DO YOU WANT JOINT CMG (COMPUTER MODELING GROUP) COURSE

Presentation for Pertamina, June 2003 Computer Modelling Group Ltd. David Hicks, European Area Manager

Course Schedule

Day 1

Introduction; Data structure; Installation; Licensing;

Launcher

Using Results

Results Graph and 3D

Day 1/2

Grid design concepts and model building

GridBuilder

Day 3

Fluid modelling, PVT, SCAL

Winprop, Builder

Course Schedule

Day 4

Initialisation and Aquifers

GridBuilder, Builder, IMEX, Results

Day 5/6/7

Wells and the dynamic model

GridBuilder, Builder, IMEX, Results

Conversion to advanced simulators

GEM/STARS

Day 8/9

General modelling

IRAP and Petrel models in IMEX/GEM

Pertamina data

Course Schedule

Day 10

General discussion and questions

Go Home!

CMG Organisation

Research group coming out of University of Calgary

Formed in 1977, went public in 1997

CMG Software Benefits

Integrated modelling environment

Don’t require several programs

No worries about requiring further “add ons”

Integrated with G&G packages - RESCUE

Low entry barriers

Easy to use, PC friendly, working environment

Graphical QC of data and powerful analysis tools

Low cost

High level support from engineers and developers

All specialists in simulation

CMG Software Products

Builder ‘Keywordless’ model creation

IMEX Blackoil simulator

GEM Compositional simulator

STARS Thermal/ Advanced process simulator

Results Output graphics and reports

Winprop Phase behaviour modelling

Formex Gas network and contract optimisation

Wellwhiz Hydraulic fracture modelling

EnAble Assisted HM and risk analysis

Workflow Geological Model Well Trajectories Perforation History Production History PVT SCAL Economics Analysis Risk Analysis SIMULATOR IMEX / GEM / STARS EnAble BUILDER RESULTS

WORKFLOW Project Directory Tree File Drag and Drop Run Scheduling Area

Builder

Easy generation and visualisation of models

Low learning threshold

Most users do not require any formal training

Allows rapid update and alteration as new data arrives

Graphical interface and 3D visualisation provide:

better error checking

better understanding of the reservoir

faster results

PVT and SCAL

Log and perforation display Perforation History

IMEX – Blackoil Simulator

Standard development scenarios

Primary depletion

Waterflooding

Enhanced production methods

Gas lift optimisation

Gas condensate modelling

Polymer and solvent injection

Immiscible and pseudo-miscible gas cycling and WAG processes

Geomechanics and subsidence

Vertical, Deviated, Horizontal and multilateral wells

IMEX II - Parallel version + 64 bit technology

Fractured Reservoirs

Supported by all simulators

Dual porosity

Dual permeability

Refinement of matrix

Vertically - SUBDOMAIN

Radially - MINC

GEM only

3 pseudo cap pressure methods for vertical segregation

Matrix Fracture diffusion (SPE 16672 da Silva and Belery)

Vertical homogenisation of fluid in frac due to natural convection and additions for re-infiltration (similar to SPE 35309 Saidi). Project with IMP for large NFRs

FORMEX

Forgas supplied by Neotec

Provides surface facility and contract management

Pipeline pressure loss and compression management

Forgas

Tank-like gas reservoirs

Formex

Fullfield reservoir model

WellWhiz

WellWhiz supplied by Fractec

Provides dedicated interface

For fracture engineers

With no simulation knowledge

Access the power of IMEX

Multi-phase non-Darcy effects

GEM - Compositional simulator

Fullfield Compositional EOS simulator (PR & SRK) with thermal and solid capability

Usual EoS simulator processes

Additions for UGS and gas mixing SPE 59438

MPFA and flux limiting routines

CBM/ECBM

Asphaltene and wax deposition

Complete water phase interaction – 3 phase flash

Geochemical module – interaction between rock chemistry and CO2, H2S etc

VAPEX and CO2 sequestration

Winprop

Phase behaviour modelling tool

QC plots in Excel

Simple interface, but powerful engine

Multiphase flash (LLVS)

Complete fluid laboratory analysis

Outputs simulator fluid descriptions

Solids, waxes, asphaltenes

K values or EoS

STARS - Thermal and Advanced Process simulator

Unique functionality for modelling many thermal /chemical processes:

Steamfloods, cyclic steam, SAGD

Foam, WAG, FAWAG

Near wellbore treatments - water shutoff gels; polymers etc; lab studies

Scale and solids deposition

Air injection processes; LTO, THAI, CAPRI…..

Combines with ARC Sand for CHOP

Foamy oil; low salinity water injection

Vapex; microbial; ASP; any chemical flooding

Finite element geomechanics module

Results

Display of simulator output

Combined 3D and line plotting

Generation of pseudo well logs

Bubble Plot and Flux/Velocity vector display

Fully flexible calculator

Produce difference plots comparing sensitivities

3D immersive visualisation rooms

Easy export to:

Microsoft Office – AVI movies

Any other software product

Streamline visualisation

Currently only 2003.10 IMEX

System Setup

Hardware

System Requirements

Pentium PC (Recommend Pentium III 500MHz or higher)

Windows 2000, NT4, XP

128Mb of RAM

SVGA monitor

CD ROM and minimum of 250Mb free disk space

Ethernet card with TCP/IP installed and configured

AGP graphics card with OpenGL support

Must come with OpenGL ICD files and have enough memory to support double buffering

Hardware

Memory requirements

For IMEX around 3Kb per active cell

32 bit up to 2Gb, some systems up to 3Gb

“ Practical” limit of 500,000 active cells

1 million cells can be run with 3Gb option

64 bit technology

Allows multi million cell models

112 million active cell model run already

Visualisation

“ Practical” limit of 2-3 million cells for 3D pictures

Installation Structure

CMG directory usually stored under C:Program Files

Product

Version

Documentation and examples

Installation

Three setup types

Network Server

Network Client

Standalone

Need to have administrator privileges to install

Close all currently running applications

Run setup.exe from the installation CD

A “wizard” will guide you through the installation

If using Network setup you should know your server name

A hostid is required for the machine where the license resides

License Manager

A hostid is generated based on the unique ID of the ethernet card used by the PC

A password is generated by CMG that enables the software only on that machine

The passwords are entered using the Configure License Utility

lservrc file in CMGSecure directory

A Network license uses a demon program and must be restarted when passwords are updated

CMG uses SentinelLM to manage licenses

License Manager

Variables used by SentinelLM

LSHOST

Used to specify which computer is the license server

Several server names can be listed here

LSFORCEHOST

Similar to LSHOST but takes precedence

This will force the software to look for licenses on this particular machine

LSDEFAULTDIR

Used to specify the directory where the lservrc file is located

Usually contained in CMGSecure this file contains the passwords

License Manager

CMG_HOME

This points to the CMG directory so that the exe files can be located

Standalone Installation

LSFORCEHOST should not be set

LSHOST should be set to no-net

All this will be set automatically during the installation

The simulators can also be run on UNIX

A UNIX/PC based license manager can run both UNIX and PC license requests

Exercise

Installation of the CMG software

(Standalone license type)

CMG Launcher

Microsoft Explorer look file location

Tree diagram of directories

File display and filtering

Template directories set up by default as projects

Projects

Data can be ordered into separate projects

Projects can be set up on networked drives

Templates

Each software product has a template directory

CMG Launcher

Manuals

Online manuals accessed through the Help menu

Easy to use search facility

File filter

Allows filtering of common CMG file conventions

User defined filtering also allowed

Directory structure

C:Program FilesCMGIMEX2002.10exemx200210.exe

CMG <Program> <Program Version> Tpl , Doc , Exe

CMG Launcher

Icons and modifiers

Drag file onto icon to input file into the program shown

Can add icons for any other non-CMG programs

Additional switches

-3 Results 3D

-g GridBuilder

-x Results Graph

-s 3D rendering speed up

-p Show cell connections

CMG Launcher

Run command interpreter

Can be useful for simulators

Capturing screen output in a file

Additional command line switches….

> test.log

Batching and job monitor facility

Can set up job queues

Datasets can be timed to run at night during low usage times

Right mouse click on the job area to access menus

Log file generated

Exercise

Setting up Icons and using the Launcher

Simulator Data Organisation

All files are of the form <dataset name.extension>

e.g. model1.dat

Basic file extensions

.dat - Simulator input file that contains all the information the simulator requires to perform its flow calculations

.inc - Additional input files referred to in the .dat file

.out - File output by the simulator containing information on the model in ASCII text

.irf - Header file output for graphical post processing

.mrf - Binary data file containing the simulator results

Additional file extensions

.rrf - This file is output, and added to, at user defined intervals, allowing the simulator to restart at any of the time points referred to in this file

.ses - Template file for post processing line plots

.3tp - Template file for post processing reservoir displays

.fhf - Historical data can be stored in this file type for superimposing on simulated results to aid history matching

Ses.inc - Last entry display for GridBuilder

.log - Diary output: timestepping, rates, convergence

The .dat file contains several data areas in the following order:

Input / Output Section

Reservoir Description Section

Fluid Model Section

Rock Fluid Section

Initial Conditions Section

Numerical Section

Well and Recurrent Data Section

Input / Output Section

Controls for filenames used

Usually defaulted and based on .dat file root

Dimensioning statements

Unit system definition

Controls for frequency and content of output

Limit .out file size

OUTPRN GRID NONE

NOLIST

2001 version auto limit

Reservoir Description Section

Description of the model’s geological framework

Type and dimensions of the gridding used

Spatial location of the reservoir

Layering of flow units including gross/net thickness

Fault locations and barriers

Petrophysical properties (  k)

Dual or single porosity models can be chosen

Fluid Model Section

Type of fluid to be modeled

Blackoil; Condensate; Gaswater; Polymer; Miscible

PVT description of the fluid behaviour

Surface phase densities

FVF, GOR viscosity and compressibility data

Rock Fluid Section

Similar rock types are grouped together

Relative permeability tables

Capillary pressure curves

Rock wettability

Hysteresis effects

Initial Conditions Section

Allocation of saturation and pressure to the model

Generally assume equilibrium conditions to allocate values

Based on fluid contact positions and reference pressure

Completes the static description of the model

Fluid-in-place values can be derived

Numerical Section

Matrix solution method can be chosen

Iteration convergence limits can be set

Can be tuned to reduce simulation runtime

Most datasets require no modification apart from perhaps DTMAX

Well and Recurrent Data Section

The static simulation model is now perturbed over time

Wells are added to provide material and disturb the initial pressure regime

Historical data is used to tune the static model parameters

The wells are controlled to simulate possible scenarios

Facilities constraints

Well intervention and workover strategies

Static reservoir properties can also change over time

Fault transmissibility

Pore volume

Each section has its own set of keywords

Keyword ordering can be important

e.g. Need to define a well exists before you can allocate perfs

Some keywords can be used in more than one section

Any text on a line following ** will be ignored

The data in the input file is read by scanning along lines until either the end-of-line or ** is encountered

Wildcards * and ?

