PROCEEDINGS, Twenty-Eighth Workshop on Geothermal Reservoir Engineering

Stanford University, Stanford, California, January 27-29, 2003

INTEGRATING WELLBORE MODELING AND PRODUCTION HISTORY TO UNDERSTAND

WELL BEHAVIOR

Jorge A. Acuna

Unocal Corporation
1160 North Dutton Ave. Suite 200
Santa Rosa, California, USA.
E-mail: jacuna@unocal.com

production conditions at different points in the

ABSTRACT history of the well were different in many cases.
Ideally the well models should be able to reproduce
A study of well behavior in the Awibengkok not only current conditions but also historic behavior.
reservoir, Indonesia, demonstrated how production
history, pressure and temperature surveys could be This paper describes a project undertaken with the
combined with wellbore modeling to provide insight goal of creating well models capable of reproducing
into well behavior. Wellbore models were the historic deliverability behavior of the wells with
constructed for every producing well and a wellbore the same set of PI’s. This approach relies on wellbore
simulator was run in batch mode to automatically modeling and requires good estimates of pressure and
produce calculated total mass and enthalpy every flowing enthalpy for the individual feed zones of
three months for the entire production history of each each well at each point in time. Well models that
well. The final result was a set of wellbore hydraulic match production history should have a better chance
models that match production history for each well. of matching future behavior.
The calibrated models were then used for short-term
deliverability predictions as an alternative to well
OBTAINING PRESSURE AND ENTHALPY
decline analysis.
EVOLUTION FOR FEED ZONES
Keywords: wellbore modeling, well deliverability, The first step in trying to match the historic variation
decline analysis. in well deliverability was to construct the pressure
and flowing enthalpy variation in time for each feed
INTRODUCTION zone. Feed zone location was defined though
interpretation of flowing spinner tests.
Awibengkok is a geothermal field located on the
island of Java, Indonesia [Stimac et al., 2000]. The Awibengkok is a high-permeability reservoir that is
field had a nominal installed capacity of 110 MW liquid-dominated in the western part and currently
from start-up in March 1994 to September 1997. At has a well-developed steam cap in the eastern part.
that time additional units were put on line to take the To define the pressure evolution in the liquid western
total nominal capacity to 330 MW. part of the field we used available surveys to obtain
the pressure drop with time at the same elevation for
Throughout nine years of operation periodic pressure different groups of wells. Pressure at any elevation
and temperature surveys, enthalpy measurements as for the deep liquid part was then obtained by
well as detailed production records were obtained. subtracting the pressure drop from the initial pressure
at the corresponding elevation. An approximate
Typically pressure and temperature surveys are used function of pressure with time and elevation was
in conjunction with production records to explain created.
changes in well production. Conventional decline
analysis was then used to make short term steam Pressure in the shallow steam cap was defined using
deliverability forecasts. Long-term forecasts were two sources of data. The first one is the static surveys
done with reservoir simulation techniques. for wells that produce at least partially from the
steam cap. The other source of data is the available
In routine analysis of a well, prevailing production wellhead pressure data when those wells are shut-in.
conditions are reproduced by wellbore simulation by
constructing a well model. It was noticed however Determination of flowing enthalpies in the liquid feed
that the productivity indices (PI’s) required to match zones was made using measured temperatures.
Since it has been observed that Awibengkok wells observed pressure and temperature changes in the
that initially produce liquid and then evolve to dry feed zones.
steam producers do so relatively quickly, it was
assumed that once a feed zone becomes part of the EXAMPLES AND DISCUSSION
steam cap it produces dry steam. This is consistent
with the behavior of a reservoir with a falling liquid The technique for well performance analysis
level above which feed zones produce dry steam. presented here effectively integrates information
