LCunha-Colour Template PDF

PE503 - Applied Reservoir Engineering
Presented by :
Luciane Cunha
April 25 to 29, 2005

Universidad Nacional de Ingeniería - UNI
Outline
• Introduction
• Fundamental Data and Basic Concepts in Reservoir Engineering
• Numerical Reservoir Simulation and Simulator Equations
• Stages in a Reservoir Simulation Study
• Well Testing and Analysis
• Enhanced Oil Recovery Methods
• Reservoir Management
Luciane Cunha Universidad Nacional de Ingeniería - UNI 2

Gas/Oil Reservoir
GAS
OIL
WATER WATER
Impermeable Layer
Reservoir Image
The reservoir image (geological model) is the base

to all the reservoir engineering studies. This image
is defined with respect to shape, limits, internal
architecture (heterogeneities), fluid distribution and
volumes.

Reservoir Image
Boundaries
Fault Patterns
Areal Extension
Physical
Properties
Structural Gas
Dip
Fluid Oil
Contacts
Stratification Water
Aquifer Size
Reservoir Image ARCHITECTURAL PATTERN
SHAPES and VOLUMES • correlations
• structural surfaces • sedimentar body shapes
• isopachs • facies variations
• boundaries • microscopic reservoir structure
TECTONIC PATTERN FLUIDS

• faults • interfaces
• fractured zones • composition
• micro-fractures • thermodynamical conditions
Reservoir Engineering Objectives
Optimize the Economic Exploitation of

Hydrocarbon Reservoir
⇒ How much hydrocarbon exist?

⇒ How much can be recovered?
⇒ How fast it can be recovered?
⇒ Original hydrocarbon in place

⇒ Reserves (estimated from alternative recovery methods)
⇒ Production rate (initial, variations)

Reservoir Simulation - Outline:
1. What is Reservoir Simulation?

2. Historical Developments
3. Different Reservoir Simulators
4. How the Simulators are Used
5. Modeling Concepts
6. Why Use Simulators
7. Options and Features of Modern Simulators

1. What is Reservoir Simulation?
Reservoir Engineers use different means to study and analyze reservoir performance
and predict future production.
Reservoir performance is a function of its behavior with pressure and time.
Reservoir Simulation is a blend of:
engineering
physics
chemistry
mathematics
numerical analysis
computer programming, and engineering experience and practice.
Reservoir Simulation is a powerful technique for reservoir management and refers to
the construction and operation of a model whose (hopefully) approaches that of a
reservoir under actual reservoir condition.
It allows the engineer to predict reservoir performance – provided it is used
correctly – under different scenarios.
The kind of reservoir simulators that are used today on digital computers are not
the only tools an engineer has used to mimic the behavior of real reservoirs.
Functional Model
Representative Model
- Input - - Model - - Output -

Exploitation Production
Scheme Performance
Forecasting

Numerical Reservoir Flow Simulation
Type of Simulators:
- Physical
- Mathematical
a) Analytical
1) Well Testing Equations (Diffusivity Equation)

2) Decline Curves
3) Material Balance Equations
b) Numerical Simulators

Analytical Models:
MBE
Tank Model
- Isotropic -> Kx = Ky = Kz
- Homogeneous Reservoir -> φ, Kro, Krw, Krg, So, Sw, Sg, Bo,
Bg,Rso,µo, µg are same throughout the tank
- Flow from A to B -> Darcy’s Law
MBE, which is based on tank model does not flow from one point in the
reservoir to another point.
MBE is a powerful tool for reservoir performance analysis, particularly
when there is negligible pressure difference (i.e. pressure gradient) in
the reservoir.
Analytical Models:
Diffusivity Equation
1 ∂  ∂p  φµc ∂p
r =
r ∂r  ∂r  k ∂t
φµc
-> Hydraulic Diffusivity
k
Solution of diffusivity equation gives the pressure distribution in
the reservoir at different times.

Physical Models:
- useful in visualizing the flow processes and develop mathematical models

for the flow.
problem: lack generality – we need to modify the physical model for each
reservoir under study.
- very expensive to build and run.
Physical vs Numerical Models
Numerical models must describe what physically takes place in the

reservoir.
In cases where we cannot model the reservoir physically (physical law

equation accurately enough, then, we should not use numerical
methods, instead physical models should be used.)
Numerical Simulators
Conceptually, a reservoir numerical simulator model is consisted of

a series of tanks, which are connected with one another.
Reservoir being water flooded
water oil
Sw increases to the right with time

Tank 1 Tank 2 Tank 3 Tank 4 Tank n
The tanks still obey physical laws (specially the conservation of mass);
What it is different from the tank model for MBE, is that now is possible to allow
for fluid to flow from one part of the reservoir (for example, “Tank 1”) to
another point (for example “Tank 2”).
There is pressure difference in each tank and consequently there is fluid flow.
Reservoir Simulators: computer programs that solve the equations

for heat and mass flow in porous media, subject
to appropriate initial and boundary conditions

Mathematical Models
-Mathematically, a reservoir numerical model is consisted of equations
describing the physical laws, including conservation of mass and
momentum & energy, as well as laws governing the behavior (i.e. flow of
fluids in our case)
-A complete set of PDE and auxiliary conditions, which describe the flow
pressures in the reservoir.
-PDE’s are derived from:

Conservation laws
Physical principles (Darcy’s Law)
-The number and type of equations to be solved depends on:

Geological characteristics of the reservoir (single or double porosity)
Characteristics of the oil, and
Oil recovery process to be modeled.
Reservoir PDE with Numerical Computer Application
System Auxiliary equations Scheme Program
Mathematical Model Computer Model =

Reservoir Simulator
Solution of the mathematical system of equations requires knowledge of

reservoir properties (i.e., permeability, porosity), fluid PVT properties
and reservoir well geometry.
Reservoir description is seldom known with any degree of accuracy -

usually obtained from geological and geophysical data as well as well test
data and geostatistical description.
????????????????????????????
Uncertainty
Mathematical Models contain three main sources of error:
• inaccuracy of input data;
• truncation error – introduced when we approximate our original PDE’s

using a Finite Difference or Finite Element Scheme;
• Round off error of the machine on which we solve our equations.
Note:
The main advantage of a reservoir simulator is the ability to inexpensively

produce the reservoir and a variety of different operating conditions.
Thus the optimum strategy for producing the reservoir can be
determined without equipment investment and without actually
producing any oil.
2. Historical Developments - Reservoir Simulators – “Timeline”
Traditional Reservoir Engineering (1930 - …)

•Computations with slide rules and mechanical calculators;
•Representation of reservoir by a single block;
•1-D analytical solutions for linear two-phase flow and radial single-phase flow.
Early Reservoir Simulation (1955 – 1970)

•Computations with digital computers;
•Primarily developed and used in research labs;
•Primary limitation was computer speed and storage;
•Limited ability to handle large systems of nonlinear equations;
•Poor reliability and lack of confidence in technology;
•High cost of development and use.

2. Historical Developments - Reservoir Simulators – “Timeline”
Modern Reservoir Simulation (1970 - …)

•High level of confidence in technology;
•Steady decrease in hardware cost;
•Large number of blocks with local grid refinement and irregular shape;
•Efficient methods for solving nonlinear equations;
•Robust methods for solving large systems of linear equations;
•Multicomponent fluid description and improvements in the handling of wells;
•Improvements in the understanding of complicated processes;
•Use of graphics and workstations and availability of supercomputers.

3. Different Reservoir Models (Simulators)
-One-phase vs multi-phase simulators
multiphase can also be subdivided into black oil or compositional simulators. Using one-
phase or multi-phase depend on what flow process(es) taking place in reservoir being
studied.
-Geometry
1-D
well
1-D
x
Radial
Linear
3-D
-Mostly 3-D. Linear predominates in fractured reservoirs while radial predominates

near the wellbore. Hence, we can setup equations combining both linear and radial
geometry. No single flow model occurs in the reservoir. What occurs in the reservoir
is a combination of several geometrical simulator models

4.1- Forecasting of reservoir performance

-Sensitivity studies for evaluating reservoir development options on elements of the
field (IOR process selection)
-Evaluation of field pilots and scale-up
-Full field studies for depletion planning at various stages of field development (number
and location of wells, injection/production rates, composition of producing fluids, etc.)
-Reservoir management
-Assessment of uncertainty in forecasting reservoir performance
4.2- Improving reservoir description through history matching

-Identification of major flow units and barriers
-Identification of near well rock properties
4.3- Analysis of experiments

-Relative permeability
-Minimum miscibility pressure
-Core displacement tests

4.4- Understanding of flow mechanisms

-Unstable displacements
-Flow in heterogeneous media
-Block effective properties
-Matrix/fracture transfer
4.5- Development of simple models and correlations

-Coning
-Cyclic steam injection
-Steam displacement
-Inflow performance relationships (IPR’s)

5. Modeling Concepts
The basic steps required for a reservoir simulation study are:
Select one or
more images for simulation
Input to simulator
Select grid & average
data to get block
properties
Images of Input Output Analysis of

Reservoir Data Results
History Matching
Data
In general, an engineer interested in doing a reservoir simulation study would follow the following
steps:
- Develop study objectives
- Develop or select an appropriate simulator
- Review, collect and estimate appropriate data
- Make preliminary runs to establish model parameters and limitations
- Match available history
- Predict performance under different operating scenarios
- Analyze results and prepare a report
- Plan additional work
6. Why Use Simulators?
• No other way to solve the problem
• Cheaper or more reliable than other methods
• Complement other more traditional techniques
• Increase profitability through improved reservoir management
• Assess economic and technical risks through sensitivity studies
• Enhance credibility with third parties
• Predict consequences of reservoir development and management decisions
• Establish relative merits of alternative operating strategies
• Resolve arbitration and utilization disputes
• Monitor reservoir performance
• Respond to safety, environmental and regulatory concerns
• Improve communication among interested parties
• Train engineers and operators
• Choose the optimum IOR scheme for a given reservoir
• Assess the impact on ultimate recovery of changing to a different IOR scheme
• Establish data needs during various stages of field development
• Assess the impact of assumptions on the analysis of well tests
• Optimize well location and well completion
• Assess possible advantages of horizontal wells over vertical wells
• Troubleshooting
7. Options and Features of Modern Reservoir Simulators
a) Preprocessors
b) Postprocessors
c) Phase and component
d) Gridding options
e) Solution Techniques
f) Special features
g) Well control
h) Group control
i) Injection control
j) Sale gas production control
k) Cross flow between zones
l) Non-Darcy effects
m) Initialization
n) Aquifer models
o) Pseudofunctions
p) Control of simulator performance

FLOW OF INCOMPRESSIBLE FLUIDS
General Concept:
- Normally used to mean a liquid (gas at a high pressure);

- Simplest case of simulation -> “steady state” = independent of time.
Units:
British and SI units
Example: Darcy’s Equation
SI Units British Units Modified SI Units
Ak∆P Ak∆P Ak∆P
q= q= q=
µL µL µL
q(m 3 / s) q(B / d) q(m 3 / s)
A(m 2 ) A(ft 2 ) A(m 2 )
∆P(Pa ) ∆P(psi) ∆P(kPa )
µ(Pa.s) µ(cp ) µ(mPa.s) (or cp)
L(m) L(ft ) L(m)
k(m 2 ) k(darcyx1.127) k(µm 2 / 10 6 )
Note:
1
1kPa =psi ; 1µm 2 = 1.1013darcy ;
7
viscosity; 1 m 3 / s = 543440 B / d
Grids and Grid Numbering:
- Problem unknowns: pressures and fluid saturations at each block of a grid.
- We need to have a system for identifying each block:
¾ Simplest system: j=4

(1,4) (2,4) (3,4) (4,4) 13 14 15 16
(1,3) (2,3) (3,3) (4,3)

j=3 9 10 11 12
j=2 (1,2) (2,2) (3,2) (4,2) 5 6 7 8
j=1 (1,1) (2,1) (3,1) (4,1) 1 2 3 4
i=1 i=2 i=3 i=4 Natural Order

y
j
x
i
Each block is identified by two numbers, i and j, with i increasing in the normal x-direction and j
increasing in the y-direction.
¾ Other Ordering Schemes: 2 9 5 12 4 7 10 12 9 8 7 6
7 3 10 6 2 5 8 11 10 11 12 5
1 8 4 11 1 3 6 9 1 2 3 4
D4 D2 Cyclic
• For three-dimensional grid -> we need three numbers to identify a
given block. z
y
k
j
x
i
• Linear Flow Between Two Adjacent Blocks (i) and (i+1):
qi->i+1
ki ki+1
Ai Ai+1
i i+1
Pi Pi+1
∆xi ∆xi+1
Positive x-direction
Consider, first, flow in the right-half of block i. Writing Darcy’s equation,
k i A i p i − p i +1 / 2
qi→i+1 =
µ ∆x i / 2
Consider now, flow in the left-hand of block i+1. Writing Darcy’s equation,
k i +1 A i+1 p i +1 / 2 − p i+1
q i →i + 1 =
µ ∆x i + 1 / 2
The flow rate is the same in the two blocks. Eliminating p i +1 / 2 from both
equations above we can obtain the following equation:
qi→i+1 = Ti+1/ 2 (pi − pi+1 )

where 2Aiki Ai+1ki+1 1
Ti+1/ 2 =
Aiki ∆xi+1 + Ai+1ki+1∆xi µ
Note: For 2-D Flow the transmissibility will be Ti+1 / 2, j and a “j” subscript will be added
to all of the terms in the right side of the above equation.
2 A i, jk i, j A i+1, jk i+1, j 1
Ti+1/ 2 , j =
A i, jk i, j ∆x i+1, j + A i+1, jk i+1, j ∆x i, j µ
If flow from block (i,j) to (i,j+1), vertically above, we have:
(i,j+1)
(i-1,j) (i,j) (i+1,j)

qi,j→i,j+1 = Ti,j+1/ 2 (pi,j − pi,j+1 )
and
2 A i, jk i, j A i, j+1k i, j+1 1 (i,j-1)
Ti, j+1/ 2 =
A i, jk i, j ∆yi, j+1 + A i, j+1k i, j+1∆yi, j µ

Flow Between Two Circular Blocks: r4
4
r3 3
2
r2
1 ri-1/2
ri+1/2
r1
rw ri
¾ r1 is chosen arbitrarily;
re
¾ ri = α i−1r1
rN+1 − rN α N−1r1 (α − 1)
re = =
rN+1 ln α
where α is calculated from the equation ln
rN
where N is the block number (N=4 in figure above)
¾ once the block “centers” are located, block boundary ri+1/2 can be calculated
ri+1 − ri ri − ri−1
ri+1/ 2 = and ri−1/ 2 =
r r
ln i+1 ln i
ri ri−1
applying Darcy’s law
qi→i+1 = Tri+1/ 2 (pi − pi+1 ) 2π 1

Tri+1/ 2 =
1 r 1 r µ
ln i+1/ 2 + ln i+1
k ihi ri k i+1hi+1 ri+1/ 2
For vertical transmissibility, Tz, in a two-dimensional system, where blocks such as those
shown below are stacked upon each other the vertical transmissibility for blocks (i,j) and
(i,j+1) is given by
ki,k+1
i,k+1 ∆zi,k+1
ki,k
ri-1/2,k
i,k ∆zi,k
ri+1/2,k

2π(ri2+1 / 2 − ri2−1 / 2 )k i , jk i , j+1 1
Tz i , j+1 / 2 =
k i , j ∆ z i , j+ 1 + k i , j+ 1 ∆ z i , j µ
Note: Having derived the transmissibilities for linear and circular systems, we can always
write the flow equation in the same form, without further regard for the system, so long
as the correct transmissibility expression is used.
Potential:
There are cases where it is more appropriate (and accurate) to consider Φ
(potential) instead of P (pressure).
For example, if the blocks are situated at different heights, we should account for
the hydrostatic head.
This is accomplished through the use of “potential” differences.
By definition
Φ = P - ρgD,
Φ = potential;
P = pressure;
ρ = fluid density;
g = gravitational acceleration
D = depth with respect to a reference level (below sea level)
q = Ti+1 / 2 (Φ i − Φ i+1 ) ⇒ q = Ti+1 / 2 [(Pi − ρgDi ) − (Pi+1 − ρgDi+1 )]
FLOW OF INCOMPRESSIBLE FLUIDS (cont.)
Basic Flow Equation for Incompressible Flow in One Dimension:
- Derive mass balance equation for an incompressible fluid;

- Derivation is general (other cases of flow; higher
dimensions).
Consider a block, i, and two adjacent blocks, i-1 and i+1.