Table interpolation using *INT

Data Array Reading

Property (qualifier) read_option data (modifier)

Property

*POR *PERMI

Qualifier

*MATRIX *FRACTURE *RG <block> *EQUALSI

Read_option

*ALL; *CON; *IVAR; *JVAR; *KVAR; *IJK

Data

Modifier

*MOD <block> + - * / = <value>

Summary

Should now be aware of:

Who CMG is

What tools they develop

What areas the tools can be applied to

You should understand:

How to install the software

How its access is controlled

The general data organisation and structure

You should know how to operate the Launcher

Gridding Computer Modelling Group Ltd. David Hicks, European Area Manager

Simulation Model Gridding

Types of grid available

Grid selection and design criteria

Grid Calculations

Modelling the geology

GridBuilder

Aquifers

Typical Models Used in Reservoir Simulation

Model Grids

Models do not all have to be full field 3D

Radial

Welltest analysis

Coning studies

1D

Slim tube analysis for miscible floods

Gravity effects

Symmetry elements

Multiple reservoirs

Common Types of Grid

Radial

Cartesian

Block Centred

Point Distributed

PEBI

CVFE

Curvilinear

Common Types of Grid

These can also be locally refined

Local refinement allows greater spatial resolution localised to areas of interest

Reduces number of cells required and hence runtime

Can get reduction in accuracy due to boundary effects

Radial Grids

Cartesian Grids

Block Centred

Simple grid cell construction based on cell centre

Easy data input

More mathematically accurate, orthogonal, grid

Boundaries not modelled exactly

Faults may be required to zig-zag

Corner Point

Complicated cell construction based on cell corners

Need pre processor for data input

Non-orthogonal grid can lead to numerical error

Accurate boundary location

Gridlines follow faulting

Cartesian Grids

Block Centred

Difficult to distinguish dip from faulting

Zig-zag faulting can distort flow pattern

Vertical faults only

Corner Point

Structural dip distinct from faulting

More accurate flow pattern along and across fault planes

Vertical and sloping fault planes allowed

Cartesian Grids

Corner Point Format 1

Specify top and bottom of a coordinate line

Place cell corner positions along these lines

6 values per line + 8 corner values = 14 values required for each cell

Corner Point Format 2

Specify each corner location for all cells

24 values required for each cell

Format 2 is more flexible but requires more input and can produce insolvable grids

Grid Selection

What type of grid is required to investigate the problem?

Areal orientation of grid

Axes should be aligned along permeability max/min.

Axes should be aligned along fault trends

What sort of layering is required?

Is there an aquifer that needs modelled?

Representation of aquifer in the grid (discuss later)!

Grid Selection

What size of cell do we need to capture the heterogeneity

Areal size of cell + Number of Layers => Number of gridblocks

Number of gridblocks => Computer storage

Trade-off between

Adequate definition for problem to be resolved

Computer time required

Grid Selection

Typical Blackoil simulator IMEX

Memory requirement ~3kB per active cell

Model run time

Would like to run several models per day

Would like to run 1 or 2 large models overnight

What type of computer do I need?

Super computer; Workstation; PC; Calculator?

Grid Selection

The type of grid selected and the number of gridblocks depends upon

Study objectives (what are you trying to model?)

Number of wells and size of study area

Heterogeneities (primarily permeability variation)

Numerical effects

Flow mechanism (gravity segregation, piston-like displacement, gas cap drive, etc.)

Grid Selection

Study objectives

How much accuracy do you need in the problem?

Matching flooding breakthrough times will usually require a large amount of gridblocks in the direction of flow (Numerical Dispersion Effects)

Solution gas drive or expansion drive models can be coarse models because the drive mechanism is dispersed

Gas cap, water drive systems, or fluid injection schemes where gravity segregation occurs require more vertical grids to account for moving fluid contacts

Grid Selection

Infill drilling strategies will need enough gridblocks between existing wells to place the infill wells

For gas caps use coarse grid for the cap itself. Use fine grid where the gas cap will move down. Use locally refined grid near wellbore to capture coning effects

For aquifer it is usually best to use aquifer function rather than large gridblocks if the aquifer permeability is low. If the aquifer is high permeability use a single large gridblock to replicate the reservoir

Grid Selection

Number of wells and size of study area

Large fields, by their nature, require large numbers of grid cells

Models can be restricted in size by using natural no flow boundaries to break up the model into isolated parts

Sector studies can also be used, but influx/outflux to the surrounding area can be a problem. Suitable boundary conditions have to be found

Coning or horizontal well modelling is best investigated by using local grid refinement rather than a universal fine grid. Most of the detail you want to examine is near wellbore

Grid Selection

Heterogeneities (permeability variation)

The engineer must decide whether the permeabilities can be handled by averaging or by explicitly modelling heterogeneities

Heterogeneity is governed by different length scales e.g. lamena, beds, para-sequences, which can be investigated by different methods e.g. variograms, sequence stratigraphy

Upscaling geological heterogeneity is currently a large area of research and controversy

Some parameters lend themselves to simple mathematical averaging

Porosity =  i=1n (Porosity) i

Grid Selection

Others require more complex analysis

Permeability = arithmetic average for flow parallel to layering

= harmonic average for flow perpendicular to layering

= tensor for flow at arbitrary angle

In general, small scale heterogeneities must be accounted for implicitly by some averaging technique e.g. pseudos.

Large scale heterogeneities (megascopic, gigascopic) are directly modelled, as they tend to be at a scale directly resolvable by simulation grids

Geostatistical analysis can be applied at all scales.

Exercise

Follow the hand out exercises to build two different grids:

Orthogonal

Non-Orthogonal

Further exercise

Build a radial grid model around one of the well locations for a single well model

The well radius should equal the inner ring radius

Problem Grids

Some grids are insolvable

Can cause convergence difficulties

Timestep limitation

Can prevent simulation running

Incorrect transmissibilities

Connections may be incorrectly formed

Large amounts of data mean that we need tools to QC grid

2D and 3D visualisation

Problem Grids

Problems usually in XY plane

Large Z aspect ratio and pinchouts

Vertical anomalies not such an issue

Assumptions have to be made

Inaccuracies

Not such a problem at boundaries.

Aquifer influx

Solutions to Problem Grids

Re-grid

Manually edit grid nodes

Manually remove problem cells

NULL

Automatically remove problem cells

PVCUTOFF

CORNER-TOL

PINCHOUT-TOL

Grid Calculation

Things the grid supplies to the simulator

Pore volume

Use gross layer thickness

(  x*  y*  z) *  * NTG * Modifier

 can vary with pressure (rock compressibility)

Modifier can be altered in time - beware of pressure effects

Problem grids make gross volume (  x*  y*  z) difficult, or impossible, to calculate

Grid Calculation

Depth

Gives position relative to fluid contacts for initialisation

Allows gravity effects to be modelled:  Pgravity =  g  h

Effects transmissibility calculation

Grid Calculation

Transmissibility

Simplistically :

Transmissibility = interface area * permeability *multiplier

Calculation varies depending on grid type chosen – BC or CP

NTG term modifies horizontal interface area, but not vertical

Problem grids make cell connections and transmissibility difficult, or impossible, to calculate – only CP grids

Grid Calculation

What cells are active in the model

Null arrays completely remove unwanted cells

Can act as barriers to flow

Zero porosity cells contain no fluid, but facilitate heat flow

What connections exist between cells

Can depend on grid type !

Grid Calculation

Block centered grids

Direct flow from (i, j, k-1) to (i+1, j, k) not included

Often have to add connection manually if wanted

Point distributed grids

Mutual cell overlap area used

NNC connections made automatically

Modelling Faults

Scale of Fault

Can it be seen from seismic or well tests?

Does it have significant throw?

Model implicitly (multipliers) or explicitly (part of grid)

Fault plane transmissibility

Simulators assume perfect sand to sand connection

Modelling Faults

Should the fault plane slope or is vertical adequate?

Block centred grids often have to zig-zag faults

Point distributed grids allow most fault traces to be followed exactly

History matching process can alter ideas about faulting

Fault Transmissibility

Default

All fault connections are sand to sand

Transmissibility only modified by the overlap area

GridBuilder Data menu

Defined map faults allow individual selection

Without any map all fault connections selected

Exercise

With the class instructor do the following:

Discuss the transmissibility across the fault present in the Tutorial exercise

Make the fault in Tutorial model sealing

Further exercise

Build a cartesian grid based on the same top and thickness maps as the first exercise

Note the new keyword FAULTARRAY in the saved dataset. Discuss its significance.

Edit the grid and rotate to make the fault zig-zag

Modelling Shale

Issues to consider

Is it dispersed or does it form distinct layers?

How much hydrocarbon is contained in the shale?

Is any of it recoverable?

Beware of cut-offs removing mobile fluid.

Is it semi-permeable and are there windows?

How thick is it?

How laterally extensive is it?

Modelling Shale

Several ways to model dispersed shale horizons

NTG ratio

Will effect horizontal transmissibility

Will not effect vertical transmissibility

Kv/Kh ratio will include some dispersed shale effects

Kv reflects the tortuosity caused by dispersed shale

Geostatistically created transmissibility arrays

Modelling Shale

Several ways to model extensive shale horizons

Explicit layer of cells

Allows direct modelling of flow and hydrocarbon volume

Transmissibility barrier

Quicker modelling of thin horizons

Gaps in grid

Correct pressure centre for sand cells

Acts as a barrier to flow

Exercise

Look at ways of adding shale barriers to the Tutorial exercise.

