You are on page 1of 14

SPE 62917

Advances in Multilayer Reservoir Testing and Analysis using Numerical Well Testing
and Reservoir Simulation

R. R. Jackson, SPE, and R. Banerjee, SPE, Schlumberger

Copyright 2000, Society of Petroleum Engineers Inc.


This is known as a multilayered reservoir system without
This paper was prepared for presentation at the 2000 SPE Annual Technical Conference crossflow or commingled. For cases where communication
and Exhibition held in Dallas, Texas, 1–4 October 2000.
can take place between layers and formation crossflow can
This paper was selected for presentation by an SPE Program Committee following review occur, this is known as multilayered reservoir with
of information contained in an abstract submitted by the author(s). Contents of the paper,
as presented, have not been reviewed by the Society of Petroleum Engineers and are crossflow. In cases that are more complex, a well may
subject to correction by the author(s). The material, as presented, does not necessarily
reflect any position of the Society of Petroleum Engineers, its officers, or members. Papers produce from a composite reservoir consisting of
presented at SPE meetings are subject to publication review by Editorial Committees of the
Society of Petroleum Engineers. Electronic reproduction, distribution, or storage of any part
commingled zones that may contain crossflow layers.
of this paper for commercial purposes without the written consent of the Society of Characterisation of layered reservoirs4 is important from
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
a reservoir evaluation and management standpoint because
contain conspicuous acknowledgment of where and by whom the paper was presented. of the influence of layering on primary and secondary oil
Write Librarian, SPE, P.O. Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-
952-9435. recovery. Layered reservoirs pose special problems; for
example differences in layer permeabilities can lead to
Abstract unbalanced depletion and poor recovery. Waterflooding
projects in such reservoirs can lead to poor sweep efficiency
This paper presents multilayer testing and analysis because unswept oil is usually bypassed in the lower
techniques to obtain layer permeabilities and skin factors permeability zones. For commingled reservoirs, contrasts in
using pressure and flowrate transient data from sequential well performance and production could be due to large
flow tests acquired with production logging tools. The variations in permeability-thickness product or skin.
analysis of these tests by application of numerical well test The problem addressed here is the challenge of
analysis is highlighted. An integrated workflow is shown determining reservoir parameters in multilayered reservoirs,
and a new analysis technique is presented which i.e. estimating layer permeabilities, skin factors and
incorporates numerical reservoir simulation and an formation pressures from well test data.
automated history matching procedure. A gradient method Attempts to identify reservoir parameters for individual
has been successfully applied to modify simulation model layers have applied some of the following methods. [1]
input parameters to match synthetic and field examples of Isolation testing and drillstem testing (DST) of individual
multilayer transient tests. The study also highlights the layers by selective perforation, or by combinations of
potential for the application of this method and for straddles and/or isolation packers; [2] conventional buildup
automated history matching techniques to pressure transient and variable-rate drawdown testing; and [3] production
and flowrate transient analysis and multilayer test analysis logging.
and interpretation. Isolation and selective methods can be problematic and
they are expensive to perform. Conventional well tests
(buildup and/or drawdown) and their analysis reveal the
Introduction total system behaviour, and often give no indication that
layering phenomena are present. Buildup tests are often
Hydrocarbon reservoirs are commonly heterogeneous and affected by crossflow between layers and from wellbore
they often comprise of units or formations where layers may storage effects; pressure data alone may not provide any
have distinct values for thickness, permeability, porosity and information about crossflow or wellbore dynamics.
skin factor. In well testing terminology, these reservoir However, there are some cases where conventional well
formations are called multilayer systems. The effects of tests do work 1,3.
layering on conventional pressure-transient tests and their Testing and analysis techniques using production
interpretation have been described by many authors1-3. In logging have been developed to obtain individual layer
some situations, the layers may have such low permeabilities and skin factors. This approach has strong
permeabilities that they act as barriers to vertical flow and theoretical basis5-12. This testing technique, which has been
inhibit reservoir crossflow between more permeable units. applied in many locations around the world 10, 13-16, involves
2 JACKSON & BANERJE SPE 62917

