IPA13 SG 057 Velocity Model.

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

IPA13-SG-057

PROCEEDINGS, INDONESIAN PETROLEUM ASSOCIATION


Thirty-Seventh Annual Convention & Exhibition, May 2013

SEISMIC TIME TO DEPTH CONVERSION AND UNCERTAINTY ANALYSIS FOR


HORIZON PREDICTION IN A PROPOSED WELL-SITE OF SUNGAI GELAM
FIELD, JAMBI SUBBASIN

Muhamad Wildan P.*


Imam Muda Gunawan**
Andry Pujiriyanto**
Sudarmaji*

ABSTRACT INTRODUCTION

Seismic time to depth conversion is an important The recording method of seismic reflection data is
part of seismic interpretation work flow. This in the domain of time (mostly in two-way time)
research was conducted to determine the best (Sismanto a, 1996), whereas the interpretation of
method of depth conversion as a tool for horizon seismic data generally demands results using the
prediction in a proposed well-site KYT well. In depth. Therefore, the seismic time to depth
this research, the seismic time to depth conversion conversion is one of the most important parts in the
process was divided into nine different methods flow of the overall seismic data interpretation. In
clustered into two groups of methods. The determining the value of depth, slope, thickness of a
difference in those methods lies on the variation of reflective plane (reflector) understanding seismic
velocity model building. The velocity model wave velocity is essential (Sheriff, R.E., and L.P.
building was built from two velocity sources: Geldart, 1995). One key aspect in the conversion of
checkshot velocity and seismic stacking velocity. seismic data into depth is velocity modeling, which
Those two velocity sources were processed using a determines the relationship between the depth and
geostatistical approach to obtain the velocity model. the seismic time (TWT) (Etris, E.L. et al., 2001). A
As the process results in several numbers of time- velocity model can be developed using many
converted data, it needs to be uncertainty analyzed different methods depending on the purpose and the
to obtain the best method and the best depth values. availability of existing data sources. In seismic data,
The uncertainty analysis in this method was applied processing in the form of migration also involves
to four horizon Sand reservoirs (Top ARC, Top velocity modeling, but the migration process is not
ASH, Base ASH, and Top TALC) by using the a process of conversion into the depths. One
percentile classification approach, where the nine migration process aims to eliminate diffraction
time-converted data were classified into three effects that can lead to misinterpretation. By
percentiles: 10th percentile (P10) or Low Case, the undertaking this process, the best structural seismic
50th percentile (P50) or Base Case, and the 90th section is obtained. Modeling the velocity of the
percentile (P90) or High Case. migration process usually only involves the stacking
velocity derived from the NMO (normal moveout),
The results of the process show that the time-depth whereas the velocity data is not obtained from
conversion method using the velocity model of the measurements (hard data), but from the NMO
seismic stacking velocity with moving average correction of trial and error (Sismanto b, 1996).
interpolation shows the best result. This is justified Thus, to obtain the results of the seismic section in
by the minimum value of the horizon depth error to the depths, a process of its own, namely the
the same horizon in the well marker data. This conversion to depth using a velocity model that
uncertainty analysis is an effective and relatively involves stacking velocity, checkshot velocity (well
inexpensive tool to guide and to predict horizon velocity measurement), and a combination of both
depth in a drilling process to minimize the error. is used (Nam, N. and Lee, H.S., 2008). Converting
depth can also be used to eliminate the ambiguity
inherent in the structure of time and to test the
* University of Gadjah Mada validity of the structural pattern appearance.
** Energi Mega Persada

,3$WK$QQXDO&RQYHQWLRQ3URFHHGLQJV
In the process, velocity modeling for depth the linear regression. The form of equations derived
conversion can be derived using a variety of input by T-D curve is:
data, such as checkshot velocity, stacking velocity,
and a combination of both, and can be based on Z(t) = Xt + C (1)
certain geostatistics principles, linear velocity
estimation, and interpolation (Abrahamsen, P., where t is the TWT of the seismic horizons, and Z is
1996). No single method is the main option in the the horizons depth (Figure 4). There are two ways
depth conversion process. Each method has its own to convert a seismic time map into a depth map by
advantages and disadvantages as seen from several the T-D curve derived equation: the first one is the
aspects, such as data availability, time constraints, direct conversion, where the equation from the T-D
cost, and so on. With that, ambiguity can arise in curve is used directly to obtain the depth horizon by
the process of converting the depth because the substituting Equation (1) to obtain the depth map
depth obtained by different methods often shows (Z). Two velocity models are obtained by this
varying results. Therefore, we need to undertake method (namely, method A and B) are dissimilar in
analysis to minimize the errors that can be caused the limit of the plotted data. The second way to
by the differences in the results obtained from the convert a seismic time map into a depth map using
various methods. One method that can be used to the T-D curve is to build the three-dimensional (3D)
solve such problems is the method of uncertainty velocity model from Equation (1). The velocity is
analysis to obtain the classification percentile of the obtained by the following relation:
results obtained from the method used in the depth
conversion process. For their analysis, the depth Vavg (t) = 2000 .. (2)
map can be specified for a particular target horizon
that is considered the most likely value.
where the Vavg(t) is the average velocity, Z is the
horizon depth, and t is the seismic TWT (Figure 5).
METHODS
Two velocity models are obtained by this method
a. Velocity Model (method C and D). In the case of the T-V curve, the
mathematical equations are determined by plotting
The target of this research is in the Gelam field and the Time versus Velocity (average velocity) data of
focused on a two-way time (TWT) of 11001460 the six wells, and the equation is derived similar to
ms. Reservoir rocks are found in the Air Benakat the T-D curve by using the linear regression
formation, which consists of shaly sands. The fault method. The general form is shown in the following
traps are present in this target zone, with two major equation:
faults. The horizon targets are divided into four
sections: Top ARC, Top ASH, ASH Base, and Top Vavg (t) = Xt + C ....(3)
TALC (Figure 1). In the case of velocity modeling,
two main sources of velocity data were used: where the Vavg(t) is the average velocity, and t is
checkshot velocity (well velocity) and seismic the seismic two way time (TWT). Figure 5 shows
velocity (stacking velocity, with the sampling time the velocity model built in the 3D cube by
250 ms). A velocity model was made of each source extracting Equation (3). This method also results in
and the combination of both. Figure 2 shows the two velocity models: methods E and F. The
velocity modeling flow chart. The block of this difference in those methods lies on the limit of the
research has six productive wells: X-A, X-B, X-C, plotted data, where in the E method, well data
X-D, X-E, and X-F (see Figure 3). The checkshot (average velocity versus TWT) plotted from the
velocity was measured originally only in well X-B, datum (zero) to the lowest horizon, Top TALC,
while the checkshot velocity in other wells was whereas, in the F method, the data are plotted from
generated by a well-seismic tie process. Generally, the highest horizon Top ARC to the Top TALC
two groups of velocity models are divided into nine horizon. In the other case, the method of the
types of models (namely method A, B, C, D, E, F, velocity model in group 2 (seismic stacking
G, H, and I). Velocity model group 1 (checkshot velocity) uses a geostatistical approach to make the
velocity) was derived from mathematical equations velocity model cube. In this group, the velocity
generated by the relationship of the T-D curve cube was built by using three principles of
(Time-Depth curve) and the T-V curve (Time- geostatistics: simple kriging (method G), moving
Velocity curve) of the six wells. The mathematical average (method H), and collocated cokriging
equations for each curve are the main source for (method I). The checkshot data are also needed in
converting the seismic time data derived by using this group, as they are used as guidance for the
vertical velocity interpolation when building the process. Each group has its own step that is
model. Therefore, this second group is a different for group 1 and group 2, so only the depth
combination of checkshot velocity data and seismic maps are included in the same group that will show
velocity data in the process of velocity model identical results.
building.
The area of this study is approximately 25 km2 and
b. Uncertainty Analysis consists of six wells (X-A, X-B, X-C, X-D, X-E,
and X-F). But the wells tend to cluster near each
From the results of nine types of velocity models, other. In conjunction with the velocity modeling,
nine different types of depth map for each horizon this is a disadvantage because the estimated value of
were obtained. Therefore, uncertainty analysis was the velocity model that uses input data, such as the
performed to classify the results of the nine depth well velocity, the point, or area whose position is
maps for each horizon to obtain the most reliable far from the location of the well, tends to have a low
value. The percentile classification method has been accuracy. This is evidenced by the results obtained
used to classify the overall results of the depth maps from the several methods. Figure 6 is an example of
of each seismic horizon. Percentile classification the depth structure map on horizon Top ASH.
was divided into three groups:
Methods A and F (Figure 6) are the methods by
1. P10 or Low Case, to the deepest depths of the which the velocity model is based on checkshot