Modelling Boundaries

Point distributed grids allow exact positioning of boundaries

However cells can be severely distorted

Lease boundaries, faults, etc can be positioned exactly

Modelling Boundaries

NULL arrays needed for irregular shapes as grids need constant I J and K dimension

Block volume modifiers can also be supplied

Truncated Layers

The concept of pinchouts:

Grid must have constant I, J, and K dimensions

K layer does not disappear but becomes vanishingly small

Cell must be nulled at some point and connection made across it (Non Neighbour Connection)

PINCHOUT-TOL

PINCHOUTARRAY

Summary

The reservoir needs to be discretised to allow

Solution of the mathematical model

Fluid positions to be observed

This discretisation leads to numerical errors

Smearing of flood fronts

Methods are available to limit this error

Chose the grid type to reflect the problem you are trying to solve

Geologically realistic scenarios can be obtained

Trade-off between speed and accuracy

GridBuilder Features

Forms of Geological Data

Scattered data points

Not on regular grid, sparse (e.g. picks at wells)

Contour maps of 2D surface

Sets of connected points forming line with value, may contain faults and well locations

WINDIG, ZMAP CNTR, CPS-3

Mesh maps of 2D surface

Regular, orthogonal “grid” of data, value at each point, may contain fault lines and well locations

CMG, ZMAP GRID, EarthVision, CPS-3

3D geological model cell data

May use millions of cells

Can handle more complex 3D geometry

Some geological and geostatistical programs directly create simulation grids

(RC) 2 - Geosize

Roxar - IRAP RMS

Schlumberger - Petrel

CMG keywords and formats supported

RESCUE format

Grid may also be imported from model built for non-CMG simulator

Grid Construction

Number of simulation layers may be greater than number of geological layers

One or more top maps or bottom maps

Thickness maps

Problem of “poor” top maps with overlap

Overshoot/Undershoot

Specifying Properties

Select Property and the layer, PVT/RTYPE/Sector to be defined

Can use Value ; Map ; Formula

Editing Properties

Properties once allocated need to be calculated

Once defined the property can be graphically edited

Apply multiplier to each block value separately Multiple blocks and rectangles by holding down <Ctrl> key

Specifying Properties

Simple Input Array

CON

KVAR

IJK

Can use well perforation button to set IJK to halo around selected wells

Sectors

Sectors define individual reporting areas

Defined either by:

Numbered array

Distinct sector name

Report to text or graphics output files

Sectors

Named Sectors defined under Data menu

Selected areas highlighted

Probe shows sector name

Splitting Layers

Can divide layers in 2 different areas

Tools Menu in ModelBuilder

Layers can be combined here also

Splitting Layers

GridBuilder

Edit Grid

Do not delete property and well data

Select a cell

Select Grid : Split Grid Plane

Can also use to divide grid in I or J planes

Same approach for submodels or refinement

Array Property Calculator

User enters formulas to define property arrays from other arrays using:

Logic, arithmetic, and logarithmic functions

Have an easy to use interface for constructing formulas

Save and restore of formulas in dataset

Grid and property statistics calculation

Array Property Calculator Formula shown here Available operators Independent variables used in calculations

Enhanced Accuracy Using LGRs

Local Grid Refinement

LGRs often used around wells

Grid block areal dimensions are much larger than the wellbore diameter

results in lack of resolution in area where change is most rapid

wells can be positioned incorrectly if not centred on block

converging / diverging flow around the wellbore results in steep saturation and pressure gradients

coarse Cartesian grids may not provide enough resolution for accurate calculations of WOR, GOR and breakthrough times

phase segregation may also occur as pressure drops below dew/bubble point

especially important with condensates where oil can be trapped in the formation

e.g. coarse grid nodal pressure may be above P dew while the pressure close to the well is below P dew

Near well refinement can be accomplished using Cartesian or Hybrid refinement

Hybrid grid with four radial (three “rings”), and four theta subdivisions, required for 3D grids.

Inner radius must equal Rw

Only define the completion once (not for all 4 segments)

LGRs are defined in relation to their parent cell

Each LGR has its own unique numbering

Combination of parent cell and LGR cell for location

e.g. 2 3 2 / 1 1 2

Multiple levels of refinement can be defined

Prevents sharp cell size change when high amounts of refinement required

Multiple LGR nesting with hybrid grid at final level

e.g. 2 3 2 / 1 1 2 / 2 2 1 ……..

LGR properties are inherited either from:

The parent cell

By individual definition

POR RG 12 8 1 CON 0.1912375

Builder does this for the user automatically

Viewing LGRs

By default view middle LGR layer

Menu option to see other layers

Can edit LGR properties in same way as main grid

Exercise

Create a simple poro-perm correlation

Null cells below an oil-water contact

Add an LGR to the Tutorial exercise

Individual Exercise

Define separate sectors for each layer in the Tutorial dataset, so that we can report fluid volumes on a layer basis

Additionally split the reservoir into 2 sectors. One on the left of the fault, the other on the right of the fault.

Initialisation Computer Modelling Group Ltd. David Hicks, European Area Manager

Reservoir Initialization

Allocation of saturation and pressure profiles

Can be done automatically assuming equilibrium

Can be allocated manually

Usually for lab experiments and special situations

Dynamic situations need further intervention

Tilted Contacts

Multiple initialisation regions

Different rock types with individual Pc curves

SWINIT map of water saturation

Works by cap pressure curve scaling

Automatic allocation of values requires:

A datum pressure and depth

WOC and/or GOC positions

Contacts represent a level where Pc equals a defined value (default is zero)

*DATUMDEPTH

Reporting purposes only

Average pressure usually based on floating datum approach

Density data allows pressure profile to be generated

Capillary pressure data allows saturations to be allocated

Contact Variation

Multiple contact positions can be accommodated

Need 1 PVT region for each initialisation region in IMEX

Multiple PVT or EoS regions

Instantaneous change in PVT properties

API tracking for blackoil

1 EoS accounting for vertical compositional variation

Thermal effects can be a main influence on vertical system

Gravity segregation should be captured by EoS

Normal Initial Saturation Distribution

Rock type variation will cause a variable saturation profile as Swcon, Sgcon, and Pc vary.

S w , S g , S o calculated from gravity-capillary equilibrium consideration

Gas Cap

Can set the oil saturation in the gas cap depending on fill history

Applies when Po-Pg > largest tabulated Pcog

Initially water saturated rock fills with gas

*GASZONE *NOOIL

Sw derived from Pcow. Sg = min(Sg(table), 1-Sw). So = 0

Gas displaces oil originally in place

*GASZONE *OIL

Sw derived from Pcow. Sg = min(Sg(table), max(1-Sw-So(min),0) where So(min)=max(Sorg+Swc-Sw, 0). So = 1-Sw-Sg

Only applicable when using *DEPTH_AVE

*BLOCK_CENTER always uses *NOOIL

Bubble Point

Can be set as a constant value or user defined array

Gravity segregation of components can be accounted through varying the value vertically

If a gas cap is present then the bubble point should be consistent with GOC

Bubble point and reservoir pressure sets the initial oil and gas properties - FVF,  , GOR

Vertical layering can limit the positional accuracy of the fluid contacts and transition zones

These factors will all influence the OOIP

Solutions to the layer resolution problem

More model layers

Vertical LGR

Saturation averaging

*DEPTH_AVE

*EQUIL (default) - Pc adjusted to preserve gravity equilibrium

*NOEQUIL - No phase pressure correction applied

Solution choice depends on questions to be answered by the model e.g. horizontal well cusping

Depth Versus Initial Saturation for Various Number of Layers

Reservoir Initialisation

Automatic initialisation assumes constant contact depth. This is not necessarily true.

Tilted contacts may be present if an active aquifer is flowing under the reservoir

Initialise several contact positions and hold fluid in place with pseudo capillary pressures

Restart files can be used to initialise the model once dynamic equilibrium is achieved

To check that equilibrium has been reached run the model for several timesteps and see if any fluid movement occurs

Can set Initial Sw in GridBuilder (*SWINIT)

Pcow curves scaled to generate equilibrium condition

Reservoir Initialization (GEM)

Eight options

*USER_INPUT

User specified initial conditions. *PRES, *SW, *ZGLOBAL

*VERTICAL *BLOCK_CENTER *WATER_OIL

Under-saturated reservoir, where pressure and composition result in only the oil phase being present. If 2 phases exist then problems may occur as gas redistributes itself. *DWOC, *REFDEPTH, *REFPRES, *ZGLOBAL

Gravity-capillary equilibrium calculations are performed to calculate all grid block pressures and water saturations. So and Sg are determined by flash

*VERTICAL *BLOCK_CENTER *WATER_GAS

*DWGC *REFDEPTH, *REFPRES, *ZGAS

Gravity-capillary equilibrium calculations are performed to calculate P and Sw. Sg is then determined by subtraction.

*VERTICAL *BLOCK_CENTER *COMP

Initialization with composition as a function of depth. *DWOC, *REFDEPTH, *REFPRES,*CDEPTH, *ZDEPTH

Gravity-capillary equilibrium calculations are performed to calculate Sw. So/Sg split determined by flash calculation. If single phase determined then *CDEPTH is used to say whether it is oil or gas.

*VERTICAL *BLOCK_CENTER *WATER_OIL_GAS

Saturated reservoir where gas cap is in equilibrium with oil leg. Two compositions are required representing the average oil and gas zone compositions. *DWOC, *DGOC, *REFDEPTH, *REFPRES, *ZOIL, *ZGAS

Gravity-capillary equilibrium calculations are performed to calculate all grid block pressures and all oil, gas and water saturations. Flash calculations are performed to determine the oil phase and gas phase compositions. Global composition is calculated by mixing the oil and gas phase by the gravity- capillary-equilibrium determined saturations. Thus the global composition in the reservoir may vary with depth.

Replace *BLOCK_CENTER with *DEPTH_AVE for other 3 options.

GEM Fluids-in-Place

Initial FIP can be defined by:

By default GEM will use last EOS defined (highest numbered EOSSET) and “standard” T and P

Specified FVFs - Bo, Bg, Rs (same as “blackoil”!)

