|
|
Fault Identification and Modeling for Production Enhancement (PDF)
A.P.
Yang1, T.G. Harris2, and C. Li1
1Applied
Computer Engineering, Inc. 10101 Harwin, Suite 330, Houston, Texas 77036
Office:
713-988-2231, fax: 713-981-5519, info@gridstat.com
2
Frontera Resources, 3040 Post Oak Blvd, Suite 730, Houston, TX
(Presented
at: AAPG Geostatistics Hedberg Conference, Dec. 3-6, 2000, The Woodlands, Texas)
This case study discussed here is a porosity prediction in a shaly sand formation relying on seismic data calibrated with limited well data. A 3D geostatistic reservoir characterization and seismic inversion software (GridSTAT) was used. This paper demonstrated that geostatistics inversion of seismic to porosity is very useful in optimizing new drilling locations and in identifying target zones ahead of well path.
For
prediction ahead of the well path, VSP data from a nearby well is
used as a starting time depth curve. Well log is from the current well.
Adjustments were made to the time depth curve using sufficiently large time
window. With an optimized time depth curve and wavelet, geostatistical inversion
was carried out beyond the current well depth.
To
model porosity in 3D, the time depth relationship in each well is adjusted until
the synthetic matches the seismic and the lateral velocity variation is
reasonable. Then a common wavelet is extracted for all the wells. With an
optimized time depth model and wavelet, porosity inversion was carried out in 3D
and porosity log at any location within the seismic cube can be predicted.
Seismic
data are not only valuable in defining formation geological structures, but also
very useful in characterizing formation petrophysical properties (such as
porosity, lithology, and others) distributions away from wells if the seismic
data are calibrated with petrophysical logs at correct depth.
Geological
structures from seismic
For
exploration areas, drilling is usually expensive and geological information is
often limited and in-accurate. Well
locations are usually determined based on geological structures interpreted from
seismic. With the first exploration
wells drilled, the geological structure from seismic interpretation needs to be
improved by integrating with the newly available well data.
Petrophysical
properties from seismic
Another
important application of seismic data is that seismic can be inverted into 3D
distributions of petrophysical data. In
the geostatistics inversion described in this study, porosity log is simulated
and converted to pseudo impedance in time, reflectivity is convoluted with
wavelet, and the resulted synthetic seismic trace is compared with the actual
seismic trace near the well. If the coefficient of correlation (COC) between the
synthetic and seismic is less than the desired value, another simulation is
generated until the final results are satisfied.
These
3D petrophysical data distributions are very important because it can be used to
optimize new drilling locations, and to provide guidance if wells need to be
deepened.
In this work, data used include post-stack migrated seismic amplitude in time domain, VSP time depth curve, well locations (x and y) and KB, sonic log, and effective porosity log. Total wells covered in this study are 27.
Initial
time shift scanning
To
derive geological attribute from seismic, the most critical part is to have a
correct time depth conversion. As
shown in Fig. 1a, the time depth curve was adjusted by an amount of time shift
(Ts), which usually varies with depth. Automatic
scanning of Ts is a unique feature of GridSTAT.
Two factors were considered in order to find the appropriate Ts for each
depth interval. One is that
coefficient of correlation (COC) between the synthetic and seismic (Fig. 1b)
should be high and another is that the high COCs should be consistent along the
depth. In Fig. 1a, the selected
time shifts are connected by a solid line. The results are also verified from
visual inspection of the synthetics.
Average
velocity distribution
Seismic
velocity is a function of formation lithology, which is controlled by
depositional environment. Thus, it
is appropriate to assume that the velocity varies areally. Each well is
calibrated with separated time depth curve so that a 3D velocity model is
formed. Fig. 2 shows the area
distribution of two-way time at a subsea depth of 2030 m.
Prediction
of porosity log ahead of well path
Prediction
of petrophysical log ahead of well path can be very useful for exploration
wells. Fig. 3 shows such an
example. For well ACE1, drilling
was stopped at a depth of 2370 m. This
well was evaluated using the above mentioned inversion technique and several
high porosity zones were identified. Drilling
was then continued and the high porosity zones predicted were then confirmed.
Prediction
of logs for new wells away from the existing wells
Selection
of new well locations is always a challenge.
However, with the help of geostatistics inversion technology, a seismic
cube can be inverted into a cube of petrophysical properties.
With this 3D distribution of petrophysical properties, new drilling
locations can be identified with more confidence, as shown in Fig. 4.
From
this study, the following are concluded.
Seismic data can be inverted into petrophysical data if the seismic data
are calibrated with the petrophysical data with correct time depth conversion.
Time shift (Ts) for the time depth curve varies with depth and an
automatic Ts scanning technique has been developed and presented.
Using suitable geostatistics inversion, petrophysical logs ahead of well
path can be predicted.
New well locations can be identified with more confidence with the help of a 3D distribution of petrophysical properties inverted from seismic data.
Kern River Field produces heavy oil from a fluvial depositional environment. Resistivity logs are used to define structure/stratigraphy and sand/shale lithology. There are many wells in the field and steam injection has affected more than half of the resistivity logs. Geological cross sections are used to design perforations and steam injections and so on. Before GridSTAT was used, conventional approaches were used to manually correlate wells, identify faults, and draw cross sections. Since 1991, GridSTAT has been used by geologists to identify bad data, correct the temperature effect in resistivity logs, correlate the wells, identify faults, and generate cross sections and three-dimensional models. It is now possible to produce a cross section in more detail and in much shorter time. A study was done to compare the GridSTAT cross sections produced by one geologist with the hand-drawn cross sections produced by another group of geologists (in much longer time, of course). They found that away from the neighborhood of faults, GridSTAT cross sections are similar to the hand-drawn sections in more than 90% of the time. When GridSTAT was used in the automatic mode to locate the faults, there are more differences. The conclusion was that GridSTAT can be used to generate cross sections without loss of accuracy if user intervention is used when fault intersection at wells is involved.
Temperature surveys are taken to monitor the progress of steam injection in the Kern River Field. GridSTAT has been used to build four-dimensional models of temperature to visualize the heating of the reservoir (by animating the change in temperature over time) and identify where the steam may be leaking to upper layers and where there is oil left (cold region). Structure/stratigraphy information is used from the resistivity lithology model. The lithology model from resistivity is overlaid to the temperature model to confirm that the higher temperature is in the sand and to help concentrate viewer's attention in the sand.
The Mabee field in west Texas is a carbonate reservoir with over 800 wells. There are limited amount of core data from about 80 wells. Because of diagenesis, the correlation between permeability and porosity is poor. Coefficient of correlation is about 0.6. GridSTAT was used to build a three-dimensional model of permeability to help identify infill drilling locations. Three horizon markers are used to define the reservoir stratigraphic frame work. Core permeability is used as hard data. Log porosity is calibrated to core permeability and then used as soft data. Co-kriging is then used to produce the permeability grid. With a permeability cut-off, a permeability-thickness (K-H) map is produced. Infill well locations are picked from the high K-H areas and production rates has been so satisfactory that less number of infill well are required to reach the production target.
When wells are relatively far apart and seismic data are available, GridSTAT can be used to build reservoir models using both well data and seismic data. In an offshore field with a dozen wells and 3D seismic data, a 3D porosity model was built using GridSTAT with details and accuracy that could not have been realized with either well data or seismic data alone. In this case, the porosity logs are used as hard data. A top horizon and a base horizon were picked in the seismic and in the wells. The seismic amplitudes were converted to impedance. In GridSTAT, transforms were performed to convert impedance to pseudo porosity and calibrated to the porosity logs. The second step in GridSTAT is to fine-tune the seismic depth, because the uncertainty in seismic depth is about 50 ft which is not good enough for porosity modeling. Then co-kriging is used to generate the 3D porosity grid with transformed and adjusted seismic data as soft data, after transforming to stratigraphic coordinates. The grid is then exported to eclipse format for reservoir simulation after depth-shifting back to the original depth.
A heterogeneous tight gas reservoir in Colorado was evaluated with integrated approach using old and new well logs, geological interpretation, initial production rates, core porosity and permeability, capillary pressure, and geostatistical modeling techniques.
Old logs (GR and Neutron) were corrected for casing and normalized. Vsh and porosity were calculated from GR and Density respectively. Sw was calculated with resistivity and porosity using a shaly sand formula. Sw was also calculated from capillary pressure and matched the resistivity based Sw. The free water contact determined is consistent with observations from well production. 3D model of Vsh, Por, and Sw were built. For wells with neutron log only, porosity was calculated from neutron log with 3D model based Sw for gas correction. The model was validated by removing some wells from the model then comparing the model predictions with actual logs. Further validation was done by comparing the geological model prediction with production data. Careful study of the cutoff values for Vsh, Por and Sw resulted in a more reliable gas in place. Recovery factor was studied with reservoir engineering methods combining material balance and fluid flow calculation in heterogeneous reservoir. The gas in place and recovery factor were compared with production data analysis including decline analysis.
The results of the study provided support for increasing the reserve of the reservoir, as production also indicated the original underbooking of the reserve. The geological model also helps designing infill well location.
|
Send Email or Comments |