from pressure, temperature and spinner surveys
through the use of hydraulic models of the wells.
WELLBORE SIMULATION AND
This makes it possible to check the internal
PERFORMANCE MATCHING PROCEDURE
consistency of the subsurface data and combine them
Once the pressure and flowing enthalpy evolution of with surface production data such as enthalpy, flow
the individual feed zones has been prepared a rates and wellhead pressures.
wellbore model of the different wells is defined in
which well geometry and depth of feed zones are Figure 2 shows a well with a single liquid feed zone
given. The calibration consists of adjusting the at –900 ft. For this type of well the cooling trend is
productivity indices of the individual feed zones until directly reflected by a declining enthalpy. For the
calculated steam production and enthalpy at historic case of several feed zones that remain liquid at
wellhead pressure matches historic values. approximately the same enthalpy, the method is
incapable of determining the relative contributions of
The wellbore simulator utilizes the Duns and Ros each feed. The same is true for groups of steam
correlation [Hasan and Kabir, 2002]. Productivity producing feeds. For this case spinner surveys are
indices are corrected by reservoir fluid mobility and required.
also by a correction factor derived from energy
balance considerations. This handles the case when a When a group of feed zones flashes in sequence as
feed zone changes from liquid to steam. the falling liquid level gets deeper with time, the well
will show an increase of enthalpy with time. This
For the case of Awibengkok, an algorithm was case was observed in Awibengkok and it is shown in
written that reads the appropriate pressure and Figure 3. In this case the well was about to die in
flowing enthalpy of each feed zone and then runs the 1998. Soon after, however it increased steam
wellbore simulator to obtain, for the historic production dramatically and now is an excellent
measured wellhead pressure, the steam flow rate at producer.
separator conditions and the well enthalpy. The
wellbore simulator runs are made for each well every As mentioned before, this method can be applied to
three months steam deliverability prediction. This is illustrated in
Figure 4. Well deliverability prediction requires
The measured steam flow rates are representative of extrapolation of reservoir pressure and enthalpy in
the three-month period and taken from the daily the feed zones. This is not difficult in the case of well
measurements. The enthalpy values are derived from defined pressure and cooling trends Of particular
measurements made every three-months for most interest in the case shown in Figure 4 is the effect on
wells using tracer dilution techniques [Hirtz et al., well deliverability of a feed zone changing from
1993]. liquid to steam in 2002. The increase in steam
production with time is an insight given by this
The technique for calibrating individual wells is technique that cannot be obtained with conventional
illustrated in Figure 1. A sequence of three iterations decline curve analysis.
in steam flow rate and enthalpy match is shown. The
upper graph represents the steam flow rate match and The prediction shown in Figure 4 was made with data
the lower one the enthalpy match Part (a) is the up to early 2001. The predicted increase in enthalpy
match obtained with the initial PI’s. In part (b) the and steam production in 2002 was actually observed
relative proportion of PI’s of the shallow steam and it is also shown. The measured steam flow rate
producing feeds are adjusted with respect to the and enthalpy evolution for this well during the
lower liquid producing feeds to obtain a good match prediction time is much more gradual and starts a few
of enthalpy. In part (c) the absolute values of the PI’s months before than expected. This can be expected
are adjusted while keeping the proportionality from our assumption of a sharp transition from liquid
between steam and liquid producing feeds constant. to steam as well as our uncertainty with respect to the
This improves the match of flow rates while actual location of the liquid level in our wells based
preserving the enthalpy match. Also shown in Figure on shut-in well surveys. In any event we found the
1 are the predicted values of steam flow rate and result encouraging because it shows the ability of this
enthalpy. Predictions are made by extrapolation of method to predict steam deliverability increases.
 60
350