Reservoir volumes directly translated to surface volumes

Initial “average” separator condition

All reservoir fluids flashed through the separator

This is preferred as it produces results consistent with well produced volumes

Wells can have different separator conditions

Otherwise they will be referenced to the separator defined in the initialisation section

Exercises

Exercise to see effect on OOIP

Change number of layers

Block centred vs depth ave initialisation

Look at Pc and PVT alteration effects - density

Fluid Analysis Computer Modelling Group Ltd. David Hicks, European Area Manager

Reservoir Fluid Description

Reservoir model = Geology + Fluid

3 main ways of describing reservoir fluids

Simple “Black-oil” look-up table

Equilibrium K-value tables (pressure, temperature, composition)

Equation of State

The reservoir fluid and how we plan to exploit it determines the type of fluid description required

“ Component Properties” section in Builder

Reservoir Fluid Description

“ Black-oil” PVT description (IMEX)

Primary depletion

Waterflooding

Immiscible gas injection (solvent model allows pseudo-miscible)

EOS PVT description (GEM)

Miscible gas injection (solvents/CO2)

Volatile oil systems

Gas condensate systems

K value PVT description (STARS)

Temperature variation

Black-Oil PVT

Data required to describe a “black-oil”

“ Black-oil” PVT data describe 3 components and 3 phases

components are oil, water and gas

phases are oil, water and gas

gas component can exist in both the oil phase and gas phase

oil component can exist in both the oil phase and gas phase

PVT properties vary only as a function of pressure at a given reservoir temperature

Compositional PVT data varies both as a function of pressure and composition

Black-Oil PVT

PVT data is entered in tabular form

Enter Rs, Bo, Eg,  o,  g as functions of pressure

Rs - solution gas ratio

Bo - oil formation volume factor

Eg - gas expansion factor

 o - oil viscosity

 g - gas viscosity

Instead of Eg, gas properties can be described by:

Bg = 1 / Eg or Z = CP / EgT (C = constant)

Black-Oil PVT

Entered data should be smooth with respect to pressure

Sharp kinks may result in numerical instabilities that may be difficult to relate back to PVT

Two phase models can be run: oil-water; gas-water.

However, if 3 phases are present then a description of the fluid response both above and below the bubble point must be provided

Experimental Analysis

Three main, and one optional, experiments are required:

Flash Expansion - of fluid sample to determine bubble point pressure

Differential Liberation - to determine Bo, Rs, Bg and viscosity

Flash Separation Test - to enable modification of differential liberation data to match field separator conditions

Swelling Test - to provide properties when gas is added

Black-Oil PVT

Results of fluid analysis

Below bubble point pressure

process in reservoir simulated by differential liberation

process from bottom of well to stock tank simulated by separator test

fluid properties calculated by combining data from differential liberation and a separator

Black-Oil PVT

Above bubble point pressure

fluid properties calculated by combining data from flash expansion and a separator test

note that results of differential liberation and flash expansion are the same above bubble point pressure

IMEX allows both

direct PVT input *PVT and *PVTG

or Diff Lib input *DIFLIB and *PVTG

Requires flash Pb, Bo, Rs values

Does the following calculations for the user automatically

Black-Oil PVT

Conversion of differential liberation data to separator conditions:

Black-Oil PVT

Additional information required by simulator

oil FVF slope above bubble point

oil viscosity slope above bubble point *VOT

IMEX

IMEX conventions

PVT table describes saturated behaviour only

Can supply compressibility within the tables or as separate entries

*CO/*BO or *COT/*BOT describes undersaturated oil behaviour

Can have different CO/BO values for different Pbub

Bubble Point Pressure and Swelling Curve

Black-Oil PVT

Water compressibility assumed constant *BW, *CW, and *REFPW

Water viscosity *VWI and *CVW

Condensates

Similar conventions for condensates

Vapourised oil in gas, Rv, instead of dissolved gas in oil, Rs

*PVTCOND supplies saturated behaviour of P vs Rv

*E/B/ZGUST supplies undersaturated behaviour

Compositional simulator (GEM) provides more accurate information

Missing Data - Water

IMEX has an internal representation of water if actual properties are unknown

Builder allows access to these values.

Multiple water properties can be assigned to different parts of the reservoir

Missing Data - Oil & Gas

Builder allows access to simple PVT correlations

Empirical Correlations

Main properties which are determined:

Bubble Point; Gas Solubility; Volume Factors; Density; Compressibility; Viscosity

Many correlations are particular to specific areas

Some of the more widely used correlations are applicable to specific data ranges

Blackoil Correlations

Standing

105 data points from 22 different Californian crude oils

Lasater

Bubble point correlation using 158 data points from 137 Canadian, U.S., and Venezuelan crude oils

Vasquez and Beggs

Solution GOR and FVF correlation from 6004 data points

Glaso

45 North Sea data points used

Blackoil Correlations

Marhoun

Bubble point correlation using 160 data points from 69 Middle Eastern crude oils

Other people have used combinations of the above

Be aware of:

The range of values used in generating the correlation

The area of applicability

The size of the sampled population

The correlations are used to estimate reservoir fluid properties from basic field data

API Tracking

Not available, yet, through Builder interface

Allows mixing of varying fluid types unlike PVT regions

Can model compositional gradients (lateral and vertical)

*MODEL *API-INT

*APIGRAD (oil table)

*PVTAPI (gas table)

 o(stc) =  o(stc) *Vh +  o(stc) *Vl

(Vh - heavy volume fraction)

Rs = Rs(mix) obtained by table interpolation

Exercise

PVT generation Example 1

Generate black oil data using the described experiments in Winprop

GEM

GEM uses EoS to describe fluid interactions

Required for hydrocarbon interaction effects to be modelled correctly

PVT is a strong function of composition and not just pressure

Typical processes

Gas recycling, volatile oils, and miscible flooding

Also UGS and CBM, plus special processes

GEM

Components can be defined

From internal library of common oil components

Can define EoS manually in Builder. User defined components

Need a fluid analysis package to tune EoS to observed lab data

e.g. Winprop

Winprop

Winprop allows a mathematical model of the fluid to be created

This has many advantages

Fluid doesn’t degrade

Limitless supply of fluid

Conduct many different lab experiments

Fluid variation modelled by physical processes

Disadvantages

Only as good as the fluid samples it is tuned to

Just a mathematical approximation

EoS

Why do we need an EoS to describe our fluid?

Blackoil assumes constant composition

Oil is made of many different components

Certain processes rely on variation in these components

Miscible flooding

Gas stripping schemes

Condensate, volatile oil and critical fluids phase behaviour closely tied to P, T, composition interaction

T originally assumed constant but is now becoming more of an important influence

Water Solubility

Water solubility is also another issue

Transport of CO2 through aquifer

Removal of lighter hydrocarbons

UGS

GEM uses Henry’s Law to define component solubility in water

Can be tuned like any other EoS in Winprop

Default values in GEM

Phase Determination

Phase determination in the reservoir is important as this determines the rel perm curve and cap pressure

OK if EoS predicts more than one phase - high density assigned as oil phase, low density gas.

If EoS predicts single phase behaviour. How to determine whether it is oil or gas?

User specified - *PHASEID *OIL or *PHASEID *GAS

Phase Determination

*PHASEID *CRIT - To avoid expensive critical point calculations, EoS critical properties are computed from mixing rules.

Gas at supercritical conditions

Oil at subcritical conditions (molar vol.<critical molar vol.)

*PHASEID *DEN - Phase classified by whether its mass density is closer to an internally determined reference gas density or oil density.

*REFDEN - The phase identity is determined by comparison of the phase density with a user specified reference density

The above keywords apply only after the start of the simulation. (Initialisation covered later)

Separators

May be difficult to match EoS across large range of P&T.

GEM allows multiple EoS descriptions

Can also split into reservoir and surface EoS

Can have several EoS. Each describing a different separator stage.

Can also produce INL as well as OIL and GAS

WINPROP

Winprop File Extensions

.out ASCII file containing calculation results.

.gem PVT data for GEM.

.gmz Composition versus depth data for GEM

.str PVT data for STARS

.imx PVT data for IMEX, or generic simulator

.rls Output of component properties from the regression or lumping and splitting procedure.

.srf Output for plotting.

.xls Excel™ file created from an .srf file.

Main Menu

Start with 3 default forms

“ U” in Stat column, undefined

Inc - include/exclude form

Several datasets can be open

Can copy forms between datasets

Can only have one present if you want to open a form

Component Selection

EoS requires

P c and T c , acentric factor (  ) and BICs. MW also required for mass density

Vol. Shift  a and  b can also be defined to enhance prediction

Entered into Component form

Library components

User component with known properties

User component with known SG, Tb, and MW

Use <cntrl> or <shift> for multiple selection

Need minimum of 2 of the 3 properties

Missing property estimated using physical properties correlation

Valid ranges

Twu: Tb<715C, SG<1.436

Goosens: MW 76 to 1685, density 0.63-1.08 g/cc, Tb from 33 to 740C

Riazi-Daubert: Tb<455C, MW from 70 to 300

All give similar results below C20

Riazi-Daubert not so good above C20

Goozens good at predicting MW up to C120

Once SG and NormBPt are known the critical values can be calculated from the selected correlation

Other correlations range as before

Lee-Kessler valid upto Tb 650C

Lee-Kessler recommended for accentric factor

Compositions will be normalised to 1 or a warning issued

Several compositions can be defined.

Each remains current until next composition or end of dataset

Secondary composition usually represents injected fluid

Winprop Lab Experiments

Constant Composition Expansion

Differential Liberation

Constant Volume Depletion

Separator Test

Swelling Test

Single Phase Compressibility

Recombination

Winprop Calculation Functions

Flash calculations to get phase split

Asphaltene and wax modelling

Saturation pressure and temperature; critical points; cricondenbar -therm determination

2 phase and 3 phase envelope creation

3 phase envelope gives range of P&T for second liquid phase e.g. CO 2 dense fluid

Winprop Calculation Functions

Miscibility modelling

Aids design of gas or solvent injection processes

Analyse P, T and composition variations

First contact (MMP) and multi-contact process modelling

Ternary diagrams aid analysis

Compositional grading

Process Flow

Flash Calculations

Winprop Flash calculations

Two-phase vapor-liquid

Three-phase vapor-liquid_1-liquid_2

Three-phase vapor-liquid-aqueous

Four phase flash calculation (fluid phases only)

Multiphase flash calculations with a solid phase

Isenthalpic flash calculation

Isenthalpic flash calculations correspond to finding the temperature, phase splits (phase mole fractions) and phase compositions, given the pressure, composition and enthalpy of the feed, together with the net enthalpy added to the system.

Requires an additional energy balance equation.

Mixing Oil and Gas

The feed composition used for all calculation options can be

A mixture of the primary composition and the secondary composition

The feed from the previous calculation option

The vapor composition from the previous calculation option

The liquid composition from the previous calculation option

The composition from Phase n from the previous calculation option

Altering the EoS

After calculations which alter the components or their properties e.g.