classification results, velocity, while method G uses the velocity data
calibration between wells (checkshot) and the
2. P50 or the Base Case, the classification results seismic velocity data (Vstacking). The red circle on
for the depth which is considered the most the index is the area far away from the location of
likely, and wells. Method G returned depth values ranging
13601400 m (yellow closure). Methods A and F in
3. P90 or High Case, the classification results for the same area produced a map of the structure with
the shallowest depths. a depth range between 14001480 m. This is
obviously due to the different velocity values in
c. Proposed Well each velocity model (methods A, F, and G). Unlike
the case with the area near the well site, the results
The main purpose of this research is to predict the of the three methods suggest identical closures
certain horizon depth in the proposed well location. (greenyellow), which indicate the velocity input
This well, named the KYT well, is located in the data are in the same range.
upper part of the field. This proposed well is a
distance of about 250 m to the east from the nearest The depth structure maps of the method results of
well, the X-D well (see Figure 2). The uncertainty the same group show identical closure. Figure 7 for
analysis was performed for the depth map results method A and method B for the Top TALC horizon
within this well location. This uncertainty analysis shows the results of the depth map of the same
is a method to determine the most reliable result group with a closure resemblance.
derived from the above criteria (percentile Figure 7 (a) is the depth structure map obtained
classification). A justification of the results was also from method A of the horizon Top TALC, and (b)
performed by carrying out a blind test to calculate is the depth structure map derived by method B, the
the deviation between the depth map result obtained same mathematical function but with a different
by the depth conversion process in the existing data limit (starting point).
wells and the marker data of such wells.
Methods A and B have identical closure, but show
RESULTS AND ANALYSIS different depths. For method A, the western and
eastern zones, show values of approximately 1560
The depth structure maps derived from the same 1580 m, meanwhile for method B in the same area,
group methods show similar results and depths range from 16001680 m. The different
characteristics (Figures 68). The results of the results from are caused by the velocity model
depth map in the same group of methods, group-1 obtained from the wells, and in zones far from the
and group-2, will show a similar structure (closure), well, the velocity value has a lower accuracy.
whereas, the closure outcomes between different
groups do not have such a striking resemblance. Figure 8 shows the results of the methods G and H
This is closely related to the velocity modeling (group 2) displaying identical closure, but also
similar values in the depth in almost all zones. That From these results, it appears that methods A and B
is understandable because the velocity model is in group 1 are likely grouped into percentiles P10
based on seismic velocity, and the data points are and P90, which show the depth results obtained are
scattered throughout the whole horizon area. Thus, the highest and lowest limits. The depth map results
estimating the velocity value in an unknown zone of such class are assumed high-risk results. Methods
can be done by interpolating the velocity with a F and H tend to produce the P50 percentile, which
higher resolution of sampled data than the methods means having the most likely depth value.
for group 1.
CONCLUSIONS
The appearance of similarities of the closure
patterns, differences, and similarities of depth From the results, several conclusions can be
values always appears consistent in almost derived. In the case of this study, where there are
horizons. Such consistency is derived from only a few well data and the wells are located close
characteristics of the methods used in this study. one to another, group method 2, namely the method
Methods A and B, with the percentile results that of seismic velocity model with moving average
tend to fall into percentiles P10 and P50, show the interpolation (method H) is the best method to use.
results of the horizon depth map within the KYT This is because it produces a small deviation and
well has a high degree of risk. Method F, based on consistent data at all horizons. It is also important to
percentile classification, tends to obtain the P50 or be note that the uncertainty analysis of several
the base case, which can be interpreted as the result different velocity models derived from various
with the highest degree of probability of truth. In methods in terms of data sources (checkshot
addition, the results of the blind tests, method H velocity data, seismic velocity data, and
(moving average) have relatively small deviations combination of both) and in terms of the steps taken
and are consistent across all horizons. Table 3 (using mathematical equations and geostatistical
shows the deviation values of each method approach) is a good method to classify the horizon
generated from the blind test for horizon Top ASH depth maps by performing percentile classification.
within the X-A and X-D wells. The X-A and X-D This uncertainty analysis method can be used as a
wells are located from the KYT wells as far as 1.6 tool to predict the depth of the horizon and as the
km and 260 m, respectively. Table 4 shows the controller in the drilling process of new wells to
deviation value of method H (moving average) and minimize the errors.
the method F is relatively small in the Top TALC
horizon. However, in the case of the depth map ACKNOWLEDGEMENTS
result, method H has a relatively more accurate
depth distribution than the result obtained by The authors are very grateful to Mr. Aris Setiawan
method F (checkshot velocity only) in an area away as the General Manager of TAC Pertamina EP
from the well. This is because the velocity EMP Gelam who gave us permission to use the data
estimation of model H has more tightly sampled for accomplishing this paper and the facility for the
data. authors to complete the research.

