Sunteți pe pagina 1din 15

IPA11-G-203

PROCEEDINGS, INDONESIA PETROLEUM ASSOCIATION


Thirty-Fifth Annual Convention & Exhibition, May 2011

PREDICTING RESERVOIR PROPERTY UNCERTAINTY THROUGH AVA MONTE CARLO


MODELING AT KERISI AND HIU FIELDS, NATUNA SEA

Riki Tasrianto*
Yasril Kahar*
John Hughes*

ABSTRACT INTRODUCTION

ConocoPhillips Indonesia is currently producing oil The Kerisi and Hiu Fields are located in Block B,
and gas from Kerisi and Hiu Fields, Block B, Natuna Sea (figure 1) approximately 200 miles
Natuna Sea. Kerisi and Hiu Fields are independent northeast of Singapore in water depths of around
faulted four-way dip closures containing an oil leg 300 ft. The fields produce oil and gas from the
overlain by a gas cap. Production is primarily from Oligocene Gabus sandstone reservoir at depths
the Oligocene Gabus Massive sand. This study between 3500 ft and 5500 ft sub-sea. The reservoir
addresses a predictive technique to evaluate the sandstones typically exhibit an anomalous seismic
upside potential from secondary reservoirs within AVA (amplitude verses angle) response. We present
the Gabus formation excluding the currently an analysis to characterize predicted reservoir
producing sands. properties of untested bright amplitudes within the
Gabus formation using seismic rock physics and
Although the Kerisi and Hiu structures are relatively Monte Carlo simulation stochastic AVA forward
simple, the Gabus Formation was deposited in a modeling.
stratigraphicaly complex fluvio-deltaic environment.
The reservoir is generally high porosity and has a
Class II to Class III AVA response that varies The Gabus Massive sand is the primary producing
depending on reservoir properties and fluid type. reservoir in the Kerisi and Hiu Fields. The Gabus
Notably, lateral changes in gross sand thickness and formation was deposited during post-rifting tectonic
net to gross of the package cause corresponding phase, in a fluvial-deltaic setting. In general, the
changes in the seismic response. reservoir quality is excellent with average effective
porosity around 25% and average permeability
The 3D seismic volume shows several other isolated around 400 mD.
bright events in the Gabus interval, outside of the
primary producing reservoir, but with a similar
AVA expression. Some of these have been The Kerisi and Hiu Fields lie adjacent to each other
penetrated by development wells which provide a in a similar structural setting. Both are four way dip
means of calibrating their seismic expression to their closures within antiformal structures formed by
reservoir properties. inversion of early Tertiary half grabens. Structural
inversion occurred during a Miocene compression
This work examines the non-unique AVA response phase, at which time the half graben fault system
of these anomalies and rather than making a underwent major slip reversal to form a reverse fault
deterministic prediction, uses Monte Carlo system that bounds the south and east limits of the
simulation to establish the thickness and net to gross structures. The Kerisi anticline is further dissected
ranges consistent with the anomaly. These reservoir by a series of smaller normal faults trending SE-
properties can further be used to determine the NW, the largest of which has a maximum
viability of future drilling and completion targets to displacement of approximately 450 feet and divides
optimize field production. Kerisi Field into two distinct regions (“western” and
“eastern” areas). At Hiu, the major reverse fault
system trends from SE-NW to SW-NE. Figure 2
shows the general depth structure of the Kerisi and
* ConocoPhillips Indonesia Hiu Fields.
METHOD measured logs. During the iterative process, a
number of physical bounds were setup, so that
Our workflow involved establishing well to seismic the final modeled logs are physically realizable
ties and then scenario modeling AVA data for and consistent. Figure 4 shows the final fluid
varying reservoir properties these were then substituted curves for Kerisi-4 results.
matched to surface seismic near and far offset angle
