You are on page 1of 5

Estimation of Petro-physical Parameters based on Interpolation

between Cross-sectional Well Logs


Taimur Hassan
GeoGraphix R&D
LMKR (Pvt.) Limited
Islamabad, Pakistan
taimoorhassan@lmkr.com

Bilal Hassan
Department of Electrical Engineering
College of EME, NUST
Rawalpindi, Pakistan
bilz@live.com

organization. If a drilled well is not producible it


costs a lot of financial loss for an organization.
Many techniques have been published for
estimating permeability and porosity from well log
data. Akhilesh K. Verma et al. presented an
implementation that measures permeability and
porosity of the formation using artificial neural
network on well log data [1]. Zhou et al. presented
a technique which uses four layer feed forward
neural network and linear regression to estimate
porosity in quantitative log analysis [2]. Leila
Aliouane et al. presented estimation of petrophysical parameters using multi-layer perceptron
and radial basis function [3]. Interpolation
algorithms such as triangulation, minimum
curvature etc. is used to draw geological surfaces
[4]. LMKR log analysis software also provides
interpolation techniques in cross sectional well logs
to identify lithology between logs [5]. However in
we present an implementation which is based on
interpolating two cross sectional well log data to
estimate a new log data in between them and then
to calculate formation parameters such as: water
saturation, porosity and permeability in an
undrilled zone from the estimated well log data.

Abstract:
This paper presents a quick, resource efficient
and computer simulation based implementation
for the estimation of petro-physical parameters
between two nearby wells by using four
interpolation algorithms i.e. Ordinary Kriging,
Nearest Neighbor, Moving Average and Weighted
Average on cross sectional well logs. Many
researchers have presented their work on
estimating formation parameters based on
existing well logs using multilayer perceptron
and radial basis function etc. But those
estimated parameters are calculated from the
existing logs of drilled formation. However we
used interpolation algorithms between two well
logs (in cross sectional view) to estimate new
well log data in between wells and then
formation parameters are determined using the
estimated well log data. The results of all four
algorithms are computed and compared with
each other to find average estimation error for
estimated formation parameters. This paper
discusses an effective, less time consuming and
resource efficient estimation of a formation logs
to indicate a producible formation by
calculating petro-physical parameters. This
implementation was developed on Microsoft
Visual Studio 2012 using C#.
Keywords: Ordinary Kriging, Nearest Neighbor,
Moving Average, Weighted Average Interpolation,
Microsoft Visual Studio 2012 and C#.

The paper is organized in such a way that literature


review and introduction are displayed in section I.
Section II deals with the proposed implementation
and comparison of all four interpolation algorithms
is discussed in section III. Conclusions and
references are described in section IV and section
V respectively.

INTRODUCTION

PROPOSED IMPLEMENTATION

Today, as most of the oil in the world has been


discovered, advanced techniques and methods are
used to discover oil and gas which was hidden by
legacy tools. To obtain a producible reservoir in
formation, a borehole is drilled. Drilling a borehole
for acquiring oil and gas is an expensive operation
for
exploration
and
production
E&P

As stated above, we proposed an estimation of


formation log data in an undrilled zone based on
the interpolation on cross sectional well logs of
nearby boreholes and then using those logs we
calculated porosity, permeability and water
saturation. We used four interpolation algorithms
for this purpose, compared their results and

c
978-1-4799-3486-7/14/$31.002014
IEEE

437

calculated an average error in estimation for each


formation parameter. The interpolation algorithms
that we used are: Ordinary Kriging, Nearest
Neighbor, Moving Average and Weighted Moving
Average interpolation. The flow of our proposed
implementation is shown in Figure 1.

Where:
is the sill
is the Euclidean distance
is the range

Weights are computed as:




Where

Figure 1: Flowchart of our proposed


implementation
The well logs stored in Log ASCII Standard LAS
format are read through a LAS reader. Then these
logs are cross correlated to remove dip effect for
analysis purposes as shown in Figure 2.

Figure 2: Correlating Res-D of Well ID: 49-00530999 and Well ID: 49-005-31631
Then one of the interpolation algorithms are
applied on selected curves of both well logs to
estimate a new curve in between them. Formation
parameters such as water saturation, porosity and
permeability are then calculated after estimating
resistivity and sonic logs respectively through four
interpolation algorithms which are:

The algorithm to compute kriging interpolation


points is as follows:

Kriging interpolation is applied on all well log


curves between well: 49-005-30999 and 49-00531631 and the results are shown in Figure 3, 4, 5
and 6.

Figure 3: Ordinary kriging interpolation on sonic


(Delta-T) logs.

Ordinary kriging Interpolation:


Ordinary kriging interpolation was been used in our
implementation to estimate formation parameters.
We used rational quadratic model to compute semi
variogram values which is described by equation
below [7]:

438

2014 First International Conference on Networks & Soft Computing

Figure 4: Ordinary kriging interrpolation on


resistivity (Res-D) loggs.

Figure 8: Nearest Neighbbor interpolation on


resistivity logs

Figure 5: Ordinary kriging interrpolation on


spontaneous potential loogs.

Figure 9: Nearest Neighbbor interpolation on


spontaneous pottential logs

Figure 6: Ordinary kriging interpolaation on gamma


ray logs.
Nearest Neighbor Interpolation:
Nearest neighbor interpolation is used
u
to estimate
well curves by computing Eucllidean distance
between both input curves and chooosing minimum
distance point. The distances aree computed by
Euclidean distance formula which iss:

The results of nearest neighbor interpolation


i
on
well curves are shown in Figure 7, 8,
8 9, 10.

