1. Thesis Topic An Integrated Petrophysical Study Using Well Logging Data for Evaluation of a Gas Field in The Gulf of Thailand
Committee member : Dr. Pham Huy Giao (Chairman) : Dr. Noppadol Phien-wej : Dr. Le Hai An
Presented by Thoedpong Witthayapradit ST Date May 2009

2. Outline Introduction Objectives Scopes of work Literature review
Methodology
Results and Discussions
Conclusions and Recommendations
Q & A

3. Abbreviation ANN Artificial Neural Network GR Gamma Ray log
DT Acoustic log
NPHI Neutron porosity log
RHOB Bulk density log
RESD Resistivity log- deep investagation
RESM Resistivity log- medium investigation
RESS Resisitivity log- shallow investigation
IP Interactive Petrophysics
MATLAB MATrix LABoratory
MMSCF Million Standard Cubic Feet
MTJDA Malaysia-Thailand Join Development Area

4. Introduction
The understanding of reservoir rock characteristics by formation evaluation is needed to determine producible potential of petroleum extracted from reservoir.
Artificial Neural Networks (ANNs) have been developed to predict some petrophysical parameters. The study of well logging interpretation and flow unit characterization enhances the ANN prediction performance.
Fig 1 Application of ANN of well logging (

5. Objectives Application of the backpropagation ANN
The objectives in this study is ;
Application of the backpropagation ANN
technique to predict the porosity and permeability
of gas reservoir in North Malay basin, Gulf of
Thailand by using of well logging and core data
that can be used for formation evaluation in the
study area.

6. Scopes of Work
Collection of data, i.e. well logging, core analysis and geological data
Perform a well logging interpretation using Interactive Petrophysics software.
Flow unit characterization from core analysis data.
Review on the ANN methods and its application in parameter prediction based on the well logging data.
ANN training and testing with data sets from all available data, data selected from interpreted reservoir zone and data from flow unit characterization.
Perform a trained ANN analysis using MATLAB on data sets in integration with the core analysis data to predict porosity and permeability.
Compare porosity and permeability from derived well logging, core analysis and ANN prediction

7. Literature Review Natural gas situation and statistic in 2007
The literature review is following:
Natural gas situation and statistic in 2007
Area of study
Formation evaluation
Coring and core analysis
Well logging
Flow unit characterization
Artificial Neural Network (ANN)

8. Fig 2 Proved natural gas reserves at end 2007 (http://www.bp.com)
Natural gas situation and statistic in 2007
Fig 2 Proved natural gas reserves at end (
Fig 3 Distribution of proved natural gas reserves (

9. Fig 4 Natural gas consumption by area (http://www.bp.com)
Natural gas situation and statistic in 2007
Fig 4 Natural gas consumption by area (
Fig 5 Natural gas production by area (

10. Fig 6 Natural gas consumption per capita (http://www.bp.com)
Natural gas situation and statistic in 2007
Fig 6 Natural gas consumption per capita (

11. Natural gas situation and statistic in 2007
Fig 7 Natural gas price historical (

12. Area of study
The gas field is geologically located in the North Malay Basin.
North Malay basin characteristics
Depth : Thick sediment > 8 km
Geological age : Tertiry
Geological structure : horst and graben
fault
Source rock : Shale and siltstone
Reservoir rock : Sandstone
Fig 7 Malay Basin province (Michele,2002)

13. Area of study
Fig 8 The comparison of stratigraphy from difference area in Malay Basin
Source: Michele B. (2002),

14. Formation Evaluation
The use and interpretation of several tools and methods that are capable of locating and evaluating the commercial significance of petroleum in the rock. (Lynch,1962)
The formation evaluation tools
Coring and core analysis
Drilling fluid and cuttings analysis logging
Well Logging
Electric logging
Radioactive logging
Acoustic velocity logging
Drill Stem Testing (DST)

15. Coring and Core Analysis
Coring is integration part of formation evaluation which direct
measurements are made.
Table 1 petrophysical parameters which can be
determined from core measurement
Routine Core Analysis (RCAL)
Special Core Analysis (SCAL)
Porosity
Grain density
Absolute permeability
Water and oil saturation
Water salinity
Lithologic description
Core Gamma Log
Core Photography
Capillary pressure
Relative permeability
Wettability
Petrography- thin sections and SEM
Damage assessment tests
Electrical properties- n, m and F
Mechanical properties
Acoustic velocity
NMR
Mineralogy- QXRD
Completion testing
Fig 9 Core sample (