the sequential measurement of flowrate and pressure end of this transient the tool is then passed across all the
transients from an individual layer, or a group of layers, layers to determine the flow profile at this stabilised rate.
after a rate change using a production logging tool. For the next part of the test, the tool is positioned at the top
For the reasons outlined above, the use of downhole of the topmost layer and the test proceeds as before. The test
flowrate and pressure measurement for layered reservoir can be concluded with an optional pressure buildup
testing and analysis are discussed and applied in this paper. followed by a shut in production log survey. The typical
We review the acquisition techniques for multilayer activities involved for the acquisition of an example three
transient tests and compare existing interpretation methods, layer multilayer test are summarised in Appendix B.
and propose a new integrated workflow and analysis method
that incorporates numerical well test analysis. The method
involves reservoir simulation, which is coupled with an Convolution and Deconvolution
automated history matching procedure.
Several techniques have emerged in the last decade to The logarithmic convolution or sandface rate convolution
automate history matching in reservoir simulation. The method (SFRC) which has been well documented in the well
history-matching problem is generally transformed into a testing literature21-28 is used here to estimate average
problem of minimisation of an objective function. For permeability and skin factors from simultaneous
practical cases the gradient method has been successfully measurements of downhole flowrate and pressure. It has
used17- 19. Our aim is to use these techniques in well test and been shown that for infinite acting radial flow a linear plot
multilayer transient test analysis. of normalised pressure versus logarithmic convolution time
gives a straight line. The slope and intercept of this line is
used to calculate permeability and skin. This analysis
Multilayer Transient Test Procedure method is applied to each sequential transient in a multilayer
transient test, for estimation of average layer permeability
A multilayer transient (MLT) test or layered reservoir test and skin factors.
(LRT) involves the acquisition of both sandface flow and An alternative approach is to use deconvolution 25, 27, 28,
wellbore pressure versus time during successive drawdown which involves inverting the convolution approach.
tests with the sensors in a static position, combined with Deconvolution can be used to compute the pressure
production log surveys across all layers to determine the response of a well as if the well was producing at a constant
stabilised flow profile at each rate 5-12. This test technique rate, with or without constant wellbore storage or
combines stabilised and transient flowrate and pressure exponential flowrate effects.
measurements, and extends the concept of flow-after-flow Parameter estimates obtained by these methods can then
or isochronal tests to multiple layers. be used as initial input values for simultaneous
The simplest layered reservoir testing technique is interpretation using an analytical model combined with non-
known as selective inflow performance (SIP) testing 20. The linear least squares estimation 7-9 and type curves 29. It can
result of a SIP test is a series of inflow performance also be used in numerical well testing approach, which
relationship (IPR) curves for each layer and estimates of the combines reservoir simulation with automated history-
average layer pressures (see Appendix A). It is not possible matching techniques. This topic is developed and discussed
to determine layer permeability and skin factors uniquely in later sections of this paper.
from SIP analysis – transient data are needed.
During the acquisition of a multilayer transient (MLT)
test, after each transient and the acquisition of logging Integrated Workflow for Analysis
passes across the completed reservoir section at that
stabilised rate, the production logging tool (PLT) is As discussed previously, an LRT or MLT relies on a
repositioned at the top of a different formation and then the combination of measurements including production log
surface flowrate is changed. The resulting sandface surveys and pressure and flowrate transients acquired with
flowrates and pressures are then measured. the production logging tool in a stationary position.
A typical MLT test sequence is illustrated in Fig.1. Interpreting these types of test is complex because it
Typically the PLT is positioned at the top of the bottom involves a large volume of data, and the identification of a
layer, the well flows at Rate 1 (or it is opened up from the reservoir model with a large number of potentially unknown
shut in condition) and both sandface rate and pressure are parameters.
measured. The surface rate is then changed to Rate 2 and the A recommended approach for the analysis and
transient flowrate and pressure data are recorded. At the end interpretation of multilayer tests is summarised and
of this transient, the production logging tool is then passed discussed as follows:
across all layers to determine the flow profile at this
stabilised rate. This concludes the data gathering to derive [1] Analyse the associated production log data using
the reservoir characteristics for the bottom layer. Following conventional methods. Determine the flow profile and
this sequence of operations the tool is then positioned at the zonal/layer contributions at each stabilised flow rate.
top of layer 2 and the surface rate is then changed to Rate 3. [2] Use PLT data for Selective Inflow Performance (SIP)
This time the measured rate is the sum of layers 1 and 2. analysis to establish the layer IPR, and for estimation of
This is different from the first transient when only layer pressures.
production form one layer (bottom) was measured. At the
ADVANCES IN MULTILAYER RESERVOIR TESTING AND ANALYSIS USING NUMERICAL
SPE 62917 WELL TESTING AND RESERVOIR SIMULATION 3

[3] Prepare the pressure and flowrate transient data for example shown in Fig.2, and discussed in the final sections
analysis. of the paper this procedure would be applied three times
[4] Apply conventional diagnostic methods using log-log with a one-, two- and three-layer models for each successive
and derivative methods and conventional well test analysis. interpretation step.
[5] Sandface rate convolution (SFRC) analysis of the Once a satisfactory reservoir model and parameter
pressure and flowrate transients for initial parameter estimates for the lowest zone are established the
estimation. interpretation proceeds with the subsequent layers. For
[6] Sequential and simultaneous analysis using a layered subsequent tests, the measured flow rate is cumulative for
model, with analytic or numerical methods using a reservoir the layers. Under these circumstances SFRC analysis of
simulator or both methods in combination. cumulative flowrate and wellbore pressure provide
[7] Simulation and automated history matching and parameter estimates of ‘average’ values. Using the single-
sensitivity analysis of the MLT test. layer theory, the thickness-averaged permeability, kavg, of
the reservoir can be evaluated using the following equation,
As summarised above, the first stage in the analysis of a
multilayer test begins with the interpretation of the kavg = (k1.h1 + k2.h2)/(h1 + h2) , ……………………...(1)
associated production log survey data. Using conventional
production log interpretation techniques, the production where, ki and hi refer to permeability and thichness of layer i
profile and individual layer flowrates are determined for the respectively. Evaluation of average skin factors, Savg, using a
well at each of the stabilised flow rates. From this stabilised similar approach can be written as,
flow profile data we can produce a selective inflow
performance (SIP) plot and determine the individual layer Savg = (S1.Q1 + S2.Q2)/(Q1+Q2) , ……………………...(2)
IPR and also estimate the layer pressures.
The second step consists of preparing the data in a where, Si and Qi are the skin and rate for layer i respectively.
suitable format for analysis. This includes, for example, the Equation 2 is not applicable where skin factors are negative;
splicing of several transients to a composite file, so, care must be taken when taking this approach for initial
incorporation of the well production and test history, and the estimates (i.e. treatment as a homogeneous system) because
correction of pressures to a common datum. Once the it could lead to incorrect results.
pressure values have been corrected to a common datum,
they can then be used for subsequent history matching. Simultaneous MLT Interpretation
Before any multilayer analysis - conventional diagnostic An alternative approach to the sequential method is the
methods using log-log and derivative methods and simultaneous method. In this technique the flow transient
conventional well test analysis (eg. Horner, MDH) should data are matched simultaneously using an analytic or
be performed. For analysis of multilayer tests, two methods numerical model using non-linear regression. Values for
have been generally applied 5-11. The methods, which are permeability and skin are varied until a satisfactory match is
known as the sequential method and the simultaneous obtained with all the measured flowrate and pressure
method, are summarized in the sections that follow. transients. Sensitivity analyses can be carried out to assess
the uniqueness of the solutions and the accuracy/confidence
Sequential MLT Interpretation of the solution.
The sequential method evaluates each layer sequentially, The simultaneous approach has the advantage that
beginning with the transient data acquired for the lowest superposition effects caused by changes in the surface rate
layer and then proceeding to subsequent layers. (and previous flow history) required to initiate each
Using measured flow transients acquired with a PLT transient, are taken into account in the interpretation.
positioned above the lowest test interval as the inner- Because simultaneous interpretation method uses data
boundary condition - interpretation of the bottom layer acquired for the entire duration of the MLT test and the
becomes a single-layer interpretation problem. From the transients may be affected by outer boundary conditions -
bottom layer, the reservoir model and the dominant flow correct parameter estimates from this method require
regimes and system geometry are identified. Initial specification of the initial layer pressures. These values can
parameter estimates for permeability and skin are made be estimated from SIP test analysis.
using logarithmic convolution methods 21-28. The SFRC plot With the simultaneous approach, a single multilayer
is used to compute layer permeability and skin from the data model is used. This model can be either analytical or
that follows a radial flow regime. Values of k and S matched numerical. During the analysis the bottom layer parameters
for this layer are then fixed for analysis of the next transient are first computed and fixed, and properties for the other
which uses data from the second to lowest zone. Again as layers are searched for using non-linear least squares
per the first transient, the pressures are matched to computed estimation and automatic regression techniques30,31,34. In
pressures by use of a two layer model. It is also valid to use order to take advantage of the flexibility offerered by
flowrate changes and match with pressure changes 29. reservoir simulation, numerical well testing techniques may
This sequential method is applied to transients above be used (see next section).
each layer. For each transient all the previously determined
values are fixed, and a single well model is used with as
many layers as the number of layers tested. For the MLT
4 JACKSON & BANERJEE SPE 62917