Write a mass balance for block i, into which fluid is entering from
block i-1 at a rate and from which fluid is flowing into block i+1
at a rate .
Consider source/sink term.

qi-1->i qi->i+1
Ai-1 Ai Ai+1
ki-1 ki ki+1
i-1 i i+1
Pi-1 Pi Pi+1
∆xi-1 ∆xi ∆xi+1

injection rate, q*i
qi-1 -> i qi -> i+1
(i-1) (i) (i+1)
qi* (positive=injection/negative=production)
For Block i the mass balance is:
Net throughput + Net change over time ∆t = 0
Mass leaving - Mass entering + Mass at t+∆t – Mass at t = 0
ρ SC q i→i+1 − ρ SC (q i−1→i + q i* ) + 0 − 0 = 0
where internal injection of the fluid
ρ sc q i→i+1 − ρ sc q i−1→i − ρ sc q i* = 0
ρ sc = density of the compressible fluid @ standard conditions
Let’s write the expression above in terms of transmissibility, but

first let’s write some definitions:
P = pressure Pa
Φ = potential, Pa = P - ρgD
ρ = density, kg/m3
g = 9.80665 m/s2
D = depth below a reference plane in m (positive downward)
Utilizing the transmissibilities, and given that the potentials in the blocks
are Φi-1 , Φi , and Φi+1 , then the mass balance equation becomes:
Ti+1 / 2 (Φ i − Φ i+1 ) − Ti−1 / 2 (Φ i−1 − Φ i ) − qi* = 0
Multiplying by –1
Ti+1/ 2 (Φi+1 − Φi ) − Ti−1/ 2 (Φi − Φi−1 ) + qi* = 0
which is the standard form of the incompressible fluid flow

equation
Note: this equation is valid for radial flow, provided the

transmissibility term is calculated accordingly. In cases where the
gravity contribution is small, P can be substituted for Φ.
Flow Equations Assuming No-Flow Boundaries:
Consider a reservoir represented by a system of five blocks. Each block
has a different value of k, A, ∆x, D. It is given P1 and q*4.
q*1 q*4
(positive) (negative)
Injection point Production point
D1 D2 D3 D4 D5
k1 k2 k3 k4 k5
A1 A2 A3 A4 A5
1 2 3 4 5
P1 P2 P3 P4 P5
T1/2= 0 ∆x1 T3/2 ∆x2 T5/2 ∆x3 T7/2 ∆x4 T9/2 ∆x5 T11/2
1. In reservoir simulation, usually we make the boundaries don’t flow

boundaries.
How we can do that?
Making the transmissibility of the boundaries T1 / 2 and T11 / 2 = 0
2. Write the general equation for all blocks, putting i=1,2,…,nx

Block 1 (i=1): T3 / 2 (P2 − P1 ) − T1 / 2 (P1 − P0 ) + q 1* = 0
No-flow boundary at i=1 then T1 / 2 = 0 and
T3 / 2 (P2 − P1 ) + q1* = 0
Block 2 (i=2): T5 / 2 (P3 − P2 ) − T3 / 2 (P2 − P1 ) + q 2 = 0

*
Since there is no production or injection in block 2 q 2* = 0 and
T5 / 2 (P3 − P2 ) − T3 / 2 (P2 − P1 ) = 0
*
Block 3 (i=3): T7 / 2 (P4 − P3 ) − T5 / 2 (P3 − P2 ) + q3 = 0
Since there is no production or injection in block 3 q 3* = 0 and
T7 / 2 (P4 − P3 ) − T5 / 2 (P3 − P2 ) = 0
Block 4 (i=4): T9 / 2 (P5 − P4 ) − T7 / 2 (P4 − P3 ) + q *4 = 0
Given that there is production in block 4 we have to consider a negative sign

for q *4 and then
T9 / 2 (P5 − P4 ) − T7 / 2 (P4 − P3 ) − q *4 = 0
Block 5 (i=5): T11 / 2 (P6 − P5 ) − T9 / 2 (P5 − P4 ) + q 5* = 0

No-flow boundary at i=5 then and since there is no production or injection
in block 5 q 5* = 0 and then
− T9 / 2 (P5 − P4 ) = 0

There are 5 equations
T3 / 2 (P2 − P1 ) + q1* = 0
T5 / 2 (P3 − P2 ) − T3 / 2 (P2 − P1 ) = 0
T7 / 2 (P4 − P3 ) − T5 / 2 (P3 − P2 ) = 0
T9 / 2 (P5 − P4 ) − T7 / 2 (P4 − P3 ) − q *4 = 0
− T9 / 2 (P5 − P4 ) = 0
There are 5 unknowns:

P2 , P3 , P4 , P5 , q 1*
Adding all the equations we have: q1* − q *4 = 0 Æ this is called Material Balance check
Flow in Two-Dimensions q*i,j (external injection)
(i,j+1)
q(i-1,i->i,j) q(i,j->i,j+1) q(i,i->i+1,j)
(i-1,j) (i,j) (i+1,j)

y
q(i,j-1->i,j)
x
j (i,j-1)
i
Mass Balance for Block (i,j) for time ∆t
Remember the Notation:
q i→i+1 = Ti+1 / 2 (Pi − Pi+1 )
Leaving – Entering + (Mass at t+∆t – Mass at t) = 0
ρ osc qi, j→i, j+1 + ρ osc qi, j→i+1, j − ρ osc [qi, j−1→i, j + qi−1, j→i, j ] + ρ osc q * i, j + M − M = 0
where
M = φ∆x∆yhρ osc
obs: the corners don’t enter in the equation, they are taken in account in a different way.
Ti+1 / 2, j (Φ i, j − Φ i+1, j ) + Ti, j+1 / 2 (Φ i, j − Φ i, j+1 ) −
− Ti−1 / 2, j (Φ i−1, j − Φ i, j ) − Ti, j−1 / 2 (Φ i, j−1 − Φ i, j ) − qi*, j = 0

Put x in the T notation when the flow is in x direction and y when the flow
is in y direction
Tx i+1 / 2 , j (Φ i, j − Φ i+1, j ) + Ty i, j+1 / 2 (Φ i, j − Φ i, j+1 ) −
− Tx i−1/ 2, j (Φ i−1, j − Φ i, j ) − Tyi, j−1/ 2 (Φ i, j−1 − Φ i, j ) − qi*, j = 0
Changing the sign we get the standard form
Tx i+1 / 2, j (Φ i+1, j − Φ i, j ) − Tx i−1 / 2, j (Φ i, j − Φ i−1, j ) +

Tyi, j + 1/ 2 (Φi, j+ 1 − Φi, j ) − Tyi, j −1/ 2 (Φi, j − Φi, j −1) + qi*, j = 0
To solve a 2-D problem, write the above equation at all blocks i=1,2,…,nx, solve for
Pi,j at blocks where it is unknown, and for at j=1,2,…,ny blocks where pressure is
already specified.
If all boundaries are closed, make transmissibilities at the boundaries zero.
y y y y y
Flow in Three-Dimensions
x x
x x
y y y y y
Tx and Ty (boundaries)=0
i,j,k+1
Tx i+1 / 2, j,k (Φ i+1, j,k − Φ i, j,k ) − Tx i−1 / 2 , j,k (Φ i, j,k − Φ i−1, j,k ) + i,j+1,k
i-1,j,k i+1,j,k
i,j,k
Ty i, j+1/ 2,k (Φ i, j+1,k − Φ i, j,k ) − Ty i, j−1/ 2,k (Φ i, j,k − Φ i, j−1,k ) +
i,j-1,k
Tz i, j,k +1/ 2 (Φ i, j,k +1 − Φ i, j,k ) − Tz i, j,k −1/ 2 (Φ i, j,k − Φ i, j,k −1 ) + qi*, j,k = 0 i,j,k-1

Weinstein, Stone and Kwan Notation:
(Weinstein, H.G., Stone, H.L., and Kwan, T.V.: “Simultaneous Solution of Multiphase
Reservoir Flow Equations,” SPEJ (June 1970),99.)
Which are the advantages of this notation?
•allows a more compact representation of the flow equations;

•permits the implementation of various solution methods rather easily.
Generally,
For 1-D:
Ti+1/ 2 (Φi+1 − Φi ) − Ti−1/ 2 (Φi − Φi−1) + qi* = 0
DiΦi−1 + EiΦi + FiΦi+1 = qi

where,
qi = −qi*
Di = Ti − 1 / 2
Fi = Ti + 1 / 2
Luciane Cunha
Ei = −(Di + Fi ) 47
For 2-D:
Tx i+1 / 2, j (Φ i+1, j − Φ i, j ) − Tx i−1 / 2, j (Φ i, j − Φ i−1, j ) +
Tyi, j+1/ 2 (Φi, j+1 − Φi, j ) − Tyi, j−1/ 2 (Φi, j − Φi, j−1) + qi*, j = 0
Bi, jΦ i, j − 1 + Di, jΦ i − 1, j + Ei, jΦ i, j + Fi, jΦ i + 1, j + Hi, jΦ i, j + 1 = qi, j

where,
qi, j = −qi*, j
Bi, j = Tyi, j−1/ 2
Di, j = Txi −1/ 2, j
Hi, j = Tyi, j + 1 / 2
Fi, j = Txi + 1/ 2, j
Ei, j = −(Bi, j + Di, j + Fi, j + Hi, j )
Remember,
2 Ax i, jkx i, j Ax i+1, jkx i+1, j 1
Tx i+1/ 2 , j =
Ax i, jkx i, j ∆x i+1, j + Ax i+1, jkx i+1, j ∆x i, j µ
Ax = cross-section area perpendicular to x-direction;

kx = absolute permeability
Luciane Cunha
in x-direction.
Universidad Nacional de Ingeniería - UNI 48
For 3-D:
Tx i+1 / 2 , j,k (Φ i+1, j,k − Φ i, j,k ) − Tx i−1 / 2 , j,k (Φ i, j,k − Φ i−1, j,k ) +
Ty i, j+1/ 2 ,k (Φ i, j+1,k − Φ i, j,k ) − Ty i, j−1/ 2 ,k (Φ i, j,k − Φ i, j−1,k ) +
Tz i, j,k +1/ 2 (Φ i, j,k +1 − Φ i, j,k ) − Tz i, j,k −1/ 2 (Φ i, j,k − Φ i, j,k −1 ) + qi*, j,k = 0
Z i, j,k Φ i, j,k −1 + Bi, j,k Φ i, j−1,k + Di, j,k Φ i−1, j,k + Ei, j,k Φ i, j,k +
Fi, j,k Φ i+1, j,k + Hi, j,k Φ i, j+1,k + Si, j,k Φ i, j,k +1 = qi, j,k
where, qi, j,k = −qi*, j,k
Zi, j,k = Tzi, j,k −1/ 2
Bi, j,k = Tyi, j −1/ 2,k
Di, j,k = Txi −1/ 2, j,k
Fi, j,k = Txi + 1/ 2, j,k
Hi, j,k = Tyi, j + 1/ 2,k
Si, j,k = Tzi, j,k + 1/ 2
Ei, j,k = −(Zi, j,k + Bi, j,k + Di, j,k + Fi, j,k + Hi, j,k + Si, j,k )
Solution of the Flow Equations: Direct and Iterative Methods:
E 2 Φ 2 + F2 Φ 3 = q2 (2a)
D 3 Φ 2 + E 3 Φ 3 + F3 Φ 4 = q3 (2b)
D4Φ 3 + E4Φ 4 = q4 (2c)
Eqs. (2a), (2b) and (2c) are 3 simultaneous eqs. and 3 unknowns, Φ 2 , Φ 3 , Φ 4 . The
solution to this system of equation can be obtained by,
a) Direct Solution Methods (give “exact” solution)
¾ Gaussian Elimination or Gauss-Jordan Method

• most powerful method available;
• slowest method;
• requires a great deal of storage (memory).
¾ Thomas’s Algorithm
• variation of the Gauss-Jordan Method;
• very accurate;
• faster than the standard Gauss-Jordan Method;
• matrix structure specific.
b) Iterative Methods (start out with a guess of the answer, an successively improve the
guess to any desired precision)
¾ PSOR (Point Successive Over-Relaxation) (general);

¾ Gauss-Siedel;
¾ ADIP (Alternating Direction Implicit Procedure) (2-D/3-D problems);
¾ SIP (Strongly Implicit Procedure);
¾ SOR (Line Successive Over-Relaxation).

Direct Methods – Thomas’s Tridiagonal Algorithm:
This method is a variation of Gaussian Elimination
Tailored to system of equations of the type:
E 2 Φ 2 + F2 Φ 3 = q2 (2a)
D 3 Φ 2 + E 3 Φ 3 + F3 Φ 4 = q3 (2b)
D4Φ 3 + E4Φ 4 = q4 (2c)
Note: Characteristics of these eqs. are:
• each eq. contains 3 unknowns (first and last eqs. contain only 2
unknowns);
• unknowns are ordered in such a way that for a given eq., 2 of the
unknowns occur in the previous eq. and 2 occur in the next eq.
Tridiagonal Structure
x x = x
x x x = x
x x x = x
x x x = x
x x x = x
x x = x
•requires minimum storage;

•Thomas’s algorithm is designed to solve a system of eqs. having
tridiagonal structure (such eqs. arise from 1-D problems, and from
“linearized” (as in ADIP) multi-dimensional problems);
In order to develop Thomas’s algorithm, we have to write eqs. (2a) to (2c)
in a somewhat more general form, where the number of equations is N and
i is any one of the equations:
b1x 1 + c 1x 2 = d1
a2 x1 + b2 x 2 + c2x3 = d2
a3x 2 + b3 x 3 + c3x 4 = d3
a4x3 + b4x 4 + c4x5 = d4
...
a i x i−1 + bi x i + c i x i+1 = di
...
a N−1x N−2 + b N −1 x N −1 + c N − 1x N = d N −1
a N x N −1 + bN x N = dN
Note:
-The first equation does not have an “a1”;
-The last equation does not contain a “cN”.
-The subscripts on a, b, c in any equation are the same as the number
of the equation.
The solution algorithm is as follows:
c1 d1
¾For i=1 calculate w1 = and g1 =
b1 b1
¾For i=2,3;…;N-1 calculate

ci
wi =
b i − a i w i −1
¾For i=2,…,N calculate
d i − a i g i −1
gi =
b i − a i w i −1
Then:
¾For i=N calculate x N = gN
¾For i=N-1,…,1 calculate x i = gi − w i x i+1

Iterative Methods:
In the iterative methods, some initial solutions are assumed and then we
successively improve the solutions.
Iterative Methods – Gauss-Seidel Iteration Method:

¾ allow a solution of a system of equations through successive improvement of an
initial guess of all Φ’s (usually taken to be zero);
¾ requires less storage (only of the nonzero quantities);
¾ usually very rapid.

For our Incompressible Flow Problem …
E 2 Φ 2 + F2 Φ 3 = q2 (2a)
D 3 Φ 2 + E 3 Φ 3 + F3 Φ 4 = q3 (2b)
D4Φ 3 + E4Φ 4 = q4 (2c)
Solution Procedure:
The idea is to use each equation in order to improve the value of one of the Φ’s -> so-
called “diagonal” element.
Eqs. 2a will be use to calculate improved values of Φ2
Eqs. 2b will be use to calculate improved values of Φ3
Eqs. 2c will be use to calculate improved values of Φ4
To better understand the solution procedure lets write the equations the following way:
Φ new
2 = (q 2 − F2 Φ old
3 ) / E2 (3a)
Φ new
3 = (q 3 − D 3 Φ new
2 − F3 Φ old
4 ) / E3
(3b)
Φ new = (q 4 − D 4 Φ new ) / E4 (3c)

4 3
Start with initial guesses
Φ2= Φ3= Φ4=0
Use (3a) to calculate a

“new” value for Φ2
Use (3b) and Φ2new just calculated

to obtain a “new” value for Φ3
Use (3c) and Φ3new just calculated

to obtain a “new” value for Φ4
Perform a convergence test:
ν +1
- determine yes
d max
= max Φ iν +1 − Φ iν = max Φ new
i
i
i
− Φ iold end
ν +1
- compare d max with “tolerance”
,ε
no
Φ old
4 = Φ new
4 (Φ ν = Φ ν +1 );
Φ old
3 = Φ new
3 ;
Φ old
2 = Φ new
2

Iterative Methods - PSOR (Pointwise Successive Over Relaxation):
-Important modification of the Gauss-Seidel Method;
-Is essentially the Gauss-Seidel method, with a modification introduced into

the calculated iterates in order to “speed-up” convergence;
-Tolerance requirement is achieved in fewer iterations than for Gauss-Seidel

Method -> this is accomplished by use of an iteration parameter wb
( 1≤ wb ≤ 2 )

The general form for PSOR method is the following:
Φ iν +1 = (1 − w b )Φ iν + w b [(q i − D i Φ iν−+11 − Fi Φ iν+1 ) / E i ]
where wb = weighting factor.
wb can be determined from the “spectral radius”, 1≤ wb ≤ 2 .
2
wb =
1+ 1− ρg
where ρg = spectral radius of the seidel matrix can be determined as:
ν +1
dmax
ρ g = lim ν
n→ ∞
Note:
dmax
if wb = 1 -> Gauss-Seidel
if wb ≠ 1 -> “negative weight” on the previous iterate and greater than one, on the new Gauss-Seidel
iterate.

Solution Check:
It is necessary to check the validity of the solution obtained. Two different ways we can
use are:
¾ Material Balance Check
¾ Residuals
Material Balance Check:
widely used in reservoir simulation.

Material balance check can be stated as follows:
MB = Net change of mass over a time interval / Net mass throughput = 1
In incompressible flow, there is no time involved and then
MB = Net mass throughput = sum of qi* all terms = 0
Nx
MB = ∑q
i =1
*
i =0
The sum must be close to zero, because what goes into the system (+) must come out
(-).
Poor MB -> answers are certainly doubtful
Good MB not necessarily guarantee the validity of the answer.
Residuals:
R i = Ti+1 / 2 (Φ i'+1 − Φ i' ) − Ti−1 / 2 (Φ i' − Φ i' +1 ) + qi*
Where Φ i' +1 , Φ i' , Φ i' −1 are the “solutions”
For a “perfect” solution, Ri=0.

