IADC/SPE 81634

Reservoir Characterization during UBD: Methodology and Active Tests

Erlend H. Vefring, RF-Rogaland Research, Gerhard Nygaard, Telemark University College, Rolf Johan Lorentzen, Geir
Nvdal and Kjell Kre Fjelde, RF- Rogaland Research
Copyright 2003, IADC/SPE Underbalanced Technology Conference and Exhibition
This paper was prepared for presentation at the IADC/SPE Underbalanced Technology Conference and Exhibition held in Houston, Texas, U.S.A., 2526 March 2003.
This paper was selected for presentation by an IADC/SPE Program Committee following
review of information contained in an abstract submitted by the author(s). Contents of the
paper, as presented, have not been reviewed by the International Association of Drilling Contractors or the Society of Petroleum Engineers and are subject to correction by the author(s).
The material, as presented, does not necessarily reflect any position of the IADC, SPE, their
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper
for commercial purposes without the written consent of the International Association of Drilling
Contractors or the Society of Petroleum Engineers is prohibited. Permission to reproduce in
print is restricted to an abstract of not more than 300 words; illustrations may not be copied.
The abstract must contain conspicuous acknowledgment of where and by whom the paper was
presented. Write Librarian, SPE, P.O. Box 833836, Richardson, TX 75083-3836 U.S.A., fax
01-972-952-9435.

Abstract
In this paper methodologies for reservoir characterization during underbalanced drilling is presented. In these methodologies we are using a transient wellflow model coupled to a transient reservoir model, and use estimation techniques to estimate reservoir properties. Our focus is to estimate the permeability and reservoir pressure along the well, using measured
data usually available while drilling. The measured data are
outlet rates, pump pressure and downhole pressure. The liquid
injection and gas injection rates are used as input to the model.
The methodologies are applied to synthetic cases.
Introduction
Underbalanced drilling is becoming increasingly popular.
During an underbalanced drilling operation the well pressure
should be kept below the formation pressure at all times. Since
the well is producing while drilling, the well may be tested
real time. Estimation of near wellbore characteristics of the
formation gives important information when using smart completions since highly productive zones can be located.
Several recent papers1-4 have addressed well-testing during
underbalanced drilling. Methodologies have been presented in
these papers where the permeability profile in the near wellbore region is estimated based on the assumption that the total
flow rate from the reservoir is known and that the reservoir
pressure is assumed to be constant and known. An alternative
technique1 has been presented where the assumption of known
and constant reservoir pressure is not needed, but known total
flow rate from the reservoir is still needed. The total flow rate
from the reservoir is however not usually measured during an
underbalanced drilling operation. As it is pointed out in Ref. 1
it is not straightforward to determine the total flow rate from
the reservoir on the basis of the surface flow measurements.
Compressibility of the fluids in the system may lead to loading
and unloading in the wellbore, and mass flow rate at the sur-

face is therefore affected by production, injection, and the

change in mass stored in the wellbore as a function of time. It
is therefore a need to develop methodology for reservoir characterization during underbalanced drilling that applies only
data measured during the operations.
Some recent papers5,6 have addressed the challenge of
calibrating a well flow model real time according to measured
data. The advantage of having such a calibrated well flow
model is that reliable predictions for the well conditions (like
bottomhole pressure) can be given at all times. A limitation of
these papers is that no reservoir models have been included.
In a previous paper7 a least square methodology to estimate
near-well reservoir properties was presented. The present paper is an extension of this work. In the present paper the ensemble Kalman filter is introduced as an alternative method to
estimate reservoir properties during underbalanced drilling.
The importance of active tests during the drilling operation is
also investigated.
The outline of the paper is as follows. The well flow model
and the reservoir model are first described. Then two methodology for interpreting the reservoir properties from the measured data are described. The described methodologies are then
applied to synthetic cases.
The description of the well flow model, reservoir model as
well as the estimation methodologies, closely follows the
presentations in Refs. 5-7, but are included here for the convenience of the reader.
Dynamic Well Flow Model
A dynamic model for describing the transient behavior of the
two-phase flow conditions in LHD and UBD operations can
be expressed with basis in the drift-flux formulation of the
two-phase flow conservation laws8. Due to the complexity of
the model, a numerical solution strategy is required.
The numerical scheme solves a set of three conservation
equations, one for the mass of each phase and one for the mixture momentum. The mixture energy equation is not taken into
account. Instead a fixed temperature profile in the well is used,
which can be calculated in advance or provided by the data
acquisition system.
Conservation Equations. The drift-flux formulation of the
conservation equations is given by

( G G ) + ( G G v G ) = m ,............ (1)
t
z

E. H. VEFRING, G. NYGAARD, R.J. LORENTZEN , G. NVDAL AND K. K. FJELDE

( L L ) + ( L L v L ) = m , ....... (2)
t
z

( L L v L + G G v G ) + L L v 2L + G G v G2 + p
t
z

dp
= ( L L + G G ) g sin . ........... (3)
dz F

The transient drift-flux model is a system of non-linear

partial differential equations, which is hyperbolic in an ample
region of physical multiphase parameters9. The model describes the fully transient behavior of both pressure pulse
propagation and mass transport.
Closure relations for flow in drillstring. In the standard
drift-flux approach, the closure of the system is achieved by
specifying density models for each phase and a slip relation
between the phases. Generally the slip relation can be presented in the following form:
v G = C1 ( G v G + L v L ) + C 2 , ............. (4)

where C1 = 1 and C2 = 0 is adopted in the present work. In

addition, it is necessary to provide an appropriate model for
the frictional pressure loss term in the momentum equation. A
frequently used expression for this term is:
2f
dp
v v ,.....(5)
=
dz F

(d 2 d 1 )

The Equations (1)-(3) together with (4) and (5), constitute a

model for the downward two-phase flow in the drillstring.
Closure relations for flow in annulus. Mechanistic models
have become quite popular for describing steady-state twophase flow in producing wells. The mechanistic models provide information about flow patterns, pressure drops, gas
volumetric fractions and phase velocities based on knowledge
of the superficial velocities, densities, viscosities, interfacial
tension and well geometry. This relationship can be expressed
as
calculate
M(d1 , d 2 , L , G , , L , G , v Ls , v Gs )

( G , v G , v L , pressure drop ) . ................ (6)

Mechanistic models have also been developed for twophase flow in annuli10. Furthermore, it is possible to integrate
mechanistic steady-state procedures into fully dynamic twophase flow models to provide the necessary information regarding phase velocities and pressure loss terms 11,12. Recently
these models have been extended to include counter-current
flow13,14, which is essential as backflow can occur in UBD and
LHD operations when gravitational forces exceed pressure
drop in the annulus.
Dynamic reservoir model
The dynamic reservoir model is based on the transient constant terminal rate reservoir model15 and is similar to the reservoir model used in Ref. 1. When drilling, the reservoir inflow is calculated in small segments of 1 meter length. We
model the influx from the reservoir to each segment number j
in zone i through the equation

q i, j (t i, j ) =

IADC/SPE 81634

4 K i s(p res, i p w )
. ...(7)
4K i t i, j
(2S + log(
))
e cr w2

Several consecutively segments where the permeability and

reservoir pressure are equal are referred to as a zone. The reservoir consists of several zones, and we assume that the permeability and reservoir pressure varies in different zones. The
other quantities such as porosity, etc., are kept constant over
the whole reservoir. The reservoir zones are connected to the
well as shown in Fig. 1.
A least square methodology to estimate near-well
reservoir properties
The unknown reservoir properties may be estimated using the
well flow simulator and measurements of pump pressure,
downhole pressure and outlet rates. Using the simulator we
compute values of the observables for a given set of reservoir
properties. The estimation of the inflow profile is done by
searching for reservoir properties that reconcile the measured
data.
To estimate the unknown reservoir properties (permeability and pressure), we make the simplifying assumption that
these properties are constant in n specified zones of the well.
This means that the inflow profile can be expressed by a vector c containing 2n parameters, or by assuming knowledge of
the reservoir pressure for each zone n parameters. Hence, the
simulated values are functions of the same parameters.
The estimation of the reservoir properties is done by finding the vector c = cest such that
2

yi ,obs f i (c)

..(8)

i
i =1

is minimized. In the expression above, yi , obs is the ith measm

f i (c ) is the simulated value for the corresponding

measurement with inflow given by the parameter vector c
and i is the standard deviation of the measurement i .

urement,

To avoid unphysical solutions, certain limitations must be

put on the parameters. These limitations are implemented using linear inequality constraints on the parameter vector, c .
The Levenberg-Marquardt method is applied to compute
the unknown parameters represented by c , which minimizes
the least squares expression (8) given the constraints. The implementation of the Levenberg-Marquardt method is based
upon the description given in Ref. 16 and the handling of constraints upon methods described in Ref. 17.
Error analysis. The measurement errors will give rise to an
uncertainty in the estimated inflow profile. This uncertainty
can be estimated by linearization18,19 of the vector valued func-

tion F (c ) = f i (c )

]im=1

around the estimated point c = cest

(linearized covariance analysis). In the linearized covariance

analysis we assume that the errors are Gaussian and uses a
linearization of F (c ) around the estimated point cest in the
computations.