16. Formation Measurement
Well Logging
Well logging can measure and record formation properties continuously versus depth.
Table 1 Measurement of formation
Formation Measurement
Parameters
Electrical resistivity - Bulk density - Natural and induced radioactivity - Hydrogen content - Travel time of sonic wave
Porosity (primary and secondary) - Permeability - Fluid saturation
- Hydrocarbon type - Lithology - Formation dip and structure - Sedimentary environment - Elastic modulus
Fig 10 Well logging measurement (

17. Flow unit characterization
The flow units are the resultant of the depositional environment and diagenetic process. Many authors defined flow unit as follows
(Tiab et al,2004);
- A flow unit is a specific volume of reservoir, composed of one or
more reservoir quality lithologies
- A flow unit is correlative and mapable at the interval scale.
- A flow unit zonation is recognizable on wire-line log.
- A flow unit may be in communication with other flow units.

18. Flow unit characterization
Flow unit characterization factors
Reservoir quality index (RQI)
Based on the Kozeny-Carman equation (1939), Amaefule et al. (1993) introduced the concept of reservoir quality index, by considering the pore-throat, pore and grain distribution, and other macroscopic parameter
Kozeny-Carman equation:
(1.1)
Where:
k = permeability, µm2
ø = porosity, fraction
sVgr= specific surface area per unit grain volume, µm-1
KT = Kpsτ = effective zone factor
τ = tortuosity of the flow path
Divide with ø;
(1.2)

19. Flow unit characterization
If the permeability is expressed in milidarcies and porosity as a fraction, the equation 1.2 becomes:
(1.3)
Where:
RQI = reservoir quality index
k = permeability, mD
øe = effective porosity, fraction
Flow Zone Indicator (FZI)
The flow zone indicator is defined from equation 1.1
(1.4)
Thus equation 1.2 can be written as:
(1.5)
Where, øz is the ratio of pore volume to grain volume;

20. Artificial Neural Network
An ANN is an information processing system that roughly replicates the behaviors of a human brain by emulating the operations and connectivity of biological neurons (Tsoukalas and Uhrig, 1997). WN W2 f x1 y xN x2 W1
Fig 11 Schematic of ANN model

21. The use of Artificial Neural Network
ANN is wildly used to solve problems such as classification, function approximation and pattern recognition. ( Lawrence, 1994, and Haykin ,1999)
It is not suitable if the solution exists.
ANN is good than other method (Master,1993) when;
Data is subject to large errors
The pattern data is deeply hidden
Data is unpredictable non-linearity

22. Back-propagation ANN (Rumelhart et al.,1986)
Back-propagation Networks have a largest number of successful application especially for prediction.
This method is a supervised learning technique for training multilayer neural network.
The weights are adjusted during training to minimize error between known output and model output.

23. Application of ANN for well logging
As the relationships of the well logging data and the reservoir properties are unknown, the neural network is proposed to predict the formation parameters (Helle et al., 2001).
The advantage of this method are;
User does not need a deep geological knowledge of the area.
It is faster interpretation than conventional interpretation

24. Methodology 1. Data collection
Data source : Gas fields, Malay basin, Gulf of Thailand
Data : 3 vertical well logging data (data in LAS format)
: 3 Routine Core Analysis (RCA) data : 3 Geological Data
Software : Interactive Petrophysics version3.3
( well logs interpretation)
: MATLAB version 6 with Neural Network Toolbox
( ANN model and simulation)

25. Methodology 1. Data collection
Collected data consisted of core analysis,i.e., permeability and porosity, and well logging data, i.e. GR,SP, DT, NPHI, RHOB and RESD,RESM and RESS
Table 3 Number of core data available
Core interval
Well-1
Well-2
Well-3
Depth (ft)
No. samples 1 37 41 69 2 21 84 3 9 - 56 4 80 87 5 76 81 6 88
Total
223
125
384

26. Methodology Table 4 summary of well logging type run in Well-1,2 and 3
Well-1
Well-2
Well-3
Start (ft)
236.5
69.5
112.5
Stop (ft)
7880.5
9673
9824
Step (ft)
0.5
Log type
BS (in)
CALI (in)
DTCO (us/f)
DTSM (us/f)
GR (GAPI)
NPHI (v/v)
POTA (%)
RESD (ohm-m)
RESM (ohm-m)
RESS (ohm-m)
RHOB (g/cm3)
SP (mV)
THOR (ppm)
URAN (ppm)
Available for coring wells
Not available