(a) 50
300

STEAM RATE (kg/s)

STEAM RATE (kph)
40
250

30
200
M EAS U R E D
20
150 CALC U LATE D
M EASU RED
10
100 CALCULATED

1994 1995 1996 1997 1998 1999 2000 2001

1994 1996 1998 2000 2002 2004 2006

1800
1600
1400
ENTHALPY (kJ/kg)

1200
1000
800
600
M EASU RED
400 CALCU LATED

200
1994 1996 1998 2000 2002 2004 2006
YEAR

(b) 50

40
STEAM RATE (kg/s)

M EASU R ED
CALCU LATED

30

20

10

1994 1996 1998 2000 2002 2004 2006

1800
1600
1400
ENTHALPY (kJ/kg)

1200
1000
800 M EASU R ED
600 CALCU LATED

400
200
1994 1996 1998 2000 2002 2004 2006
YEAR
(c)
40
35
STEAM RATE (kg/s)

30 M EASU R ED
CALCU LATED
25
20
15
10
5
1994 1996 1998 2000 2002 2004 2006
1800
1600
1400
ENTHALPY (kJ/kg)

1200
1000
800 M EASU R ED
600 CALCU LATED

400
200
1994 1996 1998 2000 2002 2004 2006
YEAR

Figure 1. Sequence of calibration procedure for a well. Upper curve is steam flow rate, lower curve is enthalpy. (a)
Initial PI’s, (b) Enthalpy match by adjusting ratio of shallow (steam) and deep (liquid) PI’s and (c) Flow rate match
by adjusting absolute value of PI while preserving shallow-deep ratio. Prediction based on extrapolation of feed
zones pressure and temperature changes also shown.
 The methodology proposed here allows the engineer
60 to check the consistency of subsurface and
production data through wellbore modeling. This
STEAM RATE (kg/s)

50

40 approach makes it possible to determine whether

30
causes other than pressure and temperature decline
M EASUR ED
CALCU LATED
are to blame for well decline behavior.
20

10 It is also possible to separate the effect of pressure

1994 1995 1996 1997 1998 1999 2000 2001 decline from temperature decline in well behavior.
1600
This is particularly useful when evaluating the
consequences of changes in exploitation strategy that
may result in different pressure or temperature
ENTHALPY (kJ/kg)

1200 decline rates.

800 40
M EASUR ED HISTORY PREDICTION
CALCU LATED 35

STEAM RATE (kg/s)

30
400
25
1994 1995 1996 1997 1998 1999 2000 2001
YEAR 20
15
Figure 2. Historic deliverability match for a well with 10
M EASUR ED
CALCULATED

a single liquid feed zone. 5

M EASUR ED

1994 1996 1998 2000 2002 2004 2006

CONCLUSIONS
1800 HISTORY PREDICTION

1600
ENTHALPY (kJ/kg)

Predictions of well performance made with this 1400

technique provide an evaluation of expected well 1200
behavior superior to decline curve analysis in 1000
Awibengkok. In particular, increases in well enthalpy 800
MEASUR ED
and deliverability with time as a result of flashing of 600 CALCULATED
MEASUR ED
liquid feed zones (Figures 3 and 4) can be predicted. 400
1994 1996 1998 2000 2002 2004 2006
50 YEAR

40 Figure 4. Deliverability match for a well showing a

STEAM RATE (kg/s)

predicted increase in steam productivity once one of

30 the deepest feed zones evolves to dry steam. Squares
20
show the actual values observed after the prediction
MEASU RED
CALCULATED
was made.
10
The practical application of this methodology to a
1994 1995 1996 1997 1998 1999 2000 2001
large field containing more than 30 wells was
possible through a computer application that runs the
2800
wellbore simulator automatically. The application
2400 selects the appropriate values of pressure and flowing
ENTHALPY (kJ/kg)

2000 M EASURED
CALCU LATED
enthalpy of the feed zones for each well at a given
1600 time and select the historic wellhead pressure for the
1200 same given time. It also plots automatically the
measured and calculated steam flow rates and
800
enthalpies to facilitate manual calibration of PI’s.
400
1994 1995 1996 1997 1998 1999 2000 2001
YEAR During the past year this methodology has become
the standard for short-term steam forecasts in
Figure 3. Historic deliverability match for a well Awibengkok with satisfactory results. It has been
showing the evolution of a well from liquid to steam observed that extrapolation of pressure and
in a period close to a year. temperature trends for well models that match
historic behavior can be made more confidently than
steam deliverability forecasts.
ACKNOWLEDGEMENTS REFERENCES

Thanks to the management of Unocal and Unocal Stimac, J. S. and Sugiaman, F. The Awi 1-2 Core
Geothermal Indonesia (UGI) for granting permission Research Program Part I, Geologic Overview of the
to publish this paper. Thanks also to Ken Riedel, Awibengkok Geothermal Field, Indonesia.
Henrikus Amperanto and Gil Batayola from UGI for Proceedings World Geothermal Congress, 2000.
their support and contributions to the project from
which this paper was derived and to Ken Williamson Hasan, A.R. and Kabir, C.S. Fluid Flow and Heat
and Mauro Parini for their review and support. Transfer in Wellbores, Society of Petroleum
Engineers, Richardson, Texas, 2002.

Hirtz, P., Lovekim, J., Copp, J., Buck, C., and

Adams, M. Enthalpy and Mass Flowrate
Measurements for Two-Phase Geothermal
Production by Tracer Dilution Techniques.
Proceedings 18th Workshop on Geothermal Reservoir
Engineering, Stanford University, 1993.