Splitting/Lumping

Regression

You can Update Component Properties

Used if happy with the results of lumping, splitting, or regression

Modifies the original EoS to the new parameter values

The new values were those present in the .rls file

Cubic Equations of State

Corresponding States

All gases behave ideally as pressure approaches zero

PV = RT (v is molar volume)

Due to intermolecular forces real gases do not behave ideally. Particularly at high P

PV = ZRT (Z is compressibility factor)

Using the Principle of Corresponding States –”Substances behave similarly when they are at the same relative proximity to their critical points” - Z only depends on: T r = T/T c P r = P/P c

Corresponding States

This implies all substances behave similarly at their critical point and should have equal values for Z c

Z c = P c V c /RT c

The real value for Z c is not the same for all compounds

---------------------

For pure compounds the following approximation is valid for the vapour pressure P v :

log(P v /P c ) = A - B(1/(T/T c ))

At the critical point : log(P v r ) = A(1-1/T r )

Acentric Factor

The vapour pressure curves of all compounds plotted in reduced form should all lie on the same line. This does not happen

P v r curves are the same for simple spherical molecules at Tr = 0.7

A third factor  , the acentric factor, was added to account for deviation from this ideal case

 = -log(P v /Pc) (@Tr=0.7) - 1.0

Acentric Factor

 = 0 for simple spherical non-polar molecules

 increases with increasing molecule size

 , T r , P r are used in phase equilibria calculations to define the vapour pressure

Typical Values

Mixtures

Mixtures of pure components require a further term to be introduced

The BIC is empirically determined by minimising the difference between predicted and experimental data of binary systems

Therefore it is considered to be a fitting parameter rather than a rigorous physical term

dii = 0 and dij = dji

The value of dij increases as the molecules become more dissimilar in size or increase in polarity

So to solve the EOS for a mixture we require Pci, Tci,  i (the values for each pure component), and the empirically determined interaction coefficient dij

Binary Interaction Coefficients

Hydrocarbon - hydrocarbon interaction coefficients can be estimated using the following correlation

v c i is critical molar volume

d ij = 0 when n=0

n ~ 1.2 matches the paraffin-paraffin d ij of Oellrich et al

n is an adjustable parameter

d ij is not temperature dependent

BICs in WinProp non - HC }

Cubic EOS Calculation Pressure Temperature Composition EOS Equilibrium Phase split Phase properties and volumes

Volume Translation

Using 2 parameter EOS, predicted liquid molar volumes generally showed a systematic deviation from experimental data

Reduced values, and accentric factor give the 2 parameters

The deviation was seen to be almost constant over a wide pressure range

Applying a constant correction term can improve the predicted liquid density

Volume shift concept introduced by Peneloux et al.

Volume Translation

Results in 3 parameter EOS

The volume translation technique of Peneloux et al (1982) is used to improve the density prediction of PR and SRK EOS

The volume shift can be applied independently of the V-L equilibrium ratio, to improve the EOS density/volume prediction

This parameter can be regressed separately

Tuning the EOS

EOS are based on a combination of theory and observed results.

The EOS rarely matches the laboratory PVT results

Failure to characterise the plus fraction fully

Data errors

Inadequacies of a cubic EOS

Its parameters can be adjusted to better match the observed data

This is done by regression

Minimising some difference function between the EOS and the data

Tuning the EOS

Check data is consistent, and reject poor data

Composition mole fractions sum to 1

Material balance checks on results observed - Recombination; CVD; Separator tests

The same data can be obtained different ways

Don’t just average the values obtained

Some experiments are better than others

Gas Chromatography provides the most reliable information on the relative concentration of light components

Distillation provides more reliable data for the heavier components

Mole or Mass?

Component concentration is usually measured in a mass (or volume) basis

Results are generally reported in a mole basis using a measured or generalised SCN molecular weight

Advantageous to work in mass rather than mole fractions

Allows unreliable MWs to be adjusted while retaining the original composition

Parameters to alter

Several different approaches to the variables that can be altered to obtain a match

Coats and Smart used  a  b and the BICs

They used only a 2 parameter EOS, which can have large liquid property prediction errors

Resulted in having to changed “well defined” parameters e.g.  a (C1)  b (C1)

3 parameter (volume shift) EOS allows better liquid prediction

Parameters to alter

Better to change T c and P c directly rather than indirectly through 

Allows monotonicity checking

Increasing MW leads to increasing T c and decreasing P c (in general)

3 parameter EOS gives extra flexibility allowing

P sat matching by variation of critical properties

Independent matching of saturation density (liquid) or Z-factor (gas) using the volume shifts

Tuning BICs

Evidence suggests that adjustment of the BICs to create an “un-symmetric” pattern is dangerous

A good match can be obtained which gives spurious results at other temperatures

Winprop regresses on a coefficient to preserve symmetry

Non-HC BICs can be regressed on individually

Tuning the EOS

Volatile fluids e.g.condensates should have the plus fraction split as the fluid properties are highly dependent on the heavy end components

The plus fraction or pseudo components have the parameters which are most in doubt

Light component properties should rarely be changed

Usually limited to BICs

Light component pseudos

Other factors

Critical Volumes or Z-factors are only needed for the LBC(JST) viscosity correlation and so can be regressed independently

Volume shift parameters can also be regressed on independently to match the liquid volume/density

Weighting factors allow certain results to have more influence e.g.

P sat and then saturation density should have a high weight

Individual CVD compositions should have low weight

Grouping

Avoid regressing on properties of individual components with small mole fraction

Grouping can help avoid this

Grouping can also preserve relationships between components

e.g. split plus fraction components can be varied together

Property trends should be preserved

Condensates

Condensate dew point is an important parameter

However, beware of placing too much weight on its value

Many condensates show delayed liquid drop out below P dew

Thus matching P dew can lead to over prediction of initial liquid volumes

Dew point can be difficult to measure accurately

Often only a gradual change is seen

Easier to observe near the critical point

Better to place more emphasis on matching liquid volumes rather than dew point

WinProp Options with Regression

Saturation pressure and temperature calculation

Two and three phase flash calculations

Critical point calculation

Differential liberation

WinProp Options with Regression

Separator calculations

Constant volume depletion

Swelling calculation

Constant composition expansion

Available Regression Parameters

The following affect all properties:

P c Critical Pressure

T c Critical Temperature

 Accentricity

T b Normal Boiling point

 a EOS alpha coefficient

 b EOS beta coefficient

The last 3 items all indirectly alter the critical values

Only affect Saturation Pressure

d ij Individual BIC’s

BIC hydrocarbon interaction parameter

Available Regression Parameters

Only affect Mass Density

Molecular Weight

Volume Shift

Only affect Viscosity

viscosity correlation parameter

exponent for viscosity correlation

critical volume in viscosity correlation

Properties to be Matched

Saturation pressure

Phase densities

Mole or volume fraction of a phase

K-values of all components

Phase viscosities

GOR, relative oil volume, swelling factors, liquid dropouts, etc.

Regression

Minimise, using a least squares approach, the function

f(  ) = k=1  nm [ r k (  ) ] 2 (nm is number of measurements)

Parameters are sorted in order of sensitivity

Sensitivity is checked each iteration and parameters to be regressed are altered

The process stops when:

The, user defined, number of regression parameters are within limits

The maximum number of iterations is reached

Procedure for PVT Matching

1. Split the C 7+ (or C 6+ ) fraction into a suitable number of components. Then use these C 7+ components (usually 3 to 5 components) with the standard components to develop a ten to fifteen component model for the reservoir fluid

2. Perform regression with fluid model. Use PVC3, and T c and P c of the heaviest C 7+ components and methane, and the volume translation factors for all C 7+ components. If there is a lot of CO 2 present in the mixture also include the binary interaction coefficients between CO 2 and the C 7+ components as regression parameters

3. Include all available experimental data in the regression calculation. These include saturation pressure, constant composition expansion, differential liberation, swelling tests, separator tests, and phase equilibrium data

4. Perform regression. If a good match is achieved then stop; otherwise try some of the following changes:

Change the weight. Some of the experimental data, usually saturation pressure has a high weight

Add T c and P c of more C 7+ components to the regression variables

Change the default values of maximum and minimum values of the regression variables that are at their limits

5. If the above steps do not improve the match go back to Step 1 and redo the splitting of the C 7+ and repeat steps 2 to 4

6. Once a good match of the experimental data is achieved the EOS parameters can be used in GEM

7. If the number of components is too large for compositional simulation lump the components into fewer components. Then perform regression on the lumped components using only the T c of the heaviest component as a regression parameter. To get a good match of the phase behavior and an oil-solvent system the critical point on the P-x diagram for the many component and few component system should be the same

Light-Oil Fluid

a) Reported composition

Light-Oil Fluid

b) Modelled fluid composition

Light-Oil/CO 2 P-X Diagram

Properties of Original and Regressed Light-Oil Components

Exercise

Condensate Case Study.

Rock Fluid Interaction Computer Modelling Group Ltd. David Hicks, European Area Manager

Rock-Fluid Interaction

Simulation models use cores to help determine

Porosity

Permeability

Model Layering

SCAL

Relative Permeability

Capillary pressure

Wettability

Hysteresis

Pick out your relative permeability end points first as these control the recoverable volumes.

If you only have endpoints Builder can fill out the rest of the curve

Otherwise data can be copied and pasted into tables, or entered via an Excel reader

Endpoints can also be scaled on a gridblock basis

Usually two-phase data determined

SWT : Sw Krw Krow

SGT : Sg Krg Krog

SLT : Sl Krg Krog

Capillary pressures are also entered as part of these 2 phase tables

IMEX allows scaling of critical and connate separately

Three phase measurements are very difficult to conduct

Usually, simulators use Stone’s model for interpolating 3 phase data from 2 phase data

Is quite popular since it fits measured data quite well

Stone 1 and 2 available

Also segregated flow model

Averaging Relative Permeability Data Note: If there is a number of relative permeability experiments plot the data on normalized scale Then draw best fit line through it and denormalize the data based on averaged end points Builder provides a simple interface for this functionality and also for cap pressure averaging Normalized Relative Permeability Curves

Builder Smoothing Functions

Capillary-Pressure Data

Cap pressure determines transition zone size

Fluid saturation is strongly effected by the relationship between Pc and permeability

It therefore becomes necessary to evaluate the various sets of capillary-pressure data with respect to the permeability of the core sample from which they were obtained

Capillary-Pressure Data

The J-function correlating term uses the physical properties of the rock and fluid and is expressed as

IMEX allows the user to supply a J function directly if surface tension is supplied

This can be applied to only one, or both, of the 2 phase systems

Otherwise cap pressures are read directly

Wettability

Oil wet (oil is preferential wetting phase)

Water wet (water is preferential wetting phase)

Intermediate wettability

Both oil and water are wetting phase

Mixed wettability

Reservoir consists of regions of oil wet and water wet reservoir rock

IMEX/GEM allow 1,2, and 4 STARS has 3 intermediate wet models

Wettability

Gas is always nonwetting in the presence of oil or water

Wettability

For oil-wet systems

Earlier water breakthrough

Lower initial Sw for given pore volume

Condition kro = krw occurs at lower values for Sw

Higher values of krw at most values of Sw

Lower values of kro at most values of Sw

Wettability

West Texas carbonates, good example of oil wet system

Hysteresis

Experimental studies show dependence of relative permeability and capillary pressure on saturation and saturation history

Depending on whether wetting phase saturation increased (imbibition) or decreased (drainage) different capillary pressure or relative permeability curve observed

Hysteresis effects more pronounced in non-wetting phase relative permeability than wetting phase relative permeability

Due to trapping of the nonwetting phase by the advancing wetting phase

Hysteresis

Hysteresis effects have also been noted for capillary pressure

Capillary hysteresis has a noticeable influence on water coning behavior

Pcow hysteresis and Krg can be important

Pcog hysteresis and wetting phase relative permeability not very important

Hysteresis

In IMEX/GEM

Drainage and imbibition cap pressure data entered directly along with a dimensionless transition parameter

A single rel. perm endpoint for the max imbibition residual can be associated with each rock type

Hysteresis effects modelled for oil and gas using approach similar to Killough, 1976. Based on user supplied Sgrmax and Sormax values

Can specify different hysteresis parameters for different rock types

Simulator Hysteresis Availability

*EPSPC

This determines the transition between the imbibition and drainage curves for oil-water capillary pressure.