27. Methodology 2. Well logging interpretation Zoning Lithology
Tools : GR, Caliper, Resistivity
Tools : Neutron and Density cross-plot was used to
identify formation lithology
Lithology
Shale volume
Tools : GR
Shale Index = Vsh for linear trasformation
Porosity
Tools : Density and Sonic
For shaly-sand, Correction is made by Vsh
Rw determination
Tools: SP log (temperature and mud resistivity were known)
Sw determination
Tools : Resistivity
Using Archies’equation
Permeability
estimation
Permeability-porosity relationship from core analysis

28. Flow unit characterization
The flow unit characterization is studied from core analysis of each intervals. The following steps were done as follows:
- The RQI and øz are plotted in log-log scale
- The flow zone indicator, FZI, was determined from the intercept where øz =1.
- FZIs from each core intervals were used to determine the flow unit of the reservoir as the plots that lie on the same straight line have similar pore throat characteristics and constitute a flow unit.
- Flow unit identification and construction well correlation.
In addition, the permeability-porosity relationships from core analysis can be plotted in semi-log scale. The slopes of the plot of each core interval are also used as supporting criteria for flow unit characterization. The permeability-porosity relationships of all core intervals each well were also used for well logging interpretation to determine the permeability from derived porosity.

29. ANN construction Table 5 Summary of ANN training and testing cases
Data case
Well logging Case 1
Well logging Case 2
Well logging Case 3
Data set
Number of sample
% Testing data
All well logging
and core available
DT,GR,NPHI,
RESD,RHOB
DT,NPHI,RESD,
RHOB
Training
Well-1
179 -
Well-2
100
Testing 44
19.83 25
Selected from
reservoir zone
139 52 34
19.41 12
flow unit
111 40 27
19.25 9

30. Target output : Core analysis from a selected well
Construction of ANN
Input data : Selected well logging data (GR ,resistivity,RHOB, NPHI and DT)
Target output : Core analysis from a selected well
Software : MATLAB Version 6.0
Design the ANN architecture
Training the ANN with training data set
Testing the ANN model with a testing data set
Comparing the ANN models and selecting the optimal one
Using the selected optimal model to predict the parameters

31. Methodology Input data (Training) Target output Core analysis
ANN model
Input layer
Hidden layer
Output layer
Input data
(Training)
W11,1
W21,1 GR
Target output
Core analysis
RESD
Porosity
Permeability
NPHI
RHOB
W2 s2,s1
Sonic
W1 S1,R

32. Methodology Input data (Testing) ANN output (predicted) ANN model
Input layer
Hidden layer
Output layer
Input data
(Testing)
W11,1
W21,1 GR
ANN output
(predicted)
RESD
Porosity
Permeability
NPHI
RHOB
W2 s2,s1
Sonic
W1 S1,R

33. Results and Discussions
Well logging interpretation
- Interpretation result of Well-1, 2 and 3
- Summary table of reservoir zones.

34. Results and Discussions
Fig 12 Well logging interpretation result of Well-1

35. Table 6 Summary of reservoir zone in Well-1
Depth
Thickness
Permeability øe Sw
Vcl
Top (ft) -
Bottom (ft)
(ft)
(mD)
(fraction) 2
3732.5
3770
37.5
1252.6
0.31
0.16
0.29
3784.5
3825.5 41
0.26
0.32
0.34
3832
3860.5
28.5
352.99
0.24
0.28
0.2
3918
4007.5
89.5
159.75
0.25
0.23
0.1 3
4138
4209.5
71.5
102.51
0.18
0.07
4227
5042
815
328.39
0.36
0.22
5055
5541.5
486.5
34.84
0.21
0.05
5664
5694.5
30.5
44.49
0.15
0.01
5805.5
5893.5 88
25.56
0.37
0.19
Total
1688

