PaperID AU327
Author Swati Goel , Oil and Natural Gas Corporation , India
Co-Authors Deepak Kapoor, Sangeeta, P S Tomar, H K Rawat
Permeability prediction from well logs using clustering methods
and its application in designing of Flow zone indicators.-A case
study in Cambay Basin.
Abstract
Accurate prediction of permeability is an important factor for the determination of oil and gas reservoir
quality and its flowing capability. It is difficult to visualise solutions to many reservoir engineering
problems without having accurate permeability value. One such aspect is producibility which can be
addressed by flow zone indicator. Flow zone indicator is parameter group which predicts the
producibility of a reservoir on the basis of its porosity and permeability. Though porosity data is
available for every well, not all wells have permeability data recorded or computed. The conventional
methods for permeability determination are core analysis and well test techniques which are
expensive and time consuming. This paper aims to predict permeability of reservoirs utilizing
clustering method i.e. multi-resolution graph-based clustering (MRGC). The results were validated
with wells having permeability data and a very good match was obtained. The paper shows that
utilizing this methodology permeability can be predicted for wells even where no permeability data is
available. Based on this predicted permeability, flow zone indicator for various reservoirs were
constructed to assess their producibility. Heterogeneous reservoirs of Kalol formation in Cambay
basin have been studied for this paper. The results obtained were quite encouraging and matched
well with the production testing data of these wells.
Introduction
Oil and gas well completion relies heavily on the right porosity and permeability to ensure the oil and
gas can be extracted. Porosity and permeability are the reservoir rock most significant physical
properties. Porosity and permeability are geometric properties and are the result of its lithological,
structural and compositional behaviour (composition). However, for assessing the producibility, of a
reservoir rock, the most fundamental reservoir rock property is its permeability. Though porosity data
is available for every well, not all wells have permeability data recorded or computed. The
conventional methods for permeability determination are core analysis and well test techniques which
are expensive and time consuming. This paper aims to predict permeability of reservoirs utilizing
clustering method i.e. multi-resolution graph-based clustering (MRGC). The results were validated
with wells having permeability data and a very good match was obtained. The paper shows that
utilizing this methodology permeability can be predicted for wells even where no permeability data is
available. Based on this predicted permeability, flow zone indicator for various reservoirs were
constructed to assess their producibility. Heterogeneous reservoirs of Kalol formation in Cambay
basin have been studied for this paper.
Methodology
The study has been carried out in two parts. In part A permeability has been predicted utilizing
clustering method - multi-resolution graph-based clustering (MRGC). The results were validated with
wells having permeability data and a very good match was obtained. Once the methodology was
validated, the technique to predict permeability was extended to other wells where no permeability
data was available. In part B of this study, the predicted permeability along with porosity has been
used for the construction of flow zone indicator charts to assess the producibility of heterogeneous
reservoirs in Kalol formation. Study was taken up for 10 wells in this paper and the results are quiet
encouraging and match well with the initial production testing data. The methodology in two parts is
discussed below.
A. Prediction of permeability
Permeability has been predicted using the MRGC clustering technique. MRGC is based on non-
parametric K-nearest neighbour and graph data representation. MRGC is a tool which analyses the
structure of the complex data and partition natural data groups into different shapes, sizes, and
densities. MRGC automatically determines the optimal number of clusters. Clusters produced by
MRGC are organized in a hierarchical manner so that the high order clusters are the subcluster of low
order clusters. Consequently, depending on the needed resolution, we can select an appropriate
number of clusters. In the lowest order, 14 clusters and in the highest order, 26 clusters were
partitioned. Then, each model was used to predict permeability. The predicted permeability on the
basis of each model was compared with NMR log derived permeability. The model with the maximum
value of R2 was selected as the optimal number of clusters for MRGC model. 10 wells have been
taken up for the studies, out of these two wells were having NMR derived permeability data (KSDR).
Conventional log suit consisting of RHOB, NPHI, Deep resistivity and GR were used and MRGC
method was used for synthetic Permeability estimations. Estimated permeability was then validated
with recorded KSDR from the NMR log, and the constructed MRGC model was used for the
permeability prediction (KSDR_PRED) in the test wells.
Ye and Rabiller proposed multi-dimensional dot-pattern recognition (MRGC) as a new clustering
method for electro-facies analyses. The minimum number of clusters (14 clusters) is optimal for using
in the prediction (Figure 1). MRGC model with 14 clusters was used to predict permeability in blind
test well and a regression coeff. Of 0.89 was observed.
Fig-1- Regression analysis of Well-A between KSDR and K_PRED
.
Fig-2: shows the overlap of predicted and NMR derived permeability
B. Construction of Flow Zone Indicators
Taking predicted permeability for 10 wells Flow Zone indicators have been constructed and its utility
in predicting the producibility of the reservoirs has been analysed.
Flow Zone Indicators
The hydraulic flow unit is defined as the representative elementary volume of total reservoir rock
within which geological and petrophysical properties that control fluid flow are internally consistent
and predictably different from properties of other rocks. (Amaefule et al, 1993). Samples that have
same FZI will be classified into the same Hydraulic Flow Unit (HFU) or rock type. Different authors
have proposed various FZI charts. All these charts have been used for the construction of Flow zone
indicators and their analysis. These charts are basically Permeability vs Porosity plots. The
construction of the charts and their analysis are discussed in brief below:
i. Winland R 35
Winland developed an empirical equation relating porosity, permeability and pore aperture based on
mercury injection capillary pressure data. Winland developed the following equation:
Log is to base 10. “r” is the modal pore throat radius in microns. Because the process is based on
mercury injection which requires all water to be removed (including clay bound water), it is possibly
implied that the input porosity is total porosity PHIT.
ii. Pittman
Pittman modified the equation to
iii. RQI vs. PHIZ - FZI
These are Rock Quality Index vs inter-particle porosity of PHIZ plots overlaid with flow zone indicator
charts. Amaefule (1993) defined RQI (Rock Quality Index) and FZI (Flow Zone Indicator), from
theoretical equations involving Carmen Kozeny. RQI is defined as:
(Eq. 2)
where is in md and is a fraction
For the flow zone indicator FZI the relationship is:
where Log is to base 10
PHIZ is the pore volume to grain volume ratio or normalized porosity:
On a crossplot of RQI versus PHIZ (both scales logarithmic), FZI numbers are represented by straight
lines.
iv. FZI
These are permeability vs porosity plot overlaid with FZI chart lines.
v. KPHI
These are permeability vs porosity plots overlaid with KPHI chart lines.
Fundamental to Amaefule’s work is the ratio of . This is also present in the original work
documented by Leverett in 1943 on the J function3. Of interest, is the similarity in shape of the chart
lines between the plot with Winland and Pittman, even though is from fundamental
theory and Winland/Pittman are from empirical relationships.
vi. Lucia Rock Class
These are PHIZ vs permeability plot overlaid with Lucia rock class charts.
Lucia working in carbonates during the 1990s discovered that on a Log-Log porosity permeability
crossplot, that coarser grains and fabrics were represented more in the north western region of the
plot and conversely finer grains and fabrics in the south eastern area of the crossplot. Although his
work was on carbonates, the general principles apply to clastics as well. The coarser grained rocks
have a lower rock fabric number (RFN) and finer grained rocks a higher RFN. The lines correspond
to rocks of similar grain size not pore throat radii. Lucia’s subdivision often relates more directly to
core sedimentary facies than the other reservoir rock types.
NoteThe chart lines curve on a porosity permeability Linear Log plot.
Lucia developed the following global permeability transform:
(Eq. 5)
where:
is permeability in md
Log is to base 10
A = 9.7982
B = 12.0838
C = 8.6711
D = 8.2965
= PHIE (clay bound water removed, but capillary bound water in fines remaining)
Lucia defined rock classes according to the RFN:
Class 1 – RFN <1.5
Class 2 – RFN >=1.5 < 2.5
Class 3 - RFN >=2.5 < 4.0
Class 4 - RFN >=4.0
Discussion
Fig.- 3 shows different flow zone unit charts namely- Winland R35, Amaefule-FZI, LeverettK/PHI,
Pittman R35, Amaefule-FZI (logarithmic) , Lucia RFN (Rock Fabric Number). For the validation
purpose Flow Zone Indicators were constructed using log recorded permeability as well as the
predicted permeability and it is observed that the flow zone units matches well in case of predicted
permeability and recorded permeability. As shown below:
Fig-3 (a) FZI from recorded permeability from NMR log Fig-3 (b) FZI from Predicted permeability
Using the above mentioned approach predicted permeability was used in a blind test Well-B, Flow
zone units are depicted in (fig-4). In (fig-4) highlighted green color shows the tested zone in Well B
which is a part of Kalol formation. The above mentioned interval was recommended for testing. On
testing it was found to be hydrocarbon bearing but required Hydrofracturing job because of the poor
producibility. And our study as well shows that the K_PRED values are also less. Flow zone unit
indicators also shows the poorest flow zone. Variation and frequency of defined flow units have been
used to achieve a satisfactory understanding of the level of heterogeneity and the possible effect on
well production.
Fig-4: Results of permeability prediction and flow zone indicators in a test Well-B
Conclusion
The paper shows that utilizing this methodology, permeability can be predicted for wells even where
no permeability data is available. Based on this predicted permeability, flow zone indicator for
formation reservoirs can be constructed to assess their producibility. Study was specifically carried
out for heterogeneous reservoirs of Kalol formation of ten wells in Cambay basin. The predicted
permeability along with flow zone indicator for different reservoirs could predict their producibility. The
results match well with the initial production testing data for these wells.
References
Amaefule,J.O., Altunbay, M., Tiab, D., Kersey, D.G., and Keelan,D.K. 1993.“Enhanced Reservoir
Description: Using Core and Log Data to Identify Hydraulic (Units) and Predict Permeability in
Uncored Intervals/Wells.”, Paper SPE 26436, presented at SPE Annual Technical Conference and
Exhibition, Houston, Texas, 3-6 October.
F. Khoshbakht* and M. Mohammadnia “Assessment of Clustering Methods for Predicting
Permeability
in a Heterogeneous Carbonate Reservoir”
Leverett,M.C., 1941. "Capillary behaviour in porous solids". Transactions of the AIME (142): 159–172.
Lucia,F.J., 1999, “Carbonate reservoir characterization.” Springer-Verlag, Berlin
Pittman,E.D., 1992, “Relationship of porosity and permeability to various parameters derived from
mercury injection-capillary pressure curves for sandstone”: APPG Bulletin, v.76, p.191-198.
S.-J. Ye, P. Rabiller, "A New Tool for electrofacies Analysis: Multi-Resolution Graph-Based
Clustering", SPWLA, 41st Annual Logging Symposium Transaction, paper PP.
Acknowledgements
The authors are thankful to ONGC management for granting permission to publish this paper. The
authors are grateful to ED-HOI-KDMIPE, Dr. Harilal for his encouragement. The views expressed in
this paper are of the authors only and may not necessarily be of the organization to which they
belong.
