTER 3.
EXPLICIT, IMPLICIT, AND
CRANK-NICHOLSON
METHODS
We Want to Solve Flow Problems that are
Functions of Space and Time
P(x,t)=?
1D diffusivity equation describes this flow problem
∂ 2 p 1 ∂p
=
∂x 2
α ∂t
PDE is subject to initial and boundary conditions
P( x, 0) = Pinitial • Boundary conditions can be

• Dirichlet (constant pressure)
∂P • Neumann (constant flow)
P(0, t ) = PB1 or (0, t ) = 0
∂x • Recall that flux is proportional to
∂P pressure gradient (Darcy’s Law)
P( L, t ) = PB 2 or ( L, t ) = 0
∂x
Idea: Use Taylor Series to Approximate Derivatives
Derivative Finite Difference Approximation Type Error

∂P P ( t + ∆t ) − P ( t ) Forward Difference O ( ∆t )
∂t ∆t
∂P P ( t ) − P ( t − ∆t ) Backward Difference O ( ∆t )
∂t ∆t
∂P P ( t + ∆t ) − P ( t − ∆t ) Centered Difference O ( ∆t 2 )
∂t 2∆t
∂2 P P ( x + ∆x ) − 2 P ( x ) + P ( x − ∆x ) Centered Difference O ( ∆x 2 )
∂x 2 ∆x 2
Discretize Our Domain in N Grid-Blocks
1 2 3 i-1 i i+1 N-1 N
L
∆x =
N
• Each block has a finite size, ∆xi. Although it can vary block to block, for
now we’ll assume ∆x1=∆x2= ∆xi=∆x
• Each block has constant fluid (e.g. µ, Bw) and rock properties (k, φ)
• In a “cell-centered” approach, we solve for discrete pressures, Pi,

located at the center of every block
Substitute Finite Difference Approximation in Each
Grid Block ∂ 2 p 1 ∂p
=
∂x 2
α ∂t
Use centered difference approximation for space
∂ 2 P Pi −1 − 2 Pi + Pi +1
≈
∂x 2
(∆x) 2 Subscript indicates
block index
Pressure at block “i” depends on pressures at neighbors “i-1” and “i+1”
Use forward or backward difference approximation for time
∂P Pi n +1 − Pi n
= Superscript indicates
∂t ∆t time index
Pressure changes in “i” from time “n” to time “n+1”

Explicit versus Implicit Solutions
Explicit Method Implicit Method
Pi n +1 − Pi n 1 Pi −n1 − 2 Pi n + Pi +n1 Pi n +1 − Pi n 1 Pi −n1+1 − 2 Pi n +1 + Pi +n1+1

= =
∆t α (∆x) 2 ∆t α (∆x) 2
Spatial derivative evaluated at “n” Spatial derivative evaluated at “n+1”
time level time level
Pressure at “n+1” time level depends only Pressure at “n+1” time level depends on
on known pressures at “n” time level unknown pressures at “n+1” time level
Explicit Solution uses Pressure at Old Time Level
Called explicit method because one can solve explicitly for Pin+1
Pi n +1 − Pi n 1 Pi −n1 − 2 Pi n + Pi +n1
=
∆t α (∆x) 2
Rearranging for pressures at n+1 gives,
α∆t
Pi n +1 = Pi n + [ P n
i −1 − 2 Pi
n
+ P n
i +1 ]
(∆x) 2
An initial condition is provided,
α∆t
Pi1 = Pi 0 + [ P 0
i −1 − 2 Pi
0
+ P 0
i +1 ]
(∆x) 2
• Solve for P1 in every block i=1 to N using initial condition

• Use pressures in P1 to solve for pressures at P2
• Continue until final time step
More on Explicit Solution
α∆t
P =P +
1
1
1
0
[ P0
0
− 2 P1
0
+ P 0
2 ]
(∆x) 2
α∆t 0 In explicit solution, each

P21 = P20 + [ P1 − 2 P2
0
+ P3
0
] equation is independent (they
(∆x) 2
are coupled with the other
α∆t 0 equations)
P31 = P30 + [ P2 − 2 P3
0
+ P 0
4 ]
( ∆x ) 2
α ∆t
PN1 = PN0 + [ PN
0
−1 − 2 PN
0
+ PN
0
+1 ]
( ∆x ) 2
What to do at the boundaries?
α∆t
P =P +
1
1
1
0
[ P0
0
− 2 P1
0
+ P 0
2 ]
(∆x) 2
These grid blocks

don’t exist!!
We can use the boundary conditions defined for problem
0 1 2 3 i-1 i i+1 N-1 N N+1
x=0 x=L
∂P ∂P
P(0, t ) = PB1 or (0, t ) = 0 P ( L, t ) = PB 2 or ( L, t ) = 0
∂x ∂x
Boundary Conditions: Dirichlet (Constant Pressure)
0 1 2 3 i-1 i i+1 N-1 N N+1
P(0, t ) = PB1 P( L, t ) = PB 2
Dirichlet condition states P is known exactly at boundary. But grid pressures are
defined at center (not edge) of the block. Options?
1. Approximate pressure in block #1 (or block #N). This is only O(∆x) accurate:
P1 = PB1
2. Better approximation: edge pressure are average of two neighboring grid
pressures. O(∆x2) accurate!
P0 + P1
PB1 = ⇒ P0 = 2 PB1 − P1
2
New eqn block #1:

P1
n +1
−P 1
n
=α
( 2Pn
B1 − P1n ) − 2 P1n + P2n
∆t (∆x) 2
Boundary Conditions: Neumann (Constant/No Flux)
N N+1
Recall Darcy’s law relates flux to
pressure gradient
−k dP dP
ux = ( L, t ) = 0
µ dx dx
dP
“No flux” implies zero pressure gradient at the boundary u x ( L, t ) = 0 ⇒ ( L, t ) = 0
dx
Use a finite difference formula for pressure gradient at the boundary
∂P PN − PN +1
≈ = 0 => PN = PN +1 Often called “reflection” technique
∂x ∆x
Finite difference equation at boundary then becomes:
n +1
P −P n PNn−1 − 2 PNn + ( PNn ) PNn−1 − PNn
N
=α
N
=α
∆t (∆x) 2 (∆x) 2
Stability of Explicit Method
Explicit method is only stable (errors do not propagate over time) if
• η is dimensionless
• Might want to make ∆x small for better accuracy, but that could make it unstable
• Would have to use really small time steps (∆t) to maintain stability which is a lot
of computations
• Implicit method is unconditionally (always) stable; no restriction on size of ∆x!

Implicit Method
1 2 3 i-1 i i+1 N-1 N
In the implicit method, the spatial derivatives evaluated at the “n+1” time level
∂p ∂ p 2
Pi n +1 − Pi n Pi −n1+1 − 2 Pi n +1 + Pi +n1+1
=α 2 =α
∂t ∂x ∆t (∆x) 2
Re-arranging and grouping “like terms” gives:
α∆t n +1 α∆t n +1 α∆t

− P
i −1 + 1+ 2 Pi − Pi +n1+1 = Pi n
(∆x) 2 (∆x) 2 (∆x) 2
Pressures at n+1 are unknown known

Implicit Solution for Pressure Diffusivity Equation
Implicit Eqn:
α∆t
where: η= dimensionless
(∆x) 2
This means that the pressure of block i at the next time level n+1 depends on the
pressure at the neighbors i-1 and i+1 also at time level n+1
Need more equations because we have 3 unknowns and 1 equation per grid
block
Implicit Solution for all Grid Blocks
We can write the same equation for all blocks from i=1 to i=N
block #1 − η P0n +1 + (1 + 2η ) P1n +1 − η P2n +1 = P1n

block #2 − η P 1
n +1
+ (1 + 2η ) P 2
n +1
−η P
3
n +1
=P2
n
block #3 − η P2n +1 + (1 + 2η ) P3n +1 − η P4n +1 = P3n
block #i − η Pi −n1+1 + (1 + 2η ) Pi n +1 − η Pi +n1+1 = Pi n
block #N − η PNn−+11 + (1 + 2η ) PNn +1 − η PNn++11 = PNn
Boundary conditions can be used to eliminate two imaginary grid pressures

and then we are left N linear equations and N unknown pressures
Dirichlet Boundary Condition Revisted
0 1 2 3 i-1 i i+1 N-1 N N+1
P(0, t ) = PB1 P( L, t ) = PB 2
Approximate edge pressure as average of two neighboring grid pressures.

P0 + P1
PB1 = ⇒ P0 = 2 PB1 − P1
2
New eqn block #1:
−η P0n +1 + (1 + 2η ) P1n +1 − η P2n +1 = P1n
−η ( 2 PB1 − P1n +1 ) + (1 + 2η ) P1n +1 − η P2n +1 = P1n
(1 + 3η ) P1n+1 − η P2n+1 = P1n + 2η PB1

Neumann Boundary Conditions Revisted
N N+1
Use “reflection technique” at the

boundary
dP
∂P PN − PN +1 ( L, t ) = 0
≈ = 0 => PN = PN +1 dx
∂x ∆x
New eqn blockfor

#N:
−η PNn−+11 + (1 + 2η ) PNn +1 − η PNn++11 = PNn
−η PNn−+11 + (1 + 2η ) PNn +1 − η PNn +1 = PNn
−η PNn−+11 + (1 + η ) PNn +1 = PNn

We Can Express Equations in Matrix form
*Dirichlet on left and Neumann on right were used in this example
Dirichlet BC Dirichlet BC
1 + 3η −η 0 ... 0 P1n +1 P1n 2η PB1

−η 1 + 2η −η 0 ... ... ... ...
0 ... ... ... 0 Pi n +1 = Pi n + 0
... 0 −η 1 + 2η −η ... ... ...
0 ... 0 −η 1 + η PNn +1 PNn 0
tridiagonal matrix unknown pressures old pressures boundary conditions
Neumann BC
• PDE problem has been transformed into linear system of equations, APn+1=b
• Matrix is tridiagonal, symmetric, and diagonally dominant
• Solve N×N system of equations for each time step. Need good solver for linear
system of equations
Transmissibility Form of Problem
−η Pi −n1+1 + (1 + 2η ) Pi n +1 − η Pi +n1+1 = Pi n [ ≡ psi]
Recall definition of η and α (diffusivity constant)
α∆t k ∆t
η= =
(∆x) 2 µφ ct (∆x) 2
Multiply both sides of balance equation by a constant:
A∆xφ ct Viφ ct
=
Bw ∆t Bw ∆t
Original equation for block “i” now has units of flow rate
kA V φc kA kA V φc ft 3
− Pi −n1+1 + i t + 2 Pi n +1 − Pi +n1+1 = i t Pi n ≡
µ Bw ∆x Bw ∆t µ Bw ∆x µ Bw ∆x Bw ∆t day
Transmissibility Form
kA V φc kA kA V φc
− Pi −n1+1 + i t + 2 Pi n +1 − Pi +n1+1 = i t Pi n
µ Bw ∆x Bw ∆t µ Bw ∆x µ Bw ∆x Bw ∆t
T Bi T T Bi
Identify a few terms here:
Viφ ct Refers to the volume accumulation. How much

Bi = the fluid is expanded/contracted when pressure is
Bw increased/decreased
kA Refers to transmissibility. By Darcy’s law,

T= measures how much fluid flows in or out of the
µ Bw ∆x grid block
Bi B
−TPi −n1+1 + + 2T Pi n +1 − TPi +n1+1 = i Pi n
∆t ∆t
*Note: Don’t confuse Bw (fluid formation volume factor) with Bi (accumulation). They are different
Implicit Solution for all Grid Blocks
Consider a 4-block system with Dirichlet BC on left, Neumann on right
B1 n +1 n +1 B1 n
block #1 + 3T P1 − TP2 = P1 + 2TPB1
∆t ∆t
n +1 B2 n +1 n +1 B2 n
block #2 − TP 1 + + 2T P2 − TP3 = P2
∆t ∆t
n +1 B3 n +1 n +1 B3 n
block #3 − TP 2 + + 2T P3 − TP4 = P3
∆t ∆t
n +1 B4 n +1 B4 n
block #4 − TP 3 + + T P4 = PN
∆t ∆t
Boundary conditions can be used to eliminate two imaginary grid pressures

and we are left N linear equations and N unknown pressures
System of Equations Written in Matrix Form:
B1 0 0 0 3T −T 0 0 P1n +1 B1 0 0 0 P1n 2TPin

1 0 B2 0 0 −T 2T −T 0 P2 n +1 1 0 B2 0 0 P2 n 0
+ = +
∆t 0 0 B3 0 0 −T 2T −T P3n +1 ∆t 0 0 B3 0 P3n 0
0 0 0 B4 0 0 −T T P4 n +1 0 0 0 B4 P4 n 0
B T P n+1 B Pn Q
1 1
T+ B P n +1 = BP n + Q
∆t ∆t
• T (transmissibility) matrix is tri-diagonal for 1D problems
• B (accumulation) matrix is diagonal
• Q (source) vector includes boundary conditions and wells (discussed

later)
All Units are Now Rates
1 n +1 1
T + B P = BP + Q
n
∆t ∆t
3 3 3 3
L L L ft
= + ≡
T T T day
Explicit vs. Implicit in Transmissibility form
Explicit Method Implicit Method

Bi n +1 B Bi B
Pi = TPi −n1 + i − 2T Pi n + TPi +n1 −TPi −n1+1 + + 2T Pi n +1 − TPi +n1+1 = i Pi n
∆t ∆t ∆t ∆t
1 1 1 1
TP n + BP n +1 = BP n + Q TP n +1 + BP n +1 = BP n + Q
∆t ∆t ∆t ∆t
1 1 1 1
BP n +1 = −T + B P n + Q T+ B P n +1 = BP n + Q
∆t ∆t ∆t ∆t
−1
P n +1 = P n + ∆tB −1 ( Q − TP n ) n +1 1 1
P = T+ B BP n + Q
∆t ∆t
Both the explicit and implicit method are O(∆x2) and O(∆t) accurate. The explicit method is conditionally
stable while the implicit method is unconditionally stable
Mixed Methods: Combination of Implicit and
Explicit Methods
Mixed methods weight explicit method by 0<θ<1 and implicit by 1-θ
1 1
θ TP n + BP n +1 = BP n + Q Explicit Method
∆t ∆t
+
1 1
(1 − θ ) TP n+1 + BP n +1 = BP n + Q Implicit Method
∆t ∆t
1 1
θ TP n + (1 − θ )TP n +1 + BP n +1 = BP n + Q Mixed Method
∆t ∆t
After some rearrangement we get,
1 1
(1 − θ )T + B P =
n +1
B − θ T Pn + Q * If θ=1/2, it is called “Crank Nicholson”
∆t ∆t
Crank-Nicholson is More Accurate in Time
Both explicit and implicit methods are O(∆x2) accurate in space

because a centered difference was used but are only O(∆t) accurate in
time because forward/backward difference was used
∂P P n +1 − P n
= + O(∆t )
∂t ∆t
Equation is a forward difference for the explicit method (current time is

“n”) but a backward difference for the implicit method (current time is
n+1)
Alternatively, we could take the current time to be “n+1/2” and then the
approximation is a “centered” difference. The spatial derivative would be
a hybrid of the “n” and “n+1” time level
n+ 1 n+ 1
∂P 2
P n +1
−P n
∂ P
2 2
1 Pi −n1 − 2 Pi n + Pi +n1 1 Pi −n1+1 − 2 Pi n +1 + Pi +n1+1
= + O(∆t 2 ) ≈ + + O(∆x 2 )
∂t ∆t ∂x 2 2 (∆x) 2
2 (∆x) 2
Crank-Nicholson is second-order accurate in both time and space!!

Implicit and Analytical Solutions to 1D Diffusivity
Equation
P − PB 2 x
PD = ; xD =
Pinitial − PB 2 L
αt
tD = = 0.1
L2
N = 30; ∆t D = 2.4 E − 5
Error is Proportional to ∆x2 for Implicit Method
Implicit
Method
slope-= 2.0
1
∑(P − PDi ,numerical )
2
error = Di , analytical
N i
P − PB 2 x
PD = ; xD =
Pinitial − PB 2 L
αt
tD = = 0.1
L2
N = 40 : 300; ∆t D = 2.4 E − 7
Error is Proportional to ∆t for Implicit Method
Implicit
Method
slope-= 1.0
1
∑(P − PDi ,numerical )
2
error = Di , analytical
N i
P − PB 2 x
PD = ; xD =
Pinitial − PB 2 L
αt
tD = = 0.1
L2
N = 5000; ∆t D = 1.5 E − 6 :1.0 E − 03
Summary
• PDEs describing flow in transport in porous media can be solved

numerically using finite difference approximations
• Explicit method is relatively easy to solve because new pressures are

calculated “explicitly”, but solution is only conditionally stable (if time
steps are too large, or grids are too small it becomes unstable)
• Implicit method requires solving a system of N equations, but is

unconditionally stable. Both explicit and implicit are second order
accurate in space and first order in time
• Both the implicit and explicit equations can be written in “rate” units,
introducing transmissibility (T), accumulation (B), and source (Q)
terms
• Mixed methods are “hybrids” between implicit and explicit. Crank

Nicholson is a special mixed method (θ=1/2) that is second order
accurate in BOTH space and time
CHAPTER 4.
“CONTROL VOLUME”
APPROACH
Control Volume Approach
Goal: Write a mass balance on each grid block “i” at reservoir conditions
(RC) and attempt to recover finite difference matrix equations
mass in - mass out + injected/produced = accumulation
i-1 i i+1
ρ RC qi −1/2 ∆t − ρ RC qi +1/2 ∆t + ρ RC qi RC ∆t = mi n +1 − mi n
mass into CV mass out of CV mass injected/produced accumulation n to n+1
Control Volume Approach
ρ RC qi −1/2 ∆t − ρ RC qi +1/2 ∆t + ρ RC qi RC ∆t = mi n +1 − mi n
In the mass balance equation, mass of fluid in block “i”:
mi = ρ RCφVi (mass of fluid in pore volume times fluid density)
kA
qi −1/2 = ( Pi − Pi −1 ) = Ti −1 2 Bw ( Pi − Pi −1 ) (Darcy’s law)
µ∆x
qi RC = BwQi SC (formation factor converts from standard to reservoir conditions)
Substituting these variables in gives,
( ρ RCφ ) − ( ρ RCφ )
n +1 n
ρ RC BwTi −1/2 ( Pi − Pi −1 )∆t + ρ RC BwTi +1/2 ( Pi − Pi +1 )∆t + ρ RC BwQi SC ∆t = Vi
i
Dividing by ρ SC ∆t gives the following balance equation

n +1 n
Vi φ φ
Ti −1/2 ( Pi − Pi −1 ) + Ti +1/2 ( Pi − Pi +1 ) + Qi SC = −
∆t Bw Bw
i
Doing Some Algebra on the Right Hand Side
Expand time difference to include extra term in brackets

n +1 n n +1 n
φ φ φ φ n +1 φ n +1 φ 1 1 1
− = − n
− n
− = φ n +1 n +1
− n
+ n (φ n +1 − φ n )
Bw Bw Bw Bw Bw Bw Bw Bw Bw
these terms cancel
And recalling linear approximations for fluid and rock compression

1 1
n
≈ 0
1 + c f (Pn − P0 ) φ n ≈ φ 0 1 + cr ( P n − P 0 )
Bw Bw
Change in formation volume factor and porosity then simplify

1 1 1 1 cf
0 (
− ≈ 1 + c f ( P n +1
− P 0
) − 1 + c f ( P n
− P 0
) = P n +1 − P n )
Bw n +1 Bw n
Bw 0
Bw 0
Bw
(φ n +1
− φ n ) ≈ φ 0 1 + cr ( P n +1 − P 0 ) − φ 0 1 + cr ( P n − P 0 ) = φ 0 cr ( P n +1 − P n )
Therefore,
n +1 n
φ φ c f φ n +1 φ 0 cr
Bw
−
Bw
=
Bw 0 (P n +1
−P n
)+ B n (P n +1
− Pn )
w
And Some More Manipulation
n +1 n
φ φ c f φ n +1 φ 0 cr
Bw
−
Bw
=
Bw 0 (P n +1
−P n
)+ B n (P n +1
− Pn )
w
If Bw0 and ϕ0 are evaluated at a reference pressure,P0, near Pn then Bw0 = Bw and
φ0 =φ
n +1 n
φ φ φ φ ct
Bw
−
Bw
=
Bw
( c f + cr ) ( P n+1 − P n ) = Bw
(P n +1
− Pn )
Mass balance then becomes the same as was found from finite differences
Viφ ct n +1
Ti −1/2 ( Pi −1 − Pi ) + Ti +1/2 ( Pi +1 − Pi ) = ( Pi − Pi n ) − Qi SC
Bw ∆t
Explicit, Implicit, and Mixed Methods
Implicit Method
n +1 n +1 n +1 n +1 Viφ ct n +1
Ti −1/2 ( P
i −1 − Pi ) + Ti +1/2 ( P
i +1 − Pi )= ( Pi − Pi n ) − Qi SC
Bw ∆t
−1
1 1
P n +1 = T + B BP n + Q
∆t ∆t
Explicit Method
Viφ ct n +1
Ti −1/2 ( Pi −n1 − Pi n ) + Ti +1/2 ( Pi +n1 − Pi n ) = ( Pi − Pi n ) − Qi SC
Bw ∆t
P n +1 = P n + ∆tB −1 Q − TP n ( )
Mixed Method
1 1
(1 − θ )T + B P n +1 = B − θ T Pn + Q
∆t ∆t
CHAPTER 5.
HETEROGENEITIES
Heterogeneities in Reservoir Simulation
Real reservoirs are often not homogenous in permeability
i-1 i i+1
ki −1 ≠ ki ≠ ki +1
Consider flow from center of block “i” to “i+1” which is governed by Darcy’s Law
i i+1
i+1/2
ki + 1 A q µ Bw ∆x
q= ( P − P ) ⇒ ( Pi +1 − Pi ) =
2
µ Bw ∆x i +1 i ki + 1 A
2
Averaging Permeability: Harmonic Average
∆x/2 ∆x/2
q µ Bw ∆x q µ Bw ∆x
( Pi +1/2 − Pi ) = ( Pi +1 − Pi +1/2 ) =
2 ki A 2ki +1 A
i i+1 Pressure drop in green “half block” Pressure drop in yellow “half block”
i+1/2
Add pressure drops across green and yellow “half blocks”

q µ Bw ∆x q µ Bw ∆x
( Pi +1/2 − Pi ) + ( Pi +1 − Pi +1/2 ) = +
2 ki A 2ki +1 A
Simplifying yields:
−1
q µ Bw ∆x 1 1 A 1 1
( Pi +1 − Pi ) = + ⇒ q= 2 + ( Pi +1 − Pi )
2A ki ki +1 µ Bw ∆x ki k i+1
k
i+ 1
2
Inter-block permeability and transmissibility can then be defined as:

−1
1 1 ki + 1 A
ki + 1 =2 + ; Ti +1/2 = 2
2 ki k i+1 µ Bw ∆x
Non-Uniform Grid Size?
We may use varying grid sizes when we want more resolution in some
areas of the reservoir (e.g near wells)
ki + 1 A
Interblock transmissibility can now be defined as: Ti +1/2 = 2
µ Bw ∆xi + 1
2
Where the interblock grid size is an arithmetic means and interblock permeability
is a harmonic mean:
∆xi ∆xi +1
+
∆xi + ∆xi +1 ki + 1 = 2 2 = ∆xi + ∆xi +1
∆xi + 1 = 2 ∆xi ∆xi +1 ∆xi ∆xi +1
2 2 + +
2ki 2ki +1 ki ki +1
What if Area Changes Too?
Cross sectional area (Ai) can vary from one grid block to the next
Mass balance gives:
q µ Bw ∆xi ∆xi +1 1 ∆xi + ∆xi +1