36. Results and Discussions
Fig 13 Well logging interpretation result of Well-2

37. Table 7 Summary of reservoir zone in Well-2
Depth
Thickness
Permeability øe Sw
Vcl
Top (ft) -
Bottom (ft)
(ft)
(mD)
(fraction) 1
4010.5
4026.5 16
103.93
0.23
0.52
0.39
4030.5
4058
27.5
69.27
0.21
0.64
0.4
4064
4070 6
159.35
0.24
0.46
0.37 3
4102
4108
24.88
0.17
0.29
4118
4145.5
89.59
0.22
0.32
0.31
4186
4211.5
25.5
46.02
0.19
0.57
0.42
4303.5
4326
22.5
30.69
0.18
0.6
4337.5
4359.5 22
63.42
0.61
0.35
4398
4555.5
157.5
77.29
0.3
4568.5
4579.5 11
69.47
0.5
4603.5
4630.5 27
42.64
0.63 4
4639.5
4666
26.5
58.27
4673
4715.5
42.5
60.78
4725
4791 66
55.82
0.65
0.28
5009.5
5024.5 15
48.96
0.2
5057.5
5076
18.5
41.46
0.67 5
5119
5152 33
40.66
0.68
5179
5188.5
9.5
106.73
5350.5
5386.5 36
27.05
0.27
5438
5479.5
41.5
37.24
0.58
0.25
5954
5978.5
24.5
0.26
Total
661.5

38. Results and Discussions
Fig 14 Well logging interpretation result of Well-3

39. Table 8 Summary of reservoir zone in Well-3
Depth
Thickness
Permeability øe Sw
Vcl
Top (ft) -
Bottom (ft)
(ft)
(mD)
(fraction) 1
1952.5
1961.5 9
0.35
0.23
2087.5
2104.5 17
0.38
0.32
0.15
2113
2162 49
0.37
0.18
2367.5
2377.5 10
0.36
0.17 5
4187
4237.5
50.5
176.28
0.25
0.2 6
4259.5
4271.5 12
124.49
0.24
0.47 7
4474.5
4500.5 26
487.2
0.28
0.48
0.09
4563
4586 23
664.35
0.29
0.11
5132
5175 43
108.72
0.22
0.16
5263.5
5286
22.5
37.16
5320.5
5353
32.5
0.27 11
5646
5683.5
37.5
33.13
0.43
5749
5792.5
43.5
863.88 13
6086
6099
48.07
0.19
6144
6178 34
109.85
6191.5
6200
8.5
259.51
0.26
0.3
0.04
6219
6231.5
12.5
13.13
0.14
Total
443.5

40. Results and Discussions
2. Flow unit characterization
- Permeability and porosity relationship
- FZI determination
- Flow unit characterization
- Well correlation

41. Results and Discussions
2. Flow unit characterization (continued)
- Permeability and porosity relationship
Fig 15 Porosity-permeability relationship of
all core interval in Well-1

42. Results and Discussions
2. Flow unit characterization (continued)
- Permeability and porosity relationship
Fig 16 Porosity-permeability relationship of core interval No.1 in Well-1
Fig 17 Porosity-permeability relationship
of core interval No.3 in Well-1

43. Results and Discussions
2. Flow unit characterization (continued)
- FZI determination
Fig 18 Flow unit characterization of all core interval in Well-1

44. Results and Discussions
2. Flow unit characterization (continued)
- FZI determination
Fig 19 Flow unit characterization of
core interval No.1 in Well-1
Fig 20 Flow unit characterization of
core interval No.3 in Well-1

45. 3. Flow unit characterization (continued) - Flow unit selection
Table 9 Summary of permeability and porosity relationship of each core interval
Well
Core interval
Depth
Slope
Intercept
Well-1 1
3864.3 -
3922.5
0.65
-5.717 2
3945.2
3970
0.501
-1.678 3
3971.5
3979.5
0.367
-0.065 4
4094.3
4183.6
0.851
-11.28 5
4184.2
4271.6
0.574
-4.066
Well-2
4016.3
4069.3
0.388
-1.148
4070.5
4118.5
0.549
-2.154
4119.5
4154.5
0.565
-3.895
Well-3
3930.5
4020.5
0.577
-4.183
4022.6
4026.6
4085
0.954
-9.111
4168.5
4255.6
0.864
5139.4
5220.6
0.87
-9.225 6
5560.4
5648.3
0.857
-9.408

46. 2. Flow unit characterization (continued)
Table 10 Summary of flow unit characterization of each core interval
Well
Core interval
Depth
Slope
FZI
Well-1 1
3864.3 -
3922.5
1.842
4.595 2
3945.2
3970
1.827
4.926 3
3971.5
3979.5
0.843
1.146 4
4094.3
4183.6
2.877
14.021 5
4184.2
4271.6
1.966
5.382
Well-2
4016.3
4069.3
0.7
1.244
4070.5
4118.5
1.542
4.35
4119.5
4154.5
1.089
1.779
Well-3
3930.5
4020.5
1.709
4.051
4022.6
4026.6
4085
2.089
9.392
4168.5
4255.6
2.707
13.602
5139.4
5220.6
2.378
13.122 6
5560.4
5648.3
2.048
7.403