stacks to derive reservoir property uncertainty
ranges (figure 3). 3. Log Trend Curve Generation,
Trend curves were fitted to the oil, gas and
The database used in the analysis consists of 7 wells brine fluid substituted Vp, Vs, and density logs
in the Kerisi and Hiu area, together with near and far from Kerisi-4. The trend curves were used to
offset angle stack 3D seismic data. The objective construct “pseudo logs” at the prospective
was to model the AVA response by varying anomalous amplitude locations. This was
reservoir properties such as gross thickness, net to achieved by taking an offset well log and using
gross ratio, and fluid phase, calibrate these to a Monte Carlo approach to simulate varying
surface seismic data and thereby predict the gross thicknesses and sand net to grosses for the
reservoir properties at undrilled areas exhibiting prospect interval using the trend curve values
anomalous AVA responses. A Monte Carlo for Vp, Vs and density. Figure 4 shows the log
simulation approach was applied to provide ranges trend curve for Kerisi-4 well and figure 5
of uncertainty for the reservoir properties. The compares the Kerisi-4 derived trend curves and
simulation results were then compared with the the Hiu-3 well log data. This comparison
surface seismic AVA response in pseudo AB (AVA provides quality control for the validity of the
intercept–gradient) cross plot space. The simulations log trend curve.
were performed iteratively and were terminated
when the simulated result matched the seismic
response. 4. Fluid Discrimination Analysis.
Cross-plotting of acoustic and elastic terms, in
Compressional, shear sonic and density well log this case acoustic impedance (AI) verses
data were used to perform rock physics analysis. Poisson’s ratio (figure 6) shows that gas or oil
The rock physics analysis involved 4 steps: 1. fluid sands can be distinguished from brine sand, but
and grain property definitions, 2. fluid substitution, gas sand cannot be discerned from oil sands.
3. log trend curve generation, 4. fluid discrimination This implies that seismic AVA data could
analysis: differentiate brine sands from oil or gas bearing
1. Fluid and Grain Property Definitions. sands assuming the reservoir is resolved within
Fluid density and Bulk modulus properties were seismic bandwidth.
calculated using equations derived by Batzle
and Wang (1992). Input parameters include Based on the Kerisi-4 data (figure 4) it is
hydrostatic pressure, formation temperature, gas observed that:
specific gravity, oil gravity and brine salinity. • Sands are relatively fast compared with
These were taken from the Kerisi-4 well and shales in terms of both compressional and
used to calculate gas-oil-brine density and Bulk shear velocities. This applies to hydrocarbon
Modulus for the fluid properties. and brine fluid substituted logs.
• Sands are relatively low density compared
2. Fluid Substitution with shales. This applies to hydrocarbon and
Fluid substitution was performed using brine fluid substituted logs, although
Gassman’s equation (Gassman, 1951, op cit. hydrocarbon presence enhances the density
Mavko, 2003) to derive brine-gas-oil saturated contrast between sands and shales.
compressional velocity (Vp), shear velocity • Sands and shales are similar in terms of
(Vs) and Density logs. The dry rock Poisson’s acoustic impedance. This applies to
ratio method using the Gauss Newton algorithm hydrocarbon and brine fluid substituted logs.
was employed to estimate the dry rock Bulk • Sands show a relatively low Poisson’s ratio
modulus. Based on fluid and grain properties compared with shales. This applies to
inputs, grain and elastic moduli of the rocks hydrocarbon and brine fluid substituted logs.
were then calculated and iterated until the Hydrocarbon presence enhances the
modeled insitu compressional velocity (Vp), Poisson’s Ratio contrast between sands and
shear velocity (Vs), and density matched shales.
Based on the above observations it can be and net to gross were simulated within reservoir
qualitatively concluded that lithology is best physical properties, and correlated with the surface