Figure 10: Nearest Neighbbor interpolation on


gamma rayy logs.
Moving Average Interpolattion
Moving Average interpolaation is used in our
implementation to estimate well log data between
two input well curves. Thiis algorithm works by
computing average with eqqual weights between
well logs. The interpolated point is calculated as:

The results of moving averagge interpolation are


shown in Figure 11, 12, 13, 14.
1

Figure 7: Nearest Neighbor interpoolation on sonic


logs

Figure 11: Moving average interpolation on sonic


logs

Figure 12: Moving averaage interpolation on


resistivity logs

2014 First International Conference on Networks & Soft Computing

439

Figure 13: Moving average interrpolation on


spontaneous potential loogs

Figure 17: Weighted movingg average interpolation


on spontaneous pootential logs

Figure 14: Moving average interpolation on gamma


ray logs

Figure 18: Weighted movingg average interpolation


on gamma raay logs

Weighted Average Interpolation:

Porosity, Permeability an
nd Water Saturation
Calculation:

We used weighted average interppolation in our


implementation to estimate formattion parameters
based on following formula.

Where has 70% weighhtage in our


implementation. The results of weeighted average
interpolation are shown in Figure 155, 16, 17, 18:

Estimated well log data (Dellta-T and Res-D curve)


produced by interpolation allgorithm in our design
is used to calculate porossity, permeability and
water saturation.
Porosity in our design is calculated
c
using sonic
logs [8] i.e.

Permeability is measure by using Timur empirical


model [9] i.e.

Water Saturation is calculated using Archie


equation [10] i.e.
Figure 15: Weighted moving averagge interpolation
on sonic logs

COMPARIISON

Figure 16: Weighted moving averagge interpolation


on resistivity logs

440

The formation parameters measured through all


four interpolation algorithm
ms are compared with
each other to estimate an error in their results.
Table 1 shows the comparisoon of each interpolation
algorithms indicating averagge porosity estimation
error.

2014 First International Conference on Networks & Soft Computing

Ordinary
Kriging
0.000

Kriging
NN
MA
WA

Nearest
Neighbor
0.037

Moving
Average
0.018

0.037
0.000
0.031
0.018
0.031
0.000
0.014
0.041
0.020
Table 1: Porosity estimation error

Weighted
Average
0.014
0.041
0.020
0.000

REFERENCE

Table 2 shows the comparison of each interpolation


algorithms indicating average permeability
estimation error.

Kriging
NN
MA
WA

Kriging

NN

0.000

0.001

Moving
Averag
e
0.001

Weighted
Average
0.000

0.001
0.000
0.001
0.001
0.001
0.001
0.001
0.000
0.000
0.000
0.000
0.000
Table 2: Permeability estimation error

Table 3 shows the comparison of each interpolation


algorithms indicating average water saturation
Ordinary
Kriging
0.000

NN

MA

WA

0.263

0.070

0.060

0.263

0.000

0.193

0.014

MA

0.070

0.193

0.000

0.014

WA

0.060

0.014

0.014

0.000

Ordinary
Kriging
NN

these algorithms can be used to find multiple well


log data between two wells to generate a complete
calculation of porosity, permeability and water
saturation in an undrilled formation.

[1] Akhilesh K. Verma, Burns A. Cheadle,


Aurobinda Routray, William K. Mohanty and Lalu
Mansinha,Porosity and Permeability Estimation
using Neural Network Approach from Well Log
Data, GeoConvention 2012: Vision
[2] Zhou, Cheng-Dang, Wu, Xi-Ling, Jianghan
Petroleum Inst., Neural Network-Based Formation
Parameters Estimation from Well Logs in
Quantitative Log Analysis: A Comparative Study,
SPE Asia Pacific Oil and Gas Conference, 8-10
February 1993, Singapore.
[3] Leila Aliouane, Sid-Ali Ouadfeul, Noureddine
Djarfour and Amar Boudella, Petrophysical
Parameter Estimation from Well-Logs Data Using
Multilayer Perceptron and Radial Basis Function
Neural Networks, T. Huang et al. (Eds.): ICONIP
2012, Part V, LNCS 7667, pp 730-736 2012,
Springer-Verlag Berlin Heidelberg 2012.
[4] LMKR GeoGraphix IsoMap (trademark) utility.

estimation error.

[5] Powerful multi-well Petrophysical analysis


and mapping, LMKR GeoGraphix PRIZM
(trademark) log analysis software and LMKR
XSection utility.

Table 3: Water Saturation estimation error


By looking in Table 1, 2 and 3 we see that one
algorithm is good for estimating one parameter
while its not so good for estimating the other
parameters so it depends on a user for which
parameters they are most interested in and choosing
the correct algorithms for it by seeing the above
table.
CONCLUSION
We present a quick, reliable and less time
consuming implementation requiring minimum
number of resources that can provide a good
estimation of formation parameters in an undrilled
formation between two offset wells. The estimation
is based on interpolating two cross sectional well
logs. The results are computed and compared with
each other and errors in estimation of formation
parameters are shown in Table 1, 2 and 3. In future,

[6] Cressie, N.A.C. 1993, Statistics for spatial


data, Wiley, New York, 900 pp.
[7] Meer, F. D. van der. Introduction to Geostatistics. ITC lecture notes, 72 pp.
[8] Ekwere J. Peters, Advanced Petro-physics.
[9] Shahab Mohaghegh, Bogdan Balan and Samuel
Ameri, Permeability Determination from Well
Log Data, SPE Formation Evaluation, September
1997.
[10] David E. Johnson, Kathrynne E. Pile, Well
Logging in non-technical Language.

2014 First International Conference on Networks & Soft Computing

441

You might also like