Note: the calculation of residuals are the best test of the accuracy of the solution
obtained. However, their calculation is very time-consuming for real problems.
FLOW OF SLIGTHLY COMPRESSIBLE FLUIDS
Compressibility:
 1 ∂V  1 dV
C = −  =− is isothermal
 V ∂P  T V dP
m
density = ρ =
V ρ  d(m / ρ)  dρ
C=−  
m  dρ  dP
1 dρ dρ
C= ⇒ = CdP
ρ dP ρ
ρ P
lnρ ρ = cP P
ref ref
ρ
ln = c(P − Pref )
ρ ref
ρ
= e c (P −Pref ) if c is small (10-6 to 10-7 kPa-1)
ρ ref
e c (P − Pref ) ≈ 1 + c(P − Pref ) ρ = ρref (1 + c(P − Pref ))

Note:
1 1 dZ
Gas compressibility = C = −
P Z dP
Assumptions:
Mass is constant, but volume changes as a function of pressure.
density of slightly compressible fluids changes with pressure (i.e. it also changes with time)
ρ = ρ(P) or ρ = ρ(P, t )
Flow behavior of slightly compressible fluids will be time dependent. What this means
in practice is that the pressure distribution in the reservoir will change with times,
n n+1 n+ 2
i.e. Pi , Pi , Pi , etc
Where n, n+1, n+2, etc… denote different times.
∆t = time step (or time elapse between 2 states of the reservoir)
n = n*∆t Pin (initial pressure) Future (2 years from today)
n+1 = (n+1) *∆t
Today (n=0) ∆t = 1 day
∆t = 1 day (365*2) (365*2) 730 “calculations”
”
730 “calculations
∆t = 2 months (12/2)*2 12 “calculations”

calculations”
∆t = 1 year 2 2 “calculations”
calculations”

Mass Balance for Block i Over Time ∆t q*
i
i-1 i i+1
qi-1->i qi->i+1
Block i:
Pore Volume = φ i A i ∆x i = φ i ∆Vi
Mass at t = φ i Vb i ρ t
Mass at t+∆t = φ i Vb i ρ t + ∆t
t → Pin pressure in block i
t + ∆t → Pin+1 Net throughput + Net change = 0

Net throughput = (Mass Leaving) – (Mass Entering)
Net change = (Mass t+∆t) – (Mass at t)
Overall,
ρ scqi→i+1∆t − ρ sc qi−1→i ∆t − ρ sc qi* ∆t + (φi Vbiρ t + ∆t − φi Vbiρ t ) = 0
where,
ρ sc q i→i+1∆t − ρ sc [q i−1→i + q i* ]∆t = Flow Term (LHS)
φ i Vbi ρ t + ∆t − φ i Vbi ρ t = Accumulation Term (RHS)
ρ sc Ti + 1 / 2 (Φ i + 1 − Φ i )∆t − ρ sc Ti − 1 / 2 (Φ i − Φ i − 1 )∆t + qi* ρ sc ∆t =

= φi Vbi (ρ t + ∆t − ρ t )
Pressure must have a time level on it

n -> time level
t → Pin t + ∆t → Pin+1
n+1
for ρ = ρ sc [1 + c(Pi − Psc )]
ρ sc Ti + 1 / 2 (Φ i + 1 − Φ i )∆t − ρ sc Ti − 1 / 2 (Φ i − Φ i − 1 )∆t + qi* ρ sc ∆t =
Now
= φi Vbi [ρ sc (1 + c(Pin + 1 − Psc )) − ρ sc (1 + c(Pin − Psc ))]
cφ i Vbi n+1
Ti+1 / 2 (Φ i+1 − Φ i ) − Ti−1 / 2 (Φ i − Φ i−1 ) + qi* = (Pi − Pin )
∆t
Note: There are TWO time levels (n) and (n+1), for the RHS term, there are, yet,
no time levels specified for the potentials (or pressure) on LHS term.
Anything@ time level (n) is known

Anything@ time level (n+1) is unknown
Must specify a time level for LHS terms and have a time level on F, choice n or n+1
So: n+1 or n
cφ i Vbi n+1
Ti+1 / 2 (Φ i?+1 − Φ i? ) − Ti−1 / 2 (Φ i? − Φ i?−1 ) + qi* = (Pi − Pin )
∆t
1- Time Level n – Explicit Scheme Method
cφ i Vbi n+1 *
Ti+1/ 2 (Pin+1 − Pin ) − Ti−1/ 2 (Pin − Pin−1 ) = (Pi − Pin ) + Ti + 1 / 2 ρg(Di + 1 − Di ) − Ti − 1 / 2 ρg(Di − Di − 1 ) − qi
∆t
cφi Vbi n + 1
Ti + 1 / 2 (Pin+ 1 − Pin ) − Ti − 1 / 2 (Pin − Pin− 1 ) = (Pi − Pin ) + ζ i
∆t
where, ζ i = Ti + 1 / 2ρg(Di + 1 − Di ) − Ti − 1 / 2 ρg(Di − Di − 1 ) − qi*

Given initial conditions, i.e. Pressure at t=0 everywhere, i.e. at all i we want
pressures at the next time t = ∆t (note that the time step does not to be constant)
So we have:
P1 =10 P2 =10 P3 =10 PN =10
Now we want to calculate P for the next time.
Calculate for i = 1,2,3,…,N
Write the above equation as: ∆t

Pin+1 = Pin + [ Ti+1 / 2 (Pin+1 − Pin ) − Ti−1 / 2 (Pin − Pin−1 ) − ζ i ]
cφ i Vbi
i = 1,2,…,n
Pin → known Pin+1 → ?
t0 = 0
t1 = ∆t1 n=1
t2 = t1 + ∆t2 n=2
Calculation is very simple. However, the size of the time step ∆t is usually restricted
(to be small) to ensure stability of the solution.
Stability:
In some problems, the round off errors accumulated during the arithmetic operations to obtain
the solution grow with time, to the extent that their magnitudes become “significant” and
the solution becomes meaningless. In these cases, the system is said to be unstable.
On the other hand, there are systems when the round off errors accumulate but their
magnitudes are kept with in bounds. The solution is therefore ok and these systems are
said to be stable.
Conditionally Stable:
Solution will be stable if conditions are satisfied. For example, the time step sizes (∆t) is
set in such a way that the round off errors accumulation is kept under control.
∆t -> function of:

• how large the space is;
• transmissibilities;
• linear & non-linear terms.

Determination of Stable ∆t:
There are many criteria to analyze stability solution. Between them we can list the most
commonly applied which are:
1. Karplus’s Method -> does not include boundary conditions
2. Neumann’s Method -> does not include boundary conditions
3. Matrix Method -> include boundary conditions
Karplus’s
Given the equation
cφ i Vbi n+1
Ti+1 / 2 (Pn
i+1 − P ) − Ti−1 / 2 (P − P ) + q =
i
n
i
n n
i −1 (Pi − Pin )
*
i
∆t

Express the equation in the following form
a(Pin++11 − Pin ) + b(Pin− 1 − Pin ) + c(Pin+ 1 − Pin ) = 0
where a, b and c are coefficients.
Note: the second pressure term in each of the groupings is Pi,n i.e. pressure in block
i at time level n. The first term can be anything.
Then
- if a, b, c, … are positive, the scheme is stable;
- if one or more are negative, then the equation might be conditionally stable if:
a + b + c + ... ≤ 0

Considering our main flow equation…
cφ i Vbi n+1
Ti+1 / 2 (Pin+1 − Pin ) − Ti−1 / 2 (Pin − Pin−1 ) = (Pi − Pin )
For stability, ∆t
cφ i Vbi cφ i Vbi
Ti+1 / 2 + Ti−1 / 2 − ≤0 Ti+1 / 2 + Ti−1 / 2 ≤
∆t ∆t
for any i
cφ i Vbi
∆t ≤
Ti+1 / 2 + Ti−1 / 2
Procedure: first calculate ∆t for every i, choose min.

2- Time Level n+1 – Implicit Scheme Method
cφ i Vbi n+1
Ti+1/ 2 (Pin++11 − Pin+1 ) − Ti−1/ 2 (Pin+1 − Pin−+11 ) = (Pi − Pin ) + ζ i
∆t
Check Stability
Add Pin − Pin in each parenthesis in the flow term. So,

Ti+1/ 2 (Pin++11 − Pin ) − Ti+1 / 2 (Pin+1 − Pin ) −
cφ i Vbi n+1
− Ti −1 / 2 (Pin+1 − Pin ) + Ti−1 / 2 (Pin−+11 − Pin ) − (Pi − Pin ) = 0
∆t
For stability…
cφ i Vbi
Ti+1/ 2 − Ti+1/ 2 − Ti−1/ 2 + Ti−1/ 2 − ≤0
∆t
and then
cφ i Vbi is true for all ∆t -> this means the implicit method is unconditionally stable.
− ≤0
∆t
As a summary we can say that
an Explicit formulation is Conditionally stable and
an Implicit formulation is Unconditionally stable.
FLOW OF SLIGTHLY COMPRESSIBLE FLUIDS (cont.)
Material Balance Check
cφi , j Vbi , j n+1

Cumulative Mass Change i , j∑ ∆t
( Pi , j − Pin, j )
MB = = =1
Cumulative Input / Output ∑ qi,j *
i,j

MULTIPHASE FLOW
General Concepts:
Three Phases - Oil, Water and Gas
Gas -> (free gas, in solution in oil, in solution in water)
Oil -> dead oil, insoluble in water
Water -> insoluble in oil, ignore any water vapor in gas phase
Rock and Fluid Properties:

Rso, Rsw -> gas solubility or solution GOR in oil, water
Rs
Bo = oil formation volume factor rm3/ m3
Bw = water formation volume factor
Bg = gas formation volume factor Bo
µ o , µ w , µ g = viscosity of oil, aqueous and gas phase,
respectively
µo
po pb
Bg
µg
pg
Obs: Usually the PVT information is given to the reservoir simulators in a table
format. The simulators evaluate the PVT properties by means of linear interpolation.
•Relative Permeability k ro , k rw , k rg
k effective
Definition k r =
k absolute
•Capillary Pressure
Pcow = Po − Pw Pcgo = Pg − Po
MULTIPHASE FLOW (Cont.)
Rock-Fluid Properties Determination:
Capillary Pressure
The capillary pressure is a function of:
•type of fluids;
•rock type;
•relative quantity of fluids (saturation);
•history of the saturation process (drainage or imbibition).
drainage curve
PCow (S w ) = p o − p w
imbibition curve
(oil: non wetting; water: wetting) entry pressure
PCog (S g ) = p g − p o 0
Swc Sw 1
wetting phase
(gas: non wetting; oil: wetting) critical saturation of
the non-wetting phase
residual saturation of
the non-wetting phase
Two phase system
Calculation of
drainage
drainage/imbibition
imbibition
0 Swc Swr Soc 1

Sw
kf = phase f effective permeability = function of Sf
Standard procedure is to use relative permeabilities:
krf = kf / k and 0 < krf < 1
Three phase Systems:
•water is the wetting phase
•gas is the non wetting phase
•oil is intermediary
In this respect we have that krw = f(Sw)
krg = f(Sg)
kro = f(Sw,Sg)
Since the experimental test to measure three-phase relative permeabilities is
extremely difficult to execute, usually the three-phase relative
permeabilities are generated from two-phase experiments (oil-water and
oil-gas).
1. Steady State Method

At steady state, output streams are identical to input. Calculate S o , S w , S g
by Material Balance
2. Unsteady State Method
Stone’s Model - Method I (1970):

Start with two-phase oil-water and oil-gas curves
For 3 phases flow
Stone’s Model - Method II
Aziz’s Model I and II

MULTIPHASE FLOW (Cont.)
Derivation of the System of Equations for the Multiphase Flow Problem
Net throughput + Net change = 0
Mass leaving – Mass entering + Mass at time t+∆t – Mass at time t = 0
Oil
 φi VbiSno+i 1 φi VbiSnoi 
[ρoscqoi→i + 1 − ρoscqoi −1→i + q ∆t + *
]
 Bno+ 1
oi −
B n
ρosc = 0

 i oi 
 φ V S n+1
φ V S n

Toni+1/ 2 (Φ noi++11 − Φ noi+ 1 ) − Toni−1/ 2 (Φ noi+ 1 − Φ noi+−11 ) + qo* i +  =0
i bi oi i bi oi
−
 ∆tBno + 1 ∆ tBn 
 i oi 

Water
 φi VbiSnw+i 1 φi Vbi Snwi 
T n
(Φ n+1
−Φ n+1
)−Tn
(Φ n+1
−Φ n+1
)+q +
*
− =0
w i+1 / 2 w i+1 wi w i −1 / 2 wi w i−1  ∆tBnw+ 1
wi
∆tBnwi 
 i 
Gas
[ρ gsc q gi→i+1 + ρ gsc R so q oi→i+1 + ρ gsc R sw q wi→i+1
− ρ gsc qgi−1→i − ρ gscR soi qoi−1→i − ρ gscR sw i q wi−1→i + qg* i ]∆t
 S ngi+1 S ngi R so ni+1S o ni+1 R so ni S o ni R sw ni+1S w ni+1 R sw ni S w ni 

+ ρ gsc φ i Vbi  n+1 − n + − + − ρ osc = 0
 Bg B gi B oin+1
B oin+1
B win+1 n+1
B wi 
 i 
Tgni+1/ 2 (Φngi++11 − Φngi+1) − Tgni−1/ 2 (Φngi+1 − Φngi+−11 ) +

− R soi+1 / 2 Toni+1 / 2 (Φ no+i+11 − Φ no+i 1 ) − R nsoi−1 / 2 (Φ no+i 1 − Φ no+i−11 ) −
− R swi+1/ 2 Twn i+1/ 2 (Φ nw+i+11 − Φ nw+i 1 ) − R nswi−1/ 2 (Φ nw+i 1 − Φ nw+i−11 ) + qg* +

i
φV  Sngi+1 Sngi R so ni+1S o ni+1 R so ni S o ni R sw ni+1S w ni+1 R sw ni S w ni 

+ i bi  − + − + − =0
∆t  Bng+1 Bng B oin +1
B oin +1
B win +1 n+ 1
B wi 
 i i 
Note:
qg* total = qg* free + R soqo* + R sw q*w
General Problem Formulation
Equations: mass balances 3
Unknowns (at a producing well):

Po , Pw , Pg , S o , S w , S g , q o* , q *w , q g*
Saturation constraint
So + S w + Sg = 1
Capillary Pressures:
Pcow (S w ) = Po − Pw Pcgo (S g ) = Pg − Po
Mobility Relationships:
* 2πrk rw kh ∂Φ w 2πrk rgkh ∂Φ g 2πrk ro kh ∂Φ o
qw = − *
qg = − *
qo = −
µ wBw ∂r r = rw µ gB g ∂r µ oB o ∂r
r = rw r = rw
q o* k ro µ w B w
=
q *w k rw µ o B o
Solution Procedure:
• IMPES (Implicit Pressure Explicit Saturation)

• FULLY IMPLICIT
• ADAPTIVE IMPLICIT
IMPES
1a. expand the accumulation terms of all three equations to extract the
saturations terms;
1b. multiply each equation by a suitable multiplier, in such a way that when
the equations are added, the saturations terms cancel;
1c. left with one equation in pressure only;
1d. solve for the grid;
1e. solve for using original mass balances.

SOLUTION OF NONLINEAR EQUATIONS
Introduction
Newton’s Method - This is the standard approach for solving the fully implicit
finite difference equations for multiphase flow in reservoirs.
Consider first a simple nonlinear equation in one unknown:
f(x)=0
Expanding f(x) about x0 using Taylor series expansion, we have
 df 
f (x ) ≡ f (x 0 ) +   (x − x 0 ) + K
 dx  x0 higher order terms = 0
f (x 0 )
We can solve the above expression for ∆x ≡ x − x 0 as ∆x = − x − x0
 df 
 df   
f (x k +1 ) ≡ 0 = f (x k ) +  k +1  dx  x0
 (x − x ) + K
k
 dx  xk
f(x)
f(x0)
f(x) = 0
x1 x x0
1
∂f
f (x ) = 0 = f (x 0 ) + (x − x 0 ) + K
∂x x0
∂f
f (x k +1 ) = 0 = f (x k ) + ( x k +1 − x k ) + K
∂x xk

k +1 f (x k )
x =− + xk
 ∂f 
 
 ∂x  x k
Problem: If the initial guess is poor and the function is very curved, Newton’s
method will diverge.
Taylor Series Expansion in Multiple Dimension
Consider a nonlinear function f(x), where x = { x 1 , x 2 ,K , x N }
and f (x ) = 0
N
∂f
f (x) = f (x ) + ∑
0
By Taylor Series Expansion, we have ( x i − x i0 ) + K
i =1 ∂x i x i0
0
Note: for reservoir simulation problems, a good first guess ( x ) is usually
the values of the dependent variables at the old timestep.
Black-Oil Model – Fluid Flow Equations in Difference Form
The finite difference approximation for a standard black-oil model are:

Water:
(∆Tw ∆p w )i − (∆Tw γ w ∆D)i + (∆Tw ∆p w ) j − (∆Tw γ w ∆D) j + (∆Tw ∆p w ) k − (∆Tw γ w ∆D) k =
Vijk  φS 
∆ t  w  + q ww ijk
∆t  Bw  ijk
(1)
Oil:
( ∆To ∆p o ) i − ( ∆To γ o ∆D) i + ( ∆To ∆p o ) j − ( ∆To γ o ∆D) j + ( ∆To ∆p o ) k − ( ∆To γ o ∆D) k =
Vijk  φS 
∆ t  o  + q owijk (2)
∆t  Bo  ijk
Gas:
(∆Tg ∆p g ) i − ( ∆Tg γ g ∆D) i + (∆To R go ∆p o ) i − (∆To R go γ o ∆D) i + (∆Tg ∆p g ) j − (∆Tg γ g ∆D) j +
( ∆To R go ∆p o ) j − ( ∆To R go γ o ∆D) j + ( ∆Tg ∆p g ) k − ( ∆Tg γ g ∆D) k + (∆To R go ∆p o ) k − ( ∆To R go γ o ∆D) k =
Vijk  φS g S 
∆t  + φR go o  + (q ww ijk + R goijk q owijk )
∆t  B B o  (3)
 g ijk

In the equations above we have
(∆Tp ∆p p ) l = Tpl + 1 / 2 (p pl + 1 − p pl ) − Tp l−1 / 2 (p pl − p pl − 1 )
where subscript l=i,j,k and p=o,g,w.
( ∆Tp γ p ∆D) l = Tpl + 1 / 2 γ pl + 1 / 2 (D l+1 − D l ) − Tp l−1 / 2 γ pl − 1 / 2 (D l − D l−1 )
where subscript l=i,j,k and p=o,g,w.
( ∆To R go ∆p o ) l = R gol + 1 / 2 Tol + 1 / 2 (p ol + 1 − p ol ) − R gol − 1 / 2 To l−1 / 2 (p ol − p ol − 1 )

where subscript l=i,j,k.
(∆To γ oR go ∆D)l = γ ol +1 / 2 R gol + 1 / 2 Tol + 1 / 2 (D l + 1 − D l ) − γ oi −1 / 2 R gol −1 / 2 To l−1 / 2 (D l − D l −1 )

where subscript l=i,j,k.

In the equations above we have
n+1 n
 φS   φS   φS 
∆ t  o  =  o  −  o 
 Bo  ijk  B o  ijk  B o  ijk
n+1 n
 φS   φS   φS 
∆ t  w  =  w  −  w 
 Bw  ijk  B w  ijk  B w  ijk
n+1 n
 φS g S   φS g S   φS g S 
∆t + φR go o  =  + φR go o  −  + φR go o 
 B B o   B B o   B B o 
 g ijk  g ijk  g ijk
A i+1 / 2  k x k rp 
Tpi + 1 / 2 =  
∆x i + µ B 
 p p  i+1 / 2
where
- Ai+1/2 is the cross sectional area normal to the x-direction
at block boundary i+1/2.
- ∆xi+ is the distance between grid points i and i+1.

In a similar manner the expressions for T can be obtained in y and z
directions.
Solution of Nonlinear Equations : Newton-Raphson Method
Any nonlinear system of equations may be solved by the application of

Newton-Raphson (NR) technique.
This technique is extensively used in reservoir simulation.
Reservoir simulation flow equations are of the form:
F( y) = 0,
where the objective is to find y (unknowns for all the blocks at a given
time-step) so that the above equation is satisfied.
This equation is a statement of the conservation of mass (and energy in

thermal models) and when it is exactly satisfied the mass is also exactly
conserved.

The application of the NR method is done by the following iterative scheme:
( ν + 1) (ν )
(ν )
J δy = −F ,
( ν + 1) ( ν + 1) (ν )
where J is the Jacobian matrix, δ y =y −y is the change of the
unknown over the iteration and is the residual at the end of the nth
(ν )
F
iteration.
Elements of the Jacobian are obtained by differentiating, either

(ν )
analytically or numerically, all elements of vector F (eqs. 1,2 and 3
above ) with respect to all the unknowns.
In reservoir simulation it is convenient to use a notation that associates

equations and unknowns to grid blocks.