( Pi +1 − Pi ) = + ⇒ q= ( Pi +1 − Pi )
2 ki Ai ki +1 Ai +1 µ Bw ∆xi + 1 1
+
1
2
ki Ai k i+1 Ai +1
kA
i+ 1
2
Equivalently we can say,
1 kA kA 2ki Ai ki +1 Ai +1
Ti +1/2 = =
µ Bw i+ 1
∆x i+ 1 ∆x i+ 1 ki Ai ∆xi +1 + ki +1 Ai +1∆xi
2 2 2
Fluid Properties Can Vary from Block to Block
1 kA
Ti +1/2 =
µ Bw i+ 1
∆x i+ 1
2 2
Option 1. Evaluate at some arithmetic “average” pressure between blocks

1 1
pi + 1 = ω pi + (1 − ω ) pi +1 ⇒ = @ pi + 1
2 µ Bw i+ 1
µ Bw 2
2
(V φ )i
ω = 1 2 or
(V φ )i +1
Option 2. Take arithmetic “average” of block fluid properties
1 1 1
=ω + (1 − ω )
µ Bw i+ 1
µ Bw i
µ Bw i +1
2
Option 3. Use technique called upwinding (discussed later in multiphase flow)

Change in Transmissibility Matrix
Mass balance on block “i” now includes interblock transmissibility

Viφ ct n +1
T ( Pi −1 − Pi ) + T ( Pi +1 − Pi ) = ( Pi − Pi n ) − Qi SC
Bw ∆t
⇓
Viφ ct n +1
Ti −1/2 ( Pi −1 − Pi ) + Ti +1/2 ( Pi +1 − Pi ) = ( Pi − Pi n ) − Qi SC
Bw ∆t
B matrix and Q vector is the same. T (transmissibility) matrix is still symmetric

and tri-diagonal, e.g.
T1/2 + T3/2 −T3/2 0 0

−T3/2 T3/2 + T5/2 −T5/2 0
T=
0 −T5/2 T5/2 + T7/2 −T7/2
0 0 −T7/2 T7/2 + T9/2
Boundary Conditions
For a Neumann (no flow) boundary, no flow goes through the
boundary, which is equivalent to having no interblock transmissibility
N qN + 1 = TN +1/2 ( PN − PN +1 ) = 0 ⇒ TN +1/2 = 0
2
For a Dirichlet (constant pressure) boundary, pressure is defined at the

edge, which is ∆x/2 distance away from the block center
q 1 = T1/2 ( P0 − P1 ) = 2T1 ( PB1 − P1 )

1 2
Replace flow equation between blocks with equation describing flow from
edge to center of block, all of which has permeability k1. But the distance
traveled is ∆x/2 so the effective transmissibility is 2T1
PB1
CHAPTER 6.
2D AND 3D FLOW
2D Flow problems
General diffusivity equation for 2D flow with anisotropic, heterogeneous

permeability (without gravity) is given as:
∂ k x ∂p ∂ k y ∂p φ c ∂p ~
+ = t + qsc
∂x µ Bw ∂x ∂y µ Bw ∂x Bw ∂t
Two types of 2D problems:

1. Areal view. Reservoir is flat so no gravity acting on system
(we’ll focus on these problems)
2. Side view. Gravity adds to potential gradient
Could discretize the PDE using finite differences, but here we’ll use the
“control volume” approach. Either way, we get a system of equations of
the form:
1 1
T+ B P n +1 = BP n + Q
∆t ∆t
Where T is “pentadiagonal” in 2D
Block Numbering for 2D Grids
• 2D grid systems have NX grids in the x-

direction and NY grids in the y-direction.
(Example problem: NX=3 and NY=4)
j=NY=4 10 11 12
• Label grids by i,j numbering. Block
highlighted in red is i = 2, j = 3
j=3 7 8 9
• Can also number grids using “l” numbering,
which is related to i and j. Block highlighted
in red is also block l = 8. j=2 4 5 6
j=1 1 2 3
• We’ll use both block numbering systems
interchangeably in slides that follow i=1 i=2 i=NX=3
• “i,j” numbering useful when writing mass
balance and interblock T’s
• “l” numbering useful for creating matrix
Equations for 2D Grids: Control Volume Approach
i,j+1 or l+NX Flux in/out of grid “l”
qi −1/2, j = TX i −1/2, j ( Pi −1, j − Pi , j ) = TX i −1/2, j ( Pl −1 − Pl )
qi , j +1 2 qi +1/2, j = TX i +1/2, j ( Pi +1, j − Pi , j ) = TX i +1/2, j ( Pl +1 − Pl )

qi −1/2, j qi +1/2, j
qi , j −1/2 = TYi , j −1/2 ( Pi , j −1 − Pi , j ) = TYi , j −1/2 ( Pl − NX − Pl )
i-1,j or l-1 i+1,j or l+1
qi , j −1 2 qi , j +1/2 = TYi , j +1/2 ( Pi , j +1 − Pi , j ) = TYi , j +1/2 ( Pl + NX − Pl )
Accumulation
i-1,j or l-NX Vi , j ctφi , j Vl ctφl n +1
( Pi ,nj+1 − Pi ,nj ) = ( Pl − Pl n )
Bw ∆t Bw ∆t
Mass Balance on block “i,j” or “l”

Vl ctφl n +1
TX i −1/2, j ( Pl −1 − Pl ) + TX i +1/2, j ( Pl +1 − Pl ) + TYi , j −1/2 ( Pl − NX − Pl ) + TYi , j +1/2 ( Pl + NX − Pl ) = ( Pl − Pl n ) − Ql
Bw ∆t sources
left right bottom top
accumulation
Heterogeneities and Anisotropy
• Three complications can occur in multidimensions regarding the interblock
transmissibility:
1. Heterogeneities in permeability and grid size
2. Anisotropy (kx ≠ ky)
3. Grid sizes may be different in each direction (∆x ≠ ∆y)
• Therefore, transmissibility in all 4 connecting blocks may be different, e.g.
∆xi , j + ∆xi +1, j kxi +1 2, j ( ∆yi , j h )

kxi +1 2, j = ; TX i +1 2, j =
∆xi , j ∆xi +1, j µ Bw ∆xi +1 2, j
+
kxi , j kxi +1, j
∆yi , j + ∆yi , j +1 kyi +1 2, j ( ∆xi , j h )

kyi , j +1 2 = ; TYi , j +1 2 =
∆yi , j ∆yi , j +1 µ Bw ∆yi , j +1 2
+
kyi , j kyi , j +1
• If reservoir is homogeneous, isotropic and uniform, square (∆x = ∆y)
grids employed transmissibility same in all directions
kA
TX i +1 2, j = TX i −1 2, j = TYi , j +1 2 = TYi , j −1 2 = T =
µ Bw ∆x
An Example Problem:
• 2D grid with NX=3 and NY=4

10 11 12
• “No flow” boundary conditions on 3
edges (sealed boundaries)
7 8 9
• “Constant pressure” on right edge
P=PB
(open boundary)
4 5 6
• Assume homogeneous and isotropic
permeability along with uniform,
1 2 3 square grids
“Interior” Blocks Have 4 Neighbors 10 11 12
8 7 8 9
4 5 6
Note: Block l=5 is also block
4 5 6
i=2, j=2
1 2 3
Mass balance on block #5

V5ctφ5 n +1
T ( P4 − P5 ) + T ( P6 − P5 ) + T ( P2 − P5 ) + T ( P8 − P5 ) = ( P5 − P5n ) − Q5
Bw ∆t
Grouping terms & multiplying through by “-1” gives eqn for block #5
n +1 n +1 V5ctφ5 n +1 n +1 n +1 V5ctφ5 n
TP2 + TP4 − + 4T P5 + TP6 + TP8 = P5 + Q5
Bw ∆t Bw ∆t
“No Flow” Boundary Blocks 10 11 12
7 8 9
• Block #l=11 is also i=2, j=4
j=4 10 11 12
• Neighbors to left, right, 4 5 6
bottom
• No flow BC on top
1 2 3
j=3 8
i=1 i=2 i=3
Mass Balance: There is no flow from the “top”!

V11ctφ11 n +1
T ( P10 − P11 ) + T ( P12 − P11 ) + T ( Ptop − P11 ) + T ( P8 − P11 ) = ( P11 − P11n ) − Q11
Bw ∆t
After grouping terms:
n +1 n +1 V11ctφ11 n +1 n +1 V11ctφ11 n
−TP 8 − TP
10 + 3T + P11 − TP12 = P11 + Q11
Bw ∆t Bw ∆t
“Constant” Pressure Boundary Condition 10 11 12
12 7 8 9
• Block #9 has no neighbor to its right
• But, pressure at boundary is fixed (PB)
• Q= 2T (PB-P9) 4 5 6
8 9 P=PB • “2T” appears because the boundary is only
a half ∆x away from block center
1 2 3
6
Mass Balance in “l” numbering:

V9 ctφ9 n +1
T ( P8 − P9 ) + 2T ( PB − P9 ) + T ( P6 − P9 ) + T ( P12 − P9 ) = ( P9 − P9n ) − Q9
Bw ∆t
Grouping Terms and moving known values to right hand side:
n +1 n +1 V9 ctφ9 n +1 n +1 V9 ctφ9 n
−TP6 − TP8 + 5T + P9 − TP12 = P9 + Q9 + 2TPB
Bw ∆t Bw ∆t
What about corner grids? 10 11 12
• Nothing really special about corner 7 8 9

grids
10 11
4 5 6
• Just treat each boundary separate
1 2 3
7
V10 ctφ10 n +1
T ( Pleft − P10 ) + T ( P11 − P10 ) + T ( P7 − P10 ) + T ( Ptop − P10 ) = ( P10 − P10n ) − Q10
Bw ∆t
Equation for block #10
V10 ctφ10 n +1 V cφ
−TP7n +1 + 2T + P10 − TP11n +1 = 10 t 10 P10n + Q10
Bw ∆t Bw ∆t
Final T Matrix is Pentadiagonal
2T −T −T
−T 3T −T −T
−T 4T −T
−T 3T −T −T
−T −T 4T −T −T
• Pentadiagonal, banded matrix
−T −T 5T −T
• Diagonally dominant
T= • Symmetric
−T 3T −T −T
−T −T 4T −T −T
−T −T 5T −T
−T 2T −T
−T −T 3T −T
−T −T 4T
Simple rules for homogeneous, isotropic 2D systems

• Matrix is square and the # rows= total # grid blocks
• Off diagonals are –T if that row (block) is connected to that column (block)
• Main diagonal = (# connected blocks + 2*dirichlet boundaries)T
• Blocks 5 and 8 are interior blocks and have 4 neighbors, so “4T”
• Blocks 2,4,7, and 11 have 3 neighbors and no flow boundaries, so “3T”
• Blocks 1 and 10 have 2 neighbors and no flow boundaries, so “2T”
• Blocks 6 and 9 have 3 neighbors and one dirichlet boundary, so “5T”
• Blocks 3 and 12 have 2 neighbors and one dirichlet boundary, so “4T”
System of Equations in Matrix Notation
1 1
T+ B P n +1 = BP n + Q
∆t ∆t
The B and Q arrays are listed as follows (assumes no wells)

