Title: Review of neural network analyses performed in Croatian part of Pannonian basin (petroleum geology data)
1Review of neural network analyses performed in
Croatian part of Pannonian basin (petroleum
geology data)
- T. Malvic, J. Velic M. Cvetkovic (Croatia)
12th Hungarian geomathematical congress and 1st
Croatian-Hungarian geomathematical
congress Morahalom, 29-31 May, 2008
2INTRODUCTION
- Neural networks represent very strong tools for
different prediction tasks in many sciences. - Petroleum geology, and geology overall, is one of
the field where such networks can be very
successful and relatively easy applied for
different variable, like porosity, depth,
lithology and saturation. - Up to now, the neural prediction is applied at
the three Croatian fields (Figure 1). - The Okoli field (prediction of facies) in 2006
the Benicanci field (porosity) and the Kloštar
field (lithology and saturation) in 2007.
3Figure 1 Areas analyzed by neural networks in
Croatia (in petroleum geology domain)
4OKOLI FIELD
The Okoli field, located in the Sava depression,
is selected as the example for clastic facies
prediction using neural network. The
significant oil and gas reserves are proved in
Lower Pontian sandstones. Methodology of neural
networks using in reservoir facies prediction is
presented through paper of MALVIC (2006) and
basics given i ROSENBLATT, 1958 McCORMACK, 1991
RIEDMILLER BRAUN, 1993. The analysis is based
on RProp algorithm. The network is trained
using log data (curves GR, R16", R64", PORE/T/W,
SAND SHALE) from two wells (code names B-1
B-2 Figure 2). The neural network was trained
based on selected part of input data and
registered lithology from c2 reservoir (as
analytical target) of Lower Pontian age.
Positions of facies (sand/marl sequences) were
predicted. The results indicate on
over-trained network in the case of sandstone
sequences prediction (Figures 3,4), because the
marl sequences in the top and the base are mostly
replaced by sandstone. The further neural
facies modelling in the Sava depression need to
be expanded with additional logs that
characterised lithology and saturation (SP, CN,
DEN). Then, RPORP algorithm could be reached
with more than 90 probability of true prediction
(in presented analysis this value reached 82.1).
5Figure 2 Structural map of c2 reservoir top with
selected well's positions
6Figure 3 Relations of errors in periods of
training (T), learning (L) and validation (V)
and position of Face and Best configurations
(the symbols F, B in legend) for B-1 well
Figure 4 Relations of errors in periods of
training (T), learning (L) and validation (V)
and position of Face and Best configurations
(the symbols F, B in legend) for B-2 well
7BENICANCI FIELD
The reservoir is represented by carbonate breccia
(and conglomerates) of Badenian age. Locally the
thickness of entire reservoir sequence is locally
more than 200 m. The three seismic attributes
were interpreted amplitude, phase and
frequencies making 3D seismic cube, averaged and
correlated (CHAMBERS YARUS, 2002) by well
porositites at the 14 well locations. It made
the network training. The network was of the
backpropagation type. It was fitted through 10000
iterations, searching for the lowest value of
correlation between attribute(s) and porosities
and the minimal convergence. Results are
presented in the paper of MALVIC PRSKALO
(2007).
8The best training was reached using all three
attributes together, what indicated on tendency
that neural networks like numerous inputs.
Obtained results (Figure 7) were compared by
previously interpolated geostatistical porosity
maps (done by kriging and cokriging approaches
Figures 5 6). Relatively smooth map, and
rarely reaching of measured porosity minimum and
maximum, strongly indicates on conclusion that
neural estimation is more precisely than
previously interpolations (Figure 7). The
cokriging approach included only reflection
strength (derivation of amplitude) as secondary
seismic source of information (compared by neural
inputs of three attributes). On contrary, the
neural approach favor using of all three
attributes. In this case they are all in
physical interconnection, but generally iIt
alerts us on carefully and geologically
meaningful selection of the network inputs for
any reservoir analysis.
9Figure 5 Kriging porosity map (color scale 4-10)
Figure 6 Cokriging porosity map (color scale
3-11)
Figure 7 Neural network porosity map (color
scale 5-10)
10KLOŠTAR FIELD
The filed is located in the Sava depression.
The largest oil reserves are in Upper Miocene
sandstones, i.e. in I. (Lower Pontian age) and
II. (Upper Pannonian age) sandstone series.
The basic principles of neural networks had
been studied from ROSENBLATT (1958), ANDERSON
ROSENFELD (1989) and ZAHEDI (1993), and
application on sandstones in diploma thesis of
CVETKOVIC (2007). Geophysical borehole
measurements were used as input data for the
neural network analysis. Supervised neural
networks (SNN) were trained in the sandstone
series two wells (Klo-A and Klo-B). Firstly,
input data where log data (curves SP, R16 and
R64) used for prediction of lithology. Secondly,
the neural network was used to predict saturation
with hydrocarbons.