Numerical Simulation with Automated History


rdi = wd wdi
(odi − cdi ) , ……...…….………………….(4)
Matching
σd
Several techniques have been recently developed to
automate history matching in reservoir simulation. The where, d references one set of observed data of a given type
history matching problem is generally transformed into a at a given well, i references an individual data point for the
problem of minimisation of an objective function. For dth item of observed data, odi and cdi are the observed and
practical cases the gradient method has been successfully simulated values respectively, σd is the measurement
applied and in our work we endeavor to use this techniques standard deviation for the dth dataset, wd is the overall
in multilayer well testing. The objective is to use this weighting for the dth data set and wdi is a weighting for the ith
technique to modify the model parameters such as data point of the dth data set.
permeability or transmissibility so that the pressure response From equation 4 we can see that that our data type does
generated by our simulation model matches the acquired not have to be limited to pressure. One may, in addition,
MLT test data. In principle, this procedure is similar to non- regress on water cut, gas oil ratios, and flowrates.
linear regression associated with analytical models in well Furthermore, the confidence given to each data type as a
testing30,31. However, its application in this context has whole may be varied, in addition to the weight attached to a
obvious differences. The objective function used here is given data point. Please note that the overall data-type
more generalised and the gradients used are generated by a measurement error is accounted for by the term σ. For
numerical simulator rather than an analytical model. example, in the case of pressure transient data the
As discussed in the previous sections the final stage of measurement error would be reflected by the downhole
multilayer transient test analysis involves matching the gauge resolution.
acquired pressure transient data with the response obtained The prior term is given by
from a multilayer numerical model using a regression
technique. While the technique itself is general purpose and 1
applicable to radial as well as cartesian grids, we have used f prior = ν T BT C prior
−1
Bν , ……………………….(5)
a radial model. 2
Numerical models that use a reservoir simulator have
flexibility in terms of handling spatial variation of properties where, ν is the vector of normalised parameter modifiers. B
and multiphase flow effects. However, successful is the diagonal matrix of the parameter prior weights and
application of numerical well testing requires, at a Cprior is the parameter prior correlation matix. Cprior indicates
minimum, the ability to automatically generate input data how strongly we think the parameters are correlated. The
for the simulator and, preferably, application of an values range from –1 to +1. A value of 1 indicates strong
optimisation technique. correlation, 0 signifies no correlation and –1 indicates
Regression is performed to adjust the model parameters correlation having opposite effect. In equation 5 the
to match the observed pressure and the simulated pressure. normalised parameter modifier, ν, rather than the actual
It should also be noted that this observed data can also parameter is used. For a normal distribution this is given by
include, in addition to pressure, other quantities such as
flowrates and phase ratios (e.g. GOR, and water-cut). A x−x
commercial history matching software application was used v= , ……………………………….(6)
for this purpose19 which was coupled to a reservoir32 σ
simulator and a numerical well test software33. The software
uses a refinement of the Levenberg-Marquardt algorithm. where, x is the mean and σ the standard deviation of the
The gradient required to solve the algorithm is obtained parameter modifier distribution. The parameter modifier
directly from the reservoir simulator which is calculated at distribution is a normalised distribution of a parameter and
the same time as the simulated quantities themselves. may be obtained from the petrophysict. It should be noted
Moreover, use of an analytical method to obtain the initial that the initial value of a parameter in doing the regression is
parameter estimates ensures that the objective function is obtained from analytical methods described and referenced
near the minima and so can be optimised easily with few earlier in this paper.
simulation runs. The prior term is essentially a penalty term which makes
The objective function, f, that is minimised is a modified it difficult for the solution to move away from the estimated
form of the commonly used sum-of-squares used in well mean. By default this term is ignored by setting prior
testing. weights for all the parameters to zero (B becomes a null
matrix). A modified form of Levenberg-Marquardt
algorithm is used to optimise the objective function given in
1
f = f prior + ∑
2 d ,i
rd2,i , ……...………….…………….(3) equation 5. This modification is called the model trust
region method19. This optimisation scheme requires first and
second derivative of the objective function with respect to
where, fprior incorporates a priori knowledge of the reservoir the parameters (permeability, fault transmissibility, etc).
and rdi is the weighted difference between an observed value Using the Gauss-Newton approximation we can ignore the
and a simulated one. The latter term is defined as second derivative term. This means that the problem can be
solved by only considering the first derivative of the
ADVANCES IN MULTILAYER RESERVOIR TESTING AND ANALYSIS USING NUMERICAL
SPE 62917 WELL TESTING AND RESERVOIR SIMULATION 5