V1ctφ1
Bw
V2ctφ2
0
Bw 0
V3ctφ3 2TPB
Bw
V4ctφ4
0
Bw 0
V5ctφ5
2TPB Constant P boundary
Bw Q= condition
V6 ctφ6 0
B=
Bw 0
V7 ctφ7
2TPB
Bw
V8ctφ8 0
Bw 0
V9 ctφ9
2TPB
Bw
V10 ctφ10
Bw
V11ctφ11
Bw
V12 ctφ12
Bw
Irregular Geometries and Inactive Grids
• Inactive grids allow for irregular geometries
• Mathematically, assume the grid still exists but assign a

permeability of zero
• This automatically gives harmonic mean of interblock

transmissibilities equal to zero
• Result is the T matrix is different in the inactive block and its

neighbors
2T −T −T
−T 3T −T −T
10 11 12
−T 4T −T
−T 2T −T 0
7 8 9 −T −T 4T −T −T
−T −T 5T −T
T=
0 0 0 0
4 5 6 −T 0 3T −T −T
−T −T 5T −T
0 T −T
1 2 3 −T −T 3T −T
−T −T 4T
Summary and Important Notes
• 2 different numbering systems are useful in 2D geometries “i,j” and “l”. They are
related:
• Mass balance is written on each block. Important to be careful with boundary

conditions. We obtain the usual system of equations
1 1
T+ B P n +1 = BP n + Q
∆t ∆t
• Matrix equations are “penta-diagonal” and symmetric. Thomas’ algorithm
CANNOT be used
• “No flow” boundary conditions are most common. Some commercial simulators
do not even have option of “constant pressure”
• “Areal” models view the reservoir from above. “Side-view” models are
sometimes used too, but then gravity must be included
• Heterogeneities, anisotropy, and non-uniform/non-square grids handled using

harmonic mean of permeability
• If reservoir geometry is not a rectangle, we can fill with “inactive” grids;

interblock transmissibility is then zero
CHAPTER 7.
WELLS AND WELL MODELS
Treatment of Wells in Reservoir Simulation
• Wells are treated as sources (injectors) or sinks (producers)

of mass
• Well is a line source or sink inside the grid block. Mass flux
for the source/sink is distributed over the whole grid block
• Well pressure is DIFFERENT than block pressure, which is

the average pressure in the block and we locate it at its
center
• Well and block pressure are linked to each other through a

well model. It can be constrained either by pressure (Pwf is
given) or by rates (qw is given).
Rate Need to calculate Pwf
Well constraints
Pressure Need to calculate rate

Radial Flow around the Well
We assume that flow occurs radially around a well. Radial

Diffusivity equation around a well is given by the following
equation; we assume flow is steady state
Steady state
φµ ct ∂P 1 ∂ ∂P
= r =0
k ∂t r ∂r ∂r
∂P q µB
BC #1: lim r =− w w
r →0 ∂r 2π kh
BC # 2 : P = Pref @ r = rref
Solution to ODE with BCs can be solved using separation of variables
P(r)
qw µ Bw r
P(r ) = Pref − ln
2π kh rref
Pressure is known (Pref) at reference radius, rref r
Radial Flow around Well
Consider the analytical solution for radial flow at steady state:
qw µ Bw r
P(r ) = Pref − ln
2π kh rref
Now, let the reference pressure be the well pressure,

Pwf. The reference radius is the well radius, rw.
rw
req
Define req as radius where pressure, P(req) equals
average block pressure, Pl. Evaluate the pressure at
the equivalent radius (req) and we get:
∆x
qw µ Bw req
Pl = Pwf − ln
2π kh rw Pl
Pwf
But how do we compute req?
rw req
Steady Mass Balance on Grids w/ Wells in 2D
Now consider steady state (no accumulation) flow in a grid, “l”. Flow “in-
out” from its 4 neighbors is equal to that injected/produced from a well
qw + ( q1 + q2 + q3 + q4 ) = 0
q3
kh∆x
q1 = ( P1 − Pl )
µ Bw ∆y
kh∆y
q2 = (P − P ) q4 q2
µ Bw ∆x 2 l
kh∆x
q3 = ( P3 − Pl )
µ Bw ∆y
q1
kh∆y
q4 = ( P4 − Pl )
µ Bw ∆x
If the block is square, ∆x=∆y, then the mass balance reduces to:
kh
qw = − ( P1 + P2 + P3 + P4 − 4 Pl )
µ Bw
Radial Flow around Well to Adjacent Blocks
Recall our radial solution for pressure around a well
qw µ Bw r
P(r ) = Pref − ln
2π kh rref
Now, let the reference pressure be the average block pressure of grid “l”, Pl. The
reference radius is the equivalent radius, req, which again is the radius where the
pressure is equal to the average block pressure rw
r eq
Evaluate the pressure in the adjacent block (for example q2 ∆x

Block #2), which is exactly ∆x away from block #1
∆x
qw µ Bw ∆x P2
P2 = Pl − ln
2π kh req Pl
Similar expressions can be made for 3 other

neighboring blocks
Pwf
rw req ∆x
Basic Well Model
Recall the equation from the mas balance:
kh
qw = − ( P1 + P2 + P3 + P4 − 4 Pl ) q3
µ Bw req
Derived equations for neighboring block

pressures: q4 q2 ∆x
qw µ Bw ∆x qw µ Bw ∆x
P1 = Pl − ln P3 = Pl − ln
2π kh req 2π kh req
q µB ∆x
rw
q µB ∆x P4 = Pl − w w ln q1
P2 = Pl − w w ln
2π kh req 2π kh req
∆x
Substituting into the mass balance gives:
kh 4
qw µ Bw ∆x
qw = −
µ Bw
∑
j =1
Pl −
2π kh
ln
req
− 4 Pl
Basic Well Model
Mass balance equation can be simplified:
kh 4
qw µ Bw ∆x kh 4qw µ Bw ∆x
qw = −
µ Bw
∑
j =1
Pl −
2π kh
ln
req
− 4 Pl = −
µ Bw
4 Pl −
2π kh
ln
req
− 4 Pl
And simplified further:
2 qw∆x
qw = ln
π req
Solving for req we get the famous “Peaceman Correction”

π
π ∆x −
= ln ⇒ req = ∆xe 2
⇒ req = 0.2078∆x ≈ 0.2∆x
2 req
*So radius at which the pressure is equal to the average grid block pressure is about 20% of the grid block
size
Wells can be constrained by rate or bottom hole
pressure
Rate Constraint
We can solve for well pressure
qw
qw µ Bw 0.2078∆x qw
Pwf = Pl + ln = Pl + w
2π kh rw Jl
Pressure Constraint
We can solve for well rate Pwf
2π kh
qw = − ( Pl − Pwf ) = − J lw ( Pl − Pwf )
0.2078∆x
µ Bw ln
rw
Basic Well Model in Simulation Equations
Well model can be written with a “productivity index” or “well index”

2π kh
qw = − J ( Pl − Pwf )
w J lw =
l req
µ Bw ln +s Skin factor
rw
Mass balance equation written implicitly in 1D:
n +1 n +1 n +1 n +1 1
T (P l −1 − Pl ) + T (P
l +1 − Pl ) = Bl ( Pl n +1 − Pl n ) − Ql
∆t
For a constant bottom hole pressure well:
1
T ( Pl −1n +1 − Pl n +1 ) + T ( Pl +1n +1 − Pl n +1 ) = Bl ( Pl n +1 − Pl n ) + J lw ( Pl n +1 − Pwf )
∆t
Forming the System of Equations
n +1 n +1 n +1 n +1 1
T (P
l −1 − Pl ) + T (P
l +1 − Pl ) = Bl ( Pl n +1 − Pl n ) − J lw ( Pwf − Pl n +1 )
∆t
Grouping like terms of Pressure and multiplying through by “-1”
1 1
−TPl −n1+1 + (2T + J lw + Bl ) Pl n +1 − TPl +n1+1 = Bl Pl n + J lw Pwf
∆t ∆t
We get a system of equations of the usual form, but with a new diagonal J matrix
1 1
T+J+ B P n +1 = BP n + Q
∆t ∆t
Diagonal Matrix Includes J*Pwf terms + constant

rate wells + BCs
Matrix Equations for Constant BHP Well
1 1
T+J+ B P n +1 = BP n + Q
∆t ∆t
Consider a 1D, 4-block homogeneous problem with no-flow BCs a constant BHP
well in block #3
T −T 0 B1 0
−T 2T −T 0 B2 0
T= J= B= Q=
−T 2T −T J3 B3 J 3 Pwf
−T T 0 B4 0
• B and T matrix are no different now that wells are present
• J is a diagonal matrix and diagonal elements are non-zero if a BHP well exists
in that block. In which case Ji is always positive
• Q is source vector; it includes a positive (J*Pwf) regardless of whether we have

an injector or producer BHP well. Note: This is different from constant rate wells
where the sign depends on producer or injector
Anisotropy and Rectangular Grids
If the porous medium is anisotropic and/or has rectangular grids:
1
(k k x ) ∆x + ( k x k y ) ∆y
1 1 2
2 2 2 2
2π h k x k y y
J =
w
; req = 0.28
( k y kx ) 4 + ( kx k y )
l 1 1
req 4
µ Bw ln +s
rw
kx≠ky
• Numerator of productivity index includes

geometric mean of permeability
∆y
• Equivalent radius has ratios of
permeability and includes rectangular
∆x
grid sizes
• Note that the expressions reduce to the

simpler case when medium is isotropic
and grids are square
Horizontal Wells
Horizontal wells have the following issues:
1. Much more likely anisotropic (kz ≠ kx or ky)
2. Well traverses horizontally, so dimensions changed
3. Well passes through multiple grids in x-y plane, so J
matrix has entries in each grid
2π∆x k y k z
J lw = ;
req
µ Bw ln +s 0
rw
0
1
(k k y ) 2 ∆y 2 + ( k y k z ) 2 ∆z 2
1 1 2
0
z
req = 0.28 J4
(k k y ) 4 + ( k y kz )
1 1
4 J= J5
z
J6
0
• Assumes that well is in x-direction, not
y-direction 0
• ∆z = h, the reservoir thickness in 2D 0
Well traverses through blocks 4,5, and 6

Ch 8. Multiphase Flow
in Porous Media
Review of Numerical Simulation of Single Phase Flow
1. Derived continuity and diffusivity equations
2. Derived finite difference approximations for derivatives
3. Discretized PDE to obtain algebraic equations
4. Included reservoir heterogeneities in model
5. Extended to multidimensions
6. Added wells (constant rate and BHP) into model

Multiphase Flow
Applications: water flooding, aquifer support, saturated reservoirs, gas

flooding, and more!
Modeling Approaches:
1. Compositional: Models several
individual components. Accurate but very Component 1
slow and computationally demanding
• Component : any identifiable chemical
entity
• Phase : a homogeneous region of matter
Single Phase
• Can have gas component in the oil phase! Region
• Balance on the component, not the phase
Two Phase
2. Black Oil: Models 3 components (oil, Region
gas, and water) in 3 phases: Oil (oileic), Component 2 Component 3
Water (aqueous), and Gas (gaseous)
Black Oil Models (or β-model)
Gas component in oil or flowing separately
Gas does not dissolve in water
Water and oil do not mix

Mass Balance of Water
Balance of water component:
mw − mw A∆t + mw, well ∆V ∆t = ( mw t +∆t − mw t )
x x +∆x
mass in from well mass accumulated
mass in-out water in aqueous phase water in aqueous phase
water in aqueous phase
mass of water component

Recognizing: m w = ρ wu w mw = ρ w S wφ∆V ρw ≡
volume aqueous phase
(ρ u
w w x − ρ wu w x +∆x ) A∆t + mw, well ∆V ∆t = ( ρ w S wφ∆V t +∆t
− ρ w S wφ∆V t )
mass in-out mass in from well mass accumulated
water in aqueous phase water in aqueous phase water in aqueous phase
Divide by A∆x∆t and taking limits gives differential mass balance:
∂ ( ρ wu w ) ∂ ( ρ wφ S w )
− + mw, well =
∂x ∂t
Water Formation Volume Factor (Bw)
We measure volume at surface but do the mass balance at the reservoir
Surface/Standard Conditions
P = 14.7 psi; T = 60 °F; V = Vsc
depth
Reservoir Conditions
P = PR ; T = TR ; V = VR
[Vw ]
RC
reservoir volume of aqueous phase ρ w SC
Bw = = = RC
[Vw ] ρw
SC
std. volume aqueous phase
Note: Bw (formation volume factor for water) ~1.0

Mass Balance of Water
∂ ( ρ wu w ) ∂ ( ρ wφ S w )
− + mw, well =
∂x ∂t
Divide by density at standard conditions, ρwsc, and use our definition of

formation volume factor of water
ρ w SC
Bw = RC
ρw
We get balance equation in terms of formation volume factor
∂ uw ∂ φ Sw
− + qw =
∂x Bw ∂t Bw
Mass Balance of Oil
Balance of oil component:
mo − mo A∆t + mo , well ∆V ∆t = ( mo t +∆t − mo t )

x x +∆x
mass in from well mass accumulated
mass in-out oil in oileic phase oil in oileic phase
oil in oileic phase
mass of oil component

Recognizing: m o = ρ o uo mo = ρo Soφ∆V ρo ≡
volume oleic phase
(ρ u
o o x − ρ o uo x +∆x ) A∆t + m o , well ∆V ∆t = ( ρo Soφ∆V t +∆t
− ρo Soφ∆V t )
mass in-out mass in from well mass accumulated
oil in oileic phase oil in oileic phase oil in oileic phase
∂ ( ρ o uo ) ∂ ( ρoφ So )
Divide by A∆x∆t and taking limits: − + mo , well =
∂x ∂t
Formation Volume Factors and Solubility Ratios
Fig. 4-10, Lake and Walsh (2003)

Oil Formation Volume Factor (Bo)
Reservoir Surface
(high pressure) (low pressure)
Oil formation volume factor is defined as:
res. vol. of oleic (oil + dissolved gas) phase VoRC RB oleic
Bo = = SC [=]
std. volume oleic phase ( oil only ) Vo STB oil
Mass of oleic phase at RC is mass of liquid + mass of dissolved gas at SC
mass oileic RC =mass oilSC + mass gasSC

ρoleic
RC RC
Voleic =ρoSC VoSC + ρ gSC VdgSC
Divide through by volume of oleic phase we get:
VoSC V SC
V SC
ρ SC
+ R ρ SC
ρ SC
R ρ SC
ρolRCeic =ρoSC RC
+ ρ gSC RC ⋅ oSC =
dg o s g
= o + s g
Voleic Voleic Vo Bo Bo Bo
ρoRC RC
ρ dg
Note: Bo>1(usually, 1.0<Bo<2.0) because dissolved gas comes out of solution

Mass Balance of Oil
∂ ( ρ o uo ) ∂ ( ρoφ So )
− + mo , well =
∂x ∂t
Divide by density at standard conditions, ρosc, and use our definition of

formation volume factor of oil
ρo SC
Bo = RC
ρo
We get balance equation in terms of formation volume factor
∂ uo ∂ φ So
− + qo =
∂x Bo ∂t Bo
Mass Balance of Gas
Balance of gas component:
mg − mg A∆t + m dg − m dg A∆t + mg , well ∆V ∆t + mdg , well ∆V ∆t

x x +∆x x x +∆x
mass in from well mass in from well
mass in-out mass in-out gas in gaseous phase dissolved gas in oil phase
gas in gaseous phase dissolved gas in oileic phase
(
= mg
t +∆t
− mg
t
) + (m dg t +∆t − mdg
t
)
mass accumulated mass accumulated
gas in gaseos phase dissolved gas in oil phase
mass of gas component

m g = ρ g ug mg = ρ g S gφ∆V ρg ≡
Recognizing: volume gaseous phase
mdg = ρ dg Soφ∆V mass of dissolved gas
m dg = ρ dg uo ρ dg ≡
volume oleic phase
(ρ u
g g x − ρ g ug
x +∆x
) A∆t + ( ρ u
dg o x − ρ dg uo
x +∆x
) A∆t + mg , well ∆V ∆t + mdg , well ∆V ∆t =
mass in-out mass in-out mass in from well mass in from well
gas in gaseous phase dissolved gas in oileic phase gas in gaseous phase dissolved gas in oil phase
( ρ S φ∆V
g g t +∆t
− ρ g S gφ∆V
t
) + (ρ dg Soφ∆V
t +∆t
− ρ dg Soφ∆V
t
)
Mass Balance of Gas
(ρ u
g g x − ρ g ug
x +∆x
) A∆t + ( ρ u
dg o x − ρ dg uo
x +∆x
) A∆t + mg , well ∆V ∆t + mdg , well ∆V ∆t =
mass in-out mass in-out mass in from well mass in from well
gas in gaseous phase dissolved gas in oileic phase gas in gaseous phase dissolved gas in oil phase
( ρ S φ∆V
g g t +∆t
− ρ g S gφ∆V
t
) + (ρ dg Soφ∆V
t +∆t
− ρ dg Soφ∆V
t
)
Divide by A∆x∆t and taking limits:
∂ ( ρ g u g + ρ dg uo ) ∂
−
∂x
+ mg , well + mdg , well =
∂t
( ρ gφ S g + ρ dgφ So )
Solution Gas-Oil Ratio (Rs)
Solution Gas-Oil Ratio is volume of dissolved gas at surface conditions

divided by the volume of oil at surface conditions
Gas Formation Volume Factor (Bg)
Definition of gas formation volume factor:
RC
reservoir volume of gaseous (gas + volatile oil) phase Vg ft 3
Bg = = [=]
std. volume gaseous phase ( gas only ) Vg
SC
scf
Mass of gaseous phase is mass of gas + liquids at SC:
mass gaseous RC = mass gasSC + mass oilSC

ρ gRC VgRC =ρ gSC VgSC + ρoSC VvoSC
Divide through by volume of gaseous phase at reservoir conditions
VgSC SC
VvoSC Vg ρ gSC + Rv ρoSC
ρ gRC =ρ gSC + ρoSC ⋅ SC =
VgRC RC
Vg Vg Bg
If no liquid comes out of the gaseous phase, Rv = 0 (Black Oil Model)
ρ gSC
ρ gRC =
Bg
Mass Balance of Gas
∂ ( ρ g u g + ρ dg uo ) ∂
− + mg , well + mdg , well = ( ρ gφ S g + ρ dgφ So )
∂x ∂t
Divide by density (ρgsc) at standard conditions and recall:
ρ gSC Rs ρ gSC
ρg = ρ dg =
Bg Bo
We get:
∂ ug u ∂ 1 R
− + Rs o + qg , well + Rs qo , well = φ S g + s So
∂x Bg Bo ∂t Bg Bo
Summary of 3 Phase Balance Equations
∂ uw ∂ φ Sw
water: − + qw =
∂x Bw ∂t Bw
∂ uo ∂ φ So
oil: − + qo =
∂x Bo ∂t Bo
∂ ug uo ∂ 1 Rs
gas: − + Rs + qg + Rs qo = φ S g + So
∂x Bg Bo ∂t Bg Bo
Darcy’s Law for Multiphase Flow
−kkrw ∂pw ∂z
uw = − ρw g
µw ∂x ∂x
−kkro ∂po ∂z
uo = − ρo g
µo ∂x ∂x
−kkrg ∂pg ∂z
ug = − ρg g
µg ∂x ∂x
ρ w g ≈ 0.433 psi/ft ( water )
u = velocity [L/time] k = perm [L2] D = depth [L] µ = viscosity [M/L/time] g = gravity [L/time2]
Pressure Dependent Fluid Properties
Brooks-Corey Relative Permeability Model
𝑘𝑘𝑟𝑟𝑟𝑟𝑟𝑟 𝑘𝑘𝑟𝑟𝑟𝑟𝑟𝑟
𝑘𝑘𝑟𝑟𝑟𝑟
𝑘𝑘𝑟𝑟𝑟𝑟
Brooks-Corey empirical models

nw no ng
S w − S wr 1 − So − Sor S g − S gc
krw = krw
0
; kro = krow krog = kro0 ; krg = krg0
1 − Sor − S wr − S gc 1 − Sor − S wr − S gc 1 − Sor − S wr − S gc
Capillary Pressure Curves
•
Pc ( S w ) = Po − Pw
• Most simulators include Pc

• Small (at the reservoir scale)
compared to the reservoir
pressure
• We will assume Pc =0
• Van Genuchten empirical model
1
1 −1 n
Pc = S m
−1
α
(Aziz and Settari, 1979)
Note: “n” and “m” parameters are different than in Corey-
Brooks relative permeability model
Substituting Darcy’s Law into Mass Balance:
PDEs (without gravity) describing 3-phase flow in porous media
∂ φ Sw ∂ 1 kkrw ∂Pw
= + qw
∂t Bw ∂x Bw µ w ∂x
∂ φ So ∂ 1 kkro ∂Po
= + qo
∂t Bo ∂x Bo µo ∂x
∂ 1 R ∂ 1 kkrg ∂Pg 1 kkro ∂Po

φ S g + s So = + Rs + qg + Rs qo
∂t Bg Bo ∂x Bg µ g ∂x Bo µo ∂x
Auxiliary Equations Required

krw = f ( S w )
Pc ,ow ( S w ) = Po − Pw
S w + So + S g = 1 kro = f ( S w , S g )
Pc ,og ( S w ) = Pg − Po
krg = f ( S g )
Rarely can these coupled PDEs be solved analytically. An exception is the Buckley-Leverett problem
S = saturation k = perm [L2] B = FVF µ = viscosity [M/L/time] ρ = density [M/V]

Buckley-Leverett Theory for 1D Displacement
Semi-analytical solution saturation equation in 1D
Useful for benchmarking reservoir simulation (numerical) solutions
Assumptions
1. 2-phase flow only (oil and water)
2. 1D flow
3. Incompressible fluids
4. no capillary pressure/gravity,
5. no sources or sinks
Initial and Boundary conditions

Core initially saturated with water at residual water saturation (Swi=Swr)
Constant injection rate (q) at x=0
Constant production rate (q) at x=L
Water
Water +
Oil
Buckley-Leverett Theory for 1D Displacement
Swf = saturation at water front
Injected Water “Shock Front”
Initial Water
Water front displaces as a shock wave

Buckley-Leverett Equations
Recall mass balance equation for water:

No sources/sinks
∂ φ Sw ∂ uw
=− + qw
∂t Bw ∂x Bw
Since rock and fluids are incompressible, porosity and FVF are constant
∂S w ∂u
φ =− w
∂t ∂x
Let fw be fractional flow, qw be water rate, and q the total rate (which is constant):
qw qf w
uw = =
A A
Mass balance equation for water can be simplified:
∂S w q ∂f w
φ =−
∂t A ∂x
φ = porosity Sw= wat. sat. B = vol. factor [rB/STB] x = dist. [ft] u = velocity [ft/day] fw=fract. flow A=Area [ft2] q = flow rate [ft3/day]
Buckley-Leverett Equations
Using chain rule, the mass balance equation becomes:
∂S w q ∂f w ∂S w
=−
∂t Aφ ∂S w ∂x
Note that Sw = f(x,t), so the partial (dSw)can be expanded as:
∂S w ∂S
dS w = dx + w dt
∂x ∂t
Search for velocity front of constant saturation (i.e. dSw = 0). Setting the
above expression to zero we get:
∂x ( ∂Sw / ∂t ) Sw q ∂f w
u= =− =
∂t Sw ( ∂S w / ∂x ) S Aφ ∂S w w
Velocity of constant-saturation front is proportional to the derivative of the

