Water Encroachment from 4D Seismic

Michel KemperHenry Morris

Ikon Science Ltd

• Introduction

• Seismic Data Conditioning (SDC)

• 4D Forward Modelling• 1D / 2D / 3D

• 4D Inversion

• Conclusion

Outline

Introduction

• 4D Forward modelling workflow– You have a model of saturations over production time, and synthesize

the 4D seismic response for ...• Comparison with acquired 4D seismic (validation)• Feasibility studies

• 4D Inversion workflow– You have 4D seismic and want to know fluid movement for ...

• Reservoir management / infill drilling• Reservoir Simulator updating (“Seismic History Matching”)

• Seismic Data Conditioning– However, before we use Seismic we need to QC it, and where

necessary „repair‟ or condition it.

Partial Stacks

Seismic Data Conditioning (SDC) for AVO and 4D studies

Time/Phase Alignment

Spectral balancing(Phase & Freq)

Offset Amplitude Balancing

(AVO Decay)

3D

Quality Assurance

2D1D

RMS & AGCReview

Global /single Operator Trace by traceSpatial RMS

Trace by trace, partial stack by partial stack

Global / single Operator

Reasoning Zero Phase and balance frequency

Correction for any over or under NMO

correction. (High fidelity need for

AVO)

Correct for overburden lithologiesand fluid effects which might

dampen the seismic signature below (i.e. at reservoir level

masking the true AVO).

Balance the RMS Ratios (near/mids/fars) of the seismic to

the RMS of the wells

Method used

4D

4D Forward Modelling workflow

Geological Model

Date-dependent Reservoir Fluids

4D Synthetics from SimulatorSaturations & Pressures

PETREL ModelECLIPSE

Saturations /Conceptual

model

External Data

Synthetic 4D seismic for ...1. Feasibility Study2. Compare with real seismic

Rock Model components:•Fluid substitution•Pressure changes•Seismic response formulae

LithPhie /Sat RhoVp

Oil AVO Synthetic

Brine AVO SyntheticVs

Subtle Changes

@ 55% Increase in Amplitude

AVO Difference Synthetic

4D Forward Modelling in 1D

Porosity Model Saturation

Full Stack Amplitude Difference

Sw=1

00%

Sw=2

0%AVO Map Variable Saturation

Angl

e

Trace = Shc

Sw=1

00%

Sw=2

0%

4D Forward Modelling in 2D

Porosity Model

4D Fwrd Modelling in 3D – Geo-Model

Variable Porosity throughout the reservoir

Building an accurate 3D geological model is critical

Modelled Saturation 1997

Initial Oil Saturation 1990Make a date-dependent Saturation Model

The initial Saturation Model is usually a Saturation-Height function based on core data and pre-production log data.

The later-date(s) Saturation Model shown here is a conceptual model: it is the initial Saturation Model shifted upwards.An often-used alternative is to use the Saturation property from the flow simulator, imported into the Geological Model (see next slide for an example)

4D Forward Modelling in 3D - Satn

4D Forward Modelling in 3D –Saturation from a Flow Simulator

Far Stack Synthetic 1990

Far Stack Synthetic 1997

Far Stack Synthetic Difference due to a conceptual change in Saturation

Far Stack Seismic Difference

4D Forward Modelling in 3D - Satn

Acceptable match?

Effective Pressure Difference(Conceptual Modelling)

Along the same lines that we made a date dependent Saturation model, we can make a date-dependent (effective) Pressure Model

The initial Pressure Model is usually based on pre-production pressure interpretation

In this example the later-date(s) Pressure Model is a conceptual modelAn often-used alternative is to use the Pressure property from the flow simulator, imported into the Geological Model

4D Forward Modelling in 3D - P

4D Forward Modelling in 3D - PFar Stack Synthetic Difference due to a conceptual change in Pressure

And of course we can combine the effects of changes in Saturation and

changes in PressureFar Stack Synthetic Difference due to a conceptual change in Saturation and Pressure

Injection ProductionInitial

Logs in time have different positions due to pull up or push down depending on the scenario.

4D Forward Modelling in 3D – Pull-up

Geological Model

Date-dependent Reservoir Fluids

Neural NetInference

PETREL Model4D Seismic

External Data

4D derived dated Sw property for...1. Resv Mmt / Infill drilling2. Resv Simulator calibration

Rock Model components:•Neural Net

4D Inversion workflow

4D Inversion - Geo-Model: Facies

Like in the 4D Forward Modelling workflow, an accurate geological model is critical. For Facies modelling we often use Bayesian Classification

Shale = Brown; Gas sand = Red; Coal = Yellow; Brine Sand = Blue

Most likely Facies(based on Lambda-Mu-Rho Inversion)

MR

Freq

uenc

y

LambdaRho/MuRho2D PDF’s per identified facies

Near Stack 1990

Near Stack 1997

Far Stack 1990

Far Stack 1997

4D Inversion

Near Stack Difference Far Stack Difference

Noise Anomaly

4D Inversion

FIRST: Neural Net training

At control points where you know the Sw at the dates of 4D seismic acquisition (typically Sw at wells, using e.g. CO logging) you specify to the Neural Net:

Pre-production Date of Repeat SourceDate 4D Seismic

Sw ΔSw 1st from e.g. the Expl. Well; 2nd from e.g. CO logging

Por/Perm Por/Perm Usually invariant, but need not be!

Traces @ wells Traces @ Wells 1st from early seismic dataset; 2nd from repeat seismic dataset

NN inputs: the Black valuesNN outputs: the Red value (could have a second output: ΔP)

THEN: Neural Net application

Now we are talking 3D cubes of data: Pass in the Black 3D datasets in the table above, and a 3D dataset of ΔSw is created, consistent with the geological model, the wells and the 4D seismic

4D Inversion - Methodology

See Next Slide

Seismic Difference

Impedance Difference

Saturation Difference

4D Inversion – Seismic Dataset

?

We have found that using impedance cubes give better results. This is no big surprise, as impedances represent layer properties, whereas reflectivities represent interface properties

Absolute or Relative Inversion

to Optimum Impedance Properties

Coloured Inversion Joint Stochastic

InversionModel-based Inversion Spectral Inversion

Near stack

Far stack

Impedance

4D Inversion – Inversion Methods

Original Saturation

Saturation at later date

Saturation Difference from NN

4D Inversion - Example

-

=

Sw at Initial Date

Sw at later date

(= initial Sw + ΔSw inferred using a NN)

Note the (blue) water encroachment, and especially on the left the clear targets for infill drilling!

4D Inversion - Example

Saturation Difference Saturation at later date

4D Inversion - Example

4D Forward Modelling and 4D Inversion are two sides of the same coin and should both be used in 4D studies

It is recommended to study the 4D effects in 1D first, then 2D, then 3D.

For 4D studies in 3D it is crucial to make an accurate geological model: it is the geology that drives the 4D response!

Using impedances (representing layer properties) in 4D studies is preferable over using reflectivities (representing interface properties)

But first, perform accurate QC and Seismic Data Conditioning, as ... Rubbish in = rubbish out!

Thank you!

Conclusion