discerned using an elastic term such as Poisons seismic AB plot to predict the most likely reservoir
Ratio (or Far Offset seismic data), and that thickness and net to gross at the prospective
hydrocarbon presence will enhance this response. location. In practice, the modeling was iterative, the
This is consistent with class 2 AVA behavior. modeled net to gross and thicknesses being varied
until convergence occurred between the modeled
A number of isolated amplitude anomalies were AB scatter and the observed seismic AB scatter.
identified within Gabus formation in reservoir zones
stratigraphicaly separate from the producing Gabus RESULTS
massive sand (figure 7). These amplitude anomalies
were studied to characterize the possible causative Iterative Monte Carlo AVA modeling was run for
reservoir properties by stochastic AVA modeling. each amplitude anomaly observed within the Gabus
As the seismic data is considered to be zero phase formation.
SEG positive standard polarity (decrease in
impedance equals a trough on seismic data) the Amplitude Anomaly 1 (Kerisi Structure)
maximum trough amplitude was considered
representative of the top of a hydrocarbon bearing An example of a Gabus Formation amplitude
sand and used as the primary attribute in this anomaly at Kerisi structure is shown in Figure 8.
analysis. Maximum trough seismic amplitude This amplitude was penetrated by the Kerisi-4 well.
extractions were performed for all anomalously At the Kerisi-4 well location, the amplitude
bright Gabus events, at near and far offset stacks. corresponds to a gas sand interval with 56 ft gross
The near and far offset seismic amplitudes were then thickness, 67% N/G and 21% average effective
plotted on intercept verses gradient (A-B) AVA porosity. This sand is referred to as gas sand 1.
cross-plots using near offsets to represent the Figure 8a and figure 8b show the near and far-angle
intercept term (A) and far offsets minus near offsets maximum negative amplitude extractions. The
to represent a “pseudo” gradient (B) term. Figure 9c amplitude anomaly is strongly negative at far-angles
shows an example of A-B AVA cross-plot for an and weakly negative at near-angles, and hence can
amplitude anomaly. Each of the AVA anomalies be classified as a class II - III AVA response. Figure
plot on the A-B AVA cross-plot varying as a 8c and 8d show the amplitude anomaly on seismic
function of the physical properties of the reservoir profiles.
rock reflective interfaces (Vp, Vs & density) and
reservoir properties (thickness, net to gross, Figure 9a shows the synthetic angle gather
porosity, etc.). To capture this uncertainty a generated for Kerisi-4 gas sand 1. The model angle
stochastic approach, using Monte Carlo AVA gather shows weak negative amplitudes at near
modeling, was utilized to predict the reservoir angles increasing to stronger negative amplitude at
property variation consistent with the seismic far angles. It plots as a class II AVA anomaly in A-
amplitude character on the A-B AVA cross-plot. B space (figure 9b). The modeled data shows a
qualitatively comparable response to the amplitudes
Monte Carlo AVA Modeling extracted from a region around the Kerisi-4 well
location (figure 9c). However, the surface seismic
The Monte Carlo method is a repeated random data corresponding to gas sand 1 has a broader
sampling technique useful for stochastic modeling scatter on the A-B AVA cross-plot (red ellipse on
of uncertainty within input ranges. The method was figure 9c). Figures 9d & 9e show the Monte Carlo
used to stochastically sample a range of possible AVA modeling for amplitude anomaly gas sand 1.
thickness and net to gross parameters for a reservoir Figure 9e shows that when using a gas sand trend
interval tied to an offset well. Compressional curve and allowing the sand gross thickness to vary
velocity, shear velocity and densities were between 51–75 ft, and the net to gross to vary
constrained by the calculated trend curves and AB between 40-80% the resulting A-B cross plot has a