IADC/SPE 81634

RESERVOIR CHARACTERIZATION DURING UBD: METHODOLOGY AND ACTIVE TESTS

The uncertainty in the estimated parameters is expressed

by the covariance matrix of the estimated parameters

cest which is P = (JT 1J ) .

1

m, n

f
...................................(9)
J= i
c j i =1, j =1
is the sensitivity matrix of the simulated values with respect to
a change in the reservoir properties, and is the covariance
matrix of the measurement errors. We are assuming uncorrelated measurement errors, therefore is a diagonal matrix

i2 . The covariance

matrix, P , depends both on the accuracy of the measurements

and the sensitivity of the simulated values to a change in the
reservoir properties.
The ensemble Kalman filter to estimate near-well
reservoir properties
As an alternative to the least-square methodology for estimating near-well reservoir properties, we evaluate the use of
ensemble Kalman filter with an augmented state vector. The
Kalman filter is more suitable for online estimations, as the
parameter estimates are updated each time new measurements
become available. In the least square approach presented previously, the parameters have to be estimated after all the data
are collected, i.e. after the drilling is completed.
The ensemble Kalman filter technique has been used for
estimating model parameters in a well-flow simulator for underbalanced drilling6 and to estimate the permeability in a
near-well reservoir model20,21. For the convenience of the
reader we will recall basic facts about the ensemble Kalman
filter, and discuss some details on the actual implementation
for this study. This presentation follows closely the presentation given in (Ref. 6).
The ensemble Kalman filter was first introduced in geophysical sciences22 as an alternative to the extended Kalman
filter for large non-linear models. Using the Kalman filter it is
possible to combine the information obtained from the measurements with the model to get an improved estimate of the
state vector of the system. As the state vector we use the unknown model parameters, which are the reservoir pressure and
permeability for each of the reservoir zones the well is penetrating, and discretized values for pressure, water mass flow
rate and oil mass flow rate in the well. The model parameters
are defined as
= (p res,1 , K res,1 , p res,2 , K res,2 ,.... ),..(10)
a vector of length two times the number of penetrated reservoir zones. This gives a state vector on the form
s = ( p i , Q G, i , Q L, i ) , (11)
where i runs through a spatial grid defined by the numerical
method.
Denote the state vector for the jth member of the ensemble after inclusion of the measurement by

taken into account. The jth state vector prior to the inclusion
of the next measurements is
s jf = f(s ja ) + j ,(12)
where f( s ja ) denotes the updating of the state vector done by

Here

where the ith diagonal entry is equal to

sja . Each state vector

is used as initial value to the simulator for a forward simulation that is run to the time when the next measurements are

running forward the simulator to the time were new measurements becomes available, and j is a stochastic contribution
representing the model error. The model error we use is normally distributed with zero mean and covariance matrix .
More details on the specification of will be given below.
To take into account the measurements we use the covariance matrix of the ensemble around the ensemble mean. The
mean value of the ensemble is given by
n
1
s = j =1 s jf ,.(13)
n

and the ensemble covariance matrix is

1
n
n
( s jf s ) ( sif s ) T ,..(14)

i =1 j =1
n -1
where n is the number of members in the ensemble.
To combine the information from the measurements with
the model in a proper way, we need both to know the uncertainty in the current estimate of the state and the uncertainty in
the measurements. We assume that the errors in the measurements are statistically independent, and with known variances.
This gives a covariance matrix for the measurement errors.
For proper use of the filter an ensemble of observations is
needed23. This is defined by
d j = d + j ,..(15)
R=