Let us consider the two dimensional grid shown in figure below.
There are two ways to identify a block:
-Through the i,j index (example: the central block in the grid system is block
3,2).
-Through the ordering index (example: the central block is placed in the eighth
position).
11 12 13 14 15
3
2 6 7 8 9 10
J=1 1 2 3 4 5
i=1 2 3 4 5

The ordering index is used to order elements of F and y . For the ordering
shown, the structure of the matrix J will be shown below.
1 2 3 4 5 6 7 8 9 1011 12 13 14 15
1 XX X
2 XX X X
3 XX X X
4 XX X X
5 XX X Note that:
6 X X X X The nonzero elements in a row i,j –

indicated by X – correspond to the
7 X X XX X derivatives of the equation
8 X X XX X corresponding to block i with respect
9 X X XX X to the jth unknown in .
y
10 X X X X
11 X X X For example, the elements of row 8
12 X X X X are:
13 X X X X
14 X X X X ∂F 3,2 ∂F 3,2 ∂F 3,2 ∂F 3,2 ∂F 3,2
K K
15 X X X ∂ y 3,1 ∂ y 2 ,2 ∂ y 3 ,2 ∂ y 4 ,2 ∂ y 3 ,3
Since more than one equation and unknown could be associated with each block, individual
elements are matrices or blocks. If we have three components flow then one possibility
is:  
 Foi , j 
F i, j =  Fw i , j  (4a)
 
F
 gi , j 
 po 
 i, j 
y i, j =  S wi, j  (4b)
 
S
 i, j 
g
The derivative of the equation for block (i,j) with respect to the unknowns in block (l,m) is:
 ∂Foi , j ∂Foi , j ∂Foi , j 

 
 ∂p o l ,m ∂S w l , m ∂S gl ,m  (4c)
 
∂F i , j  ∂Fw i , j ∂Fw i , j ∂Fw i , j 
=
∂ y l,m ∂p ∂S w l , m ∂S gl ,m 
 o l ,m 
 ∂Fgi , j ∂Fgi , j ∂Fgi , j 
 
 ∂p o l ,m ∂S w l , m ∂S gl ,m 
Note that the elements in a column contain derivatives with respect to the same variable while
elements in a row are derivatives of the same equation.
Equations for Fluid Flow through Porous Media - Introduction
Darcy’s Equations
Law of State
Kr and Pc
Conservation
relationships
of Mass Law
Non-linear partial
differential equations that
describe the fluid flow
through porous media
Eqs. For Multiphase Flow - Introduction
• Model simultaneously the flow of two or three phases in the

reservoir (water, oil and gas)
• Each phase can be compound of one or more component
• A conservation of mass law is applied to each component to
be considered in the modeling
• Darcy Law is modified to consider the effect of the presence
of one phase upon the flow velocity of the others
• The variables are, beside pressure, the saturations
(volumetric fraction of each phase) and the compositions of
each phase

Eqs. For Multiphase Flow – Main Models
“Black-Oil” Model:
• Three phases (water, oil and gas), three components
(water, oil and gas)
• Oil component only flows as an oil phase, water
component only flows as an water phase and component
gas can flow as gas phase or dissolved in the oil phase
• The gas quantity that is dissolved in the oil is a function
of the solubility ratio which is a function of the pressure
(determined experimentally)
• Represents very well the behavior of majority of
reservoirs and, it is because this, the most used
Eqs. For Multiphase Flow – Main Models
Compositional Model:
• Three phases (water, oil and gas), arbitrary number of
components (water + hydrocarbons)
• Water component in general only flows as a water phase
and the hydrocarbons components may flow as a gas or oil
phase
• The components distribution is done by means of a
“flash” procedure which is a function of the pressure and
compositions
• Much more complex and computationally intensive

Mathematical Model – Discretization Concept
Simulation
Cell
Functions of the Simulation Grid:
• Description of the variation of properties in space;

• Approximation in space of the differential equations:
∂P ∆P
≈
∂x ∆x
• Posting of results.

Pressure
Pressure
Time Time Steps
Pressure Variation Pressure Variation

with time in the with time in the
reservoir simulator
Mathematical Model – Internal Grid Representation
Properties and Variables vs. Nodes and Connections
Properties Variables
Centre of gravity Fluid pressure

Nodes Pore volume Fluid saturation
compressibility
Connections transmissibilities Fluid flow

Mathematical Model – Solution of the Discretized Eqs.
The non-linear system is solve by the Newton method:

• Iterative method: from an initial estimative successive
estimatives are obtained, hopefully, closer from the solution
• The estimative are obtained by approximation of the
problem by a linear problem at each iteration
• The linear system of equations must be solved at each
iteration
• The convergence tend to be slower with the increase of the
time step, reduction of the grid size and increase of flow
velocities

Some Comercial Reservoir Simulators Available
• CMG:
– IMEX (“Black-Oil”)
– GEM (Compositional)
– STARS (Thermal)
• Eclipse (“Black-Oil” - Geoquest)
• VIP (“Black-Oil” - Landmark)

*********************************************************************
** SECAO DE DESCRICAO DO RESERVATORIO **
************************************************************************
*GRID *VARI 88 48 11 ** 88 BLOCKS NA DIRECAO X
Input Data File ** 48 BLOCKS NA DIRECAO Y
** 11 LAYERS NA DIRECAO Y
*KDIR *DOWN
*DI *CON 125.0 ** DIMENSAO DOS BLOCOS NA DIRECAO I
*DJ *CON 125.0 ** DIMENSAO D0S BLOCOS NA DIRECAO J
(Part of an Example) **
*BWI 1.04300 ** FATOR VOLUME DE FORMACAO DA AGUA
*CW 55.00E-06 ** COMPRESSIBILIDADE DA AGUA DE FOR
*CO 25.23E-05 ** COMPRESSIBILIDADE DA OLEO ACIMA
*REFPW 190.0 ** PRESSAO DE REFERENCIA DA COMPRES.
*VWI 0.42 **0.37 ** VISCOSIDADE DA AGUA DE FORMACAO
*CVW 0.0 ** DEPENDENCIA DA VISCOS. DA AGUA
*************************************************************************
** TABELAS DE PERMEABILIDADES RELATIVAS
** ------------------------------------
*ROCKFLUID
**
*RPT **1
**
**SWC = 0.300000
**SORW = 0.200000
***********************************************************
*SWT
**
** PERMEABILIDADE RELATIVA AGUA-OLEO
** SW KRW KRO PCOW
0.3000 0.0000 0.7000 0.00 ** 1.000
0.3500 0.0035 0.4686 0.00 **.675
0.4000 0.0100 0.2595 0.00 **.479

Evolution of the Number of Cells in Reservoir Flow Studies
1,000,000
Númeroofdegridblocks
Células
100,000 Watts [1997]
10,000
Number
1,000
100
1960 1970 1980 1990 2000
year
Ano
Well Management: Designing & Controlling
Production Parameters

Overall Design of a Well-Management Routine
• Task Performed by the Well-Management Routine
– Simplest form: assigns user-specified well rates or pressures to individual
producing blocks in the simulator at specified times.
– More sophisticated perform the following tasks:
1. Shut-in, work over, recomplete, or redrill wells according to user-specified criteria (such
as WOR and GOR limits and minimum allowable oil rates)
2. Calculate wellbore and flow lone hydraulics
3. Initiate artificial lift (gas lift or pumps)
4. Switch flowing wells between production systems having different backpressures
5. Optimize the level and distribution of well rates to match field production-facility cpacities
and the availability of gas for gas lift.
6. Alter any of the preceding to maintain targeted reservoir or field oil or gas rates
7. Control water or gas injection rates to maintain a specified average reservoir pressure.
8. Return produced water to specified locations in the reservoir.
9. Perform coning calculations.
10.Relate the producing GOR and WOR of a well to gridblock average saturations by means
of well pseudofunctions.
11.Allow manual override by the user.
12.Check well-management-routine data for completeness and consistency.
13.Translate literal well names into well-sequence numbers that the simulator can interpret.
Overall Design of a Well-Management Routine
• Production Constraints
– Well Productivity (function of reservoir rock and fluid properties, fluid saturations
and type and effectiveness of well completions. It is affected by reservoir
permeability, fluid viscosity, oil bubble point pressure, relative permeability,
perforation quantity and quality, fraction of formation open to flow, well stimulation,
well drainage volume, and in gas wells, turbulent flow near the wellbore)
– Wellbore Hydraulics (is influenced by BHP, wellhead pressure, measured and
vertical depth, GLR, WOR, tubing size and lift mechanism)
– Surface System Hydraulics (is influenced by choke size, flowline size and
configuration, separator backpressure and the number of flowing phases)
– Surface Facility Capacity
– External production constraints
• Statutory Controls
• Operating Guidelines
• Wellbore Utility and Availability
• Economic Production
• Market Constraints
• Disposition of Produced Gas
• implest form: assigns user-specified well rates or pressures to individual producing

Logic Structure
• Typical Decision Rules
1. Shut in wells at prescribed WOR and GOR limits
2. Convert wells to artificial lift as dictated by low wellhead pressures or low
flowing well rates that result from declining reservoir pressure and/or
increasing water production.
3. Implement response to statutory limits by compliance with production
penalties for high GOR’s, limits on lease gas production, lease, platform,
and field production allowable, and concession or production-sharing
contract rules.
4. Add new completions in response to WOR and GOR performance of wells
by implementing workovers and recompletions, or by drilling new wells.
5. Allocate production among wells and producing intervals for control of
coning and fluid-contact movement.
6. Control pressure by a) scheduling the injection/production ratio and the
timing of injection-capacity increments to maintain reservoir pressure for
optimal natural flow or gas lift efficiency, or maximum recovery efficiency
and b) keeping injection pressures below the reservoir fracture pressure.

Logic Structure
• Well and Facility Classification
Well Level Classification
Group or Reservoir-Level Classification
• Hierarchy of Control

Logic Sequence
• History Mode
• Prediction Mode
– Setting Production and Injection Rates at Individual Wells,
Groups of Wells, and for the Reservoir
– Workover Algorithm
– Drilling Algorithm

Individual Well Behavior
Individual Well Capacity is governed by the following factors:
– Instantaneous well inflow performance relationship (IPR)

– Completion details
– Wellbore and surface system hydraulics
– Well stimulation

Individual Well Behavior
• Well Inflow Rates and Pressures

– Gridblock Effects
– Multiple Wells in Grids
– Allocation of Production Rates in Multiblock well Completions
– Near-Wellbore Effects
– Saturation Effects
– Inflow Performance Relationships
– Well Pseudofucntions
– Coning Functions
– Injection-Well Rates and Pressures
• Well Outflow Performance
– Water Injection
– Gas Production and Injection
– Gas/Liquid Production
Operating Conditions
Reasons why rates or pressures were maintained or

changed with time.
Data Requirements

Modeling Concepts
The basic steps required for a reservoir simulation study are:
Select one or
more images for simulation
Input to simulator
Select grid & average
data to get block
properties
Images of Input Output Analysis of

Reservoir Data Results
History Matching
Data

In general, an engineer interested in doing a reservoir simulation
study would follow the following steps:
- Develop study objectives

- Develop or select an appropriate simulator
- Review, collect and estimate appropriate data
- Make preliminary runs to establish model parameters and
limitations
- Match available history
- Predict performance under different operating scenarios
- Analyze results and prepare a report
- Plan additional work

Pre-Processors
Software Available
Builder EarthVision FloGrid

GeoSim Geolink Geostat
Gocad Grid Gridgenr
Gridstat GViz Heresim
Irap RMS Isatis Mapper
Open-GL Petrel Property 3D
ResMod (RC2) ResScale ResView
Shapes SigmaView SimGrid
SimUp SolidGeo StatMod
Storm Stratmodel SureGrid
.... .... ...

Reservoir Simulation Study Steps
1. Facies, Petrophysics and Seismic Inversion;
2. Description of the Reservoir Geometry and

Geological Structure: Horizons and Faults;
3. Geostatistical Modeling for Reservoir

Properties (Facies, phi, k);
4. Construction of the Simulation Grid and

Reservoir Model
Reservoir Simulation Study Steps
Field Development Project
• Analytical Phase
• Modeling Phase
• History match Phase
• Prediction Phase
Feasibility Study

Complimentary Topics - Trends
Challenges in Reservoir Description
for Flow Simulation and Forecasting
• What to describe?
• Which is the adequate support for description?
• How to describe property between wells?
• How to incorporate dynamic data information?
• How to deal with uncertainty?
Representation of Heterogeneity
3-D Grids
Up-Scaling
Geostatistics
Uncertainty Quantification and Risk Analysis
Determination of Critical Heterogeneities
to Fluid Flow
• Quantification of the Impact of Geological Heterogeneities

on Fluid Flow for Diverse Depositional Systems;
• Orientation and Optimization of Reservoir Characterization

Efforts to Flow Simulation (Model-Centered Approach);
• Identification of the More Adequate Techniques to

Incorporate the Effect of Heterogeneity into Flow Simulation
Models.

Introduction - Two Remarking Trends:
• In Reservoir Characterization:
Migration from 2D to 3D Methodologies
• In Simulation:
Use of more Flexible Grids

Grid Formats -Objectives
• Adequate Strategies to the Especification of Volumetric

Grids in Integrated Studies
• Strategies to the transference of information between

Characterization Grids and Simulation Grids
(Geometrical Approach x Topological Approach)
• Better Representation of the Fault Geometry in

Reservoir Simulation Models
• Development of Upscaling Techniques

Base Map to Simulation Grid Examples [Hales, 96]

Cartesian Grid Example [Hales, 96]

Orthogonal Curvilinear Grid Example [Hales, 96]

Corner Point Grid Example [Hales, 96]

Voronoi or PEBI Grid Example [Hales, 96]

Terminology in Grid Classification
• Shape (Topology and Geometry) x Properties

• Confusing Terminology
Example: Regular Irregular [Treinish, 92]
• Grid Regularity
Any characteristic that allows a more compact
representation [Haber et al., 91]
• Connectivity Relationships between Cells

Example: Strutuctured Grids x Voronoi
Classification for Practical Purposes
More Compact Representation

Type 1
Type 2
Type 3 (Standard)
More Flexible
More Regular
Structured Type 4
Corner
Type 5 Point
Type 6 Grids
3D Type 7
Grids Others
Non Structured

Specification Strategies of Volumetric
Grids in Integrated Studies
1. Independent Grids
for Characterization and Simulation
2. Overlapping Grids
for Characterization and Simulation

Conclusions
• Increasing use of Volumetric Grids for Reservoir

Representation
• Actual Stage: Substitution of Grid Type 3 by Type 4
• Grid Evolution brings new opportunities but also new

challenges for Modelling people
• Grid Specification in Integrated Studies is a Critical

Task

Upscaling

Equivalent Homogeneous Medium

Limitations of the
Upscaling Process
Fundamental Questions:
The Desire of Having the Same Results with Less Information!
Operational Questions:
Non Uniqueness with the Computation Methods.
Practical Questions:
One (or More than One) Pseudofunction per Gridblock.
Functional Questions:
The Fine Grid Results are not Reproduced...
Adequate Approach:
To Have in Mind the Approximate Character of the Solutions.

Absolute Permeability:
Computation Methods
• Simple Formulas:
Arithmetic, Geometric and Harmonic Means
Power Law Average (w)
Combined Averages (Le Loc’h)
• Single Phase Simulation:

Finite Difference
Finite Elements

Power Law Average:
1 w 1/w
Kw =( ∑ k )
n
Particular Cases:
w = 1 : Arithmetic Mean
w --> 0 : Geometric Mean
w = -1 : Harmonic Mean

Equivalent Relative Permeabilities Curves
(Kr Pseudo Functions)
Are Function of:
• Spatial Distribution of Local Properties;
• Viscous x Capillary x Gravitacional Forces;
• Initial Conditions;
• Relative Distance to the Injection Well;
• Numerical Dispersion.
Equilibrium Methods
Dynamic Methods
Geostatistical Simulation Methods
• Comparison Between Available Methods
• Development of Selection Criteria to

Identify the most Adequate Method for
Geostatistical Simulation
• Incorporation of Seismic Data
Streamline Simulators
Motivation
Conventional Simulators:
– Complex Models: high computational cost
– Limited grid dimension
Streamline Models:
– Faster runs for a similar grid
– Lower computational cost
– More refined grids

Applications
• Fast Analysis of Exploitation Alternatives

• Ranking of geostatistical images
• Upscaling
• Uncertainty analysis

Available Techniques
Streamline Based Models
Characteristics:
– Numerical calculation of the velocity field
– Streamline generation
– 1-D Saturation solution at each streamline
– Numerical dispersion and grid effect reduction

Available Techniques (cont.)
Different Treatment to Pressure and Saturation
Characteristics:
– Implicit solution for pressure a few times during the
simulation run
– Explicity solution for saturation
– More efficient and with the same numerical
dispersion level that the conventional simulators

Streamline Based Flow Simulators
• Commercials
– FrontSim (TSC - Noruega)
– StreamSim (Stanford)
– RC2 (Datta-Gupta - UT)
• Multiclient Projects
– UTSTREAM (UT)
• In-house Development
– 2-D, uncompressible, gravitational effects

FRONTSIM - Characteristics
• 3D, two-phase, compressible

• Gravitational effects
• Corner-point grid
• Well or group pressure or rate control
• Multiphase vertical flow table
• Pre and Pos-processor

FrontSim – Application Examples
• Heterogeneous 3-D case with 25000 cells

• Heterogeneous 3-D case with 225000 cells
• Real Field A Example

Conclusions
•Use for qualitative evaluations of reservoir behavior;
•Use for the ranking of geostatistical images;
•Problems as a quatitative evaluation tool (history match).

Background
Automatic History Matching Techniques
The proposed methodologies borrow heavily from

techniques of:
• optimization;
• reservoir simulation;
• geological and geostatistical modeling.

Motivation
Traditional trial and error approach:
• Very frustrating and time consuming;
• Modifications in the geological model are generally made
on arbitrary basis;
• Static and dynamic data are not integrated;
• Many attempts were made to automate the process, but
only recently some degree of success was achieved.

Conditioning Methodology
• Mathematical Model Construction;
• Objective Function Definition;
• Application of a Minimization Algorithm.

Objective Function
Deviation between simulated and observed data:
)= ∑ ( ) = (dsim − dobs ) CD−1(dsim − dobs )T

obs 2
Om ( est disim − di
(obs 2
σi )
• Good smoothness properties;
• No prior information taken into account;
• No regularization, parameter identification problem
expected to be strongly ill-posed.

