Beruflich Dokumente
Kultur Dokumente
economic point of view. This cannot be accomplished using The drainage mechanism of the reservoir is assumed to be
this technique. natural depletion mechanism.
To overcome the drawbacks presented in the above The production exhibits a certain type of decline during
methods, we need a technique which can: the period of time considered in the history match
Predict the future performance as a function of time in the
computations. That decline can be exponential, hyperbolic
or harmonic according to the behavior of the reservoir
presence of various production components including the
under consideration. This behavior is assessed by using
reservoir.
the decline curve analysis theory and the Fetkovich type
Match the prior production data in the presence of various curve4,5.
production components so that the appropriate parameters For wet gas reservoir, it is assumed that the reservoir
can be assigned for future production prediction. This is pressure is above the dew point pressure. This assumption
similar to decline curve analysis except that we need to implies that the flow is single-phase gas in the reservoir.
include the production components in the system. The well head pressure is reasonably constant throughout
Quantify the uncertainties with respect to various the period of time considered for the history match.
parameters ( e.g., reservoir permeability, skin factor, Other limitations involved in this work depend on the type of
tubing roughness, drainage area, the type of pressure drop correlation selected to compute the pressure losses across the
correlation) by generating alternate possibilities of individual component in the system.
parameters which can match the production data.
History Match. The procedure used to compute the history
Predict the future performance under existing conditions match is summarized in the following steps:
as well as altered conditions to compare the production 1. Assume that the production history is known. Thus, for
scenarios in the future. each observed production time T obs1, Tobs2,, Tobs j,,
Quantify the uncertainty in predicting the future Tobs n, the corresponding observed rate Qobs1, Qobs2, ,
performance which can be combined with the price of gas Qobsj, , Qobsn is known.
to conduct a risk analysis. 2. Assume that at time Tj the following data are known:
fluid properties as a function of pressure and
Optimize the producing well configuration so that the net temperature
profit over the life of the well is maximized. The type of decline (harmonic, hyperbolic or
The system considered in this work is shown in Figure 1. It exponential) as well as the rate of decline. If type
represents a single well producing from a gas reservoir up to is not known, assume exponential decline.
the separator. This system is divided into the following The pressure drop correlations as a function of
completion and piping components: rate for each Q.
reservoir 3. The gas in place at this time Tj is computed as:
perforations
gravel pack V * * S
tubing b g
bottom hole device Gj (1)
Bgj
subsurface safety valve (SSSV)
well head choke where
surface pipeline Vb= reservoir bulk volume
separator reservoir porosity
Sg= gas saturation
Approach Bgj= gas formation volume factor at pressure P j
This section explains the procedure used to combine the
important elements of the decline curve analysis, nodal 4. Calculate the rate Qj at which the well will produce at Pj.
analysis and material balance technique. This is done by using the nodal analysis technique. As
stated earlier, in this study the node is chosen at the
Assumptions. The major assumptions made with respect to bottom hole. The nodal analysis technique is explained in
the flow of gas in the reservoir and the piping system are: ref. 1, 2, 3.
The production system operates under pseudo-steady state
conditions. The well is flowing at a steady flow rate for a
fixed average reservoir pressure and separator pressure.
This implies that the gas well produces with a fixed
liquid/gas ratio.
SPE 52170 GAS WELL PRODUCTION OPTIMIZATION USING DYNAMIC NODAL ANALYSIS 3
5. Assume a small decrement in reservoir pressure Pj. The 9. At this point, we have the model predicted timesT 1, T2,
new reservoir pressure is then P j+1 = Pj- Pj . At this , Tj, Tj+k, and the corresponding rates:Q 1, Q2, ,
reservoir pressure , calculate the new gas in place Qj, , Qj+k, For each observed time T obs j, we calculate
V * * S the corresponding predicted rate Qj by interpolating the
Gj 1
b g (2) model predicted rates. At this point, we check how the
Bg j1 calculated flow rate Qj compares with the historical
observed production rate Qobs j at the same time.
The total amount of gas produced when the reservoir
pressure decreases from Pj to Pj+1 is: 10. If the difference between the predicted and observed rates
is significant, a regression technique is used to adjust
G Gj Gj 1 (3) some of the parameters and the procedure is repeated
(steps 3 - 9). The process terminates when a desirable
6. Calculate the rate Qj+1 at which the well will produce match between predicted and observed rates is achieved.
under the present reservoir pressure P j+1. This is done by The regression analysis technique is discussed below:
nodal analysis.
Regression Analysis. The basic objective of using the non-
7. Knowing the total amount of gas produced (G) and the linear regression in this problem is to determine the optimum
gas flow rate Qj and Qj+1 at reservoir pressures Pj and Pj+1, set, , of reservoir/completion parameters such that the
we can calculate the elapsed time T required to reach observed data match as closely as possible to the calculated
that production data from the model.
In this study, the parameters on which the regression is
For exponential decline:
performed consist of any set of three variables chosen among
1 Qj
the following parameters: permeability, skin, radius of
T ln (4) drainage, pay, perforated interval, radius of perforations,
D Qj 1
diameter of perforations, porosity, water saturation, and
Qj density of perforations. For example, one can choose a set
D Qj 1 Qj Q j 1 such that parameters ={permeability, skin, radius of drainage}.
