Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

Journal of Engineering Technology Vol.

2(1): 78-81, 2012


ISSN 2231-8798
© 2012UniKLBMI

Lithology Discrimination in Elastic Impedance


Domain Using Artificial Neural Network
M. Hermana, M.N. Mohd Sani, Z.Z.T. Harith & C.W. Sum
UniversitiTeknologi PETRONAS
Bandar Seri Iskandar, 31750, Tronoh, Perak

Corresponding email: mherjody@yahoo.com

Abstract: Interpretation in elastic impedance domain is a new method in geophysical prospect evaluation and
development. Inspired by the success in lithology and fluid discrimination in elastic domain, we used Neural
Network technique into elastic volume to predict fluid and petrophysical properties in the reservoir. The study
consist of three phases: 1) Correlation study between Elastic Impedance log and Petrophysical Log to
determine chi angle optimum, 2) Extended Elastic Impedance (EEI) equivalent generation for lithology
description from AVO attribute, 3) invert seismic volume equivalent to elastic domain using deterministic
inversionmethod and, 4). Refine the result using ANN method to get Gamma Ray (GR)–seismic equivalent
volume. The result showed that the application of ANN on elastic impedance has improving the lithology
discrimination for reservoir delineation.

Keywords: Elastic Impedance, EEI Rotation, Neural Network

1.0 INTRODUCTION (Acoustic Impedance), Poisson Ratio and Shear


Impedance.
In prospect evaluation, amplitude is a response to
different reservoir properties. The polarity is a key factor
in which the interpreter often used to determine the
hydrocarbon sand. In most of the case the hydrocarbon
sand will produce strong negative response or we can call
as bright spot. But, several occasion where we can get
similar response to that [1] :
 Hard shale / Soft shale
 Coal layers
 Good quality brine sand
Hence, amplitude interpretation usually is a difficult
problem in certain circumstances during exploration as
well as development stages.
The problem that rises can be solved when we look in
Elastic Impedance (EI) domain. Concept of EIwas Fig. 1 Similar Response of AI log between soft shale and good
introduced by Connolly[2] and improved by Whitcombe[3] quality sand. Difficult to distinguished between soft shale
by introduced the Extended Elastic Impedance (EEI) and good quality sand[1].
technique. This concept was built to full fill the
The result of AI is basically a softness that indicates
requirement of impedance response offset dependent.
lithology. The softness is due the changed from hard shale
The EI is not really a physical property of medium;
to hard sand or hard sand to hard shale. Interpretation of
however by using this concept, correlation between true
AI is best done if we have log data. However, we always
physical properties of medium with seismic can be
face the limitation of number of wells to predict the
estimated.
lithology.
There are three important rock properties:Vp, Vs and
Rho that play critical role in Seismic Interpretation. In 3D
domain, these 3 properties can be distinguished as AI

78
Journal of Engineering Technology Vol.2 (1): 78-81, 2012

Hence in this study, we will derive the seismic volume of 2.0 AVAILABLE DATA AND METHODS
well log properties (Gamma Ray) to discriminate the
lithology in elastic impedance angle dependent. The data used in this study is from Scotia Basin, from
Nova Scotia Department of Energy, which is available on
1.1 NEURAL NETWORK Opendtect Seismic repository. The Data consist of seismic
gather data, wells and interpretation of horizon.
Neural Network Technology (NNT) is a branch of
artificial intelligence that addresses pattern or shape
recognition. The algorithms mimic human brain’s
processing ability of object discrimination.

Fig. 2. Schematic NNTprocess: Input will be train by the


neuron and will lead to the result as output similar process of
human brain

NNT technique can be divided into two ways that is


supervised and unsupervised. For this study we used
supervised method.
Log signature in certain zone will be used as the
input to the training set to determine the lithology. For Fig. 4 Well Log of Penobscot field. The Porosity, Gamma Ray,
this study GR (Gamma Ray) log will be used as input for resistivity, acoustic impedance, clay volume and water
the neural network training set. saturation log respectively. In the study area is showing
From this GR log, it will be used to predict the lithology reservoir sands.
from the seismic EEI equivalent volume.
In this work, the estimation of petrophysical
parameter volumes is based on EEIcorrelation. Velocity
log and density log are used to derive EEI log. Here, S-
wave log is estimated using Greenberg Castagna’s
equation. Then the gamma ray logs are correlated with the
spectrum of EEI to calculate the optimum rotation angle.
The optimum rotation angle was chosen which is
represented by its highest coefficient correlation. The
Intercept (A) and gradient (B) of pre-stack data based on
two terms AVO equation were calculated and the
equivalent seismic volumes of EEI at optimum rotation
Fig. 3. Example of the application of Neural Network method in
Malay Basin showing improved seismic imaged. The angle was generated for appropriate petrophysics target.
lithological facies and gas water contact can be clearly separated Finally, Neural Network are applied to EEI volume
compared to the normal seismic attribute. [1]. seismic to improve the interpretation and well-seismic
correlation. Crossplot between gamma ray are constructed
to check the quality factor of separation.