plots generated. AB plots were generated in this similarly broad data scatter to the surface seismic.
manner for gas, oil and brine fluid substituted cases This implies that either the reservoir quantity or the
using the appropriate trend curves. Implicit in this seismic data quality vary laterally around the Kerisi
assumption is that the trend curves themselves were 4 well location.
held constant which implies an average description
of Vp, Vs and density and thereby a similar average The following sensitivities validate the above
description of porosity. Combinations of thickness conclusions:
1. If the wet sand trend curves are used instead of seismic data (figure 11c) as the prefered model
gas sands a weakly reflective response is results above.
modeled on the synthetic A-B AVA cross-plot
(Figure 9f). This is clearly inconsistent with the
surface seismic A-B AVA response (Figure
9c). CONCLUSIONS
2. If the gas sand trend curve is used but sand
thickness reduced to 32–50 ft, and net to gross This work describes a case study approach to
to 10-30%, the synthetic A-B AVA cross-plot quantify reservoir quantity (thickness & net to
modeling shows a weak AVA gradient and gross) based on integration between well and
again does not match the surface seismic A-B seismic data. Rather than using deterministic
AVA response (figure 9g). assumptions for thickness and net to gross, a
stochastic approach was employed to be consistent
Amplitude Anomaly 2 (Hiu Structure) with seismic amplitude observations but also capture
the inherent uncertainties in this approach. Rock
Figure 10 shows an example of an amplitude physics analysis, with fluid substitution and log
anomaly at the Hiu structure that has had not been curve trend generation were critical steps in the
tested by a well. To perform the thickness and net to analysis. While this method offers a good first pass
gross AVA Monte Carlo modeling for this at characterizing the reservoir properties it is based
amplitude, the Hiu-3 logs were used with Kerisi-4 on a qualitative comparison between modeled and
log trend curve data to splice in the stochastically observed AVA anomalies, future work to progress
varying reservoir section for input to the modeling. the analysis would be to calibrate the surface
seismic AVA data to the well models.
The observed AVA response is similar to amplitude
anomaly 1, with strongly negative amplitude far- ACKNOWLEDGEMENTS
offset amplitude and a weak near offset amplitude
thereby classified as class II AVA (figure 10a, 10b, The authors wish to thank BPMIGAS,
10c &10d). At the closest well, Hiu-3, this ConocoPhillips, Chevron and Inpex Petroleum for
amplitude correlated to wet sand interval with a their permission to publish this work. Special thanks
gross thickness of 55 ft, net to gross of 16%, and are also given to DR. Sigit Sukmono & Mark Boyd,
average effective porosity of about 5%. Synthetic whom supervised this work.
AVA modeling (figure 11b) of this wet sand at the
well location exhibits zero offset weak reflectivity
and a weak gradient term, this being consistent with REFERENCE
the surface seismic A-B AVA cross-plot response
(blue ellipse on figure 11c). To obtain the best Batzle, M. And Wang, Z., 1992, Seismic Properties
match with the AVA response at amplitude anomaly of Pore Fluids, Geophysics : Vol 57, no 11
2 (red ellipse on figure 11c), it was necessary to use (November 1992); P. 1396-1408.
the gas sand trend curves with reservoir properties
of gross thickness 51– 75 ft and net to gross of 30– Deuflhard, P. And Hohmann, A., 2003, Numerical
50% (figures 11d and 11e). Analysis on Modern Scientific Computing An
Introduction, Springer
To confirm this modeling result, a Monte Carlo
model sensitivity was run using a sand gross Mavko et. all, 2003, The Rock Physics Handbook,
thickness range 30–50 ft and net to gross of 10-30%, Cambridge University Press.
together with the gas sand trend curves. Figure 11f
shows the resulting A-B cross plot which does not Sheriff, R.S, 2002, Encyclopedic Dictionary of
show as good a match with the observed surface Applied Geophysics, SEG
Block B
Natuna Sea