Objective Function
Inclusion of a prior term:
(
Om est
) = (m − m ) ( − m )
est prior −1 est
CM m prior T
+
(d − d ) CD (d − d )
sim obs −1 sim obs T
• Good smoothness properties;

• Prior information included;
• Quadratic nature of prior term expected to improve “well-
posedness” of parameter identification problem.

Objective Function
Bayesian interpretation:
• Assuming:
– parameters and observed data are gaussian variables;
– CM is the covariance matrix for the prior model;
– σ est are the standard deviations of measured data.
i
• The a posteriori PDF for the parameters is given by
( (
π = a exp − O mest ))
( )
• Minimization of O mest gives the most probable model;
• Ideally, one would like to sample π .

Minimization Algorithm
Two broad classes:
• Gradient-based methods (steepest descent, CG, Gauss-
Newton, Levenberg-Marquardt, etc)
– fast convergence;
– only find local minima;
– require changes in simulator code to get derivatives in a
efficient way.
• Non-gradient methods (simulated annealing, genetic
algorithms, etc)
– low convergence;
– able to find global minimum;
– simulator can be treated as a black-box.
Minimization Algorithm
Gauss-Newton:
• Probably the most used method in applications;
• Requires to solve a system of order equal to the number of
parameters per iteration;
• Requires inversion of CM;
• Most computer demanding step is the construction of
derivatives matrix.

Calculation of Derivatives in
Reservoir Simulators
How to achieve it?
• “Perturbation” of parameters is easy to implement
(simulator can be taken as a black box);
• Analytic techniques can have a dramatic improvement
over “perturbation”, but require modifications in the
simulator’s code.

Comments
Î Valuable methodology to reservoir heterogeneity
representation and uncertainty minimization;
Î Emerging technology.

Value of Information Techniques for
Ranking of Appraisal Wells
Paper SPE 36631

Primary Variables Uncertainty Range
P90 P50 P10

Phi (%) 27 30 33
So (%) 72 80 88
H (m) 17,5 25 30
6 3
VOIP (10 m ) 341 614 883

Comments
• The results of the process are mainly influenced
by the probabilities adopted.
• This process does not define the appraisal

program; How many wells and which location?

Virtual Reality Centres
- integration environment -

Introduction
The first Virtual Reality Centre was

inaugurated 07/12/1994, in Berkshire,
England, at Silicon Graphics
headquarters.

Aplicability of VRCs
1. Manufacture, architecture, engeneering and civil

construction
2. Petroleum
3. Transportation
4. Universitities
5. Government
6. Services and entertainment
7. Telecomunications

Styles of collaboration environments
Immersion
Immersion
Virtual Reality
iCentre
Luciane Cunha
Collaboration
Collaboration 165
Reservoir Model Characteristics
• Non structured grids
• 3D Models
• Fault Representation
• Seismic Information
• Lateral K Variation
• Geoestatistic
• Well Test Information
• Different Rock Regions
• Absolute K Upscaling
• Pseudo-Functions Calculation (Kr)
Simulation Grid Characteristics
• Standard Block Centered Grid

• No Local Grid Refinement
• 2D Maps
• Fault Representation
• Seismic Information
• Lateral K Variation
• Well Test Information
• Different Rock Regions
• Non zero Pc
Well Testing Objectives
• Reservoir Evaluation – to reach a decision how best to produce a given reservoir
– Conductivity (kh product)
– Initial reservoir pressure
– Reservoir limits
• Reservoir Management – to monitor performance and well condition
– Changes in average reservoir pressure
– Monitoring the condition of the wells (workover and stimulation candidates)
– Track fluid movements
• Reservoir Description – to aid in forecasting reservoir performance
– Communication
– Development of the production plan
+++++++++++++++++++++++
In summary:
– change in a variable (pressure/rate) with time
– most of the tests will measure change in pressure (∆p)
To determine:
– reservoir parameters (calculate) k, S, Cd, radius of investigation;
– reservoir system description (qualitative).

Well Test Types
• Drawdown or Flow Test
– Measurement of BHP vs. time for at the constant flow rate
– Measurement performed while the well is flowing
– Test starts when the well is shut-in
• Buildup
– Well already flowing at the constant rate is shut-in and the BHP is
measured
– No production
• Injection or Injectivity Test
– Similar to drawdown test except fluids are injected at a constant rate
rather than produced
• Falloff
– Similar to buildup test except that shut-in follows injection
• Interference
– Measurement of BHP vs. time at the observation well when a disturbance
is created at an active well
• Drill Stem Test (DST)
– Combinations of drawdown and build up done at time of drilling a well

Ideal Reservoir Model
Assumptions:
• radial flow of fluid to a well in a infinite cylindrical

reservoir
• homogeneous isotropic (relatively isothermal) porous
medium of infinite extent
• constant flowrate (q)
• constant pressure at beginning of test pi
• uniform thickness (h)
• single phase fluid of small and constant compressibility

Principles of Well Testing
• Inverse problem
• Diffusivity Equation
– Conservation of mass (continuity equation)
– Flow equation (Darcy’s law)
– Equation of state
∂ 2p 1 ∂p 1 φµC t ∂p
2
+ =
∂r r ∂r 0.0002637 k ∂t
• for radial coordinates and slightly compressible fluid

John Lee, Well Testing, SPE Textbook Series Vol. 1

Diffusivity Equation
• Assumptions used in derivation of diffusivity equation (liquids)
– Porous medium of uniform thickness with homogenous, isotropic
properties
– Rock and fluid properties independent of pressure
– Small pressure gradients
– Applicability of Darcy’s Law
– Negligible gravity forces
• Solution
– Bounded Cylindrical Reservoir (Exact Solution)
– Infinite Reservoir with Line Source Well (rw=0)
– Pseudo-steady (Semi-steady) State Solution
– Infinite Reservoir with Wellbore Storage


Exact Solution
• Assumptions
– Well produced at constant rate
– Well of radius is centered in a cylindrical reservoir of radius
– Uniform pressure before production begins
– No flow across outer boundary
qBµ  2t D 
∞ 2
3 e − αnt D J 12 (α n reD )
p wf = p i − 141.2 
kh  reD 2
+ ln reD − + 2 [ 2 2
4 ∑ 2
]
n =1 α n [ J 1 ( α n reD ) − J 1 ( α n )] 
 
– Where
re kt
reD = t D = 0.000264
rw φµC t rw2
– Where α n are the roots of J 1 and Y1 (Bessel Functions)
J 1 ( α nreD )Y1 ( α n ) − J 1 ( α n )Y1 ( α nreD ) = 0

Infinite Cylindrical Reservoir
• Assumptions
– Well produces at constant rate
– The well has zero radius (i.e. line source approximation)
– Reservoir is at uniform pressure before production starts
– Well drains infinite area (i.e. p → p i as r->∞ )
qBµ  948φµCt r 2 
p = p i + 70.6 Ei  − 
kh  kt 
Where Ei = exponential integral
∞
e −u
E i (− x) = −
u ∫
x
du

How do we get values for Ei(-x)?
• For 10 .9 ≥ x ≥ 0.02 : Use a table
• For x < 0.02 (with less than 0.6% error)

Ei ( − x ) ≅ ln( γx ) γ = 1.781 → Euler’s constant
E i ( − x ) ≅ ln(1.781x )
• For x>10.9 : E i ( − x ) ≅ 0
When is it appropriate to use the Ei solution?
 φµCt rw2
5
  φµCt re2 
3.79x10  < t < 948 
 k   k 
   
For most well tests (in the region of interest) the logarithmic approximation is used

Summary
Flow in the “infinite acting” flow period can be described by the line
source (Ei) solution.
qBµ φµCt r 2
p = p i + 70.6 Ei ( −948 )
kh kt
for most times of interest we can use log approximation
qBµ 948φµCtr 2
p = p i + 70.6 [ln(1.781 )]
kh kt
Flow influenced by the wellbore (i.e. rw = 0 assumption not valid) or flow
influenced by boundaries/discontinuities can’t be handled by equation.

Wellbore Damage
In practice, most wells have a zone of reduced permeability near the
wellbore resulting from drilling and completions operations.
One way to represent this is to consider a zone of altered permeability

of ks having a radius rs directly around the wellbore.
This will cause an additional pressure drop across the zone ∆pS as
compared to the same well with no damage.
Note: pwf = pwf - ∆pS (skin)

Skin damaged area rs<<small

This additional pressure drop can be represented by the
steady state radial flow equation as:
qBµ  rS  qBµ  rs 
∆p S = 141.2 ln  − 141.2
 ln 

k S h  rw  kh  rw 
(Hawkins steady state Model)

qBµ  k   rs 
∆p S = 141.2 
 − 1  ln 
kh  k S   rw 
qBµ  k  r 
∆p S = 141.2 S where S =  − 1  ln s 

kh  kS   rw 
This is called the “thin skin” model as rs is considered to be

infinitesimally small.
Wellbore Damage (Important Comments)
S > 0 indicates damage
For S > 0, ks < k  k   rs 

 − 1  ln >0
k  r 
 S   w 
r 
We know rs > rw , and thus ln s  > 0
 rw 
The greater the contrast between ks and k, and/or the deeper into the
formation the damage extends, the larger the value of S. There is no
upper limit on the value of S.
 k  k
∴  − 1  > 0 > 0 ⇒ kS < k
 kS  kS
Some newly drilled wells will not flow at all before stimulation. For these
wells k S ≈ 0 , and S → ∞

Positive Skin Damage

Skin can also be negative. This implies that the flow characteristics near
the wellbore are better than flow characteristics in the reservoir.
For s<0, k s > k
Negative skin indicates a “stimulated” well. The deeper the extent of

stimulation, the more negative skin becomes.
In practice the lower limit on skin is –infinite. This usually represents a
well with deeply penetrating, highly conductive hydraulic fractures.
Not all stimulated wells have negative skin. If we take a well that has
S → ∞ , and stimulate it, we may end up with a well of S = 2053, which
would be a success, if the well flows at economic rates.
If a well is neither stimulated nor damaged then k = ks and S = 0.
The equation for pressure drop for skin can only be used for the
damaged/stimulated zone, it should not be used to determine
pressures at radii such that rw < r < rs

Negative Skin Improvement

Adding the expression for skin to the Ei solution gives
qBµ  948φµC t rw2 

 + ∆p s
p i − p wf = −70.6 Ei −
kh  kt 
 
qBµ  948φµC t rw2    

 − 2 k − 1  ln rs 
= −70.6 Ei − k  r
kh  kt   
   s  w 
 2 
qBµ  948φµCtrw 
pi − pwf = −70.6 [Ei − − 2S]
kh  kt 
 
Using log approximation:
qBµ  1688φµC t rw2   k

 − 2
  rs 
p i − p wf = −70.6 [ln − 1  ln ]
kh  kt   ks 
     rw 
 2
qBµ  1688φµCt rw 
p i − p wf = −70.6 [ln  − 2S]
kh  kt 
 
qBµ 1  1688φµC t rw2 

p i − p wf = −141.2 [ ln  − S]
kh 2  kt 

Skin Effect
• Change in permeability caused by drilling/completion operations or

stimulation operations
• Causes additional pressure drop (positive Value of S)
• Hawkins steady-state model
qBµ  k   rs 
∆p S = 141.2 
 − 1  ln 
kh  k S   rw 
• Expressed as skin factor,S
 k  r 
S =  − 1  ln s 
 kS   rw 
• S varies from –10 up to + infinity
−S
• Effective wellbore radius reff = rw e

Pseudo-Steady State (PSS)
– Assumes a finite cylindrical reservoir of radius re, with the
well in the center
– This system, when the well is flowing, and pressure versus
time is measured, exhibits three distinct flow regimes.
» a transition period
» an infinite acting transient response, and
» a pseudo-steady state flow period
[ The PSS flow period is characterized by the pressure change with time being
constant (dp/dt=C) at all points in the reservoir ]
PSS occurs at
φµCt re2
t > 948
k
and cannot be analyzed using the line source solution.
Consider the exact solution:

qBµ  2t D 3
∞
e − αn2 t D 2
J 1 (α n reD ) 
p wf = p i − 141.2  2 + ln reD − + 2 [ 2 2
kh  reD 4 ∑ 2
]
n=1 α n [J 1 (α n reD ) − J 1 (α n )] 
 
where
re kt
reD = t D = 0.000264
rw φµC t rw2
At large times, the exponential term and the Bessel

functions cancel each other and the term goes to zero.

The PSS solution is then
qBµ  2t D 3 
p wf = p i − 141.2  2 + ln reD − 
kh  reD 4 
or in dimensional variables,
qBµ  0.000527kt  re  3 
p wf = p i − 141.2  + ln − 
kh  φµC t re2 r  4 
 w 

Solutions with Skin Effect
• Skin effect causes additional pressure drop

• Infinite acting reservoir:
– Skin factor affects only if the pressure point is within the damaged
zone (normally the wellbore)
qBµ  948φµCt rw2  qBµ  1688φµCt rw2 

pwf = pi + 70.6 [Ei − − 2S] = pi + 70.6 [ln − 2S]
kh  kt  kh   kt 
  
• Pseudo Steady State:

qBµ  0.000527kt  re  3 
pwf = pi − 141.2  2
+ ln  − + S
kh  φµCtre  rw  4 
What if the reservoir is not cylindrical?
In 1971, Ramey and Cobb published a paper detailing a

general theory for a well in a closed drainage area
1 A 1  2.2458 
p D = 2πt DA + ln( 2 ) + ln 
2 rw 2  C A 
where CA is the shape factor (It is a geometric factor
characteristic of the system shape and well location)
and 0.0002637kt kh
t DA = pD = (p i − p )
φµCt A 141.2qBµ

PSS Equation Variations
Consider the PSS Equation
qBµ  0.000527kt  re  3 
p = pi − 141.2  2
+ ln  − 
kh  φµCt re  r  4 
by differentiating the above equation

∂p 0.0744qB
=−
∂t φCt re2h
and substituting liquid filled pore volume of the reservoir

Vp = πφre2h
we have
∂p 0.234qB
=− ⇒ determining reservoir size from
∂t Ct Vp the rate of pressure decline in the
wellbore in time
replacing pi with average pressure p within the drainage volume of

the well from material balance
∆V 5.615qB(t / 24) 0.0744qBt
pi − p = = 2
= 2
Ct V Ct ( πre hφ) φCt hre
and then

0.0744qBt 0.0744qBt qBµ   re  3 
p=p+ − − 141.2 ln  − 
φCt hre2 φCt hre2 kh   r  4 
qBµ   re  3 
p = p − 141.2 ln  − 
kh   r  4 
considering Skin
qBµ   re  3 
p = p −141.2 ln  − + S
kh   r  4 

Average pressure concept:
qBµ   re  3 
pwf = p − 141.2 ln  − + S
kh   rw  4 
Equations for generalized geometry:

qBµ  1  10.06A  3 
pwf = p − 141.2  ln 2 
− + S
kh  2  CArw  4 
where:
– A = drainage area
– CA = shape factor (drainage area and well location dependent)

Steady State (Constant Pressure, i.e. oil rate change)
When pressure at every point in the system does not vary
with time (dp/dt=0), flow is said to be steady state.
In reservoirs, SS flow can occur only when the reservoir is
completely charged by a strong aquifer, or when
production and injection are balanced.
In this case, Darcy’s law in radial form is represented by
kh(p e − p w )
q=
Bµ ln(re / rw )
1. a transition period
2. infinite acting
3. pseudo-steady state

Dimensionless Variables
• In pressure transient analysis it is convenient and customary

to present graphically or tabulated solutions to flow equations
in terms of dimensionless variables.
• In this way it is possible to present compactly solutions for a
wide range of parameters: φ, µ, S, Cs and k and variables r, p
and t.
• Dimensionless parameters and variables are not unique. I.e.
there is more than one choice for each.
• Rather, these are defined not derived quantities.
• The definitions are chosen such that we can characterize the
flow situation and its solutions.

Standard Definitions:
0.00708kh kh r
pD = (p i − p ) = (p i − p ) rD =
qBµ 141.2qBµ rw
0.002637kt based on wb radius

tD =
φµC t rw2
0.002637kt  rw2 
t DA = = tD  based on the drainage area
φµC t A A 
 
1  rD2 
p D ( t D , rD ) = − E i −
2  4t D 
1  tD 
log approximation p D (t D , rD ) = − [ln  + 0.80907]
2  rD2 

Wellbore Storage Coefficient (CS)
• Consider a shut in oil well in a reservoir of uniform unchanging pressure.

• Reservoir pressure will support a column of liquid to some equilibrium
height in the wellbore.
• If the valve at the surface is opened and flow is initiated, the first oil
produced will be that stored in the wellbore, and the initial flow rate from
the formation to the well will be zero.
• With increasing flow time, at constant surface producing rate, the downhole
flow rate will approach the surface rate and the amount of liquid stored in
the wellbore will approach a constant value.
• This phenomenon is called wellbore storage (also called after production
after flow, wellbore loading/unloading).
• Wellbore storage has long been recognized as affecting short time transient
pressure behavior.


Wellbore Storage Coefficient
• Wellbore storage coefficient (Cs) is the volume of fluid that
the wellbore will produce due to unit drop in pressure
– General: ∆Vwb
Cs = , bbl / psi
∆p
– For falling liquid level:

25.65Awb
Cs = , bbl / psi
ρ
– For liquid expansion case:

Cs = CwbVwb, bbl/ psi
Where
Vwb – volume of the wellbore
Cwb – compressibility of the wellbore fluid
Awb – cross sectional area of the wellbore
ρ - fluid density (liquid)
• Dimensionless wellbore storage coefficient:

0.894Cs
CsD = 2 (Field units)
φCthrw
Ct = total compressibility of reservoir system
(Ct = Cf + CoSo + CwSw +Cg)

*wellbore storage causes the sandface flowrate to change
more slowly than the surface flow rate
24Cs dp w
q sf = q +
B dt
dp D (t D , CsD )
q sf = q[1 − CsD ]
dt D

Wellbore Storage (Afterflow)
• Early transient response is characteristic of the wellbore, not the reservoir
• Variable production rate after the surface valve is opened
– Initial production from expansion of fluid stored in the wellbore
– Initial flow rate from sandface is zero
– With increasing flow time sandface rate approaches surface production rate
• Two common types of wellbore storage:
– Changing liquid level
– Fluid expansion
• Distorts Cartesian and semi-log diagnostic plots
• Unit-slope line is present on the log-log pD vs tD plot
• Can cause conventional interpretation impossible (if flow time is short)
• Type curves used for interpretation
– Analytical solution to diffusivity equation with wellbore storage
– For given tD and pD skin factor and CD can be estimated

Superposition
We have considered reservoir/well systems with a single well producing at
a constant rate, starting at time t=0, in an infinite or finite reservoir.
To extend the theory to more general systems we use the PRINCIPLE OF
SUPERPOSITION.
Superposition states that adding two or more solutions of a linear
differential equation results in a new solution to that differential
equation.
This new solution represents a solution of the original equation with
different boundary conditions.
For our purposes, we can state the principle of superposition as the total
pressure drop at any point in a reservoir is the sum of the pressure
drops at that point caused by flow in each of the wells in the reservoir.