G G j G j 1 In this case the regression calculations will be performed on
the following variables: permeability, skin and radius of
For harmonic decline: drainage.
In this study, the Levenberg-Marquardt algorithm6,7, has
Qj Qj 1
T
1
(5) been used. This algorithm has been selected because it does
D Q j 1 not require to provide the derivatives of the functions to
Qj Qj
minimize.
D ln An obvious choice of objective function to minimize is the
G Q j1 difference between observed and predicted rates. However,
such choice will be sensitive to outlier data. Considering that
the observed rates fluctuate significantly, it is better to choose
For hyperbolic decline:
an objective function which is more resistant to the outlier
(6) data. In our work, we used a correlation coefficient between
b
T 1 * 1 Qj 1 observed rate and predicted rate as one of the matching
b* D Qj
functions, and we also used a plot between observed and
predicted rates to have a slope of one, and the intercept of
Qj Q 1b zero. If average reservoir pressure data were available, we also
D * 1 j 1
(1 b)* G Qj make sure that it matches with predicted average reservoir
pressures.
The total calculated time when the reservoir pressure is P j+1
The Levenberg Marquardt algorithm6,7 that we use is
can be calculated as: unconstrained: i.e., variables can be chosen to minimize the
objective function with value between infinite. Obviously,
Tj 1 Tj T for our problem, we need to ensure that the values of the
variables lie in the predefined interval of uncertainty and that
8. Repeat the process from step 4 to step 7 until the total these values are meaningful. For example we may want the
calculated time is greater or equal to the observed regressed permeability value to be between K max and Kmin. In
production time. order to keep the values of the regression variables in certain
4 ARSENE B. BITSINDOU, M.G. KELKAR SPE 52170
predefined intervals, the imaging extension6,7 procedure is run are shown in Figure 3. Again, the predicted rate matches
used. very well with the observed data. Also, the predicted reservoir
pressure matches very well with the observed data3.
Future Performance Prediction Future Performance Simulations.
1. The future performance of the well under the existing Future Performance Simulations for Different Skin Values.
conditions as well as under altered conditions can be In order to simulate the effect of a stimulation job (acidizing,
calculated. The procedure is the same as described from fracturation,..) on the performance of the well, the program has
step 2 to step 8 in the History Matching section. Repeat been run with different skin factors. The skin of 116.5, 50 and
the steps till an abandonment rate is reached. 0.0 has been used in the forecast computations. The results of
this sensitivity analysis are shown in Figure 4.
2. Consider different scenarios for variations in production The improvement of the well performance as the skin factor is
procedures. These include, for example, changing the reduced is clearly displayed on the graph. The forecast
number of perforations, stimulating the well, fracturing performance declines faster as the skin is lower. For example
the well, installing the compressor at the surface. the decline rate corresponding to skin 0.0 is greater than the
one corresponding to skin 116.5. This is due to the fact that the
3. Predict the future performance under the new operating removal of the skin does not increase the reserves, but
conditions using the same procedure as explained in Step accelerates the gas recovery.
1.
Field case
4. Repeat Step 3 for alternate combinations of input Case #1: Dry Gas Well Producing at a Constant Well
parameters to quantify uncertainties in the prediction of Head Pressure. This field case represents a dry gas well, open
future performance. to production since 1989. It has been produced at a constant
well head pressure. The initial reservoir pressure is estimated
5. Compare the performance under the new scenario with the to be 2083 psia.
base case to calculate the incremental gas production as a History Match. The computer program was run with the
function of time. regression parameters selected to be radius of drainage, skin
and permeability. The results of the history match are shown in
6. Repeat step 5 for different input configuration. Figure 5 and Figure 6. An excellent production history match
is obtained. The reservoir pressure history match is also in
7. Use information generated in step 5 and step 6 to study
very good agreement with the observed field data. The
the economic feasibility of making the changes in the
calculated values of the regressed parameters as well as the
production configuration.
observed values based on well test of those parameters are
shown in Table 3. The skin exhibits a good agreement
Results between the observed value and the calculated value from well
This section presents the results obtained by applying the test. The permeability and radius of drainage calculated from
dynamic nodal analysis technique to several production the program are higher than the corresponding observed
systems. Those production systems include synthetic data as values. This may be due to the fact that the actual reservoir
well as actual field data. These results validate the dynamic drive mechanism may not be exactly natural depletion. Some
nodal analysis technique. other mechanism such as compaction drive may contribute to
the actual reservoir mechanism. The change in rate observed at
Synthetic Data time 750 days is simply due to a well head pressure
These synthetic data have been generated using the results perturbation that was very limited in time.
