Accurate reservoir modelling through optimized
integration of geostatistical inversion and flow
simulation. A North Sea case study.
A. Castoro1, L de Groot2, D. Forsyth3 & R. Maguire1
1Fugro-Jason UK, 2GDF SUEZ E&P Nederland B.V., 3Currently Saudi Aramco, Previously GDF SUEZ E&P Nederland B.V.
DEVEX 2009, 12 - 13 May, Aberdeen Exhibition and Conference Centre
2
Objectives and workflow
Input structural model, seismic data & well data
Geostatistical inversion > 20 reservoir models with range of properties
Prediction runs for well planning and field development
Engineering Data
History match & ranking of all reservoir models
Create robust reservoir model fit for prediction of field production
and development planning
Selection & fine tuning of final reservoir models
Business Decision
3
Outline
• Models from geostatistical inversion
• Initial history match and model selection
• Final models history match
• Geostatistical vs deterministic models
• Conclusions
4
Geostatistical reservoir modeling
� Interpolate between the wells
• Plausible details ����
• Accurate near wells ����
• Not elsewhere ����
Deterministic seismic inversion
� Optimize P-Impedance to minimize synthetic-to-seismic misfit
• Accurate within seismic bandwidth ����
• Unrealistically smooth ����
• Only one possibility ����
StatMod MC – FJ geostatistical seismic inversion
� Subsumes geostatistical modeling and deterministic inversion
� Does both, simultaneously and in a statistically rigorous way
� Multiple plausible realizations at high detail (e.g. 1ms × 25m)
� Yet also coherent “interpretations” of the seismic up to the km scale
Geostatistical inversion vs other modelling
techniques?
5
Reservoir properties
A1 A2 A3 A4 A5 A6 A7 A8 B3
Lithotype
Best
Good
Low
Non reservoir
LOWER RESERVOIR
UPPER RESERVOIR
6
Seismic data
Near stack seismic
Estimated wavelet
Seis
mic
-well
ties a
nd s
ynth
etics
Amplitude Phase
Seismic Synt Impedance
Seismic Synt Impedance
A2
B3
7
Geostatistical vs deterministic inversion
Impedance sections
Deterministic
(4ms)
Geostatistical
(1ms)
A8 A5 A2 A7A1 A6
8
Geostatistical inversion honours all inputs
Wells are honoured Seismic is honoured
Geostatistics are honoured
Deterministic Geostatistic
Input
Output
9
Blind well validation
• Single realization with wells overlaid
• Wells were not used as a constraint
P-Impedance
A8 A5 A2 A7A1 A6
Lithotype
Best
Good
Low
Non reservoir
1.5e+07
7e+06
1.1e+07
Lithology
Impedance
10
Outline
• Models from geostatistical inversion
• Initial history match and model selection
• Final models history match
• Geostatistical vs deterministic models
• Conclusions
11
Time
SM
3 /
da
y
• Models not matching
historical gas production
discarded
Model screening: historical field gas production
• Some models can fail
to match historical total
field gas production by
up to 2 % Historical
Model 9
Model 1
Model 4
Model 16
Historical
Model 9
Model 1
Model 4
Model 16
Simulated
Historical
Simulated
Simulated
Simulated
Time
SM
3
12
Outline
• Models from geostatistical inversion
• Initial history match and model selection
• Final models history match
• Geostatistical vs deterministic models
• Conclusions
13
Final models history match (bottom hole pressure)
Well A1
Well A3
Well A2
Well A6 Well B3
Well A7
14
Outline
• Models from geostatistical inversion
• Initial history match and model selection
• Final models history match
• Geostatistical vs deterministic models
• Conclusions
15
Geostatistical vs deterministic inversion models
Permeability (mD)
16
Geostatistical vs deterministic inversion: history match
• Historical
• Historical
Time
Pre
ssu
re (
bars
a)
Time
Pre
ssu
re (
bars
a)
Time
Pre
ssu
re (
bars
a)
Time
Pre
ssu
re (
bars
a)
Time
Pre
ssu
re (
bars
a)
• Historical
• Historical • Historical
• Historical
Time
Pre
ss
ure
(b
ars
a)
17
Outline
• Models from geostatistical inversion
• Initial history match and model selection
• Final models history match
• Geostatistical vs deterministic models
• Conclusions
18
Conclusions
• Creation of multiple equally probable reservoir models
given the geophysical and geological data
• Inverted impedance (→ porosity, etc) shown to be reliable by blind well
predictions
• Geostatistical character and correlations maintained
• History match achieved
� without the use of porosity or permeability modifiers
� without the incorporation of faults not substantiated by evidence from seismic
interpretation or well test analysis
• The reservoir model is considered robust and
suitable for use in forward simulation for well
planning and field development
19
Acknowledgements
• Thanks to GDF SUEZ E&P Nederland B.V. for permitting this study to be
shown
Questions?