fractional flow curve:
∂x q ∂f w
=
∂t Sw Aφ ∂S w
φ = porosity Sw= wat. sat. t = time x = dist. [ft] u = velocity [ft/day] fw=fract. flow A=Area [ft2] q = flow rate [ft3/day]
Dimensionless Variables
It is convenient to introduce some dimensionless variables .
xD = x / L ≡ dimensionless distance
qt pore volumes injected
tD = ≡
φ AL pore volume of core
Dimensionless velocity front can then be written as:
∂x q ∂f w ∂xD ∂f w
= =
∂t Sw Aφ ∂S w ∂t D Sw
∂S w
Integrating and recognizing that xD =0 at tD =0
∂f w
xD ( S w ) = t D
∂S w Sw
φ = porosity Sw= wat. sat. t = time x = dist. [ft] u = velocity [ft/day] fw=fract. flow A=Area [ft2] q = flow rate [ft3/day]
Fractional flow curve is function of Sw through
relative permeability
krw
qw µw
f w ( Sw ) = =
qo + qw krw +
kro
µw µo
∂f w
∂S w Sw
Steps for Generating Saturation Profile at
Snapshot in time (tD)
qt
1. Choose a time/dimensionless time to make the plot t D =
φ AL
2. Create fractional flow curve (fw vs. Sw) using relative permeability curves
krw
µw
f w ( Sw ) =
krw kro
µw + µo
𝜕𝜕𝑓𝑓
3. At a given Sw (e.g. 1-Sor), compute 𝜕𝜕𝑆𝑆𝑤𝑤 |𝑆𝑆𝑤𝑤 , numerically (slope of tangent to
𝑤𝑤
the fractional flow curve)
4. Compute xD at that Sw and tD. Plot point on Sw vs. xD curve
∂f w
xD ( S w ) = t D
∂S w Sw
5. Repeat for all Sw (Swr < Sw < 1-Sor) to make an “Sw versus xD” curve
Resulting saturation plot is non-physical; at a given
distance in the core, more than one saturation exists
Swf
xDf
In realty, a shock front occurs at some xDf at snapshot in time, tD. Saturation has a
“step change” from Swf to Swi at that distance xDf. But how do we determine xDf?
Fractional Flow Curve has an inflection point. As a
result, the derivative goes through a maximum
∂f w
∂S w S
w
Obtaining a “Physical” Solution and determine Swf
Mass balance on water in the core where Sw=Swf
ρ w qw x ∆t − ρ w qw x +∆x ∆t = ρ w Aφ∆xS w t +∆t − ρ w Aφ∆xS w t

Swi
mass into CV mass out of CV accumulation
xDf, Swf
We know from shock front that Sw goes from Swi to Swf and therefore fw goes
from fwi to fwf. Mass balance becomes:
q ( f wf − f wi ) ∆t = Aφ ( S wf − S wi ) ∆x
Using some algebra and using dimensionless variables:
∆x u ( f wf − f wi ) ∆xD f wf
= ⇒ =
∆t φ ( S wf − S wi ) ∆t D ( S wf − S wi )
Obtaining a “Physical” Solution Swi
xDf, Swf
We have two equations describing velocity at “shock front” and they must
be equal for mass conservation
∆xD f wf ∂xD ∂f w
= =
∆t D ( S wf − S wi ) ∂t D S wf
∂S w S wf
Equation suggests a secant line can be drawn connecting two points on a fw

versus Sw plot; (Swi, fwi) and (Swf, fwf)
∂f w
f wf =
∂S w
(S wf − S wi )
S wf
So right at the shock front (xD=xDf, Sw=Swf), the slope of the line tangent to the
fractional flow curve must equal the line intersecting (Swi, fwi); i.e. the point of
interest (Swf, fwf)
krw ( S w )
qw µw
Fractional Flow Curve f w ( Sw ) = =
qw + qo krw ( S w ) kro ( S w )
+
µw µo
∂f w f wf
=
∂S w S wf
S wf − S wi
Secant line connecting Swr to

the fw curve gives the value
of Swf
Swf
Steps for Generating Saturation Profile
1. Compute Swf by using secant method on fractional flow curve
𝜕𝜕𝑓𝑓
2. For all Sw > Swf, compute 𝜕𝜕𝑆𝑆𝑤𝑤 |𝑆𝑆𝑤𝑤 , numerically
𝑤𝑤
3. Compute dimensionless distance, xD at each Sw > Swf
4. Plot Sw versus xD using points obtained from #3 and use Sw=Swi for
xD>xDf
Water Saturation Profile from BL Theory
∂f w
xD ( S w ) = t d if S w > S wf
Swf ∂S w Sw
S w = S wi otherwise
xDf
Steps in Reservoir Initialization (Pw, Po, Sw)
1. Identify the water pressure at a reference point, e.g. the WOC
Pw = Pw,WOC @ z = WOC
2. At the WOC, Pc = PD( the “capillary entry pressure”):
Po = Pw,WOC + PD @ z = WOC
3. Compute Po and Pw at various depths (grid block centers) using the
hydrostatic head
Poi = Po ,WOC + ρo g ( z i − zWOC )
Pwi = Pw,WOC + ρ w g ( z i − zWOC )
1. Compute Pc at all depths: Pci = Poi − Pwi

2. Use the computed Pc’s and the Pc curve to compute saturations at each
depth
Pci = f ( S wi )
Initial Conditions for Multiphase Flow
Gas cap (oil/water at residual sat, only gas is mobile)
Gas-oil transition zone (Water at residual sat, gas and oil depend on capillary
pressure. P<Pb)
Oil zone (Water at residual sat, gas in solution P>Pb, Only oil is
movable)
Water-oil zone (Water and oil are movable, saturation depends on capillary pressure)
Water zone (Below water-oil contact)
• Goal: determine Pw, Po and Sw as a function of depth

• Oil migrated to rock displacing water. This is drainage
• Pc @ WOC determined from Pc drainage curve. Called displacement pressure Pd
• Starting point is the water-oil contact (WOC); Sw=1
• No oil at or below WOC
• Cap pressure calculated from density difference
Pc Found Using Density Differences
1. Identify Pw at a reference
Pc
ow
= Po − Pw = ∆ρow gz point, e.g. the WOC
2. At the WOC, Pc = Pd
(“capillary entry pressure”)
Oil
= ρ gz
PPressure
o o
3. Compute Po and Pw at various
depths (grid block centers)
using the hydrostatic head
WOC Pd Poi = Po ,WOC + ρo g ( z i − zWOC )
Water Pressure Pwi = Pw,WOC + ρ w g ( z i − zWOC )
Pw = ρ w gz
4. Compute Pc at all depths:
z
Pci = Poi − Pwi
Pw Po P
Pc Found Using Density Differences
Pcow
,i + 2
Pcow
,i +1 Pd
S w ,i + 2 S w,i +1
WOC
5. Use the computed Pc’s

Pd and the Pc curve to compute
z saturations at each depth
P
Pci = f ( S wi )
Saturation Profile with Depth
Water Saturation Profile Saturation in

grid-blocks
z
Get z
Compute Pc
Read Sw
From Pc
table
WOC
Swir 1 Sw
Steps in Reservoir Initialization (Pw, Po, Sw)
1. Identify the water pressure at a reference point, e.g. the WOC
2. At the WOC, Pc = PD( the “capillary entry pressure”):
3. Compute Po and Pw at various depths (grid block centers) using the
hydrostatic head
Poi = Po ,WOC + ρo g ( z i − zWOC )
Pwi = Pw,WOC + ρ w g ( z i − zWOC )
4. Compute Pc at all depths: Pci = Poi − Pwi

5. Use the computed Pc’s and the Pc curve to compute saturations at each
depth
Pci = f ( S wi )
Chapter 9:
1D IMPES for Multiphase Flow
(no capillary pressure, gravity)
Review: 2 Phase Flow: Oil/Water (No Gravity or Pc)
∂ φ So ∂ kkro ∂p
OIL: = + qo
∂t Bo ∂x µo Bo ∂x
∂ φ Sw ∂ kkrw ∂p
WATER: = + qw
∂t Bw ∂x µ w Bw ∂x
Expand the time derivative (No Gravity or Pc)
∂ φ So ∂ kkro ∂p
OIL: = + qo
∂t Bo ∂x µo Bo ∂x
Expand the time derivative using the product rule (Oil Equation)
∂ φ So φ ∂ ∂ 1 S ∂
= ( So ) + φ So + o (φ )
∂t Bo Bo ∂t ∂t Bo Bo ∂t
Chain rule to get derivative in terms of pressure (Oil Equation)
∂ φ So φ ∂So ∂ 1 S ∂φ ∂p
= + φ So + o
∂t Bo Bo ∂t ∂p Bo Bo ∂p ∂t
Recall definition of compressibilities:
∂ 1 ∂ 1 1 ∂φ
co = Bo ; cw = Bw ; cr = ;
∂p Bo ∂p Bw φ ∂p
Oil and water balance equations become
φ ∂So φ So φ So ∂p ∂ kkro ∂p
Oil: + co + cr = + qo
Bo ∂t Bo Bo ∂t ∂x µo Bo ∂x
Water: φ ∂S w φ Sw φ Sw ∂p ∂ kkrw ∂p
+ cw + cr = + qw
Bw ∂t Bw Bw ∂t ∂x µ w Bw ∂x
Determining Component and Overall Balances:
Multiplying the “oil” equation by Bo/Bw and add mass balances become:
φ ∂S w φ Sw φ Sw ∂p ∂ kkrw ∂p
Water: + cw + cr = + qw
+
Bo φ ∂So Bo φ So φS ∂p Bo ∂ kkro ∂p B
Oil: + co + o cr = + o qo
Bw Bo ∂t Bw Bo Bo ∂t Bw ∂x µo Bo ∂x Bw
=
Overall (total) balance:
φ ∂So ∂S w φ So φS φS φS ∂p Bo ∂ kkro ∂p ∂ kkrw ∂p B
+ + co + w cw + o cR + w cR = + + o qo + qw
Bw ∂t ∂t Bw Bw Bw Bw ∂t Bw ∂x µo Bo ∂x ∂x µ w Bw ∂x Bw
terms cancel φ ct
Bw
∂So ∂S w
(1) + = 0 since Sw +So =1
∂t ∂t
(2) define: ct = So co + S wcw + cr
Summary of oil, water, and “overall” balance
φ ∂So φ So φ So ∂p ∂ kkro ∂p
Oil: + co + cr = + qo
Bo ∂t Bo Bo ∂t ∂x µo Bo ∂x
Water: φ ∂S w φ Sw φ Sw ∂p ∂ kkrw ∂p
+ cw + cr = + qw
φ ct ∂pBo ∂ kkro ∂p ∂ kkrw ∂p Bo