11- RESULTS
- Relatively small prediction error values and very
good correspondence between predicted and real
values was achieved. - This points out to great possibilities in neural
network application on petroleum geology problems
and in exploration. - Accuracy of prediction can additionally be
heightened by adding more input data, primarily
more data logs that are good at defining
lithological composition and hydrocarbon
saturation such as Gamma Ray (GR), Compensated
Neutron (CN) and Density (DEN) logs.
12Figure 8 Lithology prediction in well Klo-B
(1st case)
Figure 9 Lithology prediction in well Klo-B
(2nd case)
13Figure 10 Saturation prediction in well Klo-A
Figure 11 Saturation prediction in well Klo-B
14CONCLUSIONS (Okoli field)
- This is the first neural analysis of such type in
hydrocarbon reservoir analysis in Croatia - Excellent correlation was obtained between
predicted and true position of sandstone
lithology (reservoir of Lower Pontian age in the
Sava depression) - 2. On contrary, positions of predicted and true
marlstones positions (in top and bottom) mostly
do not correspond - 3. 4. The best prediction (so called Face
machine) is reached in relatively early training
period. In B-1 well such prediction is observed
in 2186th iteration, and in B-2 well in 7626th
iteration - 5. It means that in similar facies analyses in
the Sava depression, it is not neccessary to use
large iteration set (about 30000) - 6. The input dataset would need to be extended on
other log curves that characterize lithology,
porosity and saturation, like SP (spontaneous
potential), CN (compensated neutron), DEN
(density) and some other - 7. The wished true prediction could reached 90
(Face machine could be cinfigured with 90
probability).
15CONCLUSIONS (Benicanci field)
- The neural network was selected as the tool for
handling uncertainties of porosity distribution
in breccia-conglomerate carbonate reservoir of
the Badenian age - 2. The lateral changes in averaged reservoir's
porosities are influenced by the Middle Miocene
depositional environments - 3. The best porosity training results are
obtained when all three seismic attributes
(amplitude, frequency, phase) were used - 4. The reached correlation is R20.987 and
convergence criteria Se20.329 - 5. These values can slightly (a few percent)
differs in every new training, what is
consequence of stochastic (random sampling) is
some processes of the network fitting - 6. The result indicates that neural network very
favor the numerous inputs, but also can be easily
applied in the Benicanci field for porosity
prediction.
16CONCLUSIONS (Kloštar field)
- Several artificial neural networks were trained
with the tasks of - Predicting lithology of Upper Pannonian sediments
(II sandstone series) and Lower Pontian
deposits (I sandstone series) as well as - Hydrocarbon saturation within these beds.
- 2. Radial Basus Function and multi Layer
Perceptron networks were used and excellent
corresponding of true and predicted LITHOLOGICAL
values was achieved - 3. Also, same algorithm gave excellent
corresponding relation for WATER SATURATION
between real and predicted values - 4. Acquired results show large potential of
neural networks application in reservoir
characterisation - 5. Networks are also tools for acquiring quick
results from well logs vertical and lateral
correlation with the goal of reservoir variables
prediction.
17REFERENCES
ANDERSON, J.A. and ROSENFELD, E. (1989)
Neurocomputing Foundations of Research.
Cambridge, MA MIT Press. CHAMBERS, R.L.
YARUS, J.M. (2002) Quantitative Use of Seismic
Attributes for Reservoir Characterization.
RECORDER, Canadian SEG, Vol. 27, pp. 14-25, June.
CVETKOVIC, M. (2007) Petroleum geology use of
neural networks on the example of reservoir in
Kloštar field. University of Zagreb, Faculty of
Mining, Geology and Petroleum Engineering,
Graduate thesis, mentor Prof. Dr. J. Velic, 15.
June 2007, 49 p. MALVIC, T. (2006) Clastic
facies prediction using neural networks (Case
study from Okoli field). Nafta, 57, 10,
415-431. MALVIC, T. and PRSKALO, S. (2007) Some
benefits of the neural approach in porosity
prediction (Case study from Benicanci field).
Nafta, 58, 9, 455-467. McCORMACK, M.D. (1991)
Neural Computing im Geophysics. The Leading Edge,
10/1, Society of Exploration Geophysicists. RIEDM
ILLER, M. and BRAUN, H. (1993) A direct adaptive
method for faster backpropagation learning The
RProp algorithm. Proc. of the IEEE Intl. Conf. on
Neural Networks, San Francisco, p.
586-591. ROSENBLATT, F. (1958) The perceptron
A probabilistic model for information storage and
organization in the brain. Psychological Review,
65, 386-408. ZAHEDI, F. (1993) Inteligent
systems for business, expert systems with neural
networks. Wodsworth publishing Inc.