*EPSPCG

This determines the transition between the imbibition and drainage curves for oil-gas capillary pressure.

*HYSKRG

Gas relative permeability hysteresis modelled.

*HYSKRO

Oil relative permeability hysteresis (krow) modelled.

Hysteresis

Water-oil capillary pressure

Hysteresis

Gas relative permeabilities

during drainage

during imbibition

Exercise

Complete the tutorial model so that it can be initialised

Full model.doc

Aquifers Computer Modelling Group Ltd. David Hicks, European Area Manager

Aquifer

What is an Aquifer?

Water saturated rock

A source of water outside of the hydrocarbon reservoir

Provide external pressure support to the reservoir

Actively flowing

Water expansion

They can be observed directly

They can be inferred from production history

Aquifer

Several methods of aquifer representation

Numerical

Analytical

Combination of the above

Any description should be geologically reasonable and justifiable

Reservoirs are not usually completely sealed boxes

Aquifers allow something other than a no-flow boundary condition

Aquifer

Numerical - explicitly modelled aquifer

Aquifer modelled directly as part of the simulation grid

Aquifer

Using gridblocks

Can result in a significant increase in number of gridblocks

Single phase flow allows large gridblock sizes

Complexity may reduce block size in order to model barriers or heterogeneity

Must balance accuracy with model practicality

Gridblock Modification

Individual blocks or series of cells can be isolated from the grid

e.g. chain of cells can model transient pressure effects

Special connections can be made to attach them to the relevant parts of the reservoir

Aquifer

Type of aquifer best suited to this method

Confined aquifers of limited extent

Directly observed complex aquifers

Complexity resulting in variable inflow into reservoir

Very large aquifers providing strong water support can be represented by constant pressure wells

More usually apply volume multipliers to water cells

Aquifers attached to more than one reservoir

Aquifer

Analytical models

No extra gridblocks required

Simple mathematical model quickly solved

Supplies pressure boundary condition and water influx to the reservoir model

No other reservoirs can be attached

Types of analytical aquifer models

Carter and Tracy Model

Fetkovich Model

Aquifer

Fetkovitch

Models finite aquifer without use of influence function tables

Carter-Tracey

Pressure at external boundary of aquifer does not change

Infinite acting aquifer by default

Influence function tables control the water influx for finite representations

Gravity effects not accounted for in analytical models

H ence bottom aquifer may overpredict water influx

Aquifer

Combination of gridblocks and analytical models

extend grid so that reservoir boundary blocks lie in aquifer

calculate water influx into these blocks using analytical aquifer model

Restricts the number of gridblocks required

Allows pressure variation across the aquifer contact area

Aquifer

Analytical aquifer connections

Bottom

Boundary

Region

Direction of flow *AQLEAK

Default is only inflow allowed

*ON allows flow back into the aquifer

Summary

Aquifers are usually present

Aquifers are rarely delineated well

Properties, size, and influence poorly known

Often altered during history matching

Chose an aquifer type:

To best represent observed behaviour

To minimise runtime

Exercise

Adding an aquifer to your model

Analytical

Numerical

Wells Computer Modelling Group Ltd. David Hicks, European Area Manager

Well and Recurrent Data

Section starts with the *RUN keyword followed by the start *DATE

Simulation advances using *DATE or *TIME or both

Static data can be altered with time

I/O keywords can be inserted to produce outputs at specific dates

Can also add LGRs, but not remove them (yet)

TRANS

RTYPE

Relative permeability endpoints

Well Keyword Description

The following is a complete description of a well

*WELL

*PRODUCER or *INJECTOR or *CYCLPROD(GEM)

*PWELLBORE or *IWELLBORE (only if WHP calculation needed)

*INCOMP (only for injectors)

*OPERATE

*MONITOR (optional)

*GEOMETRY (optional)

*PERF

Modifiers

*ALTER or *TARGET

Wells in Simulation

Wells are points of material transfer either into , injector , or out of , producer , the simulation grid

The rate of material transfer is governed by the well PI:

PI = Q / ( P o - P wf )

Q - Well flow rate

P o - Pressure at outer boundary of well drainage area

P wf - Well flowing BHP

Wells in Simulation

PI can also be determined from Darcy’s Law as

PI = 0.00708kh /  (ln(ro/re)+S+c) (field units)

c = 0 for steady state flow; -p/2 if PI is based on gridblock pressure and re is set to block size  x; S = skin factor; ro = external drainage radius; re = effective well radius

Single well radial models can allow the wellbore to be modelled explicitly as the inner radius of the model

Grid type *RADIAL

Most models are non-radial

Wells in Simulation

Gridblock dimension is >> wellbore radius

The well block pressure is not normally the well drainage boundary pressure

i.e. the r o term required in the Darcy eqn

How do we relate the flowing BHP of the well and the well block pressure?

Wells in Simulation

Most simulators use Peaceman’s approach

Based on an “equivalent well block radius” re

Block pressure is interpreted as a flowing pressure at this radius

Well models assume radial flow such that

P o - P wf = Q   2  kh ln (r w / r e )

(P o - Well block pressure ; r w - well radius)

Peaceman showed r e = 0.21  x for square blocks in an isotropic, uniform grid

Wells in Simulation

More general expression for anisotropic systems

PI determined from field pressure tests should be adjusted for application in the simulator model

Wells in Simulation

Wells should ideally be placed in individual gridblocks. LGR can help with this

A single pseudo well can be used if they maintain the same relative production ratio otherwise interference effects need modelled

Horizontal wells use the same Peaceman formulation

 z and k z substitutes for x or y variables depending on direction in r e calculation

Kh also represents the I or J penetration

Wells in Simulation

The well model is controlled by 2 dependent variables

Well pressure and well flow rate

The basic equation relating well inflow to pressure is:

Q j = WI j  P (j = g, o, w)

Where,

 P = P Block - P BHP - H

Injection Wells

Injector

Mobility weighted - Q j =  WI(  T )  P/ B j

Unweighted - Q j =  WI  P - WI must be entered directly

Wells in Simulation

The Well Index is a function of:

Mobility - as determined by the simulator at each time

Transmissibility - fixed values of kh, r w , r e and skin

Transmissibility

based on full 360 o influence of well on surrounding grid blocks

well centred on gridcell

user defined factors can account for deviation from this ideal

Wells in Simulation

Total inflow for a production well is :

Q j =  All Connections ( WI j  P )

SETPI allows fixed WI value from a welltest or a multiplier to be applied for whole well at a given time

IMEX has several ways to allocate this production to multiple intervals

Defining individual layer WI directly (PERF WI)

Using grid block parameters (PERF GEO or KH)

Wells in Simulation

The GEOMETRY keyword

Grid penetration direction

Gives the Kh term and cell dimension in r e calculation

Well radius

Geofac

Used in r e calculation

Wfrac

Multiplier on the 2  term to account for non-360 o flow

Skin

Wells in Simulation

IMEX uses a modification of Peaceman’s 0.21  x term

r e = geofac (  x  y /  *wfrac) 1/2

This results in a geofac value of ~0.37 to achieve Peaceman’s original result

If PERF GEO used then WI is allocated on the kh value given by the cell values

GEOA uses Peaceman’s anisotropic calculation

Any entered geofrac is ignored as it is constant value of 0.28

If PERF KH used then production will be allocated on the defined kh values given

Also have KHA

Well Inflow

Geometry Keyword

Since:

r e = geofac (  x  y/ wfrac  ) -½

= geofac  x/( wfrac  ) -½

Substituting

r e = 0.21  x gives geofac = 0.37

Wells in Simulation

*PERF *WI allows exact specification of well index

Q j =  All Connections ( WI j  P )

The  P term is provided by the individual layer drawdowns i.e. P block - P BHP

Wells in Simulation

Well flowing pressure - Bottom Hole Pressure (BHP) is calculated at a defined depth

Default datum depth is centre of first defined perf

Otherwise by adding *REFLAYER to completion in *PERF

Can also specify a BHP datum depth *BHPDEPTH

*BHPGRAD allows user defined gradient between reference layer and *BHPDEPTH otherwise the default mobility weighted fluid density is used

Wells in Simulation

Default density gradient is used to allocate individual layer P otherwise:

*LAYERGRAD allows user defined gradients between perfs

*FLOW-TO allows gravity head to be incorporated along whole well trajectory in the correct ordering (multilaterals) and through *CLOSED perfs.

Wells in Simulation

LAYERIJK

Well can be completed in several directions

*GEOMETRY only allows one direction to be specified

LAYERXYZ

GridBuilder calculation of well index from entry/exit points and measured length. Geofac will be ignored.