simulated quantity with respect to the parameters. These transients, with a pressure buildup (PBU) at the end of the
derivatives are obtained from our reservoir simulator at the MLT test. For simplicity this MLT example has no vertical
same time as the simulated quantities themselves in a single reservoir crossflow and the well drainage radius is large (i.e.
run thus saving a considerable amount of time. no boundaries are present). The oil formation volume factor
In multilayer tests the parameters we are mainly used is 1.2 and constant wellbore storage is modelled. In this
interested in are permeability are skin. Currently, for a radial scenario, the reservoir is assumed to be at virgin condition
model, we cannot use any of these parameters in the or shut in for a long period prior to the test. The initial
regression scheme outined above. However, this can be conditions and the reservoir and fluid parameters are
easily worked around by regressing on radial detailed in Table 1 and the rate schedule is in Table 2.
transmissibilty. Modifying the radial transmissibility is the
same as modifying radial permeability but with one
exception. A change in radial transmissibility has no effect HU-A HU-B HU-C
on the well connection factor which for a radial model is Layer 1 Layer 2 Layer 3
given by the equation Top (ft) 8000 8045 8085
Bottom (ft) 8035 8057 8125
Porosity, PU 9.5 18.5 4.5
cθkh
T= 2
, ………………….(7) Wellbore radius, 0.29 0.29 0.29
ft

(r2 − rw2 ) ln(r2 rw ) − 0.5 + S


r2
2 Bo (bbl/STB) 1.2 1.2 1.2
Muo, cp 1.38 1.38 1.38
Ct, 1/psi 2.1e-5 2.1e-5 2.1e-5
where, c is a unit conversion factor, θ is the segment angle C, bbl/psi 0.01 0.01 0.01
k, mD 48 100 12
of the grid block in radians, r2 is the block’s outer radius, rw
Skin, S 2.3 -1.1 4
is the well bore radius, k is the permeability, h is the layer Pi, psia 5000 5000 5000
thickness and S is the skin factor. On the other hand a
change in radial permeability effects the well connection Table 1. Reservoir and Fluid Parameters
factor. So if we change the radial permeability by the same
factor which we obtain for the radial transmissibility, in
order to maintain the match we have to modify skin so that Time Oil Rate PLT Position Layer
the well connection factor remains unchanged. In this way hrs STB/D ft
we are able to regress on both permeability and skin. Our 0 2450 8085 Above 3
assumption of constant connection factor is valid if the 12 925 8000 Above 1
initial parameter values we obtain by analytical means are 18 1850 8045 Above 2
not too far from that obtained using numerical simulation. 30 0 8000 Above 1
It should also be noted that varying the vertical 38 0 8000 Above 1
transmissibility is also equivalent to varying the vertical
Table 2. Rate Schedule and PLT Position
permeability. Thus layer connectivity may be investigated in
the analysis. Reservoir simulation is suitable for handling
For the first transient the production logging tool was
models with spatial heterogeniety and complex geometry,
stationed above the lowest layer (Fig.1.). The well was
and multiphase flow effects. In the context of this paper, the
initially shut in and then opened up to a surface rate of 2450
latter has special significance. For example, this method
STB/D. This transient and flow period which lasted for 12
allows one to regress on water-cut either independently or in
hours was then followed by a production log survey at this
conjuction with pressure.
stabilised flowrate.
For the second transient the PLT was then stationed above
all of the layers, and the surface rate was decreased to 925
Application to Simulated and Real Multilayer Tests
STB/D (Fig.1). This transient and flow period that lasted for
six hours was then followed by a production log survey at
To illustrate the application of our proposed workflow and
this stabilised rate.
methods to the analysis of multilayer tests, a synthetic
The third transient was recorded with the PLT stationed
layered reservoir test and a field example are shown. These
above the middle zone (Fig.1). The surface rate was
examples are used to show how to conduct a multilayer test
increased to 1850 STB/D and pressure and flowrate data
and analysis. The synthetic MLT test was generated
was acquired for 12 hours. This transient and flow period
analytically.
was followed by a production log survey at this stabilised
rate. Following the third transient and production log, the
PLT was then stationed above the top layer and the well was
Example 1: Simulated Case
shut in for an 8-hour pressure buildup. This step is optional
and in a real field situation if the decision is taken to acquire
This is a synthetic example of a three layer MLT test
a PBU, then the PLT should be positioned in-between layers
containing three hydraulic units with different permeability,
that are expected to crossflow. In this way, any crossflow
porosity and skin values, which are separated by layers of
can be observed and recorded – this will provide the
very low permeability. The test comprises of three flow
6 JACKSON & BANERJEE SPE 62917