47. 2. Flow unit characterization (continued) - Well correlation
Fig 21 Flow unit correlation of Well-1, Well-2 and Well-3 with GR log

48. 2. Flow unit characterization (continued) - Well correlation
Fig 22 Flow unit correlation of Well-1, Well-2 and Well-3
with core interval schematic

49. Results and Discussions
3. ANN prediction model
- ANN porosity model
- Performance error
- Selected porosity models
- Regression analysis
- Test performance error
- Well-3 porosity prediction

50. - ANN Performance error
Fig 23 Testing performance error of ANN porosity model from
well log and core data available in case 1,2 and 3
Fig 24Testing performance error of ANN porosity model from
selected reservoir zone data in case 1,2 and 3
Fig 25 Testing performance error of ANN porosity model
from selected flow unit data in case 1,2 and 3

51. - Selected porosity models
All data from well logging and core analysis, 5 well logging input data, 14 hidden neurons and 5000 epochs, MSE= 23.40
Selected data from reservoir zone, 4 well logging input data (without GR), 15 hidden neurons and 8000 epochs, MSE= 15.66
Selected data from flow unit, 4 well logging input data (without GR), 17 hidden neurons and 500 epochs, MSE= 18.85

52. Fig 26 Regression analysis of a selected ANN porosity model

53. Fig 27 Performance test of a selected ANN porosity model

54. Fig 28 Regression analysis of predicted ANN porosity
versus core data of Well-3

55. - Well-3 porosity prediction result
Fig 29 The result of ANN porosity model from selected reservoir zone data in
case 1,2 and 3 Depth 6088 – 6325 ft

56. Results and Discussions
3. ANN prediction model (continued)
ANN permeability model
- Performance error
- Selected permeability models
- Regression analysis
- Test performance error
- Well-3 permeability prediction result

57. - ANN performance error
Fig 30 Testing performance error of ANN permeability model
from well logging and core data available in case 1,2 and 3
Fig 31 Testing performance error of ANN permeability model
from reservoir zone data in case 1,2 and 3
Fig 32 Testing performance error of ANN permeability model
from flow unit data in case 1,2 and 3

58. Results and Discussions
- Selected permeability models
All data from well logging and core analysis, 4 well logging input node (without GR), 15 hidden neuron and 8000 epochs, MSE= 39,169
Selected data from reservoir zone, 4 well logging input node (without GR), 16 hidden neurons and 8000 epochs, MSE= 27,513
Selected data from flow unit, 4 well logging input node (without GR), 9 hidden neurons and 8000 epochs, MSE= 11,736

59. Fig 33 Performance test of a selected ANN permeability model

60. Fig 34 Performance test of a selected ANN porosity model

61. versus core data of Well-3
Fig 35 Regression analysis of predicted ANN permeability
versus core data of Well-3

62. - Well-3 permeability prediction result
Fig 36 The result of ANN permeability model from selected flow unit data
in case 1,2 and 3 Depth ft

63. Conclusions and Recommendations
An Integrated well logging interpretation was done for three wells in the North Malay basin to evaluate porosity, permeability and water saturation.
The flow unit characterization gives a better understanding of hydraulic flow pattern. The same flow unit might have consistent petrophysical and fluid properties. The flow units were characterized using FZI and the permeability and porosity relationships in this study, three flow units were observed, having FZI equal to 1.5, 4.5 and 13.8.
The ANN for permeability and porosity prediction was constructed and used to predict permeability and porosity from well logging in the study area. The ANN and permeability porosity models has the least performance error when input data are selected only from reservoir zone with four input data of DT, NPHI, RESD and RHOB.

64. Conclusions and Recommendations
Conclusion (Continued)
The ANN for porosity and permeability prediction for this study area were found as follows:
(i) ANN porosity model: selected data from reservoir zone, 4 well logging input node (without GR), 15 hidden neurons and 8000 epochs, MSE= 15.66
(ii) ANN permeability model : Selected data from flow unit, 4 well logging input node (without GR), 9 hidden neurons and 8000 epochs, MSE= 11,736

65. Q & A

66. Thank you for your attention