79
Journal of Engineering Technology Vol.2 (1): 78-81, 2012

In Fig. 6, we can see that rotation of EEI at optimum


angle for each appropriate log properties showing good
 Correlation study correlation with log data. In low frequency the trend of
between Elastic that properties both of logs and EEI are almost the same,
Impedance log Oppositely, in high frequency, the high spike especially
and Petrophysical on water saturation log is not represented by EEI log.
Log Fortunately this high frequency are not to be presented on
 AVO response seismic data until the resolution of seismic data are
and EEI improved.
projection The correlation is one important stage in EEI study.
 Deterministic Hence the high attention should be paid on this stage. The
Inversion EEI log normally is increase with depth. Hence, zonation
on depth should be considered during rotation, in some
cases the rotation angle of EEI with certain log properties
Neural Network in changes with depth. This will influence the correlation
EEI Volume coefficient especially the correlation with the log
properties that is not has a trend in depth. Some zonation
should be determined carefully.
Fig. 5. Flowchart of lithology prediction in EI Domain Supervised ANN method is implemented on seismic
volume of GR equivalent after deterministic inversion.
Two of Penobscot well of Gamma ray log are used for
3.0 RESULTS AND DISCUSSION training. Training phase was using 30% of well data.. The
training was stopped when normalized errors are
In EEI log correlation, the EEI logs are calculated minimum as shown in the Fig. 7.
from P and S wave velocity and density log. The
calculation showing that Gamma ray log has significant
correlation with EEI at 34 degrees with the coefficient
correlation is about 0.6. The angle at this maximum
correlation is defined as optimum angle which is used for
AVO rotation. The maximum coefficient correlation
between EEI and porosity, clay volume, resistivity, water
saturation and Gamma ray logs are: -0.92, 0.59, 0.43,
0.78, and 0.59. Comparisons between EEI at certain angle
with log properties are shown in the Fig. 6.

(a) (b)
Fig. 7 (a) Normalized Errors between train data and test data
(b) plot scattering of train data and test data (right).

Relation between Gamma ray log and Gamma ray


equivalent that was generated from EEI was improved
through application of ANN on the elastic impedance
domain. Comparison relation between both of them
before and after ANN is shown in the Fig. 8.

Fig. 6. Comparison between Log properties (blue line) and


EEIlog (red line) for appropriate petrophysical logs: porosity,
water saturation,volume of clay, Neutron and lateral logs. Fig. 8. Crossplot between log Gamma ray and Gamma ray
equivalent are extracted from seismic before (left) and after
(right) ANN.

80
Journal of Engineering Technology Vol.2 (1): 78-81, 2012

Slicing seismic data at Logan Canyon formation [5] P.M. Wong, D. Gedeon, and I. J. Taggart, “An Improved
showing that sand can be delineated clearly as shown in Technique in Porosity Prediction : A Neural Network
the Fig. 8. In vertical view, Gamma ray log is plotted on Approach,” IEEE TRANSACTIONS ON GEOSCIENCE
EEI cross section data. It is clear that seismic volume of AND REMOTE SENSING, vol. 33, no. 4, pp. 971-980,
1995.
EEI represents the lithological as also shown in Fig. 8
(right). Different lamination on EEI data is fitting with
Gamma ray log. [6] S.D. Mohaghegh, A. Popa, R. Gaskari, S. Wolhart, R.
Siegfreid, and S. Ameri, “Determining In-Situ Stress
Profiles From Logs,” SPE 90070, 2004.

[7] G.J. Hicks and A. M. F. Earthworks, “Extended Elastic


Impedance and Its Relation to AVO Crossplotting and Vp /
Vs,” in EAGE 68th Conference & Exhibition, 2006, no.
June 2006.

[8] D.N. Whitcombe, P. A. Connolly, R. L. Reagan, and T. C.


Redshaw, “Extended elastic impedance for fluid and
lithology prediction,” Geophysics, vol. 67, no. 1, pp. 63-67,
2002.

[9] P.Connolly, “Robust Workflows for Seismic Reservoir


Characterisation,” Spring, pp. 1-77, 2010.
Fig. 8 (a) Slicing of Gamma ray volume at Logan Canyon
formation
[10] S. Verma and S. Biswal, “Estimation of Effective Porosity
(b) Cross section of Gamma ray volume with Gamma
and saturation volume by Extended Elastic Impedance
ray log. approach : A case study,” in The 2nd South Asian
Geoscience Conference and Exhibition, 2011, no. 3.

4.0 CONCLUSION

Volume of rock properties of reservoir can be


approximated from seismic data using EEI technique at
certain rotation angle. Application of ANN on the elastic
impedance domain can improve the relation between
seismic equivalents with well logs, hence the
interpretation of this data can be improved especially for
lithological description.

REFERENCES

[1] D. Ghosh, M. Brewer, and N. Darman, “Geophysical issues


and challenges in Malay and adjacent basins from an E&P
perspective,” SEG The Leading Edge Offshore Technology,
vol. Apr, pp. 436-449, 2010.

[2] P. A. Connolly, “Elastic impedance,” The Leading Edge,


vol. April, 1999.

[3] D. N. Whitcombe, P. A. Connolly, R. L. Reagan, and T. C.


Redshaw, “Extended elastic impedance for fluid and
lithology prediction,” SEG expanded Abstract, 2000.

[4] S. I. ARSALAN and A. YADAV, “Application of extended


elastic impedance: A case study from Krishna-Godavari
Basin, India,” The Leading Edge, pp. 1204-1209.

81

You might also like