Professional Documents
Culture Documents
Reservoir Monitoring and Characterization
Reservoir Monitoring and Characterization
Reservoir Monitoring and Characterization
Reservoir characterization
Reservoir characterization refers to reservoir modeling activities right up to the point when a
simulation model is ready to simulate the flow of fluids. In many Commercial industries,
Softwares are available which are used in the construction, simulation and analysis of
the reservoir models (Reservoir modeling - Wikipedia). In this piece of work, I will discuss the
different methods used in seismic reservoir characterization. These methods include but not
limited to the following;
1. Characterization using spectral pattern variation (case study, CBM in china coal
seam)
The information for reservoir characterization can be provided by many sources such as
the seismic frequency spectrum (Yanghua Wang). In this method, I present a case study
for the exploitation of seismic frequency spectra pattern variation for the characterization
of potential reservoirs in the CBM fields of Qinshui basin of china.
For the above reason, its quite important to clearly characterize the reservoir by understanding
the spatial extent of the CBM in the coal seam in order to optimize production and reduce the
risk of accidents and loss of lives.
In this method, it is really important to know that there is no straight forward relationship
between CBM content and the strength of seismic profiles. This means that we can use the
quantitative characterization of the CBM spatial variation based on the various types of
amplitude either in the frequency or in the time domain. This method involves the use of three
stages which are;
a) To raise an extra dimension from say a 2D seismic profile in the time-space domain.
From this 2D data, we generate a 3D data in which the third dimension is frequency. This
is now a time-space-frequency domain data and is referred to as a time-frequency
spectrum. This process is normally achieved through the use of matching pursuit
technique instead of the famous Fourier and wavelet transform.
b) Clustering analysis. This involves reduction of dimension from 3 to 2. Meaning that we
are moving from the obtained 3D to 2D data. The resultant image is presented again in
the time-space domain. Remember that this image now has a frequency component.
c) The third and last step is thus characterization which calibrates the indexes of spectrum
pattern with known CBM production and predicts the CBM spatial distribution, the latter
being our main purpose.
Figure 2: (a and b) two seismic profiles W-E 500 and S-N 500 from
the T formation across well H2 (SEG, Geophysics VOL. 77,No.6, page M90).
From the figure 2 (a) which has a CBM layer represented by R15c indicated by a yellow arrow,
upon performing a matching pursuit technique at different frequencies like 20, 30, 40 and 50 Hz,
we come up with time-frequency spectra at each time-space position as bellow;
Figure 3: The time-frequency spectra at frequencies 20, 30, 40 and 50Hz (SEG, Geophysics
VOL. 77, No.6, page M91).
It can easily be observed that from reflection R4-R5 in red, the amplitudes of different
frequencies have good lateral continuity. However the strength of the coal seam reflection R15c
immediately above R4-R5 is much weaker. This makes it very difficult for us to base on the
amplitudes of the frequency domain data for direct analysis of the lateral heterogeneity of the
coal bed. In other words, the amplitudes from the R15c are hidden from these frequency domain
data. Since we cannot rely of amplitude, we then proceed to the next step which is clustering
analysis as discussed below.
These representative models are weight vectors that are listed in order in 2D space. The
purpose of doing this is to compress the field data in a computer working space,
something I would basically refer to as structuring of data in a software system. This
algorithm works on the principle that similar models are close to each other while non
similar models sit apart.
When running the hierarchical clustering algorithm, (Yanghua Wang, SEG, Geophysics
VOL. 77, No.6, page M93) for m neurons vectors, it constructs a m x m distance matrix
and then clusters these vectors based on distances: defining each vector as an individual
cluster, finding pair of vectors close to each other and putting them into a single cluster,
calculating the distance between this new cluster and all other cluster, repeating this
procedure and updating iteratively. Its important to note that this process is irreversible
in that once a vector is classified in to a cluster; it cannot be reclassified in to another.
Picking reflection times of a reference event (underneath the target coalbed) and
flattening the reference event, in this case the reference event time is picked from
those of R4-R5 as indicated in the figure below;
Figure 6: (left), reference event for coalbed 15 and (right), reference event for coalbed 15
with a CBM at R15c before flattening (SEG, Geophysics VOL. 77, No.6, page M94).
Figure 7: (left), reference event for coalbed 15 and( right), reference event for coalbed 15
with a CBM at R15c after flattening (SEG, Geophysics VOL. 77, No.6, page M94).
Calculating the time-frequency spectrum of each sampling point over the entire
time-space window.
Clustering analysis, as earlier discussed in the previous part of this work
Figure 8: (left), a cluster analysis for coalbed 15 and( right), cluster analysis for coalbed
15 with a CBM at R15c after flattening (SEG, Geophysics VOL. 77, No.6, page M94).
Scaling all available CBM information in the study are in to eight levels,
Finding an integer value for each clustered spectral variation pattern, so as that the
resultant potential index matches these five level indexes in an optimal sense. The
potential index is the sum of pattern values over a time interval for each trace.
Finally, assigning determined integer values to all patterns calculating potential
indices over all samples within the time-space window.
For a matter of comparison, the same analysis is carried out for coalbed 3 which does not
have methane. Finally, we obtain CBM distribution along coal seam 15 and 3, as shown
in figure 10 below;
Figure 10: Spatial distribution of CBM reservoir along coal seam 15 and 3 (SEG,
Geophysics VOL. 77, No.6, page M94).
Conclusion: