Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

IFEDC2015482

A NEW ALGORITHM TO QUANTIFY WATER CONING


IN HYDRAULICALLY FRACTURED TIGHT OIL
RESERVOIRS

2015IFEDC
Shams Kalam1, Sami Alnuaim2, Muzammil Hussain Rammay3
King Fahd University Of Petroleum and Minerals, Dhahran, Saudi Arabia
shamskalam@hotmail.com, neaimsa@yahoo.com, hussain_petro@yahoo.com,

Keywords: Numerical simulation, Correlation, Water  Corrosion problem due to corrosive water and the
coning problem, hydraulically fractured reservoirs, tight additional cost of its disposal.
oil reservoirs.  High probability of early abandonment of the
affected well.
Abstract
This paper shows a general mathematical correlation and
procedure to predict breakthrough time of water in the
hydraulic fractured tight oil reservoirs obtained through
numerical simulation modelling.
The research work exhibits a 3D simulation model which
is run for different ranges of fracture conductivity, kf*w
(md-ft), reservoir permeability, k (mD), anisotropy ratio,
kz/kx, density difference, ρw-ρo (lb/ft3), water oil contact,
WOC (ft), fracture length, Lf (ft), fracture height, h f (ft)
and oil viscosity, μo (cp) to find time to breakthrough tB.
More than 20,000 data points are obtained by around
20,000 number of simulation runs by coupling the
commercial reservoir simulator with MATLAB. Figure 1: Geometry of Vertical Hydraulic Fracture [1]
Multivariate regression technique is employed to obtain
generalized correlation for time to breakthrough in Fig. 1 shows the typical idealistic geometry of the vertical
hydraulically fractured vertical well in tight oil reservoirs fracture. Problem of water coning becomes more critical
(TOR). in the hydraulically fractured vertical well. So, there is a
The algorithm is obtained by applying multivariate need to make a coning correlation for hydraulically
regression technique with mean absolute percentage error fractured vertical wells.
less than 5 percent. Results show the good agreement with Different correlations for water coning have been
the unseen data. Correlation parameters are transformed to suggested by several authors. Meyer and Garder [2]
dimensionless form in order to get generalized correlation. proposed that coning problem is due to radial flow of the
It is observed that breakthrough time highly depends on oil and associated pressure sink near the wellbore. They
fracture height and dimensionless fracture conductivity in assumed a homogeneous reservoir with a uniform
the hydraulic fractured vertical tight oil well. permeability throughout the reservoir, i.e., kv = kh. For the
case of vertical oil wells a potentiometric model was used
1 Introduction to predict the water coning behaviour by Chierici and
Ciucci [3]. Their work shows dimensionless graphs which
captures the vertical and horizontal permeability
Water coning is a production problem in which bottom-
behaviour. Hoyland, Papatzacos, and Skjaeveland [4]
water penetrates into the perforation region in the near-
wellbore zone and decreases oil production. Some presented two methods (analytical and numerical solution)
for forecasting critical oil rate for bottom water coning in
common water coning problems are as follows:
the case of anisotropic and homogeneous sytem. In the
numerical simulation method, they ran a large number of
 Additional cost due to extra water production.
simulations. Based on those runs with more than 50
 Depletion mechanism efficiency will be reduced. critical rate values, the authors employed a regression
technique to build the relationships for isotropic and
anisotropic reservoirs. Chaperson [5] suggested a simple
relationship to predict the critical rate of a vertical well in
an anisotropic formation (kv ≠ kh). Schols developed an
empirical correlation based on results acquired from
laboratory experiments and numerical simulator.
Two of the most widely used correlations in coning study
are the Sobocinski-Cornelius Method [6] and the
Bournazel-Jeanson Method [7]. In the Sobocinski-
Cornelius Method, the authors correlated the breakthrough
time with dimensionless cone height and breakthrough Figure 3. Front View of the Simulation Model
time. They develop the correlation by using laboratory
data and modeling results. Based on experimental data,
Bournazel and Jeanson developed a methodology that uses
the same dimensionless groups proposed in the
Sobocinski-Cornelius method [5]. The only difference is
in finding the dimensionless time to breakthrough.

2 Model Description
This paper shows a three dimensional simulation model
which is made of 15 by 15 grids in X and Y directions
respectively, each with dimensions of 200ft. The reservoir Figure 4. Fracture in the Simulation Model
is layered into 25 layers of thickness 10ft. The Local Grid
Refinement (LGR) technique is used to simulate the
fracture. Single vertical well is placed in the reservoir 3 Methodology
penetrating in all the layers. Table 1 below shows the
reservoir properties used in this simulation model. The 3D simulation model was run for different ranges of
fracture conductivity, kf*w (md-ft), reservoir permeability,
Table 1. Reservoir Properties k (mD), anisotropy ratio, kz/kx, density difference, ρw-ρo
(lb/ft3), water oil contact, WOC (ft), fracture length, Lf (ft),
Parameters Values fracture height, h f (ft) and oil viscosity, μo (cp) to find
time to breakthrough (tB). More than 20000 data points
Reservoir Pressure 4500 psi
were obtained by integrating the commercial reservoir
Porosity 21.0 % simulator with MATLAB. Multivariate regression
Depth 8000 ft technique was employed to obtain generalized correlation
Bo 1.24 res bbl/STB for the time to breakthrough in a hydraulically fractured
vertical well in tight oil reservoirs.
Oil Viscosity 1.14 cp
The ranges of parameters varied for the different
Figures 2 and 3 show the top and front view of our simulation runs are as follows:
simulation model. While figure 4 shows the fracture in the
simulation model. Table 2. Ranges of Variables
Property/Parameter Ranges
Fracture Conductivity, kf*w, md-ft 10 to 10000
Reservoir Permeability, k, md 0.001 to1
Anisotropy ratio, kv/kx 0.1 to 9
Oil Density, ρo, lb/ft 3
44 to 58.5
Water Oil Contact, WOC, ft 8050 to 8090
Fracture Half length, Lf, ft 100 to 1500
Fracture Height, hf, ft 40 to 70
Oil Viscosity, μo, cp
3.52 to 0.96

Figure 2. Top View of the Simulation Model 4 Simulation Analysis


Different sensitivity runs were conducted for WOC,
fracture half length, fracture height and viscosity. Several
graphs were plotted between breakthrough time Vs k and Figure 7: Graph between tB and kf*w at different h f
kf*w at different viscosities. tB vs kf*w is also plotted at
different fracture height. 5 Algorithm (Correlation)
Fig. 5 shows the relationship between tB and k at different
viscosities which shows that higher the k, lesser will be By using the reservoir parameters, the dimensionless oil
the tB. Also tB vs kf*w is plotted at different viscosities rate “qd” can be found out which will be used in
(Fig. 6) which shows the similar behavior as that of tB vs k. computing the breakthrough time.
From the viscosity sensitivities, we can see that higher
viscosity of oil gives higher breakthrough time. k*h*ΔP
Fig. 7 shows the relationship between tB vs kf*w at qd  (01)
different hf (fracture height) which illustrates the behavior 141.2*B*μ
of fracture height i.e. greater fracture height leads to early
breakthrough of water. Where k is the reservoir permeability in mD, h is the
reservoir thickness in ft, ΔP drawdown pressure in psi, B
and μ are formation volume factor (res bbl/STB) and
viscosity of oil (cp) respectively. Time to breakthrough
‘tB’ can be calculated from td (Dimensionless time) by
using the following formula.

0.00634*k*t B
td  (02)
*μ*c t *rw2

Where,

k = reservoir permeability, mD
tB = breakthrough time, day
ct = total compressibility, psi-1
μ = oil viscosity, cp
Figure 5: Graph between tB Vs k at different Viscosity φ = formation porosity, fraction
rw= well radius, ft

td can be estimated by using the following new correlation.

t  kv hf
log  d   0.1083 +0.0007Lf  0.3272
 qd  kx h well

hf
 4.4827  0.423ln  FCD  1.0107log  k f w 
ho
0.0005  w  o   8.8565 (03)

Equation (03) can be arranged as:


Figure 6: Graph between tB and kf*w at different oil
viscosities
kv  0.3272 4.4827 
log  t d   8.8565  0.1083 +0.0007Lf  h f   
kx  h well ho 

 q * FCD0.974169 
 log  d   0.0005 (04)
 1.0107
 FC 

Where,

qd = (k*h*ΔP)/(141.2*B*μ)
FCD = (kf*w)/(k*Lf) = Dimensionless Fracture
Conductivity
FC = kf*w = Fracture Conductivity, mD-ft
kv/kx = Anisotropy ratio
Lf = Fracture Half Length, ft
hf/hwell = Ratio of fracture height to the well height in the
reservoir
hf/ho = Ratio of fracture height to the oil zone thickness
 = ρw - ρo = Density difference of water and oil, lb/ft3

6 Results
70 % of the simulated data points were used to develop
the correlation and rest of them were used to validate and
test the proposed algorithm. Figures 8 to 10 shows scatter
plot for seen, unseen and total data points showing R2
greater than 90% in all cases.
Table 3 shows MAPE, APE, SD and R2 for training (seen), Figure 10: Scatter plot for seen data
testing (unseen) and all data points which confirm that our
new mathematical model is a strong positive correlation. Table 3. Accuracy results with seen, unseen and all data
Mean absolute percentage error, average percentage error,
standard deviation and R2 using unseen data is 4.5886,
0.2869, 6.4975, and 0.9191 respectively representing good Mean Average Standard R2
agreement with the testing data. Absolute Error Deviation
Percentage (%)
Error (%)
Seen 4.5966 0.5008 6.4912 0.9223
Data
Unseen 4.5886 0.2869 6.4975 0.9191
Data
All 4.4519 -0.5208 6.3759 0.9213
Data

7 Conclusion

 A new empirical correlation for the breakthrough


time is developed by using numerical simulation
and multivariate regression technique.
Figure 8: Scatter plot for seen data  Since the new proposed correlation is
dimensionless, so it can be applied for all ranges
of the parameters used in it, i.e. it is general.
 The proposed correlation will be very helpful
since it eliminates the cost of running simulation
in the hydraulic fractured tight oil reservoirs for
the water coning problem.
 Breakthrough time of water in the hydraulically
fractured tight oil reservoir highly depends on
fracture parameters.

Acknowledgments
We would like to express our special gratitude and thanks
to KFUPM for providing us the facilities and sponsoring
this research study.

Figure 9: Scatter plot for unseen data


References
[1] Reservoir Stimulation Second Edition by: Michael. J.
Economides Kenneth G. Nolte. Figure 12-14 Geometry of
ideal hydraulic fracture
[2] Meyer, H. I., and Garder, A. O., “Mechanics of Two
Immiscible Fluids in Porous Media,” J. Applied Phys.,
Nov. 1954, No. 11, p. 25.
[3] Chierici, G. L., Ciucci, G. M., and Pizzi, G., “A
Systematic Study of Gas and Water Coning by
Potentiometric Models,” JPT, Aug. 1964, p. 923.
[4] Critical Rate for Water Coning: Correlation and
Analytical Solution Lelt A. Hoyland, SPE, Statoil, and
Paul Papatzacos, SPE, and Svein M. Skjaeveland, SPE,
Rogaland U
[5] Theoretical Study of Coning Toward Horizontal and
Vertical in Anisotropic Formations: Subcritical and
Critical Rates by 1. Chaperon, TOTAL-CFP, SPE 15377
[6] A Correlation for Predicting Water Coning Time by:
D. P. Sobocinski
[7] Fast Water-Coning Evaluation Method by: Bernard
Jeanson, SPE 3628

Appendix:

Mean Absolute Percentage Error (MAPE) is a statistical


tool to determine the accuracy of data. It is defined
mathematically as:

100 n Actuali  Pr edictedi


MAPE  
n i 1 Actuali

Average Percentage Error (APE) is another tool to find the


accuracy of data and is defined mathematically as:

100 n  Actuali  Predictedi 


APE  
n i 1 

Actuali 

Standard Deviation (SD) quantifies the deviation of data


from its mean value. It is defined mathematically as
follows:

1 n
SD   (Xi  X)2
n i 1
Where,

X = Absolute Percent Error for data point ‘i’


X = Mean Absolute Percentage Error (MAPE)
n = Number of data points

You might also like