Percentile Classification The authors would also like to thank to Mr. Imam
Muda Gunawan as their supervisor and Mr. Andri
The previous section describes that the standard Pujiriyanto (EMP Malacca Strait) for giving many
deviation value is used as the control to determine corrections and advices to the authors. We also
the relationship between the results obtained for the thank the entire management and employees of P.T
depth of each horizon. The results indicate that the EMP Gelam who give their support to the authors to
value of the depth of the nine methods has a good finish the research.
relationship, and this can be seen from the small
standard deviation on each horizon. Table 1 shows REFERENCES
the overall results for the nine methods along with
standard deviations for each horizon. The results in Abrahamsen, P., 1996, Geostatistic for Time to
Table 1 show that the percentile distribution is not Depth Conversion, Norwegian Computing Center,
specific for certain methods, but that this Box 114 Blindren, N-0314, Oslo, Norway.
classification can be used to view the characteristics
of the overall results for each group of methods. Etris, E.L., et al., 2001, True Depth Conversion:
Table 2 shows the overall percentile classification More Than A Pretty Picture, Core Laboratories
results for all four horizons. Company, CSEG Recorder.
Nam, N. and Lee, H.S., 2008, Rapid Multiple- Sismanto a, 1996, Akuisisi Data Seismik, Modul
Scenario Depth Structure Risk Analysis Case Study Kuliah, Geophysics Laboratory, Faculty of
In Cuulong Basin, Vietnam, Proceedings, Mathematics and Natural Sciences, Gadjah Mada
Indonesian Petroleum Association, Thirtieth Annual University, Yogyakarta.
Convention & Exhibition.
Sismanto b, 1996, Pengolahan Data Seismik, Modul
Sheriff, R.E., and L.P. Geldart, 1995, Exploration Kuliah, Geophysics Laboratory, Faculty of
Seismology, 2nd edition, Cambridge University Mathematics and Natural Sciences, Gadjah Mada
Press. University, Yogyakarta.
TABLE 1

OVERALL RESULTS FOR EACH HORIZON

Top Top Base Top


No Method ARC ASH ASH TALC
(m) (m) (m) (m)
1 Method A [Direct Conversion] 1356.8 1416 1452 1530
2 Method B [Direct Conversion] 1330 1410 1457.7 1562
3 Method C [Function Model Z(t)] 1331 1411.4 1446 1561
4 Method D [Function Model Z(t)] 1330.3 1411 1445 1562
5 Method E [Function Model Vavg(t)] 1332 1406.6 1438 1563
6 Method F [Function Model Vavg(t)] 1330.4 1410 1444 1564.5
7 Method G [Simple Kriging] 1330.5 1406.4 1439 1562
8 Method H [Moving Average] 1330.6 1406 1438 1562
9 Method I [Colocated Cokriging] 1333.7 1403 1436 1563.3
Difference between the lowest and the highest value 26.8 13 21.7 34.5
Standard Deviation 8.7 3.8 7.2 10.4

Group Method 1 Group Method 2

TABLE 2

PERCENTILE CLASSIFICATION OF THE TOTAL RESULTS

Top Top Base Top


No Method
ARC ASH ASH TALC
1 Method A [Direct Conversion] P10 P10 P90
2 Method B [Direct Conversion] P90 P10
3 Method C [Function Model Z(t)]
4 Method D [Function Model Z(t)]
5 Method E [Function Model Vavg(t)] P10
6 Method F [Function Model Vavg(t)] P50 P50 P50
7 Method G [Simple Kriging]
8 Method H [Moving Average] P50
9 Method I [Colocated Cokriging] P90 P90
TABLE 3

DEVIATION VALUES OF EACH METHOD GENERATED FROM


THE BLIND TEST FOR HORIZON TOP ASH WITHIN X-A WELL AND X-D WELL.

Deviation in Deviation in
No Method
X-A (m) X-D (m)
1 Method A [Direct Conversion] 12 3
2 Method B [Direct Conversion] 18 13
3 Method C [Function Model Z(t)] 20 12
4 Method D [Function Model Z(t)] 20 13
5 Method E [Function Model Vavg(t)] 11 16
6 Method F [Function Model Vavg(t)] 3 4
7 4 3
Method G [Simple Kriging]
8 4 4
Method H [Moving Average]
9 3 3
Method I [Colocated Cokriging]

TABLE 4

DEVIATION VALUES OF EACH METHOD GENERATED FROM


THE BLIND TEST FOR HORIZON TOP TALC WITHIN X-A WELL AND X-D WELL.

Deviation in Deviation in
No Method
X-A (m) X-D (m)
1 Method A [Direct Conversion] 37 35
2 Method B [Direct Conversion] 13 5
3 Method C [Function Model Z(t)] -10 -5
4 Method D [Function Model Z(t)] -4 -4
5 Method E [Function Model Vavg(t)] -10 -5
6 Method F [Function Model Vavg(t)] -2 12
7 Method G [Simple Kriging]
-1 -2

8 Method H [Moving Average]


-3 -5

9 Method I [Colocated Cokriging]


-2 -2
Figure 1 - Four horizon targets (time structures, TWT), from the shallowest to the deepest,
respectively: Top ARC, Top Ash, Base ASH, and Top TALC. A major fault separates the
northern and southern zones.
Figure 2 - The Flow Chart of the Research. The main sources used in this study are the checkshot
velocity data, seismic stacking velocity, well tops data (marker), and the seismic time
structure (horizons). The process of velocity model building is divided into two groups.
Figure 3 - Location of the six wells and proposed well KYT. The most wells are clustered in the
northern part of the block (X-B, X-D, X-E, and X-F), with a major fault separating the
south and the north.
Depth (Z) versus Time (TWT)

-200 X-F
X-E
-400 X-D
-600
X-C
X-B
-800 X-A
Z (m)

Top ARC
-1000 Top TALC
-1200

-1400

-1600

-1800
-1600 -1400 -1200 -1000 -800 -600 -400 -200 0

TWT (ms)

Figure 4 - TD Curve. The relation of Depth (Z) against two way time (TWT) of the six wells. The
colored dots (red and green) are the position of the shallowest and the deepest horizons.

Figure 5 - TV Curve. The relationship of the average velocity (Vavg) of the six wells against a two
way time (TWT). The colored dots (red and green) are the positions of the shallowest and
deepest horizons.
Figure 6 - A comparison of method A (equation derived from checkshot), method F (3D cube model
extracted from checkshot), and method G (velocity model from seismic velocity). The red
circle indicates the difference of the closure and depth values between the results.
Figure 7 - Figure (a) is the depth structure map obtained from method A of the horizon Top TALC,
and (b) is the depth structure map derived from method B. The two methods result in a
similar structure but different depth ranges.

Figure 8 - (a) is the depth structure map obtained from method G (simple kriging interpolation), and
(b) is the depth structure derived from method H (moving average interpolation). Both of
them show an identical structure (the closure) and similar values of depth in almost all
zones.

You might also like