500 km Kerisi- Hiu


25 km

Figure 1 a - Location of Block B at Natuna Sea; 200miles North East of Singapore.


b - Location of Kerisi Hiu field at Block B Natuna Sea.

High

Hiu

Low

Kerisi

450 ft
fault 3km

Figure 2 - Gabus Formation depth structures map, Kerisi-Hiu Fields, with exploration and
development wells posted.
Well Seismic
data data

Well seismic
tie

Near – Far amplitude


extraction
Rock Physic, Trend
Curves Generation
& AVA modeling

AVA interpretation

N: G / Thickness
uncertainty estimation

Figure 3 - General workflow


Depth (ft)

Shale  Wet  Oil  Gas 


Figure 4 - Fluid Substitution and Log trend curves Kerisi-4.
1100

Shale  Wet  Oil  Gas 


Figure 5 - The comparison between Kerisi-4 log trend curves compared with Hiu-3 log curve.
Poisson' ratio vs AI

0.5
Oil
Gas
0.4 Shale

Poisson's Ratio
Brine
0.3

0.2

0.1
6 7 8 9
AI (GR*KM/S/CC)

Figure 6 - Poisson’s ratio vs. AI cross-plot. Cross-plot illustrates that AI - Poisson’s ratio space can differentiate gas sands from wet sands, but cannot discern gas
sand from oil sands

K‐2  K‐1A  NE‐Hiu 1 


K‐3  K‐4  H‐2  H‐3 

Gabus massive sand amplitude

Base upper Gabus

Top lower Gabus

Figure 4.25 A-B Far stacks seismic Section. Red Ar


2 Km
A  B 
Figure 7 - Full stack seismic across Kerisi-Hiu structure. Black circle indicate amplitude anomalies within Gabus formation outside the producing Gabus massive
sand.
Far offset Near offset
Kerisi‐4  1 km Kerisi‐4  1 km

N  N 

a) b)
A  Kerisi‐4  B  A  Kerisi‐4  B 
GR  RTD   GR  RTD  
V Clay  GR  RTD  Phie 

At Kerisi‐4  
0.5 km 56 ft Gross sand 
38 ft Net sand 
67 % N/G 
21 % Ave Porosity

c) d) e)
Figure 8 - Amplitude anomaly 1: a. Max negative far angle slice, b. Max negative near angle slice, c. A-B far angle seismic section d. A-B near angle seismic
section e. Kerisi-4 log showing gas sand 1 interval with average effective porosity 21%
Kerisi-4 Thickness 56 ft
Vp  Vs  RHOB  Synthetic 
N/ G 0.67

Far – Near 
a) b) Near 
c)
Gas sand thickness 51 – 75 ft
Vp  Vs  RHOB  Synthetic 
N/G 0.4 – 0.8

Top Sand

Base Sand

d) e)
Figure 9 - Amplitude anomaly 1; a, b. AVA modeling & synthetic intercept-gradient cross-plot, using gas trend curve as the sand curve and Kerisi-4 well
data as the background curve, c. seismic pseudo intercept-gradient cross-plot d, e. AVA modeling & intercept-gradient cross-plot, using gas trend
curve and range of thickness 51–75 ft & N/G 40-80%.
Brine sand thickness 50 – 75 ft Gas sand thickness 32 – 50 ft
N/G 0.4 – 0.8 N/G 0.1 – 0.3
Gradient 

Gradient 
Intercept  Intercept 
f) g)

Figure 9 - f. Synthetic intercept-gradient cross-plot, using wet sand trend curve as the input log and range of thickness 50–75 ft & N/G 40-80 %
g. Synthetic intercept-gradient cross-plot, using gas sand trend curve as the input log and range of thickness 32–50 ft & N/G 10–30 %.
Far offset Near offset
Hiu‐3  Hiu‐3 
N  B  N 

1 km 1 km

a) b)

A  Hiu‐3  B  A  Hiu‐3  B 
GR  RTD   b) GR  RTD    V Clay  GR  RTD  Phie 

At Hiu‐3  
55 ft Gross sand 
1 KM 9 ft Net sand 
Far Near 16 % N/G 
c) d) 5  % A e 
e)
Figure 10 - Amplitude anomaly 2 a. Max negative far angle slice b. Max negative near angle slice c. A-B far angle seismic section d. A-B near angle seismic
section e. Hiu-3 log curve.
Top Sand

Vp  Vs  RHOB  Synthetic  Base Sand

Brine sand thickness 30 ft N/G 0.1

Far – Near 
a
b Near 
Vp  Vs  RHOB  Synthetic  c)
Gas sand thickness 20 - 30 ft N/G 0.15 – 0.2
Gas sand thickness 28 - 60 ft N/G 0.2 – 0.7

d) e) f)

Figure 11 - Amplitude anomaly 2 a, b. AVA modeling and synthetic intercept-gradient cross-plot c. seismic pseudo intercept-gradient cross-plot d, e. AVA
modeling & intercept-gradient cross-plot using gas trend curve with thickness range of 51–75 ft & N/G 30–50 %, f. AVA modeling intercept-
gradient cross-plot using gas sand trend curve with range of thickness 30–50 ft & N/G 10–30 %.

S-ar putea să vă placă și