opportunity to determine parameters for the crossflowing For the first transient the PLT was stationed above all of
layer(s) and to compare with the previous drawdown data. If the layers. The well was initially shut in and then opened up
the PLT is stationed above all the layers during the PBU, the to a low rate. This transient and flow period which lasted for
opportunity is lost to observe and record any crossflow. about 1 hour was then followed by a production log survey
The simulated test sequence and the pressure and at this stabilised flowrate.
flowrate data are illustrated in Fig.2. Flowrates and pressure For the second transient the PLT was stationed above the
for each layer were determined from synthetic production middle layer and the surface rate was increased to a medium
log surveys taken at the end of each transient and flow rate. This transient and flow period lasted for around 1 hour,
period. These values are plotted in Fig.3 and as a selective and was followed by a production log survey at this
inflow performance (SIP) plot in Fig.4. The pressure stabilized flowrate.
intercepts for zero rate for each layer represent the average The third transient was recorded with the PLT stationed
layer pressure for each hydraulic unit. above the bottom layer, with the surface rate increased to the
high rate. During this test the flowmeter recorded no flow
Simulated MLT Analysis – SFRC and Sequential from the bottom layer. Because there was no flowmeter data
Method this transient did not provide any specific information for
Fig.5 displays the diagnostic log-log plots of the pressure, the bottom layer. Consequently the well was analysed as a
pressure derivative and the convolution derivative for each two-layer system with the first and second transients
of the simulated transients. The logarithmic convolution or providing satisfactory pressure and flowrate data.
SFRC plots, together with initial estimates for skin and Qualitative results and parameter estimates were obtained
permeability values are displayed in Fig.6. The results of for the bottom layer from analysis of the pressure data for
analysis of the transients from application of the sequential the total system and from the pressure and flowrate data
method using the pressure data are illustrated in Fig.7 and from layers one and two.
results from the sequential analysis of the flowrate data are A radial simulation model and numerical for the two
shown in Fig.8. layers was then made using parameter estimates derived
from the convolution and sequential analytic methods
Simulated MLT Analysis – Numerical Simulation and previously described. This model was then history matched
Automated History Matching Method to the observed pressure and flowrate transient data. The
For this approach, we used a radial simulation model results for one of the transients is illustrated in Fig.13 and a
consisting of three hydraulic units with the initial properties solution diagram of the final timestep from the numerical
and PVT fluid properties summarised in Table 1. Estimates simulation solution using transients one and two is shown in
for skin and permeability values for each layer derived from Fig.14. This diagram clearly illustrates the pressure drops
the logarithmic convolution (SFRC) and sequential methods within the two layers and around the perforated intervals.
were input into our reservoir model. To test our simulation This field example shows how the multilayer testing
and history matching technique we deliberately varied some technique can be conducted and analysed using conventional
of our input parameters and model constraints and ran our multilayer test analysis methods and our numerical
reservoir simulator to achieve a mismatch between the simulation and automated history matching method.
model data and our solution. This is illustrated in Fig.9
where the model data and the varied pressure and flowrate
data are displayed together. After defining the input Results and Conclusions
parameters we then applied our simulation and regression
methods as previously described. The results are shown in Multilayer testing and analysis techniques have been
Fig.10. This example provides a good illustration of our presented and existing interpretation methods for estimating
integrated workflow and the application of automated layer permeabilities, skin factors and resevoir pressures have
history matching techniques to well test analysis and been evaluated using simulated and field examples.
multilayer transient test analysis. We propose a new integrated workflow and analysis
technique which incorporates numerical well test analysis.
This includes automatic simulation model generation,
Example 2: Field Case reservoir simulation and an automated history matching
procedure. As a result of this study the following
The test field example is a new offshore gas well which was conclusions were reached.
completed in four zones and treated as a three-layer [1] The application of numerical well test analysis using
reservoir. The formation of interest is a sand-shale sequence reservoir simulation to layered reservoir test analysis has
with the interval 4190-4214m called the top layer, the been demonstrated.
interval between 4214-4231m called the middle layer and [2] A new approach and workflow for the interpretation
the bottom layer is the interval 4231-4252m. The sequence of multilayer tests has been shown which uses an automated
of transient tests and the associated flowrates and pressure history matching technique. A gradient method has been
data are shown in Fig.11. The multilayer testing operation successfully applied to modify model parameters to match
lasted over a period of about five hours, with three surface simulation model pressure data with test data for synthetic
rate changes termed here as low, medium and high. Each of as well as field examples of multilayer transient tests.
these transients and flow periods were followed by
production log surveys at that stabilised production rate.
ADVANCES IN MULTILAYER RESERVOIR TESTING AND ANALYSIS USING NUMERICAL
SPE 62917 WELL TESTING AND RESERVOIR SIMULATION 7

[3] The study illustrates the potential for the application Modelling to Enhance the Ameland Field Development”. SPE
of automated history matching techniques in well test and Formation Evaluation, March, 5-10.
multilayer test analysis and interpretation. 16. Khalaf, A. A., El Emam, A. A., Mahmoud, Y. and Saeedi, J.
1993. “Successful Applications of the Multilayer Transient
Testing Technique: Arab-C Members of the Umm Shaif
Reservoir”. SPE 25666, ATCE 585-596.
Acknowledgement 17. Grussaute, T. and Gouel, P. 1999. “Computer Aided History
Matching of a Real Field Case”. SPE 50642.
We wish to express our gratitude to Wintershall Noordzee 18. Oliver, D.S., He, N., Reynolds, A.C.: “Conditioning
B.V. for providing us with data for the field example. We Permeability fields to pressure data”, ECMOR V, Leoben
also would like to thank Peter Hegeman for providing Sept 3-5, 1996.
assistance and discussion on layered reservoir test 19. “SimOpt User Guide”. 2000. Schlumberger GeoQuest, 238pp.
interpretations. 20. Stewart, G., Wittman, M., and Lefevre, D. 1981. “Well
Performance Analysis: A Synergetic Approach to Dynamic
Reservoir Description”, SPE 10209 24732 presented at the
SPE Annual Technical Conference and Exhibition, San
References Antonio, October 5-7.
21. Stewart, G. and Wittmann, M. J. 1983. “Afterflow
1. Lefkovits, H. C. et al. 1961. “A Study of the Behaviour of Measurement and Deconvolution in Well Test Analysis”. SPE
Bounded Reservoirs Composed of Stratified Layers”, Journal 12174.
of Petroleum Technology, March, 43-58; Trans. AIME, 222. 22. Thompson, L. G. and Reynolds, A. C. 1984. Analysis of
2. Ehlig-Economides, C. A. and Joseph, J. 1987. “A New Test Variable Rate Well Test Pressure Data Using Duhamel’s
for Determination of Individual Layer Properties in a Principle. SPE 13080.
Multilayered Reservoir”, SPE Formation Evaluation, 23. Meunier, D., Wittmann, M. J., and Stewart, G. 1985.
September, 261-283. Interpretation of Pressure Buildup Test Using In-situ
3. Sabet, M. A. 1991. “Well Test Analysis”. Gulf Publishing Measurement of Afterflow, Journal of Petroleum Technology,
Company, 460pp. 143-152.
4. Ehlig-Economides, C. A. and Economides, M. J. 2000. 24. Roumboutsos, A. and Stewart, G. 1988. “A Direct
“Formation Characterisation: Well and Reservoir Deconvolution or Convolution Algorithm for Well Test
Characterisation”. In: Reservoir Stimulation, Economides, M. Analysis”. SPE 18157, 63rd ATCE Houston, TX., October 2-
J. and Nolte, K. G. (eds.), John Wiley & Sons, 2-1 to 2-21. 5, 1988.
5. Ehlig-Economides, C. A. 1987. “Testing and Interpretation in 25. Kuchuk, F. 1990. “Gladfelter Deconvolution”. SPE
Layered Reservoirs”. Journal of Petroleum Technology SPE Formation Evaluation, March, 285-292.
17089, September, 1087-1090. 26. Ramakrishnan, T. S. and Kuchuk, F. J. 1990. “Testing and
6. Ehlig-Economides, C. A. 1993. “Model Diagnosis for Interpretation of Injection Wells Using Rate and Pressure
Layered Reservoirs”, SPE Formation Evaluation, September, Data”. SPE 20536. 65th ATCE New Orleans, September 23-
215-224. 26, 1990.
7. Kuchuk, F. and Ayestaran, L. 1983. “Analysis of 27. Kuchuk, F. 1990. “Applications of Convolution and
Simultaneously Measured Pressure and Sandface Flow Rate Deconvolution to Transient Well Tests”, SPE Formation
in Transient Well Testing”. SPE 12177. Evaluation, December, 375-384.
8. Kuchuk, F. and Ayestaran, L. 1985. “Analysis of 28. Kuchuk, F., Carter, R. G., and Ayestaran, L. 1990.
Simultaneously Measured Pressure and Flow Rate in “Deconvolution of Wellbore Pressure and Flow Rate”. SPE
Transient Well Testing”. Journal of Petroleum Technology, Formation Evaluation, March, 53-59.
February, 323-334. 29. Ayrestaran, L., Minhas, H. N., and Kuchuk, F. J. 1983. “The
9. Kuchuk, F., Karakas, M., and Ayestaran, L. 1986. “Well Use of Convolution Type Curves for the Analysis of
Testing and Analysis Techniques for Layered Reservoirs”, Drawdown and Buildup Tests”. SPE 18535, Easter Regional
SPE Formation Evaluation, August, 342-354. Meeting, Charleston, November 1-4, 1983.
10. Kuchuk, F., Shah, P. C., Ayestaran, L., and Nicholson, B. 30. Horne, R. N. 1997. “Modern Well Test Analysis: A
1986. “Application of Multilayer Testing and Analysis: A Computer-Aided Approach”. Petroway Inc., 257pp.
Field Case”. SPE 15419. 61st ATCE New Orleans October 31. Carvalho, R.S., Redner, R.A., Thompson, L.G., and Reynolds,
5-8, 1986. A.C. 1992. “Robust Procedures for Parameter Estimation by
11. Shah, P. C., Karakas, M., Kuchuk, F., and Ayestaran, L. 1988. Automated Type-Curve Matching”. SPE 24732 presented at
“Estimation of the Permeabilities and Skin Factors in Layered the SPE Annual Fall Technical Conference and Exhibition,
Reservoirs Using Downhole Rate and Pressure Data”. SPE Washington D.C., Oct 4-7, 1992.
Formation Evaluation, September, 555-566. 32. “Eclipse Reference Manual”. 2000. Schlumberger GeoQuest,
12. Hegeman, P., and Pelissier-Combescure, J. 1997. “Production 1750pp.
Logging for Reservoir Testing”, Oilfield Review, 16-20. 33. “Weltest 200 User’s Guide”. 2000. Schlumberger GeoQuest,
13. Morris, C. W. 1987. “Case Study of a Gulf Coast Layered 398pp.
Reservoir Using Multirate Transient Testing”. SPE 16762, 34. “GeoFrame: Zodiac User’s Guide”, Schlumberger GeoQuest,
62nd ATCE Dallas, TX., September 27-30, 1987. 232pp.
14. Rapach, J . M., Jefferies, R., Guedroudj, A. H., and
Westaway. P. J. 1990. “Practical Transient Multilayer Test
Design, Implementation, and Analysis of Gas Wells in the
Southern North Sea Basin”. SPE 20534, 65th ATCE New
Orleans, September 23-26, 1990.
15. Vogelij, H. N. J., Leach, M. J., and Kapteyn, P. K. A. 1993.
“Multilayer Flowmeter Testing Combined with 3D Field
8 JACKSON & BANERJEE SPE 62917

APPENDIX A measured by the PLT pressure gauges and flowmeter


(spinner); and that they are within their respective resolution
Selective Inflow Performance (SIP) and tolerance. If the transients are not of sufficient
For selective inflow performance (SIP) analysis, flow rates magnitude, then the drawdowns should be recorded with
for each layer from a production log survey acquired during two different flowrates: low and high. In this scenario the
stabilised flow at the end of each transient or flow period are drawdowns should be recorded with the first drawdown at
used. These rates are plotted as a function of wellbore the maximum rate and the second one with the low rate and
flowing pressure at a specific depth. The result of the SIP the third one with the maximum rate.
test is a series of inflow performance relationship (IPR)
curves for each layer. The point where the IPR curve
intercepts the zero rate axis is the pressure at which the layer
neither takes or produces, i.e. it is the pressure existing in
the layer at the well location when the production log survey
was acquired. These estimates of layer pressures will be
valid providing the transients and flow steps were of
sufficient duration to have reached pseudo-steady state. If
the system reached stabilised infinite acting radial flow
(IARF) the SIP is slightly modified by applying a transient
SIP analysis procedure.

APPENDIX B

Typical Activities for Acquisition of a Multilayer Test - 3


Layers
[1] Determine a minimum flowrate with which all three
layers in the well are expected to produce, and decide the
maximum flowrate with which the well can be flowed;
meeting all operational conditions. Define these three
different flowrates with the first flowrate as the minimum
rate and the subsequent flowrates being approximately one-
third increments of the difference between maximum and
minimum rates. The flowrates can be termed as low,
medium and high rates.
[2] Place the tool above the top layer and record the
drawdown with the low flowrate for 6-8 hours, or for a
sufficient time to observe IARF. At the end of this
drawdown, record up and down flowing passes with the
PLT to measure the flow profile and determine the zonal
contributions.
[3] Station the PLT above the middle layer. Increase the
production rate to the medium flowrate. Record the
drawdown for 6-8 hours or for sufficient time to observe
IARF. At the end of the drawdown, record up and down
passes with the PLT to measure the flow profile and
determine zonal contributions.
[4] Place the PLT above the top layer. Increase the
production rate to a high flowrate. Record the drawdown for
6-8 hours or for a sufficient time to observe IARF. Record
up and down passes with the PLT at the end of the transient.
[5] Place the PLT between two layers where crossflow is
expected. Close the well for shut in and record the pressure
buildup. If the test objectives include determination of any
boundary effects or reservoir limits – record for long enough
to observe these effects. At the end of the transient record
shut in up and down passes with the PLT across the
reservoir. Remove PLT from well and rig down.

In this particular example it is assumed that the


transients produced by changing the surface flowrates from
low to medium will induce a sufficient transient to be
ADVANCES IN MULTILAYER RESERVOIR TESTING AND ANALYSIS USING NUMERICAL
SPE62917 WELL TESTING AND RESERVOIR SIMULATION 9

Figure 1. Multilayer Transient (MLT) Test Sequence and Reservoir Completion.

5100 3000

5000

4900
2500
4800

4700

4600 2000
Downhole Rate (rb/d)

4500
Pressure (psi)

4400
1500
4300 Qt rb/d

4200

4100 1000

4000

3900
500
3800

3700

3600 0
0 5 10 15 20 25 30 35 40
Time (hrs)

Figure 2. Multilayer Transient (MLT) Test: Simulated Example Data Set.


10 JACKSON & BANERJEE SPE 62917

Selective Inflow Performance


IPR Plot
5000
HU A, Potential = 4898.7 psi
HU B, Potential = 4843.9 psi
HU C, Potential = 4908.2 psi
Composite, Potential = 4873.4 psi
4750

4500
Hydraulic Unit Pressure Potential (psi)

4250

4000

3750

3500

3250

3000
0 500 1000 1500 2000 2500 3000 3500 4000
HU Flowrate Contribution [at surface conditions] (bbl/d)
Figure 3. Multilayer Transient (MLT) Test: Selective Inflow Performance (SIP) Interpretation.

Pressure - Flowrate vs. Depth


7990
Profile 1 P
Profile 1 q
Profile 2 P
Profile 2 q
Profile 3 P
8010 Profile 3 q

8030

8050
Depth (ft)

8070

8090

8110

8130
3600 3800 4000 4200 4400 4600 0 500 1000 1500 2000 2500 3000
Pressure (psi) Flowrate (bbl/d)

Figure 4. Multilayer Transient (MLT)Test: Stabilised Flow Condition – Flowrates and Wellbore Pressure
ADVANCES IN MULTILAYER RESERVOIR TESTING AND ANALYSIS USING NUMERICAL
SPE 62917 WELL TESTING AND RESERVOIR SIMULATION 11

Diagnostic Plot Diagnostic Plot


TR1 TR2
1e+05 1e+04
Pressure Pressure
Derivative Derivative
Conv. Derivative Conv. Derivative

1e+04
1e+03
Delta P & Derivative Groups (psi)

Delta P & Derivative Groups (psi)


1e+03

1e+02

1e+02

1e+01
1e+01

1e+00 1e+00
1e-03 1e-02 1e-01 1e+00 1e+01 1e+02 1e-03 1e-02 1e-01 1e+00 1e+01
Delta T (hr) Delta T (hr)

Diagnostic Plot Diagnostic Plot


TR3 TR4
1e+05 1e+04
Pressure Pressure
Derivative Derivative
Conv. Derivative Conv. Derivative

1e+04
1e+03
Delta P & Derivative Groups (psi)

Delta P & Derivative Groups (psi)

1e+03

1e+02

1e+02

1e+01

1e+01

1e+00
1e+00

1e-01 1e-01
1e-04 1e-03 1e-02 1e-01 1e+00 1e+01 1e+02 1e-04 1e-03 1e-02 1e-01 1e+00 1e+01
Delta T (hr) Delta T (hr)

Figure 5. Diagnostic Plots for MLT Transient Data: Pressure, Pressure Derivative and Convolution Derivative.

Specialized Analysis Plot Specialized Analysis Plot Specialized Analysis Plot


Sandface Rate Convolved Plot - TR1 Sandface Rate Convolved Plot - TR2 Sandface Rate Convolved Plot - TR3
800.0
Test Interval C Test Interval A Test Interval B
-------------------- -------------------- --------------------
kh/mu (mD.ft/cP) = 339.3 800.0 kh/mu (mD.ft/cP) = 1062
kh/mu (mD.ft/cP) = 2183
12000.0 k (mD) = 11.71 k (mD) = 28.18
k (mD) = 34.63
Skin = 3.845
Normalized Delta P (psi)

Normalized Delta P (psi)

Normalized Delta P (psi)

Skin = 0.2103 Skin = -0.7658


600.0
600.0

10000.0

400.0
400.0

8000.0

200.0 200.0
-3.0 -2.0 -1.0 0.0 1.0 2.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0
Rate-Convolved Time Function Rate-Convolved Time Function Rate-Convolved Time Function

Figure 6. Sandface Rate Convolution (SFRC): Rate normalised change in pressure vs. the Rate Convolved Time Function.
12 JACKSON & BANERJEE SPE 62917

Figure 7. Sequential MLT Interpretation Results.

Drawdown Convolution Type Curve for Transient TR1 Multi-Rate Convolution Type Curve for Transient TR2
1e+03
measured delta P measured delta P
measured derivative measured derivative
measured convol.deriv. measured convol.deriv.
model delta P model delta P
model convol.deriv. model convol.deriv.

1e+03
Delta P & Derivative (psi)

Delta P & Derivative (psi)


1e+02

1e+02

1e+01

1e+01

1e+00
1e-03 1e-02 1e-01 1e+00 1e+01 1e-03 1e-02 1e-01 1e+00 1e+01
Delta T (hr) Delta T (hr)
Hydraulic Unit C : Fully Completed - Homogeneous - Infinite Acting Hydraulic Unit A : Fully Completed - Homogeneous - Infinite Acting
k (mD) = 11.83, Total Skin = 3.939 k (mD) = 39.26, Total Skin = 0.4588
Constant Wellbore Storage: C[spin] (bbl/psi) = 9.914e-07 Constant Wellbore Storage: C[spin] (bbl/psi) = 1.016e-06

Multi-Rate Convolution Type Curve for Transient TR3


1e+03
measured delta P
measured derivative
measured convol.deriv.
model delta P
model convol.deriv.
Delta P & Derivative (psi)

1e+02

1e+01

1e+00
1e-04 1e-03 1e-02 1e-01 1e+00 1e+01 1e+02
Delta T (hr)
Hydraulic Unit B : Fully Completed - Homogeneous - Infinite Acting
k (mD) = 100, Total Skin = -1.1
Constant Wellbore Storage: C[spin] (bbl/psi) = 1.005e-06

Figure 8. Sequential MLT Interpretation Results: Flowrate Convolution.

Drawdown Flowrate Type Curve for Transient TR1 Multi-Rate Flowrate Type Curve for Transient TR2
3500
measured flowrate measured flowrate
model response model response

300.0

3000
Bottomhole Flowrate (bbl/d)
Bottomhole Flowrate (bbl/d)

2500
200.0

2000

100.0

1500

0.0 1000
1e-03 1e-02 1e-01 1e+00 1e+01 1e-03 1e-02 1e-01 1e+00 1e+01
Delta T (hr) Delta T (hr)
Hydraulic Unit C : Fully Completed - Homogeneous - Infinite Acting Hydraulic Unit A : Fully Completed - Homogeneous - Infinite Acting
k (mD) = 11.83, Total Skin = 3.939 Multi-Rate Flowrate Type Curve for Transient TR3 k (mD) = 39.26, Total Skin = 0.4588
Constant Wellbore Storage: C[spin] (bbl/psi) = 9.914e-07 1600 Constant Wellbore Storage: C[spin] (bbl/psi) = 1.016e-06
measured flowrate
model response

1400
Bottomhole Flowrate (bbl/d)

1200

1000

800

600
1e-04 1e-03 1e-02 1e-01 1e+00 1e+01 1e+02
Delta T (hr)
Hydraulic Unit B : Fully Completed - Homogeneous - Infinite Acting
k (mD) = 100, Total Skin = -1.1
ADVANCES IN MULTILAYER RESERVOIR TESTING AND ANALYSIS USING NUMERICAL
SPE 62917 WELL TESTING AND RESERVOIR SIMULATION 13

Figure 9. Simultaneous MLT Interpretation using numerical well test analysis and simulation: Simulated Example with intermediate
results and final solution (match) shown with radial model.

Figure 10. Numerical Simulation and Automated History Matching of Multilayer Transient (MLT) Test Data: Solution for the test data using
a radial simulation model and a gradient based history matching technique.
14 JACKSON & BANERJEE SPE 62917

1 : 150 Flowrate Conversion


Spinner To Downhole q Conversion
7000
4185
6500

Pressure (psi)
6000

5500
4200 5000

4500
4000
7000

Downhole q (bbl/d)
4215 6000
5000
4000
3000

4230 2000
1000
0
25

20
4245

Spinner (c/s)
15

10

5
4260

1 : 150 0
25 25.5 26 26.5 27 27.5 28 28.5 29 29.5
Time (hr)

Figure 11. Field Example of MLT: Test sequence and layer configuration. Figure 12. Field Example of MLT Data Set.

Figure 13. Field Example of an MLT Transient with observed data


and numerical solution. Figure 14. Field Example of MLT: Solution diagram for final
timestep from the numerical simulation solution using
Transients one and two.

You might also like