Overall: = + + qo + qw
Bw ∂t Bw ∂x µo Bo ∂x ∂x µ w Bw ∂x Bw
Note: We can use any 2 of the above 3 equations to define the problem. Usually
we will use the overall balance with one component balance (e.g. water)
Focusing on the “Overall” Balance (Pressure Eqn)
Bo ∂ kkro ∂p ∂ kkrw ∂p φ c ∂p B
Overall: + = t − o qo − qw
Bw ∂x µo Bo ∂x ∂x µ w Bw ∂x Bw ∂t Bw
Using Finite Differences and the “implicit” method
Bo ∂ kkro ∂p
=
1 Bo
λo ,i −1 2
(P n +1
i −1 − Pi n +1 )
+ λo ,i +1 2
(P n +1
i +1 − Pi n +1 )
Bw ∂x µo Bo ∂x ∆xi Bw ∆xi −1 2 ∆xi +1 2
∂ kkrw ∂p
=
1
λw,i −1 2
( Pi −n1+1 − Pi n +1 )
+ λw,i +1 2
( Pi +n1+1 − Pi n +1 )
∂x µ w Bw ∂x ∆xi ∆xi −1 2 ∆xi +1 2
φ ct ∂p φ ct Pi n +1 − Pi n
=
Bw ∂t Bw ∆t
Where the mobility ratio is defined as:
ki −1 2 kro ,i −1 2
λo ,i −1 2 =
µo Bo
Focusing on the “Overall” Balance (Pressure Eqn)
Bo ∂ kkro ∂p ∂ kkrw ∂p φ c ∂p B
Overall: + = t − o qo − qw
Bw ∂x µo Bo ∂x ∂x µ w Bw ∂x Bw ∂t Bw
Using Finite Differences and the “implicit” method
1 Bo
λo ,i −1 2
( Pi −n1+1 − Pi n +1 )
+ λo ,i +1 2
( Pi +n1+1 − Pi n +1 )
+
1
λw,i −1 2
( Pi −n1+1 − Pi n +1 )
+ λw,i +1 2
( Pi +n1+1 − Pi n +1 )
∆xi Bw ∆xi −1 2 ∆xi +1 2 ∆xi ∆xi −1 2 ∆xi +1 2
φ ct Pi n +1 − Pi n Bo
= − qo ,i − qw,i
Bw ∆t Bw i
Multiplying through by Vi=A∆xi , and lumping terms gives the pressure equation
Bo Bo Viφ ct Pi n +1 − Pi n Bo
Bw
To ,i −1 2 + Tw,i −1 2 (P n +1
i −1 − Pi n +1
) B To,i+1 2 + Tw,i+1 2
+ (P n +1
i +1 − Pi n +1
) B
=
∆t
−
Bw
Qo ,i − Qw,i
w w i
Now Consider “Water” Balance (Saturation Eqn)
φ ∂S w φ Sw φ Sw ∂p ∂ kkrw ∂p
Water: + cw + cr = + qw
Using Finite Differences:

n +1
φi S w,i − S w,i
n
φi S wn ,i ( cw + cr ) Pi n +1 − Pi n 1 Pi −n1+1 − Pi n +1 Pi +n1+1 − Pi n +1
+ = λw,i −1 2 + λw,i +1 2 + qw , i
Bw,i ∆t Bw,i ∆t ∆xi ∆xi −1 2 ∆xi +1 2
Multiplying through by Bwi∆t/φ, and lumping terms gives the Saturation Equation
Bw,i ∆t
S wn ,+i1 = S wn ,i + Tw,i −1 2 ( Pi −n1+1 − Pi n +1 ) + Tw,i +1 2 ( Pi +n1+1 − Pi n +1 ) + Qw,i − S wn ,i ( cw + cr ) ( Pi n +1 − Pi n )
Viφi
IMPES: IMplicit in Pressure, Explicit in Saturation
For each time step:
1. Solve pressure equation (a system of linear equations) implicitly
Bo Viφ ct Pi n +1 − Pi n
Bw
To ,i −1 2 + Tw,i −1 2 (P n +1
i −1 − Pi n +1
) + BBo To,i+1 2 + Tw,i+1 2 (P n +1
i +1 − Pi n +1
)= B ∆t
B
− o
Bw
Qo ,i − Qw,i
w w i
B B
or in Matrix Form: T + P n +1 = P n + Q
∆t ∆t
2. Using pressure solution for n+1, solve for saturation explicitly for n+1
Bw,i ∆t
S n +1
=S n
+ Tw,i −1 2 ( Pi −n1+1 − Pi n +1 ) + Tw,i +1 2 ( Pi +n1+1 − Pi n +1 ) + Qw,i − S wn ,i ( cw + cr ) ( Pi n +1 − Pi n )
Viφi
w ,i w ,i
or in Matrix Form: S nw+1 = S nw + d12

−1
−Tw P n +1 + Q w − Ctw ( P n +1 − P n )
Viφi
Ctw,ii = S wn ,i ( cw,i + cr ,i ) ; d12,ii =
Bw,i ∆t
*** Since the solution is partially explicit it is only conditionally stable!!!
S=Saturation Bo=Volumetric Factor [rB/STB] k=perm [mD] q=rate [ft3/day] A=Area [ft2] V=Volume[ft3] T=Transmissibiltiy [(ft3/day.psi)]
B B
Pressure Matrix Eqn (Implicit) T+ P n +1 = P n + Q
∆t ∆t
Total Transmissibility: sum of the water and oil transmissibility
Bo
T = Tw + To
Bw
Tw,1 2 + Tw,3 2 −Tw,3 2 0 0 To,1 2 + To,3 2 −To,3 2 0 0
−Tw,3 2 Tw,3 2 + Tw,5 2 −Tw,5 2 0 −To,3 2 To,3 2 + To,5 2 −To,5 2 0
Tw = To =
0 −Tw,5 2 Tw,5 2 + Tw,7 2 −Tw,7 2 0 −To,5 2 To ,5 2 + To ,7 2 −To ,7 2
0 0 −Tw,7 2 Tw,7 2 + Tw,9 2 0 0 −To ,7 2 To ,7 2 + To ,9 2
ki −1 2 krw,i −1/2 A ki −1 2 kro ,i −1/2 A

Tw,i −1/2 = To ,i −1/2 =
∆xi −1 2 µ w Bw ∆xi −1 2 µo Bo
Accumulation and Source Matrices

V1φ1ct ,1
Qw,1 Qo ,1
Bw,1
Bo Qw,2 Bo Qo ,2
B= ... Q = Qw + Qo = +
Bw Bw
VN φN ct , N
Qw, N Qo , N
Bw, N
S=Saturation Bo=Volumetric Factor [rB/STB] k=perm [mD] q=rate [ft3/day] A=Area [ft2] V=Volume[ft3] T=Transmissibiltiy [(ft3/day.psi)]
Saturation Matrix Equation: Explicit
S nw+1 = S nw + d12 −1 −Tw P n +1 + Q w − Ctw ( P n +1 − P n )

Saturation at
last timestep water saturation change due to flow water saturation change to accumulation
and injection/production from well (often small and negligible)
V1φ1 Sw,1
∆tBw,1 S w1 ( cw + cr )
Sw,2
d12 = ... S nw = Ctw = Sw2 ( cw + cr )
VN φN S wN ( cw + cr )
∆tBw, N Sw,N
• Accumulation term is often small and can be neglected

• Note that d12 would be a little different if capillary pressure were present
Review IMPES method (2 phase)
Discretized mass balances for oil and water

1. Implicitly calculate pressure (overall mass balance) by solving system of equations
2. Explicitly update the saturation equation (water mass balance)
B B
T+J+ P n +1 = P n + Q S nw+1 = S nw + d12
−1
−Tw P n +1 + Q w − Ctw ( P n +1 − P n )
∆t ∆t
Explicit methods are only conditionally stable (and IMPES is partially

explicit). So time step (∆t) needs to be reasonably small
u ∆t
<1
∆x
Upwinding must be used to determine half relative perms!!!
kr ( S w,i −1 ) if Pi −1 > Pi
kr ,i −1 2 =
k r ( S w ,i ) if Pi > Pi −1
Boundary conditions and wells are treated similarly to single phase flow
ki −1 2 krw,i −1/2 A
Interblock Transmissibility? Tw,i −1/2 =
∆xi −1 2 µ w Bw
Absolute permeability: use harmonic mean as in single-phase flow:

−1
1 1 kA 2ki Ai ki +1 Ai +1
ki + 1 = 2 + =
2 ki k i+1 ∆x i+ 1 ki Ai ∆xi +1 + ki +1 Ai +1∆xi
2
Relative permeability: use technique called “upwinding”

• Harmonic and arithmetic mean can lead to errors
• Upwinding involves evaluation relative permeability in block flow comes from
kr ( S w,i −1 ) if Pi −1 > Pi
kr ,i −1 2 =
k r ( S w ,i ) if Pi > Pi −1 Pi-1, Sw,i-1 Pi, Sw,i
kr ( S w,i −1 )
k r ( S w ,i )
S=Saturation k=perm [mD] kr = rel perm q=rate [ft3/day] A=Area [ft2] V=Volume[ft3] T=Transmissibilty [(ft3/day.psi)]
Wells and well models
As in single phase flow, wells can be constant rate or BHP. Think of
each block being composed of rate and BHP wells
qw,i = qw,i ,rate + qw,i , BHP ; q o ,i = qo ,i ,rate + qo ,i , BHP
∴
qw,i = qw,i ,rate + J w,i ( Pwf ,i − Pi ) ; q o ,i = qo ,i ,rate + J o ,i ( Pwf ,i − Pi )
In matrix form:
Q w = Q w,rate + J w ( Pwf − P n +1 ) ; Q o = Q o ,i ,rate + J o ( Pwf − P n +1 )
Bo Bo
Q = Q w,rate + Q o ,i ,rate + J w + J o Pwf
Bw Bw
Q rate J
Pressure and Saturation Equations::
B B Bo
T+J+ P n +1 = P n + Q J = Jw +
Bw
Jo
∆t ∆t
S nw+1 = S nw + d12
−1
−Tw P n +1 + Q w − Ctw ( P n +1 − P n )
Bw=Volumetric Factor [rB/STB] k=perm [mD] Q=rate [ft3/day] kr = rel perm J’=Productivity [(ft3/day-psi)] h = thickness [ft] s = skin
Productivity Index for BHP wells
Constant BHP wells have Pwf specified:
Qw,i = J w ( Pwf − Pi )
Productivity indices for each phase:
2π krw k h 2π kro k h
Jw = ; Jo =
r r
µ w Bw ln eq + s µo Bo ln eq + s
rw rw
Total productivity index is the sum of phase indices weighted by FVF’s
Bo 2π k h krw kro
J = Jw + Jo = +
Bw req µ w µo
Bw ln +s
rw
Bw=Volumetric Factor [rB/STB] k=perm [mD] Q=rate [ft3/day] kr = rel perm J’=Productivity [(ft3/day-psi)] h = thickness [ft] s = skin
Courant–Friedrichs–Lewy (CFL) condition
(CFL) condition is a necessary condition for stability of IMPES
u ∆t
<1
∆x
SPECIAL TOPICS
• Capillary Pressure and Gravity
• Gas Flow
• 3D and Radial Flow
• History Matching
• Hydraulic and Natural Fractures
• Horizontal Wells
• Upscaling
• State-of-the-Art Simulation and the Future
Categories of Reservoir Simulators
Commercial
CMG
Eclipse (Schlumberger)
Intersect (Schlumberger + Chevron)
Nexus (Halliburton)
In House
Empower (ExxonMobil)
Cheers (Chevron)
PSim (ConocoPhillips)
MoReS – (Shell)
Powers – (Saudi)
Academic
UT-CHEM (UT Chemical Flooding)
IPARS (UT Multiscale Simulation)
GPAS/UT-COMP (UT compositional)
TOUGH2 (Lawrence Berkley National Labs)
Single Phase Gas Flow
Start with a mass balance (continuity equation) but the fluid is

compressible for gases, so density is a funcytion of pressure
Use gas law for density, our PDE (diffusivity equation) is now
nonlinear:
∂ p k ∂ p ∂p
( )= ( )
∂t z φ ∂x µ z ∂x
Two options to solve this problem:
1. Convert to “pseudopressure” and solve the linear problem
p p φµ c g ∂m( P) ∂ 2 m( p )
m( p ) = ∫ dp =
p µz
o k ∂t ∂x 2
2. Solve the nonlinear PDE directly. This gives a “nonlinear” system of equations. Solve
it using a multidimensional Newton Raphson method
B ( P n +1 ) B ( P n +1 )
F = T(P n +1
)+ P n +1
− Pn − Q = 0
∆t ∆t
1 1
Capillary Pressure and Gravity T+J+
∆t
B P n +1 = BP n + Q + G
∆t
Capillary pressure and gravity can sometimes be ignored if they are

small relative to the reservoir pressure
However, they are easy to handle if they need to be included

Potential (P+ρgh) is just used to account for gravity
Capillary pressure Pc=Po-Pw relates phase pressures – curve is needed
Mathematically, gravity and capillary pressure are lumped together in the

right hand side of the system of equations
G = ρ gTz + TPc
z1 Pc ,1
z2 Pc ,2
z= Pc =
z3 Pc ,3
z4 Pc ,4
3D and radial geometry
Although 2D geometries often suffice,
3D simulations are sometimes
necessary
Matrix equations are “heptadiagonal”
Gravity included
1 1
T+J+ B P n +1 = BP n + Q + G
∆t ∆t
For near-well simulations, we must use

radial geometry
Just change our coordinate system and
perform FD approximations
1 Matrix equations look same
∂u v= K ⋅∇P
ρ + u ⋅∇u = ρ g + ∇ ⋅ σ µ
∂t
∂p k 1 ∂ ∂P
= r
∂t µφ ct Bw r ∂r ∂r
1 1
T+J+ B P n +1 = BP n + Q
∆t ∆t
Fractures, Horizontal Wells, etc.
Permeability of fractures is drastically different than the reservoir
Usually too thin to justify grids just for fractures
Inaccurate to use “average” permeability for blocks with rock and fracture in it
“Dual Porosity/Dual Permeability” models sometimes used
Horizontal wells span over several grid blocks

Vertical wells had a source/sink for one block
Horizontal wells just distribute fluid over several blocks
http://www.geology.wisc.edu/courses/g115/oil/4.html
History Matching
Reservoir simulators are used as tools to predict

the future, but are only as good as the inputs
(permeabilty, porosity, etc.)
Input parameters generally come from a

geological model, well logs, core data, and /or
upscaled pore-scale results.
A model cannot be expected to predict the future History match at six wells of the watercut (blue) given the oil rate (green)
if it cannot predict the past used as input in the black oil reservoir simulator.
(http://www.worldoil.com/Simulation-workflow-for-a-fractured-reservoir-
aids-CO2-flood-planning-November-2010.html)
History matching adjusts the simulator inputs

(permeability, porosity, etc.) in order to fit existing
production data (production rates, GOR, etc.)
Manual tuning
Global optimization routines
182
History Matching Process
Parameters to change in a history match

1. Aquifer transmissibility, kh
2. Aquifer storage, φhct
3. Reservoir transmissibility kh
4. Relative permeability and capillary pressure functions
5. Reservoir porosity and thickness
6. Structural definition
7. Rock compressibility
8. Reservoir oil and gas properties
9. Water/oil contacts (WOCs) and gas/oil contacts (GOCs)
10. Water properties
Data to match
1. Pressure (e.g. Bottom-hole flowing pressure)
2. Water/oil ratios (WORs)
3. Gas/oil ratios (WGRs)
4. Water and gas arrival times
5. Fluid saturations from cores, well logs, and chemical tracer
tests.
183
Mattax, C.C. and Dalton, R.L., “Reservoir Simulation”, SPE Monograph, v13
GPAS: Natural Fracture Modeling
Fracture
Traditional Approach: New Approach:

Dual Porosity/Permeability Model Discrete Fracture Modeling
Courtesy of Kamy Sepehnoori, UT-Austin

Upscaling
Upscaling is process of averaging fine-scale processes to coarser scales

Goal is to develop a coarse permeability, porosity field, etc. that produces the same
results as the fine-scale simulation
Inevitably, information will be lost during upscaling
Upscaling continues to be an open area of research
FULLY IMPLICIT (Alternative to IMPES)
• Idea is to solve both pressure and saturation implicitly at the same time
• Two unknowns per block, pressure and saturation; therefore there are 2N
equations and 2N unknowns in the system
• Note that the T matrix is “block” tridiagonal and the D matrix “block” diagonal
• Unconditionally stable, but also more computationally demanding
( T + D ) Xn+1 = DXn + Q + G
Tw, 1 + Tw, 3 0 −Tw, 3 0
2 2 2
To , 1 + To , 3 0 −To , 3 0
T1 + T3 −T3 2 2 2
2 2 2 −Tw, 3 0 Tw, 3 + Tw, 5 0 −Tw, 5 0
−T3 T3 + T5 −T5 2 2 2 2
T= 2 2 2 2 = −To , 3 0 To , 3 + To , 5 0 −To , 5 0
2 2 2 2
−Ti − 1 Tn − 1 + Tn + 1
2 2 2
−Tw,n −1 0 Tw,n −1 + Tw,n +1 0
2 2 2
−To ,n −1 0 To ,n −1 + To ,n +1 0
2 2 2
S=Saturation Bo=Volumetric Factor [rB/STB] k=perm [mD] φ = porosity μ=viscosity [cP] q = Flow rate [ft3/day]
For the entire system we get
( T + D ) Xn+1 = DXn + Q + G
The “D” matrix is block diagonal
d1,11 d1,12
d1,21 d1,22
D1
d 2,11 d 2,12
D2
D= = d 2,21 d 2,22
Dn
d n ,11 d n ,12
d n ,21 d n ,22
The “RHS” are given as follows:

P1 q1, w
S w,1 q1,o
P2 q2, w
X= S w,2 ; Q= q2,o
Pn qn , w
* If capillary pressure is included, the pressure, P, is
S w, n qn ,o of the non-wetting phase (e.g. oil)
S=Saturation Bo=Volumetric Factor [rB/STB] k=perm [mD] φ = porosity μ=viscosity [cP] q = Flow rate [ft3/day]
FULLY IMPLICIT method requires iteration using
Newton Raphson
Formulate matrix equations in terms of residuals, R. The arrays T, D, G

etc. are functions of the solution vector X.
R ( X n +1 ) = ( T + D ) X n +1 − DX n − Q − G = 0
Step (1) Guess the solution, Xk+1
Step (2) Calculate R and J, the matrix of partial derivatives
∂R1 ∂R1 ∂R1
∂X 1 ∂X 1 ∂X n
R1
∂R2 ∂R2 ∂R2
R
J = ∂X 1 ∂X 2 ∂X n ; R = 2
Rn
∂Rn ∂Rn
∂X 1 ∂X n
Step (3) Solve J dX=-R
Step (4) Update X(k+1) =X(k) + dX and repeat

Compositional Simulation
In principle make mass balances for each hydrocarbon component, such as

methane, ethane, propane, etc. but in practice, we limit the number of
components included, and group components into pseudo-components.
Solve PDEs for concentration of each species

∂ k ∂P ∂C ∂
Ci ρ + ρ D i + Ci ρ q = φ ( Ci ρ ) k=1,...N
∂x µ ∂x ∂x ∂t
Can solve the problem fully implicit or using IMPEC (implicit pressure,
explicit concentration)
Equilibrium flash calculations using K values and an equation of state (EOS)

must be used to determine hydrocarbon phase compositions. These are
used to determine densities, etc.
zk
∑ x =∑ 1 + ν ( K
k
k
k − 1)
=1
k
Streamline Simulation
Alternative to IMPES/Fully Implicit for multiphase flow

http://www.ireservoir.com/case_streamline.html
Basic Idea: Fluids are transported over a timestep along streamlines rather
than cell-to-cell as is done in finite difference approaches
3D pressure field is solved first, Darcy’s law is used to compute velocities

and streamlines. Components are traced along streamlines
Advantages Disadvantages
• Computational speed and memory • Diffusive-dominated flows (e.g. gas
• Slightly compressible expansion and capillary pressure)
systems/convective-dominated flow
Thiele, M.R. “ Streamline Simulation for Modern Reservoir-Engineering Workflows”, JPT January 2010
Advances in Computation
All these complexities have led to bigger, more computationally difficult
problems
Some of the world’s leaders in simulation are solving hundreds of millions or

billions of simultaneous equations at each time step. Bigger, faster
computers alone won’t do.
Parallel computing is a method that splits the problem into hundreds or

thousands of smaller problems
Each smaller problem is sent to its own computer
Computers must communicate because smaller problems depend on eachother
http://en.wikipedia.org/wiki/Parallel_computing http://www.cfd-online.com/Wiki/Parallel_computing
UT-CHEM is UT-Austin’s 3D multi-component,
multiphase compositional model
Numerics
IMPEC (implicit pressure; explicit concentrations)
Flash calculations for phases in each cell
Cell-centered grids
Energy balance employed for thermal effects
Higher-order finite differences
Physics
Polymers with non-Newtonian rheology
Tracers
Surfactant/foam model
Vertical and horizontal wells
Full-tensor dispersion coefficients
Applications
Waterflooding
NAPL spill and migration
Geochemical reactions
Enhanced Oil Recovery – especially CHEMICAL EOR
192
IPARS
Integrated Parallel Accurate Reservoir Simulation System

State of the art solvers
Highly scalable
Couplings with geomechanics and chemistry
Multiblock approach: non-matching grids (structured or unstructured)
Courtesy of Mary Wheeler, UT-Austin

Where we are: Where we are going:
Numerics Numerics
Millions of unknowns Billions of unknowns
Advanced FEM and unstructured grids Advanced gridding
Parallel processing Bigger, faster computers
Applications Applications
Enhanced Oil Recovery Unconventional resources (shale oil and
Carbon Sequestration gas)
Tight-gas formations More carbon sequestration
Physics Physics
Better modeling of natural fractures Coupled hydraulic fractures with fracture
Nonlinear flow (e.g. polymer flooding) Horizontal wells
Compositional models and 3 phase Coupled flow and geomechanics
flash calculations Advanced characterization, especially for
Improved history matching shale formations
Better Upscaling
194

Reservoir Simulation Note

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Reservoir Simulation Note

Uploaded by

Copyright:

Available Formats

What is Reservoir Simulation ?

• Understand flow and transport in our

• Develop mathematical models to

• Described by coupled, nonlinear partial

• Solution to PDEs can’t be found

1. How should a field be developed to maximize economic recovery?

2. What is the best enhanced recovery scheme for the reservoir?

3. Why is the reservoir not behaving according to predictions made by

4. What is the ultimate economic recovery of the field?

5. What type of laboratory data is required?

6. Is it necessary to do physical model studies of the reservoir?

7. What is the best completion scheme for wells?

8. From what portion of the reservoir is the production coming?

1. Conversion to mathematical model requires understanding physics

2. Numerical solution to the mathematical problem requires simplifications

3. Using the math solution to find an engineering solution requires

• Break up the reservoir into individual “blocks” or “grids”

• Write algebraic equations for pressure, saturation, etc. each block

• Each block depends on the adjacent block equations so we get a

Grid block Out Out

• Equations come from a mass balance and implementing Darcy’s law:

• These equations are very complicated.

• Approximations allow us to get a problem that is easier to solve

• Write equations for fluid transport in porous media, rock/fluid properties,

• Use and perform error analysis of approximate numerical techniques to

• Develop computer programs to solve these massive numerical problems

• Use Matlab and/or program

• Find the root of a nonlinear equation

• Solve a system of equations (Ax=b)

• Curve fit/regression analysis

• Find an integral numerically

• Solve a differential equation numerically

• Great danger in using software if you don’t understand the fundamentals it

Mass In – Mass Out +/- Generation/Consumption = Accumulation

Mass in Mass out

Divide the mass balance by A∆x∆t

In 3D Cartesian coordinates the equation becomes

More general form of the Continuity Equation

- accounts for mass entering or leaving a control volume from a

- We model injection/production wells as discrete sources/sinks so they

- Taking limits without using this variable creates difficulty

PRINCIPLES TO FOLLOW AND FORMULAS TO BE USED

THE DISTRIBUTION OF WATER

AN APPENDIX RELATING TO THE WATER SUPPLIES OF SEVERAL CITIES

THE FILTERING OF WATER

THE MANUFACTURE OF STRONG PIPES OF LEAD, SHEET METAL AND BITUMEN

1 darcy = 9.869 x 10-9 cm2

VICTOR DALMONT, EDITOR,

1 darcy = 1.062 x 10-11 ft2 Quay of Augustins, 49.

Medium is homogeneous if permeability does not vary with space, but

Medium is isotropic if permeability is the same in all directions,

Medium could be homogenous and anisotropic or heterogeneous and isotropic

In 1D and horizontal flow (gravity neglected)

Replacing fluid velocity in the continuity equation

We measure volume at surface but do the mass balance at the reservoir

Formation volume factor and ρ depend on both pressure and temperature

At reservoir conditions, density is:

Replacing density in continuity equation and divide through by ρSC (a constant)

Using the product rule on the left-hand side of the equation

Chain and product rule on time derivative (right hand side)

With some manipulation:

1D diffusivity (with homogenous fluid and reservoir properties) can be written: