You are on page 1of 8

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/327239009

New Approach to Reservoir Characterization by Integrating Seismic Stochastic


Inversion and Rock Physics Model: Case Study in Malay Basin

Conference Paper · August 2018

CITATIONS READS

0 393

6 authors, including:

Amir Babasafari Deva Ghosh


Universiti Teknologi PETRONAS Universiti Teknologi PETRONAS
15 PUBLICATIONS 3 CITATIONS 201 PUBLICATIONS 659 CITATIONS

SEE PROFILE SEE PROFILE

Ahmed Mohamed Ahmed Salim Teresa Ratnam


Universiti Teknologi PETRONAS Universiti Teknologi PETRONAS
68 PUBLICATIONS 58 CITATIONS 6 PUBLICATIONS 1 CITATION

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Reservoir Modeling View project

Study on elastic wave excitation based EOR method View project

All content following this page was uploaded by Amir Babasafari on 27 August 2018.

The user has requested enhancement of the downloaded file.


New Approach to Reservoir Characterization by
Integrating Seismic Stochastic Inversion and Rock Physics
Model: Case Study in Malay Basin

Amir Abbas Babasafari 1*, Deva Prasad Ghosh 2, Ahmed Mohamed Ahmed
Salim3, Teresa Ratnam4, Chico Sambo5, Shiba Rezaei6
1, 2,3,4,5
Center of Seismic Imaging, Department of Petroleum Geoscience, Universiti Teknologi PETRONAS
(E-mail: amir_16004381@utp.edu.my, a.babasafari@yahoo.com )

ABSTRACT – Reservoir modelling is conducted through petrophysical properties distribution between the
wells utilizing mainly geostatistics. Using only well data (Geology and Petrophysics) leads to any anomalous
zones and heterogeneity being ignored. This type of interpolation is not geologically matched. To overcome this
problem, seismic data as the only laterally extended data is extensively used. On the other hand, seismic data
possesses a scaling issue. In order to honour seismic data, well log data are upscaled to cover the limitation of
seismic vertical resolution. Recent advances in broadband frequency data leads to high resolution seismic
bandwidth. Seismic stochastic inversion by far is one of the most predominant methods. After seismic data
inversion process and prior to elastic properties conversion (derived from seismic stochastic inversion) to the
reservoir petrophysical properties, an accurate classification of litho-facies distribution need to be performed.
The main aspect of the new approach is to integrate seismic stochastic pre-stack inversion together with a valid
rock physics model leads to a more obvious separation of litho-facies within reservoir scale. In this study a valid
rock physics model is intended to be built at well locations. In the next stage, stochastic inversion is run to populate
different litho-facies classes away from the wells. The results reveals how successfully pay zone (Gas Sand) is
separated from non-pay zone at reservoir interval incorporating mentioned approach.

Keywords: Seismic Stochastic Inversion, Rock Physics Modeling, Litho-Facies Classification

1. INTRODUCTION

Rock physics is a bridge between elastic and petrophysic properties of rock. Rock physics modeling is
extensively used for many purposes such as elastic log prediction, calibrating well to seismic tie, litho-
facies differentiation, fluid replacement in AVO analysis and 4D seismic data. [1]
In this study rock physics is implemented in terms of shear wave log estimation and also more
remarkable litho-facies discrimination. Afterwards, seismic pre-stack stochastic inversion is performed
on the basis of Bayesian stochastic approach using Sequential Gaussian Simulation (SGS) for sampling
step.
The studied area is located in the Malay basin (Figure 1) which includes number of discovered Gas
fields from the Group B, D and E. The field of study is intersected by several north-south normal faults
and a major east-west fault.
Thickness of reservoir layers are less than seismic vertical resolution but results of this research disclose
nevertheless significant discrimination of Gas sand interval cross validated with blind test.
Figure 1. Field Location Map

2. METHODOLOGY

First, litho-facies classification is carried out at well locations using petrophysics logs (e.g. mineral
volumes, porosity and water saturation) and 5 different litho-facies classes are defined. (Shale, Wet
sand, Oil Sand, Gas Sand and Coal).
Fluid distribution at reservoir layer E9 middle and lower intervals is much complicated due to high
lateral litho-facies changes as a result of stratigraphic hydrocarbon traps in this field. Well A
demonstrates less than 5 meters oil sand in shale background layers, well B represents shale and wet
sand sequences, well C displays two thin beds of gas sand and well D discovers wet bearing zone.
Figure 2 illustrates elastic, petrophysics, litho-facies log and seismic section passing through well C
that encompasses two thin beds of gas sand. Each gas sand layer is less than seismic vertical resolution.
(Around 7 and 11 meters).

Figure 2. Elastic, Petrophysics, Litho-facies log and seismic section passing through well C that
encompasses two thin beds of gas sand at reservoir interval

To overcome issues mentioned above first a consistent rock physics is modeled to predict elastic
attributes (P_Impedance and Vp/Vs ratio) at well locations which results in far more considerable
differentiation of litho-facies classes in P_Impedance vs. Vp/Vs cross plot. Afterwards, given
relationship is generalized through the seismic pre-stack stochastic inversion to populate elastic
properties per defined litho-facies classes. Finally, the distribution map of pay zone interval is generated
and analyzed.
2.1 Rock Physics Modeling

There is a relationship between seismic velocity (compressional and shear velocities) data and reservoir
parameters such as pore shape, pore size, pore fluids, mineral volume and … [1]
In this study we use a technique which is called Rock Physics Template (RPT), based on a theory by
Dvorkin and Nur in which fluid and mineralogical content of a reservoir is predicted on a cross plot of
P_Impedance vs. Vp/Vs ratio. [2]
As can be considered in Figure 3-left the separation confidence of 5 defined litho-facies classes are not
obvious enough using measured elastic logs (X axis: P_Impedance vs. Y axis: Vp/Vs ratio), thus
improvement of litho-facies discrimination is needed to be done. Moreover, only 2 wells possess shear
log data. To tackle mentioned matters, a valid rock physics model is addressed to be built that not only
keep the main trend of measured logs, but also make the litho-facies classification much easier on a
cross plot of seismic inversion results, P_Impedance vs. Vp/Vs ratio (Figure 3-right).

Figure 3. Cross Plot of P_Impedance vs. Vp/Vs ratio colour coded by litho-facies log for
measured elastic logs (left) and predicted elastic logs through rock physics modeling (right)

The elastic properties of litho-facies classes (Shale, Wet sand, Oil sand and Gas sand) are predicted
regularly other than Coal which is estimated approximately because of input data (discrete value of coal
mineral volume which is 0 or 1). Elastic log estimation of Coal class is modified through using Monte
Carlo Simulation.
The generated Rock Physics Model (RPM) is calibrated through adjustment of predicted elastic log data
with measured one. A quality control is carried out by overlay comparison of measured and predicted
elastic logs in vertical section.

2.2 Deterministic vs. Stochastic Inversion

To have a clear understanding of litho-facies distribution away from the wells, seismic pre-stack
inversion is a must. Considering reservoir layers which is below seismic vertical resolution, the
stochastic pre-stack inversion is selected to be performed. [3] Figure 4 shows the result of stochastic
inversion at well C.
Figure 4. Left: Overlay Comparison of the Predicted Elastic Logs (P-Impedance, S-Impedance and
Vp/Vs ratio) using stochastic inversion (red) versus elastic logs derived from rock physics modeling
(black). Right: Near, middle and far Synthetic Angle Stacks (red) versus near, middle and far Seismic
Angle Stacks (black) at well C

Since, seismic deterministic inversion produces results based on a limited frequency bandwidth (10 to
80 HZ), is not nominated for a detailed reservoir characterization. Seismic vertical resolution in
reservoir interval of this field is between 20 to 24 meters, whereas in contrast the pay zone thickness is
less than 15 meters. To generates elastic properties (P_Impedance and Vp to Vs ratio) that is compatible
with subtle geological features and thin bed resolution, stochastic inversion is utilized. In this study
totally 40 realizations of Acoustic and Shear Impedance (PI and SI) are created based on Bayesian
inference [4] and Sequential Gaussian Simulation (SGS) algorithm [5]. The mean and uncertainty
volumes of all realizations are computed.

3. RESULTS AND DISCUSSION

Stochastic inversion results, including one of the realizations and mean of P-Impedance and S-
Impedance passing through well C are shown in Figure 5. Using P-Impedance and Vp/Vs ratio of
Deterministic and Stochastic inversion, result in discrimination of litho-facies classes demonstrated in
Figure 6. The litho-facies classification originated from stochastic outcomes represent much more
suitable differentiation especially at thin layers (both pay zone and non-pay zone) from which an
acceptable match at well locations is concluded.
Figure 5. a) A Stochastic Realization of inverted P-Impedance, b) mean of all stochastic realizations
of P-Impedance, c) A stochastic realization of inverted S-Impedance, d) mean of all stochastic
realizations of S-Impedance

Figure 6. Litho-Facies Classification derived from Deterministic Inversion (left) and one of
Stochastic Inversion Realization (right) overlaid by litho-facies log data at well locations.
Litho-Facies Classification at right Figure illustrates suitable differentiation especially at thin layers
Low P-Impedance value (consisting of Shale, Wet Sand and Hydrocarbon sand) in studied area within
reservoir interval is separated via a boundary which cover all wells. (Cross plot and map in Figure 7-
top) Once Vp/Vs ratio incorporates, the low P-Impedance and low Vp/Vs values (comprising
Hydrocarbon sand only) is discriminated through two new boundaries inside the 1st one that only
includes well C. This differentiation successfully confirms litho-facies classes at well locations
distinctly at blind well D. (cross plot and map in Figure 7-bottom)
The map (Figure 7-bottom) reveals well B and blind well D are out of pay-zone boundary (cross
validated by litho-facies log at well locations which is Shale and Wet sand), well C is located inside
probable stratigraphic reservoir boundary verified by litho-facie log (two thin Gas sand layers of 7 and
11 meters). Well A is close to the mentioned boundary but not located inside despite oil sand presence.
It might be related to just less than 5 meters pay zone in this interval encompassed by Shale background
that make it difficult to be distinguished.

Figure 7. Average P-Impedance Map (top) and average Vp/Vs map (bottom) at reservoir interval
together with P-Impedance (X) vs. Vp/Vs (Y) cross-plot colour coded by litho-facies classes
Note: Colour legend for the map is continuous (PI & Vp/Vs) and for the cross-plot and litho-facies log
is discrete (5 classes)

4. CONCLUSION

The results disclose integrating rock physics modeling and seismic stochastic inversion aid to a proper
discrimination of probable stratigraphic trap boundary. Apart from well A with less than 5 meters oil
sand, litho-facies log information represent a reliable match between wells and predicted hydrocarbon
distribution map. This study investigates lateral litho-facies variation at thin reservoir layers that is
beneath seismic vertical resolution. More precise reserve volumetric calculation, increasing production
rate and field development plan optimization can be concluded. The procedure will be developed
through assessment of hydrocarbon distribution probability map.
View publication stats

ACKNOWLEDGMENT

We are thankful to Centre of Seismic Imaging (CSI) and Universiti Teknologi PETRONAS (UTP) for
valuable resources and also Petroliam Nasional Berhad (PETRONAS) for providing proprietary data
for this research.

REFERENCES

[1] P. Avseth, T. Mukerji, and G. Mavko, (2010) Quantitative Seismic Interpretation: Applying Rock Physics
Tools to Reduce Interpretation Risk. Cambridge University Press.
[2] E. Ødegaard, P. Avseth, (2003) Interpretation of elastic inversion results using rock physics templates,
EAGE.
[3] D. P. Ghosh, N. A. Ibrahim, B. Viratno, and H. Mohamad, (2010) Seismic Attributes adding a new
Dimension to Prospect Evaluation & Geomorphology Identification in the Malay and adjacent basins,
SEG Denver Annu. Meet., pp. 1307–1311.
[4] H. Buland, A., Omre, (2003) Bayesian linearized AVO inversion, Geophysics, vol. 68, no. 1, pp. 185–
198.
[5] I. Escobar, P. Williamson, A. Cherrett, P.M. Doyen, R. Bornard, R. Moyen, T. Crozat, (2006) Fast
geostatistical stochastic inversion in a stratigraphic grid, Soc. Expl. Geophys., no. 76th Ann. Mtg., pp.
2067–2071.

You might also like