where d is the actual observation and j is drawn from a normal distribution with zero mean and covariance matrix .
The observation vector d is related to the state vector s
through the equation d = Hs , for an appropriate matrix H. The
state vectors in the ensemble are updated using the gain matrix
G = RHT (HRHT + ) 1 , ...(16)
through the equation
s ja = s jf + G(d j Hs jf ) (17)
A major issue with the ensemble Kalman filter is the size
of the ensemble. The optimal size of the ensemble for our application is a subject for further research. Experience in the
oceanographic science24 has indicated that the filter may function using a size of the ensemble in the range 100 500. We
have chosen to use 100 members in the ensemble. This means
that 100 forward simulations are needed. The size of this
model is small compared to the models used in the oceanographic science, therefore it should be investigated if a reasonable performance of the filter can be achieved with smaller
ensemble size than 100.
It is our experience that proper specification of the covariance matrix for the modeling error is crucial to get good performance of the filter, and we have investigated different
forms of model noise. The covariance matrix for the model
noise, , is diagonal, i.e. the noise added to one state is independent of the noise added to any other state.
The model noise added for the time dependent variables is
very small ( 10 16 ) in all the cases. Concerning the model

E. H. VEFRING, G. NYGAARD, R.J. LORENTZEN , G. NVDAL AND K. K. FJELDE

noise added to the model parameters we have tried different

scenarios, the one described below is found as the most promising, but this is a topic for further research.
In selecting the model noise of the reservoir parameters
(pressure and permeability) of the zones, we take into account
that there is no information in the measurements about the
reservoir zones that is not penetrated. Therefore we use the
covariance matrix for the modeling error, , to activate the
estimation of the parameters in the actual zone. Since most
information about the permeability and the reservoir pressure
can be extracted from the measurements during the first time
period after opening (30-90 minutes), the estimation of the
parameters is nearly turned off after the zone is passed.
In the examples we present, all measurements are generated synthetically by running the model with a given permeability and pressure, and adding noise to the obtained values to
generate measurements. As covariance matrix, , for the
measurement error we have used the same covariance matrix
as used when generating the measurements. In a field implementation, the covariance matrix for the measurement errors
should take into account the uncertainty in the measurement
devices, but also include uncertainty in the positioning of the
measurement gauges, and inaccuracies due to the applied numerical method25.
Examples
We consider a total of four examples. The first two examples
are presented to show the performance of two estimation techniques, the Levenberg-Marquardt methodology and the Ensemble Kalman filter, both applied to a single-phase case with
a reservoir consisting of 10 different zones.
In the first two examples the inflow of liquid occurs in 10
zones, each of length 30 m. Within each zone the permeability
and the reservoir pressure are constant. Each zone is again
divided into 1m long segments and the flow from each segment is given by (7).
Table 1: Parameters used in Example 1 and Example 2

Compressibility of liquid
Viscosity of liquid
Wellbore radius
Penetration rate in the
formation
Uncertainty in pump pressure
Uncertainty in bottom hole
pressure
Uncertainty in flow rate

2.1810-9 Pa-1
0.05 Pa.s
0.067 m
0.005 m/s
0.15 % (standard deviation)
0.5 % (standard deviation)
1 % (standard deviation)

In the last two examples, water and gas is injected into the
drillstring and oil is produced from a reservoir which consists
of 3 different zones, each of length 100 m.
For all the examples, the synthetic measurements representing pump pressure, downhole pressure, liquid return rate
and gas return rates are generated by adding normal distributed noise to the simulation results.
In all the figures presenting liquid mass flow rate, gas mass
flow rate, pump pressure or bottom-hole pressure the dots

IADC/SPE 81634

Table 2: Parameters used in Example 3 and Example 4

Compressibility of injection
liquid
Compressibility of
production liquid
Viscosity of injection liquid
Viscosity of production
liquid
Wellbore radius
Penetration rate in the
formation
Uncertainty in pressures
Uncertainty in flow rates

1.010-9 Pa-1
1.1710-9 Pa-1
0.001 Pa.s
0.04 Pa.s
0.01 m
0.0083 m/s
0.15 % (standard deviation)
1 % (standard deviation)

represents synthetic measurements and the solid line represents the results from simulations with the estimated parameters.
In Ex. 1, Ex. 2 and Ex. 4 the downhole pressure is fluctuated, in an effort to oscillate the inflow from the reservoir into
the well. With reference to the equation describing the transient inflow from the reservoir (7), it should be possible to
identify the reservoir permeability if the pressure difference is
varied. This corresponds with the estimation theory within the
field of adaptive control26, where the input signal to a plant is
referred to as sufficiently rich of order n, if it consists of at
least n/2 distinct frequencies. This means that if the system is
oscillated with one frequency sinus oscillations with certain
amplitude, then two parameters can be identified simultaneously. The frequency applied to the downhole pressure is 0.5
mHz and the amplitude is 8 bar peak-to-peak.
Example 1: Permeability and pressure estimation, large
reservoir pressure variations.
In this simple example, liquid is circulated in the well, and the
same liquid is produced from the reservoir. There is no gas
injection into the drillstring. A true state is generated synthetically by using true reservoir pressure pres,i = [230, 260, 220,
220, 240, 280, 230, 230, 250, 220] bar . The true permeability values are Ki =[300, 600, 400, 100, 50, 800, 200, 300,
100, 400] mD. The choke opening is varied, giving sinus
variations to the choke pressure. These variations to the choke
pressure gives similar fluctuations in the bottom hole pressure.
From Fig. 2 we can observe the fluctuations in bottom hole
pressure. The fluctuations peak-to-peak is about 8 bar, with a
mean of 204 bar. Since the flow into the well depends on the
bottom hole pressure, the influx from the reservoir also shows
these fluctuations, as presented in Fig. 3.
The actual and estimated permeabilities are presented in
Fig. 4, and actual and estimated pressures are presented in Fig.
5. From Figs. 2 and 3 we observe that we are able to match the
bottom hole pressure and liquid flow rate at outlet. However,
we observe that the permeability and reservoir pressure not are
well estimated for the different zones. Since we are able to
reproduce the measured data with some discrepancy between
estimated parameters and true parameters, additional information is needed to obtain reliable estimates. We observe that if
the permeability of a zone is over estimated, the reservoir
pressure is under estimated and vice versa.

IADC/SPE 81634

RESERVOIR CHARACTERIZATION DURING UBD: METHODOLOGY AND ACTIVE TESTS

Example 2: Permeability and pressure estimation, small

reservoir pressure variations.
This example is equivalent to Ex. 1, but the variations in reservoir pressure are smaller. As in Ex. 1, liquid is circulated in
the well, and liquid is produced from the reservoir. There is no
gas injection into the drillstring. A true state is generated synthetically by using true reservoir pressure pres,i = [215, 220,
225, 219, 220, 221, 222, 222, 218, 230] bar . The true permeability values have not been changed, and Ki =[300, 600,
400, 100, 50, 800, 200, 300, 100, 400] mD. The fluctuations of
the bottom hole pressure are generated in the same way as in
Ex. 1. This causes the fluctuations of the liquid mass rate at
the outlet.
The actual and estimated permeabilities are presented in
Fig. 6, and actual and estimated pressures are presented in Fig.
7. Again, we observed that we were able to match the bottom
hole pressure and the liquid flow rate at outlet. In this example
we also observe that the estimates of reservoir permeability
and reservoir pressure compares fairly well with the true values.

Conclusions
Novel methodology for reservoir characterization during underbalanced drilling has been presented. The methodologies
are based on a transient well flow model coupled to a transient
reservoir model and estimation techniques are applied to estimate permeability and reservoir pressure. An important aspect
of the methodology presented here is that only the data usually
measured during an underbalanced drilling is used as available
data. The examples presented in this paper illustrate that active
tests may improve the reservoir characterization during underbalanced drilling. The results from Example 1 and 2 did not
show any significant differences between the results obtained
with least squares estimation and with ensemble Kalman filter.
The advantage of the ensemble Kalman filter is however that
the results can be obtained during drilling.

Example 3: Permeability and pressure estimation, no oscillations in downhole pressure.

In this example, water and gas is injected into the drillstring.
Oil is produced from the reservoir. A true state is generated
synthetically by using true reservoir pressure pres,i = [212,
220, 216] bar . The true reservoir permeability values are Ki
=[400, 100, 700] mD. The choke pressure is kept constant at 6
bar. The initial estimates of the reservoir pressures for all three
zones are 215 bar, and the initial estimates of the reservoir
permeabilities are 500 mD.
The ensemble Kalman filter produces on-line estimates
during the drilling process. The reservoir permeability estimates for zone 1 while drilling is presented in Fig. 8. The reservoir pressure estimates for zone 1 is presented in Fig. 9.
From Fig. 8 we observe that the permeability is approaching
the true permeability value, but the estimate drifts away after zone 2 is passed. From Fig. 9 we see that that the estimate
does not approach the true value at all.

d = Measurement vector.
d1 = Outer diameter of the annulus.
d2 = Inner diameter of the annulus.
f = Fanning friction factor.
f i (c) = Simulated value at the ith sensor, as a function of
the inflow profile.
a
=
Forecasted
value by simulator after next time step
f(s j )