Example 1: Three wells in an infinite reservoir
Well A
rAC rAB
Well B
Well C
At well A:
p i − p wf = (p i − p ) due to well A + (p i − p ) well B + (p i − p )well C
q A Bµ  948φµC t rwA
2 
p i − p wf = −70.6 {E i  −  − 2S A }
kh  kt 
 
q B Bµ  948φµC t rAB
2 
q
 − 70.6 C B µ  948 φµ C r 2 
− 70.6 Ei − E − t AC 
kh  kt  kh
i  kt 
   
Note: infinite reservoir use Ei solution. In dimensionless form:
Bµ
∆p A = 141.2 { ∑ q i p D ( t D , rD ) + S A } .
kh
Equation above assumes homogeneous thickness of reservoir at all wells

(i.e. cannot use if thickness varies in a reservoir).

Superposition
No flow
between
1 and 2
1 2 John Lee, Well Testing, SPE Textbook Series Vol. 1

Example 2: Pressure Behavior in Bounded Reservoirs
Consider a well a distance L from a single no-flow boundary.
Mathematically, this is identical to two wells at a distance of 2L apart,
flowing at the same rate q.
It can be shown, using potential theory, that a line equidistant between
the two wells is a no flow boundary (i.e. Dp=0)
The problem reduces to a simple two-well in an infinite reservoir
problem.
qBµ  948φµC t rw2 

p i − p wf = −70.6 {E i  −  − 2S}
kh  kt 
 
This technique is called the method of images and can be used to model
complex well/boundary situations.

Superposition in space x Superposition in time
Superposition in time is used to model variable-rate producing wells.
Consider a reservoir that has the following production history:

• From well 1 producing at q1 starting at t = 0
q1Bµ  1688φµCt rw2 
 − 2S}
∆p1 = (p i − p wf )1 = −70.6 {ln
kh  kt 
 
• From well 2 producing at q2-q1, starting at t = t1
(total elapsed time for well 2 is (t-t1))
(q 2 − q 1 )Bµ  1688 φµC t rw2 

 − 2S }
∆p 2 = (p i − p wf ) 2 = −70.6 {ln
kh  k (t − t ) 
 1 
• From well 3 producing at q3-q2, starting at t > t2 (elapse
time = t - t2)
(q 3 − q 2 )Bµ  1688φµC t rw2 
 − 2S }
∆p 3 = (p i − p wf ) 3 = −70.6 {ln
kh  k (t − t 2 ) 
 
∆p = (p i − p wf ) = ∆p1 + ∆p 2 + ∆p 3
Bµ  1688 φµC t rw2 

∆p = (p i − p wf ) = −70.6
kh ∑ (q i − q i − 1 ){ln
 k (t − t )
 i −1
 − 2S }


In dimensionless form
Bµ
∆p = 141.2
kh ∑ (q i − q i − 1 )p D (t − t i − 1 ) D , rD ) +S}

Productivity Index (PI or J)
From before (PSS Equation)…
qBµ  1  10.06A  3 
pwf = p − 141.2  ln 2 
− + S
kh  2  CArw  4 
q 0 .00708 kh
J= =
p − p wf  1  10 .06 A  3 
B µ  ln   − + S
2 
 2  A w  4
C r 

Radius of Investigation (or Radius of Drainage)
The radius of investigation defines a circular system with a PSS pressure
distribution from the well to the “drainage radius”.
Stated another way the radius of investigation is the distance a

significant pressure disturbance is propagated by production or
injection at a constant rate.
As time increases, more of the reservoir is influenced by the well and rinv
increases.
1/ 2
 kt 
rinv =  

t in hours , r in feet
 948φµ Ct 

Note that rinv is not dependent on flow rate
but on time and flow properties
Radius of investigation has several uses in analysis and test

design. A qualitative use is to help explain the shape of a
BU (Buildup) or DD (Drawdown) curve. Rinv provides a
guide for well test design. If we want to know the
properties of the reservoir at least “x” feet from the well,
we don’t have to guess at the length of time to run a test.

Type Curve Analysis
• Type Curves are commonly used to characterize damage and
stimulation of the tested well, and to determine formation
permeability
• Most of these curves were generated by simulating constant rate
pressure drawdown tests. However, most can be applied to buildup
tests by using an “equivalent time” transformation.
• The advantage of type curve analysis compared to conventional
analysis is that type curves may allow test interpretation even when
wellbore storage distorts most or all of the test data. They are also
useful for fractured well analysis.
• Fundamentally a type curve is a pre-plotted family of pressure-
drawdown curves.

The most fundamental of type curves is a plot of
dimensionless pressure change PD vs. dimensionless time tD
Type curves are generated by obtaining solutions to flow equations (ex.

Diffusivity equation) with specific initial and boundary conditions.
Some solutions are generated analytically others are based on
numerical solutions.
The family of type curves of this type most often used were generated
analytically by Ramey and Agarwal. They are referred to as Ramey’s
curves or Agarwal’s curves or storage skin model. They have two
parameters in that distinguish the curves from one another: skin factor,
S, and wellbore storage CD. For an infinitely acting reservoir, the
specifications of CD and S uniquely determines the value of PD at a
given value of tD
(Note that rD does not enter the equation, as it is r/rw and as we are
measuring the pressure at the wellbore rD=1)

Build-Up Test
• Procedures:
– Producing a well at the constant rate
– Shutting the well in (mostly at the surface)
– Record the pressure as a function of time
• Application:
– Estimation of reservoir permeability
– Estimation of skin effect
– Estimation of drainage-area pressure
– Detection of reservoir heterogeneities and boundaries
• Interpretation:
– Horner’s Method
– Matthews-Brons-Hazebroek (MBH)
– Modified Muskat Method
– Type Curves
Ideal Build-Up Test
An ideal Build-Up test is a test in an infinite homogenous, isotropic
reservoir, containing slightly compressible, single phase fluid with
constant fluid properties. Any wellbore damage or stimulation is
considered to be concentrated in a skin of zero thickness at the
wellbore. At the instant of shut-in, flow into the wellbore ceases
totally.
Make the following assumptions:

1) The well is producing from an infinite acting reservoir (no boundary
effects are felt during the flow or subsequent shut-in period)
2) Formation and fluids have uniform properties so that the Ei function
may be used (and also the ln approximation)
3) The well is produced at a rate q for a time tp and is then shut-in for a
time ∆t. Total time for the test is tp+ ∆t.

Horner’s Method
• Assumptions:
– Well producing from infinite acting reservoir (both flow and
build-up period)
– Formation and fluid properties are uniform
– Horner’s pseudoproducing time approximation applicable
qBµ t p + ∆t
p ws = pi − 162.6 log( )
kh ∆t
• Where:
– Semi-log plot straight line slope: m = -162.6 qBµ/kh
– Horner’s time: (tp+∆t)/∆t

This suggests that shut-in pressure pws, should plot as a
t p + ∆t
straight line function of log( )
∆t
t p + ∆t qBµ
p ws = p i + m log( ) where m = −162.6
∆t kh
In practice m is always considered to be a positive quantity.

Thus formation permeability k, can be determined from
slope m.
Conventional industry practice is to plot :
t p + ∆t t p + ∆t
p ws vs( on a semi-log plot with values of
) ( )
∆t ∆t
decreasing from left to right.

• Slope m, is determined by subtracting the two
values of pws over one cycle.
• In truth, it doesn’t matter how you plot the data,
because we have agreed to use positive value of m.
• If shut-in time ∆t, is infinite, then tp + ∆t ≈ ∆t and ( tp ∆+t∆t ) ≈ 1
t p + ∆t
The intersection of the straight line with ( ∆t ) = 1
gives p i , the initial reservoir pressure ∆t,
assuming an p * false pressure for finite reservoirs.

Skin Factor
pwf
or
pws DRAWDOWN BUILDUP
tp ∆t
At the instant of shut-in
qBµ φµcrw2
p wf = p i + 70.6 {ln(1688 ) − 2S}
kh kt p
φµcrw2
p wf = p i + m{log(1688 ) − 0.869S}
ktp
At shut-in time ∆t
t p + ∆t
p ws = p i − m log( )
∆t
Combining these two equations gives
p ws − p wf 1688φµcrw2 t p + ∆t
Skin ⇒ s = 1.151( ) + 1.151 log + 1.151 log( )
m k∆t ∆t
Conventional industry practice is to choose ∆t=1hr, and the

corresponding shut-in pressure p ws = p wf to use in the skin equation
(any ∆t and corresponding p∆ t will work).
With this simplification

p1hr − p wf k
⇒ s = 1.151{ − log 2
+ 3.23}
m 1688φµcrw
Reservoir Limit Test
• Test to estimate reservoir size and distance to boundaries
• Single boundary near the wellbore causes buildup curve slope “m”
to double
qBµ t p + ∆t
p ws = p i − 325.2 log( )
kh ∆t
• Distance (L) to the boundary can be estimated
0.000148k∆t x
L=
φµct
• Time required for the slope to double: 1.9E5φµct L2
∆t x =
k
• ∆t is long for large L and low k, alternative approach may be used:
* qBµ  − 3792φµct L2 
∆p ws = 70.6  − Ei 
kh  k∆t 
qBµ t p + ∆t
Compare to p ws = p i − 162.6 log( )
kh ∆t
qB µ
Define m = − 162 . 6
kh
For a boundary,
slope = 2m
pws 1300psi/cycle
2m
650psi/cycle
m
(∆t+tp)/∆t tx

The distance to the no-flow boundary is given by
0.000148k ( ∆t x ) (Gray’s Correlation)
L=
φµC t
For the case of multiple boundaries, each new encountered? Doubles the slope.
4m
pws
2m
(∆t+tp)/∆t

Static Drainage Area Pressure
• Average pressure in the drainage area of the well
– MBH (Matthews-Brons-Hazebroek) (p* method)
– Modified Muskat method
• MBH
– Estimation of static drainage area pressure
– Obtained by computing build-up curves for wells in various
positions and areas of various shapes and comparing
extrapolated p* with static average pressure
– Build-up curves were calculated using imaging techniques
and principle of superposition
– Series of plots of kh(p * −p ) / 70.6qµB vs. 0.000264ktp / φµct A
were prepared
Matthews-Brons-Hazebroek Method (MBH Method)
• Build-Up test analysis method for Pseudo-Steady State reservoirs
• For tp >>∆t, (tp +∆t)= tp and log[(tp +∆t)/ ∆t]=log tp -log∆t
p ws = cons + m log( ∆t )
• Calculations of permeability and skin factor are similar to Horner’s
method
• Plot of pws vs. ∆t is made on semi-log scale increasing from left to
right
• Wellbore effect is estimated similarly to Horner’s method and can be
estimated with equation:
0.14S
170,000Ce
∆t =
kh / µ
• End of semi-log straight line can be estimated from:
φµc t A
∆t = (t DA )est
0.000264k
This technique is much simpler (less calculation on the time
variable)
However:
MBH assumes that the well has reached PSS before shut-in
for buildup
PSS ∂P
r → re = cons tan t
∂t
False pressure p* may be estimated from MBH plot by using
p * = p 1hr + m log( t p + 1)
qBµ
p * = p 1hr + m log t p m = 162.6
kh
Average reservoir pressure in the drainage area can be

calculated using MBH p* and Fig 6.6 in Earlougher book

– Procedures:
• Extrapolate MTR of Horner’s plot to to (tp +∆t)/ ∆t =1 and read
extrapolated p*
• Estimate the drainage area shape
• Choose the proper curve for the shape of our area (MBH curve)
• Estimate 0.000264ktp / φµct A and find 2.303(p * −p ) / m = p DMBH
• Calculate p = p * −mp DMBH
– Advantages:
• Does not need to have data beyond MTR
• Applicable to a whole variety of shapes
– Disadvantages:
• Requires knowledge of drainage area shape and size
• k , φ, µ , c t

Modified Muskat Method
• Limiting form of general solution to diffusivity equation
– Depth of investigation reached reservoir boundaries
– Build-up followed stabilized flow
qBµ k∆t
p − p ws = 118.6 exp(−0.00388 2
)
kh φµct re
• For analysis of build-up tests this is approximated by
qBµ k∆t
log(p − p ws ) = log(118.6 ) − 0.00388
kh φµct re2
log( ∆p ) = A + B∆t
approximation valid if, 250φµct re2 750φµct re2

≤ ∆t ≤
k k
• Procedures: log( p -p ) ws
p too high
– Assuming a value for p p too low
– Plot log( p − p ws ) vs. ∆t correct p
– When straight line results the correct value of p was found ∆t

– Too low p produces downward curvature / Too high p produces
upward curvature
• Advantages (over p* method)
– Requires no estimates of reservoir properties when it is used to
establish p (except to choose the correct portion of data analysis)
– Yields satisfactory estimates of p for hydraulically fractured wells and
for wells with layers of different permeability communicating through
the wellbore
• Disadvantages:
– Does not work if the well is not centered in the drained area
250φµct re2 750φµct re2
– Shut-in times of ≤ ∆t ≤ are frequently impractically
k k
long (especially for low k)
Drawdown Test
• Used for:
– New wells
– Wells shut-in for long enough time to establish uniform pressure in the reservoir
– Wells which have to maintain production (economic reasons, etc.)
• Procedures:
– Starting with uniform pressure in the reservoir
– Producing a well at the constant rate
– Rate and pressure are recorded as a function of time
• Application:
– Estimation of reservoir permeability
– Estimation of skin effect
– Estimation of drainage volume, initial oil in place
– Detection of reservoir heterogeneities and boundaries
• Interpretation:
– Diagnostic plots
– Type curves
– Pressure derivative analysis

Drawdown Test Analysis
• Diagnostic plots:
– Plotting pwf vs. logt
– Estimate end of wellbore storage (log-log plot or type curve matching)
– Estimate the beginning of late transient (with type curves or semil log
plot)
– Estimate the slope “m” and calculate formation permeability, k
– Calculate skin factor, s
• Estimation of pore volume:
−0.234qB
Vp =
 ∂p 
C t  wf 
 ∂t 
 ∂p wf 
where:   slope of straight line on Cartesian plot of pwf vs. t
 ∂t 

( Pore volume (pseudo-steady state):
qBµ  0.000527kt re 3 
p wf = p i − 141.2 + ln[ ] −

kh  φµCt re 2
rw 4 
∂p wf 0.0744qB
=− substitute Vp = πre2hφ
∂t φµC t re2
∂p wf 0.234qB
=− ⇒
∂t C t Vp
0.234qB
Vp = − only when the boundaries are reached )
∂p wf
Ct ( )
∂t
Drawdown Test
• Idealized constant rate solution:
qBµ  1688φµCt rw2 
p wf = p i + 162.6 [log  − 0.869S]
kh  kt 
 
• Skin effect:
p i − p1hr  k 
S = 1.151[ − log  + 3.23]
 2 
m  φµCt rw 
• Permeability:
qBµ
k = 162.6
mh

Gas Well Testing
Diffusivity equation for liquids:
1 ∂p ∂ 2 p φµC t ∂p
+ 2 =
Based on: r ∂r ∂r k ∂t
law of conservation of mass
flow law (Darcy’s law) C( p − p o )
equation of state: ρ = ρ o e
Diffusivity equation for gases: 1 ∂Ψ(p) ∂ 2 Ψ(p) φµC t ∂Ψ(p)

+ =
r ∂r ∂r 2 k ∂t
Changes:
M p
. equation of state: ρ=
RT Z
p
. real gas pseudopressure: p
Ψ (p ) = 2 ∫
pB
µ (p ) Z( p )
dp


For gases flowing in an infinite-acting reservoir, the solution is:
(field units)
p sc T q g 1688φµ i C ti rw2
Ψ (p wf ) = Ψ (p i ) + 50300 [1.151 log( ) − (S + D q g )]
Tsc kh kt
C ti = C gi S g + C w S w + C f ≈ C gi S g (gas compressibility term is dominant)
p sc T q g 125.3φµ i C ti rw2
(SI) Ψ (p wf ) = Ψ (p i ) + 3.733 [1.151 log( ) − (S + D q g )]
Tsc kh kt
The term D q reflects a non-Darcy pressure loss. It takes into account the
g
fact that at high velocities near the producing well (characteristics of

high gas production rates) Darcy’s Law does not predict correctly the
relationship between flow rate and pressure drop. The absolute value of
qg is used so that the term D qg is always positive, regardless of
production or injection.