Multi-lateral Well

Go to IMEX template project in Launcher

WWM directory – mxwwm016.dat

Look at text file and drag and drop into Builder to view. Note how the diagram and keywords are related

Well Control

Wells can be controlled by either rate or pressure limits

One must be fixed, by the user, so the other can be derived

Many limits can be set but only one can be in control at any given time

Multiple *OPERATE can be defined with the most limiting one actually controlling the well

*ALTER - changes first defined *OPERATE constraint

*TARGET - changes any defined constraint

Re-defining *OPERATE resets all earlier settings

Well Control

Well initially produces at a specified constant oil rate, q osp

Secondary constraint is minimum bottom hole pressure

If P BHP  P min , then switch to P BHP = P min constraint

Well constraint violation occurs at t = t cv

The accuracy of t cv depends on the simulator timestep length

Well Control

When dealing with wells we introduce time into the simulation

Simulator timestep length depends on:

Numerical solution method used

Cell saturation and pressure change

User defined maximum

Time truncation error = dS - (dS p /dt p ) dt

dS, dt - change in current timestep

dS p , dt p - change in previous timestep

Modelling Well Treatments

Acidising

Usually modelled as a simplistic skin alteration

Could extend into reservoir via *TRANSMULT

STARS allows reaction effects on surrounding formation to be modelled

Water shutoff gels

Model as a Sw crit variation in selected completions

Could extend this into the reservoir via endpoint alteration

GEOMETRY K 0.0762 0.37 1. -2 PERF GEO 'Well 3' 11 24 1 0.9 OPEN FLOW-TO 'SURFACE' 11 24 2 1. OPEN FLOW-TO 1 KRPERF 11 24 1 SETN 1 SWCRIT 0.2 11 24 2 SETN 1 SWCRIT 0.2

Modelling Fractures

Fractures can be modelled either explicitly or implicitly

Explicit modelling

Accurate positioning

Long runtime

High flow rate in to small cells

Many extra cells required

Modelling Fractures

Implicit modelling

Fracture contained within wellblock

Little effect on runtime

Modelled by negative skin or increased apparent well radius

Can extend enhanced transmissibility to surrounding gridblocks containing the fracture

Gas Wells

Peaceman’s correlation is based on liquid flow and is also applicable to high pressure gases

This approach can cause difficulty, in matching THP, for gas reservoirs if the reservoir pressure is low

Default is to use density and viscosity at P block

In low pressure gas reservoirs the inflow equation should be altered. *QUAD modifier to *PERF uses P block for viscosity but for density it uses:

0.5*block density*P block +P well /P block

Leading to Q = C (P o 2 - P wf 2 ) n

Gas Wells

Po - formation pressure

Pwf - well flowing pressure

C is a function of permeability, net pay, gas viscosity, gas deviation factor, drainage and wellbore radius

C, n are usually empirically determined by gas well testing

*PSEUDOP modifier allows viscosity to be included

Gas Wells

High production rate gas wells can also suffer from turbulence effects as flow converges into the well

This can be corrected for by a rate dependent skin known as a D-factor :

S total = S + D|q free gas |

High velocity through a porous medium requires non-darcy effects to be modelled away from the well

Darcy Flow Inaccuracies

In a gas well producing at 20 MMscf/D from a 100 ft frac.

Fracture conductivity was a factor of 10 less than that predicted by simple Darcy flow.

In an oil well producing 4,000 BOPD from a 20 ft frac.

Fracture conductivity was reduced by a factor of 3.

In all fractured wells with high rates of production, non-Darcy effects play a very important role

Crossflow in the Well

Crossflow occurs when the well sees a change of drawdown sign between completions

Communication between layers in plugged wells

Differing pressure regimes

Usually requires a barrier in the reservoir to be crossed

Shale layers, faults etc

Phase change along the completion

e.g. gas present low in the completion cannot enter due to weight of fluid entering above

Simulator default is to model crossflow

Need well to flow for this to be modelled

Well Location Options

GridBuilder will calculate simulator well completion blocks

X,Y location in a layer given in map - use perforation manager (in ModelBuilder) to enter perforation depths

Can also import well deviation data

With associated perforation positions

Or perforate whole path

Deviation data allows more accurate inflow description

LAYERXYZ

Can also specify well blocks using point-and-click

Well Perforation Editing

Add new well perforations by clicking on grid blocks

Click and drag perforations to new blocks

Create multi-lateral wells

Add perforation by clicking on grid block Multi-lateral connections determined by GridBuilder

Perforation Positioning

Can also use PerfManager in well section of Builder

Perforation intervals can be placed using:

Log data

Direct depth input

Allocation to simulation layers

Primarily used for vertical wells or wells with small amount of deviation

Modelling Partial Completions

Wells are perforated at distances along a wellbore which may or may not coincide with the simulation layering

Add more layers/blocks to coincide with perforated interval

Often done with LGRs

Modify the kh term in the inflow equation

Modify ff factor for well

Done automatically

Can be manually entered

Coning effects modelled by rel perm modification

Critical gas / water saturations

Exercise

Adding Wells to the simulation model

Wells Computer Modelling Group Ltd. David Hicks, European Area Manager

Restart Files

What is a restart file and how is it used?

A restart file contains an image of the reservoir at a fixed point in time

It is used to reduce runtime

Commonly used at the end of a historical data period

Multiple prediction scenarios can be run from a common point in time

Don’t have to re-run the historical data period

Can reduce stress when experiencing very long runtimes

Restart Files

WRST initiates output of the restart information

Can be used in I/O section or Wells section

Restart Files

You can also generate a restart file interactively

INTERRUPT INTERACTIVE

<cntrl> C

Restart Files

REWINDable restarts

Usually restart information is held in the .irf/.mrf files

REWIND causes an .rrf file to be output

REWIND 3 – Only the last 3 restarts will be preserved

Using Restart Files

Builder interface makes it easy to chose a restart position by loading in the base run .irf file

Can also see restart list created at end of .out file

Using Restart Files

Keywords required

FILENAME INDEX-IN ‘d: utorial utorial.irf’

(This must be the first simulator keyword read)

RESTART 32

Add these to the original run and save the new file as a different name

No other changes are required

Any data before the restart date will be skipped

Viewing Restart Results

You can chain several restarts together

For viewing in Results

Load the final .irf file in the chain

All the other data will also be loaded

Exercise on using a Restart file

Horizontal Wells

Group Control

The Hierarchical Tree

IMEX/GEM provides Field, platform, group, well, and individual perforation hierarchical control

The Hierarchical Tree

Always one ‘Default-Group’ present

Attached to the ‘Field’ group

Any wells not attached to a group will be allocated to this default group

Groups have separate Injection and Production controls

*GCONP; *GCONI; *GCONM

Cannot have both wells and groups attached to a group

2 levels of groups can exist between the ‘Field’ and the wells

Why Do We Need Group Control?

Necessary for prediction runs

Can be used in history matching if rates measured at gathering stations

Do not know a-priori the following

How many wells are required to meet a certain deliverability target?

What is the deliverability of a certain well in the future?

How does changing deliverability effect well production and target production?

How do we handle increasing GOR’s, watercuts, etc.?

Why Do We Need Group Control?

Which wells need to be re-perforated automatically to reduce GOR, watercut, etc.?

How many new wells are required to come on-stream automatically if existing wells don’t meet production criteria?

Pressure maintenance, using voidage replacement - do not know a-priori how much fluid should be injected to maintain pressure

In the field there may be a limitation on the amount of water or gas that can be handled at any group or platform. Thus it is necessary to be able to specify maximum limits on production of different phases

How Group Control Works

What Group Control Can Do

Rate, GOR and watercut constraints at any level

Force the simulator to honour facilities constraints

Voidage replacement

Reservoir pressure maintenance

Also recycling of produced water or solvent

Note that any recycling needs the producers and injectors attached to the same group

Automatic well recompletion, opening, shutting, or plugging to defined levels

Gas cycling

Re-injection of produced gas

Can control percentage of production recycled

Account for gas used for fuel

Account for sales offtake

Additional gas from sources outside model (makeup gas)

Lift gas allocation to reduce fluid column density

Gas lift optimisation

Most efficient use of lift gas determined

Overall gas mass balance (produced + lift) preserved

VFP curves use GOR

Assign/Re-assign platforms, groups, wells and layers at any time

A well can only be attached to one group at a time

Group targets allocated to wells/groups based on :

The well/group’s instantaneous potential - default

User specified ratio - Guide Rate

How Group Control Works

Drilling queues

*DRILLQ & *GAPPOR

Allow targets to be met without prior knowledge of the number of wells required

Defined well ordering list with wells opening to meet target

Wells opening according to their potential

Constraints may be exceeded for 1 timestep unless it is repeated

*CONT-REPEAT modifier for *OPERATE, or *SHUTIN-REPEAT for *MONITOR & *OPERATE

Exercise

Adding Groups to the simulation model

Take an existing model and walk through adding groups and group controls with the instructor

Surface Pressure Control

Flow to the Wellhead

Simulator solves for BHP and Q

FVF links Q res with Q surf

VFP table links BHP with THP

Single phase flow has a simpler relationship between BHP and THP

Water phase density dominated pressure loss

Gas phase often friction/turbulence dominated

Multiphase flow can have different flowing regimes and variable fluid density

Flow to the Wellhead

Vertical Flow Performance (VFP) tables can be created for multiphase flow

Relate pressure drop to flow rate; WCT; GOR

Use correlations based on pipe topology

Temperature effects can also be incorporated

The pressure drop between bottom and top hole position can also be influenced by :

Obstructions; gaslift; pumps; compressors etc

Modelling Workover Strategies

Changing the production tubing will only influence the simulator BHP/THP relationship

Can model by changing VFP table used

Gaslift changes the BHP/THP relationship by lightening the fluid column

Use a large range of GOR values when creating VFP tables

As GOR increases density decreases and friction increases. Optimal gas injection rate for any well

Surface Pressure Control

Data for Tubing Head Pressure (THP) control

Manually enter pressure drops

Import Eclipse formatted tables

Single phase pressure drop calculation gas comp

Use CMG’s Wellbore Calculator

Can also couple with Forgas

Allows full surface network addition to IMEX for gas and water

Operating Constraints

WHP constraint will not appear until a pressure loss table/correlation is associated with the well.

INITIALIZE

BHP set at first Newton iteration

IMPLICIT

BHP updated each Newton iteration (default)

Manual Entry

Set up a matrix of variables

Manually enter the BHP that results from the applied THP

Pressure Control for Injectors

A single phase wellbore model (Aziz) can be used for injectors.

Can enter gas composition to C6.