Example 4: Permeability and pressure estimation, with

oscillations in downhole pressure.
This example is equivalent to Ex. 3, but here the choke pressure is oscillated. A true state is generated synthetically by
using true reservoir pressure pres,i = [212, 220, 216] bar . The
true reservoir permeability values are Ki =[400, 100, 700]
mD. The choke pressure is varied, causing fluctuations in the
downhole pressure. The initial estimates of the reservoir pressures for all three zones are 215 bar, and the initial estimates
of the reservoir permeabilities are 250 mD.
The ensemble Kalman filter produces on-line estimates
during the drilling process, and the reservoir permeability estimates during drilling for zone 1 is presented in Fig. 10. The
reservoir pressure estimates for zone 1 is presented in Fig. 11.
From Fig. 11 we see that that the estimate converges to the
true value.
The final estimates after all the zones have been drilled are
shown in Figs. 12 and 13. We observe that the estimates in Ex.
4 fit better to the true values, than for the results in Ex. 3 with
no oscillations.

Nomenclature
C1= Gas factor.
C2= Gas holdup.
c = Compressibility.
c = Parameter vector.

using initial state s ja .

F(c ) = [f i (c)]i =1 = Simulated values as function of the

m

parameters.
G = Kalman gain matrix.
g = Gravity acceleration.
H = Measurement matrix.
J = Sensitivity matrix of simulated values with respect
to changes in the parameters.
K = Permeability.
M = Mechanistic model.
m = Mass transfer between phases.
n = Number of members in ensemble.
p = Pressure.
P = Covariance matrix of error in estimated
parameters.
q i,j = Flow from segment j in zone i
R = Ensemble error covariance matrix.
S = Skin factor

s = Member of ensemble.
t = Time.
ti,j= Time since opening segment j in zone i
v = Velocity.
yi , obs = Measured value at the ith sensor.
z = Spatial coordinate.
= Volumetric fraction.
= 0.5772.
= Measurement noise vector.

E. H. VEFRING, G. NYGAARD, R.J. LORENTZEN , G. NVDAL AND K. K. FJELDE

= Measurement noise covariance matrix.

= Standard deviation of the measurement error of

the ith sensor.

= Angle of inclination.
= Viscosity.
= Density.
= Interfacial tension.

= Porosity.
= Model noise vector.
= Model error covariance matrix.
= Covariance matrix of the measurement error.

= Model parameters.

s = Inflow segment length.

Subscripts
est = Estimated.
F = Friction.
G = Gas.
Gs = Gas superficial.
i = Counting variable (zone/measurement).
j = Counting variable (segment/ensemble member).
L = Liquid.
Ls = Liquid superficial.
M = Mixture.
O = Oil.
res = Reservoir.
w = Well.
W = Water.
Superscripts
a = Analyzed.
f = Forecast.

6.

7.

8.
9.

10
11.

12.

13.
14.
15.

Acknowledgements
This work has been supported financially by the Norwegian
Research Council.

16.

References

17.

1.

2.

3.

4.

5.

Kneissl, W. : Reservoir Characterization Whilst Underbalanced Drilling, paper SPE/IADC 67690 in the proceedings for
the SPE/IADC Drilling Conference held in Amsterdam, The
Netherlands, 27 February-1 March 2001.
Hunt, J. L. and Rester, S., Reservoir Characterization During
Underbalanced Drilling: A New Model, paper SPE 59743 presented at the 2000 SPE/CERI Gas Technology Symposium held
in Calgary, Alberta Canada, 3-5 April 2000
Kruijsdijk, C. P. J. W. and Cox, R. J. W. : Testing While Underbalanced Drilling: Horizontal Well Permeability Profiles,
paper SPE 54717 presented at the 1999 SPE European Formation Damage Conference held in The Hague, The Netherlands,
31 May 1 June 1999.
Kardolus, C. B. and Kruijsdijk, C. P. J. W. : Formation Testing
While Underbalanced Drilling, paper SPE 38754 presented at
the 1997 SPE Annual Technical Conference and Exhibition held
in San Antonio, Texas, 5 8 October 1997.
Lorentzen, R.J., Fjelde, K.K., Fryen, F., Lage, A.C.V.M.,
Nvdal, G. and Vefring, E.H.: Underbalanced Drilling: Real

18.
19.
20.

21.

22.