For pseudo steady state flow (ri > re) past boundary.
(field units) p sc T q g r
Ψ (p wf ) = Ψ (p i ) + 50300 [ln( e ) − 0.75 + S + D q g ]
Tsc kh rw
(SI) p sc T q g r
Ψ (p wf ) = Ψ (p i ) + 3.733 [ln( e ) − 0.75 + S + D q g ]
Tsc kh rw
For some (most) gases at pressures above 3000psi,

Wattenbarger and Ramey have shown that gas flow in an
infinite acting reservoir can be modeled accurately by
(field units)
qg Bgiµ i 1688φµiCti S + D qg
p wf = pi + 162.6 [log( )−( )]
kh ktp 1.151
(SI)
qg Bgiµ i 11.64φµiCti S + D qg
p wf = pi + 2.15 [log( )−( )]
kh ktp 1.151
For many cases at pressures below 2000 psi flow in
an infinite acting reservoir can be modeled by:
(field units)
2 2 qg Ziµ i T 1688φµiCti S + D qg
p wf = pi + 1637 [log( )−( )]
kh ktp 1.151
(SI)
2 2 qg Ziµ i T 11.64φµiCti S + D qg
p wf = pi + 1.508 [log( )−( )]
kh ktp 1.151

Buildup equations can be developed with the use of
superposition p>3000psi or 21000kPa
(field units)
q g B gi µ i t p + ∆t
p ws = p i − 162.6 [log( )]
kh ∆t
(SI)
q g B giµ i t p + ∆t
p ws = p i − 2.15 [log( )]
kh ∆t
Skin
(field units)
p1hr − p wf k
S' = S + D q g = 1.151[ − log( ) + 3.23]
m φµ i Cti rw2
(SI) p1hr − p wf k
S' = S + D q g = 1.151[ − log( 2
) + 2.10]
m φµ i Ctirw
Buildup equations can be developed with the use of
superposition for p<2000psi or 14000kPa
(field units) 2 2 q g Ziµ i T t p + ∆t
p ws = p i − 1637 [log( )]
kh ∆t
(SI)
q g Z iµ i T t p + ∆t
p ws 2 = p i 2 − 2.15 [log( )]
kh ∆t
Skin
(field units) p
2
− p wf
2
k
S' = S + D qg = 1.151[ 1hr − log( 2
) + 3.23]
m φµ iCtirw
(SI) p
2
− p wf
2
k
S' = S + D q g = 1.151[ 1hr − log( ) + 2.10]
m φµ i C ti rw2

Deliverability Tests
Deliverability tests have conventionally been called “back pressure”
tests because they make possible the prediction of well flow rates
against any particular pipeline back pressure. The purpose of these
tests is to provide the manner in which the flowrate will decline
with reservoir depletion.
The absolute open flow (AOF) potential of a well is defined as the rate at
which the well would produce against a zero sandface back
pressure. It cannot be measured directly, but may be obtained from
deliverability tests. It is often used by regulatory authorities as a
guide in settling maximum allowable producing rates.
Deliverability tests are historically conducted three different ways.
Flow- after flow test
Isochronal test
Modified isochronal test
All three are variations of the flow-after flow test.
Flow-after flow test

The well is flowed at a selected constant rate until the
pressure stabilizes. (i.e. pseudo steady-state is reached).
The stabilized pressure and rate are recorded. The rate is
then changed and the well is flowed until pressure
stabilizes again. The stabilized rate and pressure are
recorded. This process is repeated for a total of three or
four rates.
The collected data is analyzed by either of two methods an
empirical method (Canada) or the LIT (Linear-inertial-
turbulent) method.
Empirical Method – AOF- log-log graph
– C and n
– Deliverability equation
LIT method – linear inertial turbulent method

– sometimes called theoretical method

Simplified (Empirical) Analysis
The empirical method is based on the observations of Rawlins
and Schellhardt (1936). Their observation was that a plot of
2
∆p 2 → (p − p 2wf ) vs. qst
on log-log paper is approximately a straight line for most

wells in which pseudo-steady state is reached for each flow
rate in a flow-after flow test.
The equation for this straight line is
Where 2
q st = C(p R − p 2wf ) n = C( ∆p 2 ) n
qst =flow rate at standard conditions (m3/d)
p R = average reservoir pressure obtained by shut-in of the
well to complete stabilization (kPa)
p wf = flowing sandface pressure (kPa)

C=coefficient which describes the position of the stabilized delivery line
n= an exponent which describes the inverse of the slope of the stabilized

deliverability line.
The value of n varies from 1.0 for laminar flow to 0.5 for turbulent flow
0.5<n<1, and is sometimes regarded as a measure of the turbulence of
flow (values of n -> 0.5 look for positive skin)
The straight line is extrapolated to pwf=atmospheric pressure to

determine the AOF value for the well.
The ‘constants’ C and n are not constant at all but depend upon the
pressure dependent fluid properties ( µ , CS , Z ).
Accordingly, often this type of deliverability curve is used, periodic re-

testing of the well should occur to determine current values of C and n.
LIT method – Linear Inertial Turbulent
(Theoretical) method
Outside North America, there is a general use of a quadratic
form of the flow equation called the “Forcheimer
equation” or “turbulent flow equation”. It is actually the
laminar-inertial-turbulent (LIT) flow equation
2
∆p = p R − p 2wf = aq st + bq st2
2
This equation is derived from our pressure squared flow

equation for stabilized flow (PSS or SS)
q gµ i Zi T re
p 2wf = p i2 −C [ln( ) − 0.75 + S + D q g ]
kh rw
for pi = pR and q g = q st

2 q st µ pri Z pr T re
p 2wf = pr − [ln( ) − 0.75 + S' ]
kh rw
2
p R − p 2wf = aq st + bq st2
where
µ i Zi T re µi ZiT
a= [ln( ) − 0.75 + S' ] b= D
kh rw kh
The equation can be arranged to
2
pr − p 2wf
= a + bq st
q st
2
pr − p 2wf
A Cartesian plot of vs. q st q st is a straight
line with slope b and intercept a.
The constants a & b can be determined from the flow

test for at least two rates in which and the
corresponding value of p wf are measured.
∆Ψ = Ψ − Ψwf = a' q g + b' q g2 pseudopressure

∆p = p − p wf = a' ' q g + b' ' q g2 pressure
2
∆p = p − p 2wf = aqg + bqg2
2 pressure squared

µ pi Z p T re
a = 1422 [ln( ) − 0.75 + S' ]
kh rw
µ pi Z p TD
b = 1422
kh
plot of ∆Ψ − bq g2 vs qg on log-log paper should yield a straight line
∆p 2 / q vs. qg is plotted on the straight line

2
p − p 2wf = aq g + bq g2 / q g
2
(p − p 2wf ) / q g = a + bq g -> straight line plot y=ax+b

Isochronal Tests
The objective of isochronal testing is to obtain data to establish a stabilized
deliverability curve for a gas well without flowing the well sufficiently long
to achieve stabilized conditions at each rate. This procedure was developed
for lower permeability reservoirs were it was impractical to achieve PSS at
each rate during the test.
An isochronal test is conducted by flowing a well at a fixed rate, then shutting
the well in until the pressure builds to an unchanging (or almost unchanging)
value .
The well is then flowed at a second rate for the same length of time, followed by
another shut-in period. This flow, shut-in sequence is continued usually 4-5
time. If at all possible, the final flow period should be long enough to reach
stabilized flow. If this is impossible or impractical, it is still possible to
predict the stabilized deliverability characteristics.
Analysis for the empirical method is carried out in a similar fashion to that of
the flow-after flow test. The data from the flowing period is plotted, and the
best fit line is found. A parallel line is drawn through the stabilized point. It
is this (second) line which is extrapolated to determine the AOF and the
values of ‘c’ and ‘n’.


Procedures for isochronal tests using empirical
method (SSS achieved)
1. plot the best straight line through the points

2
(p − p 2wf ) vs.q g obtained for a fixed value of time on
log-log
2. establish the slope 1/n
3. a line with slope of 1/n is then drawn through the
single stabilized point
4. establish 2
deliverability curve equation
q g = C(p − p 2wf )n
or do graphically
Procedures for isochronal test theoretical method
For SSS
2
p − p 2wf = aq g + bq g2 , ri > re
where µpZpT r µ p Z p TD
a = 1422 [ln( e ) − 0.75 + S' ] b = 1422
kh rw kh
for infinite acting
2
p − p 2wf = aqg + bq g2 ri < re
where b is the same as for SSS

µpZpT 1 1688kt
a = 1422 [ ln( 2
) + S]
kh 2 φµ C rw p tp
Procedures:
1. for constant t determine b from the plot of
2
(p − p 2wf ) / q g vs.q g
2. using the stabilized data point, determine a from
2
a = [(p − p 2wf ) stab − bq gstab
2
] / q gstab
3. stabilized deliverability curve will have the

following equation
2
p − p 2wf = aq g + bq g2
and
2
− a + a 2 + 4b(p − 14.7 2 )
AOF =
2b

Isochronal tests analysis with no SSS achieved
Constant a have to be calculated if every term in equation
µpZpT re
a = 1422 [ln( ) − 0.75 + S' ]
kh rw
1. known s,k can be determined from one of the sections

of the isochronal tests
2. two tests (either drawdown or BU) have to be
analyzed for estimation of S
3. re have to be estimated from expected well spacing or
knowledge of reservoir geometry
4. constants a & b determined that way can be
substituted into a stabilized deliverability equation

Modified Isochronal Test
The objective of the modified isochronal test is to obtain the same
data as in an isochronal test, without using the sometimes lengthy
shut-in periods. (i.e. allowing the pressure to stabilize completely
before the next flow period is commercial)
The modified isochronal test has shut-in periods the same duration as
the flow periods. The final shut-in BHP(pws) before the beginning
of a new flow period is used as an approximation to in the test
analysis procedure.
Again, either the LIT or empirical method may be used.
Note: the modified isochronal procedure uses approximations.
Isochronal tests are modeled exactly by rigorous theory. Modified
isochronal tests are not. Despite this, however, modified
isochronal tests are excellent approximations to true isochronal
tests, and are widely used as they save time and money.



Modified Isochronal Test
Obtaining the same data as isochronal without lengthening shutin
periods
Shut-in periods are the same drawdown as flow periods
Final BHP after shut-in is used as approximation to p in the test
analysis procedure
Approximate results-> good accuracy will less time.
µ i Zi T
b= D
kh
The equation can be arranged to
2
p − p 2wf
= a + bq st
q st
2
A Cartesian plot of p − p2wf vs. q st is a straight line.
qst
Gas Deliverability Tests Design
• General Considerations:
– Time to stabilization
– Sequence of flow rates
– Constancy of flow rates
• Design guidelines:
– Choice of the test
– Choice of the equipment
– Choice of flow rates
– Duration of flow rates

Computer Aided Test Interpretation
•Speeding-up traditional techniques
•Handling problems to complex for hand calculations

–Continuously varying rate
–Multiple wells
–Complex geometry
•Estimation of reservoir properties from portions of data insufficient for

conventional interpretation
•Capable of performing Non-linear regression (automatic type curve

matching)
–Matching achieved by changing of the unknown reservoir properties
until the data fit as closely as possible (in the least square sense)
–All data can be fitted simultaneously
¾Problems associated with interpreting separate portions of data
can be avoided
¾Goodness of the fit can be estimated
Pressure Derivative Analysis
• Provides a simultaneous presentation of log(∆p) vs. log(∆t) and log[∆t(dp/dt)]
vs. log(∆t) on the single graph (see figure)
• Unique curve match can be obtained

Horizontal Well Testing
Horizontal wells differ from vertical wells in a number of ways

that are important to well test interpretation
•Interval open for fluid entry is very long;

•In many cases the producing length is not known, unless
production logging was run to measure the rate at which fluid
enters the hole at different locations;
•Vertical permeability may play important role, since there is
likely to have considerable flow in vertical direction;
•There are a number of different flow regimes during the
transient;
•Depending on the value of reservoir parameters, one or more
of the flow regimes may be missing.
Steps in Evaluating Horizontal Well-Test Data
•Identify Flow Regimes
•Apply the Proper Procedures
•Evaluate Uniqueness and Sensitivity

Horizontal Well Flow Regimes
Hemiradial
Early radial
Pseudoradial or Late radial

Linear or Early linear John Lee, Well Testing, SPE Textbook Series Vol. 9
Identifying Flow Regimes in Horizontal Wells
Late linear
1
Wellbore
Storage
Horizontal Wells – Geometry Definitions

Early radial
Horizontal Well Flow
Regimes (1)
Governing Equation:
162.6qBµ   k zk x t  
p i − p wf = log 10 2 
 − 3.227 + 0.868sd 
k z k x L w   φµCtrw  
Interpretation:
162.6qBµ  ∆p  k zk x  
k zk x = sd = 1.151 1 hr
− log 10  + 3.23
merf L w  merf  φµCtrw 
2

 
End of early radial: 1800d 2zφµCt

=
125L2W φµCt
t Eerf = t Eerf
kz ky
Hemiradial
Regimes (2)
Governing Equation:
325.2qBµ   k zk x t  
p i − p wf = log 10 2 
 − 3.227 + 0.868sd 
k z k x L w   φµCtrw  
Interpretation:
325.2qBµ  ∆p  k zk x  
k zk x = sd = 2.303  1 hr
− log 10  + 3.23
mhrf L w  mhrf  φµCtrw 
2

 
1800D 2zφµCt 125L2w φµCt
End of hemiradial flow: t Ehrf = t Ehrf =
kz ky
Early linear flow
Regimes (3)
Governing Equation:
8.128qB µt 141.2qBµ
p i − p wf = + ( sc + sd )
Lw k x φCt L w k z k x
Interpretation:
8.128qB µ Lw k zk x
kx = sd = ∆p t = 0 − s c
melf L w h φCt 141.2qBµ
h  kx    πd z  
sc = ln  + 0.25 ln  − ln sin   − 1.838
 rw   kz    h 
160L2w φµCt
End of early linear flow: t Eelf =
ky
Late-Pseudoradial flow
Regimes (4) Lw
< 0.45
Governing Equation: bH
162.6qBµ   kyt   141.2qBµ

p i − p wf =  log 10 2 
 − 2.303 + ( sc + sd )
k yk x h   φµCt L w   Lw k zk x
Interpretation:
k yk x =
162.6qBµ k z L w  ∆p1hr  ky  
sd = 1.151  − log 10  + 1.76 − sc
2 
mprf h k y h  mprf  φµCtrw  
Time of occurrence:
1480L2w φµCt 2000φµCt ( L w / 4 + d y ) 2 1650φµCt d x2
t Sprf = t Eprf = t Eprf =
ky ky kx
Regimes (5) Late linear flow
Governing Equation:
8.128qB µt 141.2qBµ b
p i − p wf = + ( s p + sc + H sd )
bHh k x φCt b H k z k x Lw
Interpretation:
8.128qB µ L  bH k zk x 
kx = sd = w  ∆p t = 0 − s p − sc 
mllf b H h φCt bH  141.2qBµ 
Time of occurrence:
4800φµCt ( L w / 4 + D y ) 2 1800φµCt D 2z 1650d x2 φµCt
t Sllf = t Sllf = t Ellf =
ky kz kx
Horizontal Well Flow Wellbore Storage
Regimes (6)
Governing Equation:
25.65A wb qBt
C = Vwbc wb C= C=
ρ wb cos θ 24 ∆p
Time of occurrence:
( 4000 + 240sd )C
t Ewbs =
k xk y Lw / µ

Summary of Analysis Procedures
•Compute Kz, from early-radial or hemiradial flow regimes (need

knowledge of Kx – permeability perpendicular to the wellbore);
•Compute Ky from pseudoradial flow regime (need knowledge of Kx);
•We can compute Kx if data from early-linear or late-linear flow
regimes is available;
•To compute Kx from early-linear flow regime need to know the
EFFECTIVE completed wellbore length Lw; to compute Kx from
late-linear flow regime, need to know the reservoir length bH
•We can compute Lw from early-linear flow regime if we know Kx
•We can compute bH from late-linear flow regime if we know Kx

Parameters Important to H.W. Pressure Response
¾Ratio of the vertical to horizontal permeability
(kV/kH)
¾Vertical permeability
¾Position of the well relative to the thickness of

the formation (zw/h)
¾Effective length of the well relative to the

formation height (Lw/h)
¾Drilling & Completion damage

EOR & SIMULATION OF SPECIAL PROCESSES:
1. Basic Concepts
2. Buckley Leverett Theory
3. Overview of EOR Processes
4. Compositional Simulation
5. Miscible Displacement
6. Chemical and Polymer Flooding
7. Steam Simulation and Steamdrive
8. In-Situ Combustion

Basic Concepts: IFT
• When two or more immiscible fluids are put in one recipient, the denser ones will be at the lower part
existing an horizontal separation surface between the fluids. This does not occur in a porous media
with different capillars with different diameters, because the surface separation will not be sudden,
existing a transition zone due to capillary effects
• This phenomena results from the molecular attraction inside the fluid. A molecule inside the liquid will
be in equilibrium since it is attracted equally in all the directions by the surrounding molecules. The
same is not true with a molecule at the surface, which will not be attracted equally since it is
surrounded by molecules of different types.
• When the contact surface is between a liquid and a gas the phenomena and the properties that are
generated from this are called superficial and when between two liquids interfacials, non existing
however, a difference between both cases.
• An isolated fluid drop presents a spherical surface due to the molecules trend to move to the center and
consequently to present a minimum surface. The surface behaves as an elastic membrane that offers
resistance to the separation of molecules.
• The needed energy to form a certain surface is defined total free surface energy (E).
• The surface energy per unit of surface is called unit free surface energy (Es).
• The force that impede the surface rupture, per unit of length, is called the superficial or interstitial
tension (s).
• The force that tends to pull a surface to the center is called capillary force (Fc) and this, divided by are
is called capillary pressure (Pc).

Basic Concepts: IFT
Relationship between Superficial Tension and Unit Free Energy Surface
By definition, dF dF τ dτ dF .da dF .db
σ= 1
= 2
Es = = = 1
= 2
Es = σ = MT −2 → erg / cm 2 = dyne / cm
db da A A da.db da.db
dF1
dF2
da Capillary Force
db
dF2
dF1
dFc
Relationship between Capillary Pressure and Superficial tension
da/2
N
M
dF1
dΘ1 Capillary Pressure
Luciane Cunha dΘ1 Universidad Nacional de Ingeniería - UNI 292

Basic Concepts: IFT
Fc dFc
pc = =
A dA
dFc = 2proj.dF1 + 2proj.dF2

dFc = 2dF1 sin dθ1 + 2dF2 sin dθ2
MN da / 2
sin dθ1 = =
R1 R1
da / 2
sin dθ2 =
R2
dA = da.db
dF1 .da dF2 .db

dFc = +
R1 R2
dF1 1 dF2 1
Pc = . + .
db R 1 da R 2
1 1 2σ
Pc = σ( + ) Pc =
R1 R 2
R

Basic Concepts: Wettability
The figure below presents an scheme with the equilibrium of forces in an
oil-water-solid interface.
Contact Angle (after Amyx et al.).

By definition, the contact angle t, is the angle (varying from 0 and 180o)
measured from the densest liquid.
When this contact angle is smaller than 90o we say that the densest fluid is
preferable wet to the solid and when it is higher than 90o we say that
the least dens liquid is preferable wet to the solid.

Basic Concepts: Wettability
The adhesion tension is defined as the tension difference throughout the solid. Thus:
σso − σsa
σ A = σ so − σ sa = σ ao cos θao cos θao =
σao
This tension is responsible for the rising and depression of fluids in capillars, i.e., determines by which
fluid the solid is preferably wet.
The wettability property and the adhesion tension varies according to the fluids and solids used,
Wettability (after Amyx et al.).

Basic Concepts: Capillary Pressure
2πrσa = πr 2hγ a
2σa σa − air cos θa − air

h= =
γa γa
σa − air cos θa − air

Pair − Pa = Pc = γ a h =
γa
Capillary Rising (after Amyx et al.).

σ oa cos θoa
h=
rg∆ρ
2σoa cos θoa

Pc =
r
Capillary Rising. System with two liquids (after Amyx et al.).

Gas Gas
Water
Oil Water
Most Common Reservoir Fluid Distribution. Oil
Most Common Reservoir Fluid Reservoir Fluid Distribution in an Oil
Distribution. Wet Reservoir.
Drainage Imbibition
Drainage and Imbibition.
0 100%
Saturation
Curve of the Gradient vs Elevation in the Capillaries with Uniform and Equal Diameters.

A
B
h
0 100%
Capillary Pressure
Saturation
Curves of Elevation vs Saturation.

Drainage
Imbibition
0 Saturation 100%
Capillary Pressure Curve.

CONVERSION FROM THE LAB DATA TO THE FIELD
The conversion of lab data to the field is needed because the fluids used in the experiments differ in terms of
physical properties from the reservoir fluids.
the relationship between the capillary pressure measured in the lab and the corresponding capillary pressure
in the reservoir will be given by
PcL σ L cos θL
=
PcC
σ C cos θC
since that the capilar radius is the same.
It is intuitive that the curvature radius is a function of the saturation. Since from the Plateau law the
capillary pressure is a curve of the superficial tension and from the curvature radius, we conclude that
the capillary pressure is a function of the saturation and from the superficial tension.
Thus: PcC = σCf (SC ) PcL = σ Lf (SL )

σC
For similar equal saturations in the lab and in the field: PcC = PcL
σL
Thus we obtain another formula to transform the lab data into the field data. Even though the two
ways will generate slightly different results, both are used to the same final objective.

Basic Concepts: Relative Permeabilities,
K o A∆p
Qo =
µoL
Relative Permeability
K g A∆p
Qg =
µg L
Kro(krnw)
Drainage
K A∆p
Qw = w
µw L Imbibition Krw(krw)
Sw(Sw)%
(After Amyx et al)
Factors Affecting Relative Permeabilities,
Curve 1: Water Wet
Curve 2: Oil Wet

Effect of the Porous Media
Consolidation
Sw(Sw)%
Kg/K
Wettability 1: Unconsolidated
Ko/K
2: Consolidated
(After Amyx et al)
Saturation%
Factors Affecting Relative Permeabilities,
1: krg – parallel to the layers
2: krg – perpendicular to the layers
3: kro – parallel to the layers
4: kro – perpendicular to the layers Saturation Process
So(%)
Anisotropy of the Porous Media
(After Amyx et al)
Sw(Sw)%

Basic Concepts: Relative Permeabilities,
k rg k g / K k g
= =
k ro k o / K k o
(After Amyx et al)

Basic Concepts: Mobility Ratio (M)
λ displacing kp
M= λp =
λ displaced µp
λ w k w µ o k rw µ o
M w −o = = x = x
λ o µ w k o µ w k ro

Basic Concepts: Sweep Efficiency (Eh)
• Fraction of the total reservoir area that has been swept at a certain time
• Eh=Swept Area/Total Area
• Factors affecting Eh (Volume Injected, Mobility Ratios, Injection Pattern)
• Determination of Eh
four-spot seven-spot
five-spot
two-spot
Peripherical Pattern
Luciane Cunha Universidad Nacional de Ingeniería - UNI Irregular Pattern

307
Basic Concepts: Vertical Efficiency (Evv)
Fraction of the reservoir cross section

(parallel to the flow direction) that has
been invaded at a certain time.

Basic Concepts: Displacement Efficiency (Ed)
It is the fraction of mobile oil that has been displaced from

the swept zone at a certain time
Mobile Oil = So –Sor
Where, So (oil saturation at the beginning of injection) and
Sor (true residual oil saturation)
(So − Sor ) Vps (So − Sor )

Ed = Ed =
(So − Sor ) Vps (So − Sor )

Basic Concepts: Displacement Efficiency
Np is a function of Eh, Evv and Ed
To improve Eh and Evv consider mobility

(M=kw.µo / µw.ko) changes in kw, µo or µw)
To improve Ed, consider EOR methods to change

IFT

Buckley Leverett Theory:
1-Sor Oil
Injected Water
Swi
Connate Water
rw rBL

EOR – Miscible Displacement
Atmospheric Conditions Reservoir Conditions
(high pressure)
(CO2)(gas) CO2 (liquid)
Mixture (liquid)
Oil (liquid) Oil (liquid)

EOR – Miscible Displacement
Residual Oil
CO2
&
Oil
CO2
CO2 contacts residual oil, allowing oil to flow outside the smaller
pores.
EOR – Polymer Injection
• Polymers are long and flexible molecules;
• Increase the oil recovery by means of the increase of the horizantal and
vertical displacement efficiency of the injected fluid;
• Relative complex process. Causes the injected fluid to reach a bigger

portion of the reservoir, therefore displacing more of the remaining oil;
• Extremely expensive the polymer bank is injected followed by

water injection, which pushes the polymer far away from the injection
wells and through the oil reservoir.

EOR – Steam Injection
⇒ Main chategories: soaks (or pulses) and continuous

injection;
⇒ Mainly in heavy oil deposits;
⇒ Viscosity reduction and increase of mobility with heat. Heat

also provides the pressure that pushes the oil to flow;
⇒ Aplications are usually restricted to oil deposits located at

relatively shallow depths (400 to 2000 ft) and thickness higher
than 20 ft for sandstones.

Steam Soak Operation
Vapor cycle Soak cycle Production cycle

vapor oil
heat heat vapor

oil water
Vapor zone Heated zone

Continuous Steam Injection
Vapor generator
Injetor well Producing well
Transition from
Vapor to water
vapor
Oil
bank

EOR – In-Situ Combustion
⇒ Reservoir with highly saturated oils, intermediate thickness

(5 to 50 ft), shallow (400 to 2000 ft depth) with oils with
intermediate API degree (10 a 35 oAPI);
⇒ Main categories: upstream and reverse;
⇒ Oil viscosity reduction of the oil that is not burned and

improvement of the mobility with addiction of heat, air
injection is useful to keep the pressure that pushes the oil to
flow.

Injector well
air oil
ignição
air air oil
Burned
zone Combustion
zone

oil
air
air Combustion ignition

zone

Compositional Simulation
Reservoir containing light oils (gas condensates or volatile oils) vapor/liquid equlibrium
depends on composition as well as pressure.
In this case rigorous flash calculations must be made with equilibrium K values or an
equation of state (EOS) to determine hydrocarbon phase compositions. Viscosities
and densities can then be calculated from phase compositions.
In theory, compositional simulation is useful whenever compositional effects are

important.
In practice, accuracy problems arise in compositional modeling of miscible

displacement. Particularly when miscibility happens through multiple contact.

Multiphase Flow in Porous Media
Fluid Properties
Beta Type Fluids (Black-Oil): the PVT behavior is represented by a simple form
through volume factors and gas-oil solubility relationship. In this case the global
hydrocarbon composition inside the reservoir remains constant during the entire
reservoir life.
Phases
Components
(in the reservoir)
(surface conditions)
Oil Oil
Gas Gas
Water Water

Compositional Equations for Numerical Simulation
Considerations:
- three phases (oil, gas and water);
Multiphase Flow in Porous Media
- nc components ; Fluid Properties
- Fluids
Beta Type each component
(Black-Oil): can bybe
the PVT behavior is represented present
a simple in factors
form through volume each one
and gas-oil orrelationship.
solubility all phases.
In this case the global hydrocarbon composition inside the reservoir remains constant during the entire reservoir life.
Definitions:
a)Mass fraction of component i in gas phase (Cig)
mass of component i
C ig =
mass of gas phase
b)Mass fraction of component i in oil phase (Cio)
mass of component i
C io =
mass of oil phase
c)Mass fraction of component i in water phase (Ciw)
mass of component i
Ciw =
mass of water phase

•Rate of Mass of Component i IN :
In the oil phase =

A(C io ρ o v o ) x
+
In the gas phase =
A(C ig ρ g v g ) x
+
In the water phase =
A(C iw ρ w v w ) x

•Rate of Mass of Component i OUT:
In the oil phase =
A (C io ρ o v o ) x + ∆ x
+
In the gas phase =
A(C ig ρ g v g ) x + ∆x
+
In the water phase =
A(C iw ρ w v w ) x + ∆x

*
If q o= injection/production rate of oil phase by unit of rock volume
And similarly, q g* , q *w , we have:
•Mass injection/production rate of i in the elementary volume = (Cio qo* + Cigqg* + Ciw q *w )A∆x
•Rate of mass accumulation of component i in the elementary volume =
∂
A∆x [ φC io S o ρ o + φC ig S g ρ g + φC iw S w ρ w )
∂t

Substituting all terms in the equation of mass conservation, re-arranging, and dividing by A∆x
and taking the limit when
∆x → 0 , we have:
∂
− (Cio ρ o v ox + Cig ρ g v gx + Ciw ρ w v wx ) + Cio qo* + Cig qg* + Ciw q *w =
∂x
∂
=
∂t
[φ(Cio Soρo + CigSgρg + Ciw S w ρ w ) ]

Considerations:
-three phases (oil, gas and water);
-nc components;
-each component can be present in each one or all phases.
Mass Conservation Law
Substituting Darcy’s Equation and rewriting the previous equation in terms of differential operators we have
∂
− (Cio ρ o v ox + C ig ρ g v gx + C iw ρ w v wx ) + C io q o* + Cig q g* + C iw q *w =
∂x
∂
= [ φ(C io S o ρ o + C ig S g ρ g + C iw S w ρ w )]
∂t
 C ρ kk (∇p o − γ o ∇D) C ig ρ gkk rg (∇p g − γ g ∇D) C iw ρ w kk rw (∇p w − γ w ∇D) 

∇  io o ro + + +
 µo µg µw 
∂
+ Cio q o* + Cig q g* + Ciw q *w = [ φ(Cio S o ρ o + Cig S g ρ g + C iw S w ρ w )] I=1,2,…,nc
∂t

for isothermal processes we have:
ρ o = ρ o (p o , C 1o , C 2o ,..., C nco )
ρ g = ρ g (p g , C 1g , C 2g ,..., C ncg )
ρ w = ρ w (p w , C 1w , C 2 w ,..., C ncw )
µ o = µ o (p o , C 1o , C 2o ,..., C nco )
µ g = µ g (p g , C 1g , C 2g ,..., C ncg )
µ w = µ w (p w , C 1w , C 2 w ,..., C ncw )
k ro = k ro (S g , S w )
k rg = k rg (S g )
k rw = k rw (S w )

p o ,p g ,p w , S o , Sg , S w
Primary variables: number of unknowns: 3nc + 6
Cio , Cig , Ciw (i = 1,2,..., nc)
So + Sg + S w = 1
Pcgo (S s ) = p g − p o
Pcwo (S w ) = p o − p w
nc
∑C
i =1
io = 1,
nc
Additional Equations ∑C
i =1
ig = 1,
nc
∑C
i =1
iw =1
fig = fio
fig = fiw
I=1,2,…,nc (Phase Equilibrium Eq. /
f = fugacity in pressure units)

Special Data Requirements – Compositional Simulation
The additional data required are:
Phase-equilibrium information
Phase densitities
Phase viscosities
Compositions of the reservoir hydrocarbons and of any injected hydrocarbons
Separator conditions
Information for the calibration of the EOS (Peng-Robinson & Soave-Redlich-
Kwong) (for each component: critical pressure, temperature, specific volume,
molecular weight, acentric factor and for each pairs of components specify the
binary interaction coefficients. Phase densities are computed using the EOS.
Usually component lumping is applied (pseudo components)

RESERVOIR ENGINEERING
1. Reservoir Characterization
1a. Sedimentary Rocks

- Depth, layering, limits, thickness
- Lithology (φ, k, Cf and wettability)
- existence or not of aquifer and/or gas cap
1b. Reservoir Fluids
- water: salinity, viscosity, saturation
- hydrocarbons: composition (PVT), present phases, interaction with rock and water
2. Well Characterization
- perturbation of a situation in equilibrium (flow)
- small diameter (well spacing – recovery x economical feasibility)
- production or injection
- type of completion
- workover, stimulation
- artificial lift method
3. Objectives
- Compute volumes in-place
- regulars updates of reserve values
- production optimization
- optimization of daily production
RESERVOIR ENGINEERING (cont.)
4. Attributions of a Reservoir Engineer
- centralization of information
- identification of production mechanisms
- prediction of reservoir behavior
- exploitation plan definition
- analysis of production history
5. Main Difficulties
- sparse/bad data/information quality

- accuracy x possibility of reformulation
- maximum knowledge -> Abandonment

volume
E R (reserve)
P (production)
Ta time
I II III
Phase Delimitation Properties Analysis of Performance
Data Comparative Volumetric Production History

Types Data Data
(ELuciane
– errorCunha
margin in the recovery estimative) 334
Main Goals:
Estimative of Original Volumes

Primary and Secondary Performance Prediction
Estimative of Primary and Secondary Reserves
Steps:
Volumetric Method
Material Balance Method
Decline Curves Analysis
Prediction by means of Analytical Methods
Prediction by means of Numerical Methods

RESERVOIR Management
• The reservoir management process integrates the following steps:

(1) acquisition of data;
(2) interpretation of each data type to obtain an interpretation model for the data;
(3) integration of all available data interpretation models into a reservoir model;
(4) calculation of the reservoir model behaviour with a reservoir simulator;
(5) calibration of the reservoir simulator by history matching production data;
(6) coupling the reservoir simulator with well and surface facility simulators; and
(7) using the coupled simulators to calculate reserves and predict production for
various development scenarios.
• required understanding of
(1) the fundamental concepts of reservoir characterization, reservoir modelling,
reservoir simulation, and field management;
(2) the links between the various types of data; and
(3) the processes for integrating and processing all available information.

RESERVOIR Management
Reservoir management is the application of available technology and knowledge to a reservoir

system in order to control operations and maximize economic recovery within a given
management environment.
The most common objectives of reservoir management are:
to decrease risk
to increase oil and gas production
to increase oil and gas reserves
to maximize recovery
to minimize capital expenditures
to minimize operating costs
The understanding of reservoir management has improved greatly over the last few years and a
methodology is slowly emerging to facilitate its routine implementation. Reservoir
management used to be identified with production engineering, then became synonymous
with numerical reservoir simulation. It is now understood that it is an iterative process, which
numerical reservoir simulation and production engineering are only two components of.

The Reservoir Management Process
Reservoir management is about
(1) making the best possible decisions that will enable a company to meet specific
objectives; and
(2) implementing these decisions. The ability to make the best possible reservoir
management decisions relies mainly on the ability to predict the consequences
of implementing these decisions. This in turn depends on the ability to model
the expected behaviours of the reservoir system. The complete process includes
four stages: Reservoir Characterisation, Reservoir Performance, Well
Performance and Field Development.

Uncertainty and Probabilistic Methods
• Which are the error bars in the orginal oil and gas in
place volumes, and aquifer also?
• and on the recovery mechanisms?
• which are the consequences of these uncertainties

on the long term oil, gas and water production
prediction?
Systemic x Occasional
Subjective x Objective

Development of an Oil Field
Development Project
• Analytical Phase
• Modeling Phase
• Prediction Phase (history matching)
Economical Analysis

Main Phases in a Reservoir Engineering Study
RESERVOIR IMAGE
Exploration: core analysis, logs,
PVT, well tests
WELL CHARACTERISTICS RECOVERY MECHANISMS

TESTS – core analysis, logs Fluids, Rocks, Flow Equations
RESERVOIR EOR
PATTERN
ANALOGOUS CÁLCULOS
FIELDS MANUAIS
SIMULATION MODEL
(analysis of different cases)
PERFORMANCE ECONOMICAL
PREDICTION EVALUATION
DEVELOPMENT

LCunha-Colour Template PDF

Uploaded by

Document Information

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

LCunha-Colour Template PDF

Uploaded by

Copyright:

Available Formats

PE503 - Applied Reservoir Engineering

April 25 to 29, 2005

• Fundamental Data and Basic Concepts in Reservoir Engineering

• Numerical Reservoir Simulation and Simulator Equations

• Stages in a Reservoir Simulation Study

• Well Testing and Analysis

• Enhanced Oil Recovery Methods

Luciane Cunha Universidad Nacional de Ingeniería - UNI 2

The reservoir image (geological model) is the base

Luciane Cunha Universidad Nacional de Ingeniería - UNI 4

TECTONIC PATTERN FLUIDS

Optimize the Economic Exploitation of

⇒ How much hydrocarbon exist?

⇒ Original hydrocarbon in place

Luciane Cunha Universidad Nacional de Ingeniería - UNI 7

1. What is Reservoir Simulation?

Luciane Cunha Universidad Nacional de Ingeniería - UNI 8

- Input - - Model - - Output -

Luciane Cunha Universidad Nacional de Ingeniería - UNI 10

1) Well Testing Equations (Diffusivity Equation)

Luciane Cunha Universidad Nacional de Ingeniería - UNI 11

- Flow from A to B -> Darcy’s Law

Luciane Cunha Universidad Nacional de Ingeniería - UNI 13

- useful in visualizing the flow processes and develop mathematical models

- very expensive to build and run.

Physical vs Numerical Models

Numerical models must describe what physically takes place in the

In cases where we cannot model the reservoir physically (physical law

Conceptually, a reservoir numerical simulator model is consisted of

Reservoir being water flooded

Sw increases to the right with time

Reservoir Simulators: computer programs that solve the equations

Luciane Cunha Universidad Nacional de Ingeniería - UNI 16

-PDE’s are derived from:

-The number and type of equations to be solved depends on:

Mathematical Model Computer Model =

Solution of the mathematical system of equations requires knowledge of

Reservoir description is seldom known with any degree of accuracy -

• inaccuracy of input data;

• truncation error – introduced when we approximate our original PDE’s

• Round off error of the machine on which we solve our equations.

The main advantage of a reservoir simulator is the ability to inexpensively

Traditional Reservoir Engineering (1930 - …)

Early Reservoir Simulation (1955 – 1970)

Luciane Cunha Universidad Nacional de Ingeniería - UNI 20

Modern Reservoir Simulation (1970 - …)

Luciane Cunha Universidad Nacional de Ingeniería - UNI 21

-One-phase vs multi-phase simulators

-Mostly 3-D. Linear predominates in fractured reservoirs while radial predominates

Luciane Cunha Universidad Nacional de Ingeniería - UNI 22

4.1- Forecasting of reservoir performance

4.2- Improving reservoir description through history matching

4.3- Analysis of experiments

Luciane Cunha Universidad Nacional de Ingeniería - UNI 23

4.4- Understanding of flow mechanisms

4.5- Development of simple models and correlations

Luciane Cunha Universidad Nacional de Ingeniería - UNI 24

Images of Input Output Analysis of

Luciane Cunha Universidad Nacional de Ingeniería - UNI 27

- Normally used to mean a liquid (gas at a high pressure);

¾ Simplest system: j=4

(1,3) (2,3) (3,3) (4,3)

j=2 (1,2) (2,2) (3,2) (4,2) 5 6 7 8