Petrel 2014 Uncertainty Analysis and
Assisted History Matching
Module 2: Introduction to history matching
Course layout
Introduction to sub- Overview of history Basic statistics concept Uncertainty assessment
surface uncertainty matching approaches and definition
analysis
4000 [psi]
Pressure
5500
5000
4500
3500
3000
2500
Workflow creation and Handling of tabular Assisted history
modification Variables matching approach in
Petrel
rate [STB/d]
600
500
400
300
200
100
0
[STB/STB]
0.8
0.6
0.4
0.2
0
production
02/01/2004
Oil cut
Water
Pressure [psi]
2500 3000 3500 4000 4500 5000 5500
0 100 200 300 400 500 600
Oil production rate [STB/d]
0 0.2 0.4 0.6 0.8
Water cut [STB/STB]
Oil production rate Case_4_TM vs Observed 1 Pressure Case_4_TM vs Observed 1
Water cut Case_4_TM vs Observed 1 Oil production rate Case_4_TM_82 vs Observed 1
Pressure Case_4_TM_82 vs Observed 1 Water cut Case_4_TM_82 vs Observed 1
Learning objectives
In this module, you will learn about:
why history matching is useful for making decisions
two different approaches to history matching
the fundamental difference between the traditional and
assisted history matching approaches and their workflows
What is history matching?
The process of adjusting uncertain parameters in the static model
to get an approximate match of the simulated historical
production and injection profile with the observed dynamic well
behavior.
History matching also can involve the adjustment of laboratory
measured data to overcome the inaccuracies inherent in a
numerical solution of the fluid flow.
It is an inverse problem with non-unique solutions. Therefore, it is
essential to obtain multiple history-matched models to properly
quantify forecast uncertainty.
Why do we do history matching?
History matching process helps to find a reduced number of models
containing unique parameter sets that closely reproduce the
dynamic behavior.
Helps to identify fundamental flaws in the conceptual dynamic
model.
Helps to reduce uncertainty and increase confidence in production
forecasts through history matching.
History matching process is normally employed for the purpose of
decision making.
Note: Reservoir models are computer models, not reality.
History matching approach: Classical
Geological model Simulation model
Static
Static reservoir
reservoir description
description
Construction
Construction of
of simulation
simulation model
model
Assign
Assign fluid
fluid model
model and
and rock
rock physics
physics function
function
Make
Make history
history strategy
strategy
Simulation
Simulation model
model
Define
Define vectors
vectors to
to match
match
Parameter
Parameter screening
screening
Manually
Manually modify
modify properties
properties to
to match
match historical
historical
data
data
No
History
History
Match
Match
Yes Accepted history
matched model
Modern techniques approach: Computer assisted
history matching
Static
Static reservoir
reservoir description
description 2
1 N
Si Oi
RMS
N
i 1
Construction
Construction of
of simulation
simulation model
model Where:
• RMS is the Root Mean Square
Assign
Assign fluid
fluid model
model and
and rock
rock physics
physics function
function • S is the simulated value
• O is the observed value
Make
Make history
history strategy
strategy • N is the number of points i
• σ is a normalization parameter
Simulation
Simulation model
model
Calculation
Calculation of
of Objective
Objective Function
Function (Mismatch
(Mismatch
Define
Define vectors
vectors to
to match
match values)
values) to be use by the
to be use by the optimizer
optimizer
Parameter
Parameter screening
screening
Manually
Manually modify
modify properties
properties to
to match
match historical
historical Automatic
Automatic modification
modification based
based on
on the
the defined
defined
data
data variables and their ranges
variables and their ranges
No
History
History Match
Match
Yes
Multiple history matched models
Modern techniques approach: Assisted history matching
Enables parameterization of multiple uncertain input parameters at a
time.
Ability of history matching algorithm to samples uncertain space of
uncertainty variables for both the geological and dynamic models.
Large and complex models are easier to handle with proper cluster
management.
Objective or mis-match function must be predefined.
Requires cluster computing technology to manage the large data that
are generated.
Creates multiple history matched models that can be used to quantify
uncertainty.