IADC/SPE 81634

Time Data Interpretation and Decision Support paper

SPE/IADC 67693 in the proceedings for the SPE/IADC Drilling
Conference held in Amsterdam, The Netherlands, 27 February-1
March 2001.
Lorentzen, R. J., Fjelde, F., Fryen, J., Lage, A. C. V. M.,
Nvdal, G. and Vefring, E. H.,: Underbalanced and Low-head
Drilling Operations: Real Time Interpretation of Measured
Data and Operational Support, paper SPE 71384 presented at
the 2001 SPE Annual Technical Conference and Exhibition held
in New Orleans, Lousiana, 30 September 3 October 2001.
Vefring, E. H., Nygaard, G., Fjelde, K. K, Lorentzen, R. J.,
Nvdal, G., and Merlo, A.: Reservoir Characterization during
Underbalanced Drilling: Methodology, Accuracy, and Necessary Data, paper presented at the Annual Technical Conference and exhibition held in San Antonio, Texas, U.S.A., 29 September 2 October 2002.
Ishii, M.: Thermo-Fluid Dynamic Theory of Two-Phase Flow,
Eyrolles, (1975).
Benzoni-Gavage, S.: Analyse Numrique des Modles Hydrodynamiques D`coulements Diphasiques Instationnaires dans le
Rseaux de Production Ptrolire, Thse ENS Lyon, France
(1991).
Lage, A.C.V.M.: Two-phase Flow Models and Experiments for
Low-Head and Underbalanced Drilling, Ph.D. dissertation,
Stavanger University College, Norway (2000).
Lage, A.C.V.M., Fjelde, K.K. and Time, R.W.: Underbalanced
Drilling Dynamics: Two-Phase Flow Modeling and Experiments Paper IADC/SPE 62743 presented at the 2000
IADC/SPE Asia Pacific Drilling Technology held in Kuala
Lumpur, Malaysia, 11-13 September.
Lage, A.C.V.M., Fryen, J., Svareid, O. and Fjelde, K.K.:
Underbalanced Drilling Dynamics: Two-Phase Flow Modeling,
Experiments and Numerical Solution Techniques, Paper IBP
41400 presented at the Rio Oil & Gas Conference held in Rio de
Janerio, Brazil, 16-19 October 2000.
Taitel, Y. and Barnea, D.: Counter Current Gas-Liquid Vertical
Flow, Model for Flow Pattern and Pressure Drop, , Int. J.
Multiphase Flow (1983) 9, 637-647.
Hasan, A.R., Kabir, C.S. and Srinivasan, S.: Contercurrent
Bubble and Slug Flows in a Vertical System, Chemical Eng.
Science (1994) 49, 2567-2574.
Pettersen, .: Grunnkurs i reservoar mekanikk, Lecture notes,
Dept. of Math., Univ. of Bergen, Norway, 1990.
Mor, J. J.: The Levenberg-Marquard algorithm: implementation and theory in Numerical Analysis (Lecture Notes in
Mathematics vol. 630) ed. G. A. Watson, Springer Verlag, Berlin (1977) 105.
Gill, P. E., Murray, W. and Wright, M. H.: Practical Optimization, Academic Press, San Diego, CA (1981) 167.
Bard, Y.: Nonlinear Parameter Estimation, Academic Press,
Orlando, Florida (1974) 189.
Donaldson, J.R. and Schnabel, R. B.: Computational Experience
with Confidence Regions and Confidence Intervals for Nonlinear Least Squares, Technometrics (Feb. 1987) 29, No. 1, 67.
Nvdal, G., Mannseth, T. and Vefring, E. H.: Near-Well Reservoir Monitoring through Ensemble Kalman Filter. Paper SPE
75235 presented at the SPE/DOE Thirteenth Symposium on Improved Oil Recovery held in Tulsa, Oklahoma, 13 17 April
(2002).
Nvdal, G., Mannseth, T., and Vefring, E.H.: Instrumented
Wells and Near-Well Reservoir Monitoring through Ensemble
Kalman Filter. Proceedings of 8th European Conference on the
Mathematics of Oil Recovery (ECMOR VIII) held in Freiberg,
Germany, 3 6 September (2002).
Evensen, G.: Sequential Data Assimilation with Nonlinear
Quasi-geostrophic Model using Monte Carlo Methods to fore-

23.
24.

25.
26.

RESERVOIR CHARACTERIZATION DURING UBD: METHODOLOGY AND ACTIVE TESTS

cast Error Statistics. J. Geophys. Res. Vol. 99 (C5), pp. 10 143

10 162, (1994).
Burgers, G., van Leeuvwen, P. J., and Evensen, G.: Analysis
Scheme in the Ensemble Kalman Filter. Monthly Weather Review. Vol. 126, pp. 1719 1724, (1998).
Evensen, G.: Application of Ensemble Integrations for Predictability Studies and Data Assimilation. Published in: Monte
Carlo Simulations in Oceanography, Proceedings 'Aha Huliko'a
Hawaiian Winter Workshop, University of Hawaii at Manoa,
January 14--17, (1997).
Cohn, S.E.: An Introduction to Estimation Theory. Journal of
the Meteorological Society of Japan. Vol. 75, pp. 257 - 288,
(1997).
Ioannou, P.A. and Sun, J.: Robust Adaptive Control, PrenticeHall, Upper Saddle River, New Jersey (1996) 255.

Liquid mass rate at outlet

700
Measured
Estimated
600

500

400

l/min

IADC/SPE 81634

300

200

100

0
0

200

400

600
Minutes

800

1000

1200

Figure 3: Liquid mass flow rate at outlet, Example 1: Permeability

and pressure estimation, large reservoir pressure variations.
Permeability
900
True
LM Estimate
EnKF Estimate

800

K1

K2

...

Ki
700

600

mD

500

400

Figure 1: Reservoir inflow.

300

Bottom hole pressure

200

220
Measured
Estimated

100

215
0

10

Zone

Figure 4: True and estimated reservoir permeabilities, Example 1:

Permeability and pressure estimation, large reservoir pressure
variations.

bar

210

205
Pressure
350
True
LM Estimate
EnKF Estimate

200
300

195
250

200

400

600
Minutes

800

1000

200

1200

Figure 2: Bottom hole pressure, Example 1: Permeability and

pressure estimation, large reservoir pressure variations.

bar

190
0

150

100

50

10

Zone

Figure 5: True and estimated reservoir pressures, Example 1:

Permeability and pressure estimation, large pressure variations.

E. H. VEFRING, G. NYGAARD, R.J. LORENTZEN , G. NVDAL AND K. K. FJELDE

Permeability

IADC/SPE 81634

Reservoir Pressure, Zone 1

900

230

True
LM Estimate
EnKF Estimate

800

True
Estimate
225

700

220

600

mD

bar

500

215

400

210

300

200

205
100

200
0

10

100

200

300

400

500

600

700

Minutes

Zone

Figure 6: True and estimated reservoir permeabilities, Example 2:

Permeability and pressure estimation, large pressure variations.

Figure 9: Reservoir pressure estimates in zone 1, Example 3: No

oscillations in downhole pressure.

Pressure
350

Reservoir Permeability, Zone 1

True
LM Estimate
EnKF Estimate

550

300

True
Estimate

500
450

250

400
350

mD

bar

200

150

300
250
200

100

150
100

50

50
0

0
0

10

Zone

100

200

300

400

500

600

700

Minutes

Figure 7: True and estimated reservoir pressures, Example 2:

Permeability and pressure estimation, large pressure variations.

Figure 10: Reservoir permeability estimates in zone 1, Example 4:

With oscillations in downhole pressure.

Reservoir Permeability, Zone 1

Reservoir Pressure, Zone 1

550

230

True
Estimate

500

True
Estimate

450

225

400
220

300

bar

mD

350

250

215

200
210
150
100
205
50
0
0

100

200

300

400

500

600

700

Minutes

Figure 8: Reservoir permeability estimates in zone 1, Example 3:

No oscillations in downhole pressure.

200
0

100

200

300

400

500

600

700

Minutes

Figure 11: Reservoir pressure estimates in zone 1, Example 4:

With oscillations in downhole pressure.

IADC/SPE 81634

RESERVOIR CHARACTERIZATION DURING UBD: METHODOLOGY AND ACTIVE TESTS

Pressure

Permeability
900

230
True
Ex. 3, no osc.
Ex. 4, with osc.

800

True
Ex. 3, no osc.
Ex. 4, with osc.
225

700

600

220

bar

mD

500
215

400

300

210

200
205
100

2
Zone

Figure 12: True and estimated reservoir permeabilities, Comparisons between Ex. 3, no choke oscillations and Ex. 4, with choke
oscillations.

200

2
Zone

Figure 13: True and estimated reservoir pressures, Comparisons

between Ex. 3, no choke oscillations and Ex. 4, with choke oscillations.