Wellbore Calculator

Segmented wellbore profile

Gaslift inclusion points

Thermal effects

Files

.wbi - input interface file

Output files

.wbo - CMG format output

.wr1 - detailed flow description along tubing

.wr2 - brief calculation

Multiphase Table Format

Similar form to other simulators

Doesn’t use ALQ but has a more rigorous gas mass balance

Gas lift forms part of GOR

Preserves gas balance for optimisation

Can use *EXTP in place of a BHP value

Viewing curves

VFP curves can be plotted over IPR curves in the Perforation Manager

Exercise

Using the Wellbore Calculator

Take an existing model and walk through generating VFP curves with the instructor

Thermal Blackoil Additions

History Matching

History matching - Introduction

Purpose: testing validity of reservoir model

Differences exist between the reservoir model parameters and actual reservoir parameters

These differences can lead to errors in simulation

Modelling past performances help identify weaknesses in data

History Matching

Once an acceptable match obtained, model can be used to

predict future performance

simulate different operating strategies

sensitivity studies

modelling secondary and tertiary recovery processes

effect of well locations and infill drilling

modification of pattern to improve production

History matching is the most time consuming task in a reservoir study

History match is not unique

History Matching

Performance data to be matched

Match reservoir pressure, WOR’s, GOR’s, water and gas breakthrough times

Also match BHP’s, fluid rates

Permeability data usually limited in quantity both areally and vertically

Pressure transient analysis (PTA) and core analysis are standard technique for determining permeability in well region

Aquifer data usually quite sparse

History matching is used to define permeability distribution, aquifer porosity, transmissibility and extent

History Matching

Errors in field measurements

Injection and production rates not always reported with regular frequency

Gas production usually not measured accurately especially if gas has been flared

Oil production rates are usually the most accurate data available

Injection data less accurate due to fluid loss to other intervals from piping or casing leaks and other causes

Volumes measured at central sites are difficult to allocate back to wells

Allocation to different zones difficult due to commingled production or injection and inaccuracies in measurement

History Matching

General steps for history matching

Assemble data on performance history

Screen data and evaluate their quality

Define specific objectives of history match

Develop a preliminary model based on best available data

Simulate history with preliminary model and compare simulated performance with actual field history

History Matching

Decide whether model is satisfactory. If not, analyze results with simplified models to identify changes in model properties that will improve agreement between observed and calculated performance

Make adjustments to model consulting with geologic, drilling and production operations personnel

Again simulate part or all of past performance data to improve match

Repeat above steps until a satisfactory match is obtained

History Matching

Strategy for history matching

Match volumetric - average pressure levels. This is first step toward confirming overall compressibility of reservoir system

Complete gross match of pressure gradients to establish flow patterns

Match pressure more precisely, making changes in small groups of blocks. Reservoir description can significantly change at this time

Match contact movements, saturations, WOR, GOR on an area basis

Match individual well behavior

History Matching

Judging model acceptability

Acceptable pressure matches for individual wells in typical sandstone reservoirs are  50 psi

For high permeability reservoirs a better match must be obtained

Accuracy of a match must be comparable or better than the accuracy required in predictions

For reservoirs where field history is in early stages of development, more than one reservoir description will provide an acceptable match

For these cases, sensitivity studies with two probable reservoir descriptions should be made

History Matching

Parameters to change in a history match

Parameters to change in order of decreasing uncertainty are

aquifer transmissibility (kh)

aquifer storage (  h C t )

well indices

reservoir transmissibility

relative permeability and capillary pressures

reservoir porosity and thickness

structural definition

rock compressibility

reservoir oil and gas properties and their distribution within reservoir

WOC’s and GOC’s

water properties

History Matching

Matching pressure history

List properties of reservoir and aquifer that are most likely to affect pressure history match

Estimate bounds of uncertainty for these properties

Develop criteria to judge acceptability of pressure history match

Complete trial simulation and decide if volumetric average pressure is satisfactorily matched

If not, use material balance, PTA and geologic information to estimate changes to the model values of fluids in place, oil zone and gas cap, average aquifer properties and size

History Matching

Decide if more detailed matching is needed based on criteria previously established. Evaluate quality of match in major segments of reservoir

Evaluate match of pressure distribution in reservoir at selected times

If match is not satisfactory, analyze pressure distribution to find evidence for heterogeneous aquifer properties. Determine if differences are due to presence of sealing faults, pinchouts and poor communication between zones

Change reservoir and aquifer properties areally with use of PTA theory

History Matching

Locate regions of pseudo-steady state condition and correct pressure gradient errors by adjusting transmissibility

In general, permeability is the principal reservoir variable used to obtain a pressure behavior match

Porosity data derived from log and core data should not be changed unless the data are sparse or of poor quality

Fluid contacts and properties are better defined than porosity and usually should not be altered

Some permeability values such as those measured by a well test can be considered relatively certain

History Matching

Matching gas and water movement

Seldom possible to match water and gas movement adequately without having reasonably complete reservoir model

List properties of reservoir and aquifer most likely to influence fluid movement

Develop criteria to judge quality of match and decide whether matching behavior of group of wells rather than individual wells is sufficient

Analyze reservoir depletion process to determine if coning or cusping has influenced water and gas arrival time

History Matching

Analyze simulations conducted when matching pressure history to determine if GOR and WOR were matched satisfactorily

If not, determine if permeability stratification is more or less severe than previously indicated. It may be necessary to adjust permeability distribution in producing section to match water and gas arrival time

Ascertain whether areal permeability distribution and/or continuity of selected zones in reservoir and aquifer should be adjusted

History Matching

Decide whether relative permeability data should be modified. Avoid changing these data if obtained from measurements at reservoir conditions. If data is not considered reliable, both shapes of curves and endpoints can be changed

Evaluate sensitivity of match to errors in vertical permeability since it can be important in calculating displacement efficiency and vertical sweep efficiency, especially in the presence of shale lenses or barriers vertical communication can vary from none to quite high. Run a few cases to determine sensitivity to errors due to this

History Matching

Analyze performances of selected wells to see if gridblock definition is a major problem

Recognize that incorrect allocation of injected and produced fluids make precise agreement between field and model well behavior difficult to achieve

As change in model properties are made to match gas and water movement, continue to compare calculated and actual pressure behavior

General Progression of a Black Oil Field Study

1. Define objectives e.g. predict field behavior under waterflood

2. Set up very detailed single well model(s)

3. History match single well model and run predictions if relative permeability curves were measured properly in the lab (i.e. restored wettability, low injection rates, etc.), then try not to change them in the history match

4. Repeat steps 2 and 3 using data from a different well

5. During the history match of the 2 nd well, try to obtain a match using the same method, or engineering principles, as were used for the match of the first well, e.g. if the relative permeability curves were changed during the 1 st match, then the 2 nd match should be obtained using identical curves (possibly different S wc ’s)

General Progression of a Black Oil Field Study

6. If the wells studied (may be more than 2) could all be matched using the same method, then a field wide study should be started

7. Create pseudo-relative permeability curves for the full field model by matching the single well model studies with the full field model. A subset (sectional model) of the full field model may be used to save CPU time

8. History match the field model using the same methods that were used in the single well model studies. Try to only have one set of relative permeability curves for the entire field (or facies type). If too many different relative permeability curves were used to obtain a history match, then it becomes questionable which curves to assign to new infill wells in prediction runs

9. During the history match, occasionally run prediction cases to ensure the predictions are reasonable and they follow existing trends in the field

Modification of Rock Data

Modify an area instead of on a single-cell basis

This ensures continuity and smoothness of rock properties

Modify within reasonable range of property values

Correcting Computed Pressure Distribution

Possible Remedial Action

1. Move fluid from the high-pressure to the low-pressure zone by a change in rock permeability

2. Decrease the oil in place in the high-pressure area by either

a) decreasing porosity

b) decreasing thicknesses

c) decreasing oil saturation, or

d) all of the above

Possible Remedial Action

3. Increase the oil in place in the low-pressure area by either

a) increasing porosity

b) increasing thicknesses

c) increasing oil saturation, or

d) all of the above

Pressure Level Too High in the Reservoir

Pressure Distribution Too Discontinuous

Modifications Using Fluid Saturations

The initial fluid saturations within the reservoir are usually reasonably well known from various sources

1. Core samples

2. Electric logs

3. Laboratory experiments

Final fluid saturations (S or , etc.) are usually not well known and can be used as a matching parameter

Predictions of Oil Production Unrealistically High

Modifications of Fluid Data

Fluid data are generally well known in a simulation study. These data include the following

1. Formation volume factors

2. Viscosity

3. Compressibility

4. Solution gas data

Watch for faulty input, misplaced decimal point, incorrect exponent, units

Modification of Relative Permeability Data

Imbibition / drainage curve

Relative mobilities of individual phases affect production streams

Producing GOR is Too High

Producing GOR is Good - Oil Mobility is Too Low Low well pressure or low oil rates

Model Producing Gas Too Early or Too Late

Original Estimate of Reservoir Extent Incorrect

Case Study of a History Match

Example

Initial history match runs indicated that the pressure decline was too rapid

Since the areal extent of the oil-in-place was defined by the dry holes, and the production history showed no water production, the engineer reasoned that the rock compressibility was too low (friable sandstone rock), thereby causing the low pressure in the model

The rock compressibility was increased by an order of magnitude in order to match the pressure decline to the point where water injection started

Example

With this system, it was impossible to predict the increase of pressure after water injection started. Therefore, the engineer reasoned that the field reported rates of water injection were too low, and he increased them by 40% to get a match

All field measured parameters were matched successfully with this method, and a large amount of time was spent on this project

A better way to history match this pool would be to increase the aquifer volume at the outer edge of the grid system, and then restrict the water influx to the wells to match the zero water production rates

Example

Restricted water influx could be modelled either by low vertical permeability, or low horizontal permeability in the aquifer

This method does not require a high rock compressibility to match the initial pressure decline

Since the compressibility of the system is lower, then water injection raises the reservoir pressure higher, and the modification to the injection rates is not necessary

Sequential Approach Versus Parallel Planning

Sequential approach

Linear logic assumes a steady progression from raw data to final forecasts

data analysis

model construction

history matching

calibration

predictions

Each step can be started only after the preceding steps have been completed

This method assumes that conditions and information available at the beginning of the study will remain unchanged

Problems manifest themselves in several ways

Delays accumulate. Each problem delays progress on all subsequent steps

Critical flaws may be identified at a late stage. All work up to that stage must be repeated

When new raw data becomes available, the entire process must be repeated to produce predictions

Parallel planning

Origins are based on “rush” modelling studies

Has five assumptions

preliminary forecasts are a business necessity

problems will arise in all phases of modelling

it is not possible to predict where and when these problems will occur

new information will probably become available while the study is in progress

identification of critical flaws is of paramount importance

Uses multiple numerical models, beginning with simple “pseudo models”, and finishing with a rigorous model

Pseudo models must be the same size and configuration as the rigorous models, except they contain synthetic data segments which are convenient substitutes for actual data