from a simulation of an actual field well. They represent a gas Future Performance Prediction Using Different Well Head
condensate well which was open to production for five years. Pressure Values. Different runs of the program were
The characteristics of the reservoir as well as the description of conducted at various well head pressures to simulate the effect
the completion are summarized in Table 1. of the installation of a compressor on the future performance
History Match. The computer program was run with the of the producing system. Well head pressure values of 870
regression parameters selected to be radius of drainage, skin psia, 700 psia, 500 psia, 300 psia, and 100 psia were used in
and permeability. The result of the history match is shown in the forecast computations. The well was producing at a well
Figure 2 and Table 2. As it can be seen, the predicted rate head pressure of 870 psia. The results of these simulations are
matches very well with the observed data. Also, the predicted shown in Figure 7. As can be seen, the well performance
reservoir pressure matches very well with the observed data 3. improves as the well head pressure decreases. However the
Several runs of the program were conducted in order to increase in flow rate is not linearly related to the decrease in
assess the sensitivity of the history match with respect to errors the well head pressure. For example, the gain in flow rate
in the input historical production data. The results of these obtained from reducing the well head pressure from 870 psia
SPE 52170 GAS WELL PRODUCTION OPTIMIZATION USING DYNAMIC NODAL ANALYSIS 5
to 700 psia is about 2400 Mscf/D, whereas the increase in the The calculated values of the regressed parameters as well
well performance is only 500 Mscf/D when the well head as the estimated values (from well test) of those parameters are
pressure is reduced from 300 to 100 psia. This sensitivity shown in Table 4.
analysis is useful to the engineer in the process of deciding The regressed value of the permeability agrees with the
whether or not to install a compressor and under what value obtained from well test. The predicted radius of drainage
optimum conditions it can be operated. is greater than the observed drainage radius. This is probably
Future Performance Prediction for Different Density of due to the fact that the computer program uses volumetric
Perforation. In order to assess the sensitivity of the density of drive mechanism and it has been documented that the reservoir
perforations on the well performance, the program was run drive mechanism for case #2 is not volumetric3.
with different values of perforation densities. Perforation Future Performance Prediction Using Different Perforated
densities of 4 spf, 8 spf and 12 spf are used in the forecast Interval Values. In order to assess the sensitivity of the well
computations. The overbalanced perforation mode is used. The performance with respect to the perforated interval, the
well is actually perforated overbalanced with a perforation program is run with different values of perforated interval.
density of 4 spf. The results of the simulations are summarized Perforated interval values of 17 ft and 64 ft are used in the
in Figure 8. As can be seen, the well performance improves forecast computations. The actual perforated interval of the
slightly as the perforation density increases. However the gain well is 17 ft. The results of the simulations are summarized in
in flow rate remains marginal compared to those obtained by Figure 11.
reducing the well head pressure (by installing a compressor for As it can be seen, the well performance increases as the
example). perforated interval increases.
Future Performance Prediction Using Different Tubing
Case #2: Conversion of the Original Data from Constant Inside Diameter Values. Different runs of the program were
Flow Rate to Constant Well Head Pressure. This field case conducted with various tubing inside diameter values in order
represents a condensate gas production system. The well, open to simulate the effect of the recompletion of the well with
to production since 1989, exhibits a very high condensate yield different tubing size. Tubing inside diameter values of 1.995,
of 145 BBL/MMscf. The initial reservoir pressure is 5011 1.049 and 2.441 have been used in the forecast
psia. The PVT analysis estimates the dew point pressure at computations. The results of these simulations are shown in
5025 psia. The decline curve analysis indicates that the well Figure 12. As can be seen, the well production life is extended
produces with exponential decline. as the tubing size is reduced. For example for tubing sizes of
In order to use the computer program presented in this 1.995 and 2.441, the well dies respectively after 2184 days
work, it is required that the well head pressure be reasonably and 8544 days of production for liquid loading. However the
constant during the period of time considered in the history life of the well is extended well beyond 12000 days when it is
match computations. Case #2 does not satisfy this requirement completed with a tubing of 1.049 inside diameter.
as it is producing with constant rate but not with constant well
head pressure. For this well, the data were converted from Conclusions
constant rate to equivalent constant well head pressure. The Dynamic nodal analysis technique allows to perform sensitivity
Q1
PR 2 PWF1 2 analysis of future performance for gas wells once a satisfactory
conversion equation used is the following: P 2 P 2
match of the past production performance is obtained. The major
Q
2 R WF 2
contribution of this work is that it provides a tool to analyze the
(7) well performance changes as a function of time when the
Q1 is the actual constant flow rate corresponding to the production parameters are altered. The classic nodal analysis can
flowing bottom hole pressure Pwf1. Since Q1 and Pwf1 are only be used if the production parameters remained unchanged.
known, the flow rate Q2 can be computed by assuming a fixed The dynamic nodal analysis provides valuable means to help the
value of the corresponding bottom hole pressure P wf2. engineer in decisions making. Opening a gas well to production
This conversion technique works well if the total reservoir always involves considerable expenses whereas a model can be
pressure decline is small during the time period considered for run many times at lower cost to try many different possible
history match calculations, and the reservoir is producing scenarios in order to make technical and economical decisions.
under pseudo-steady state conditions. It should be noted that the prediction of the future performance
History Match. The computer program was run with the based on history match of well performance is not unique. There
regression parameters selected to be radius of drainage, skin are many other sets of system parameters that can match the past
and permeability. The results obtained are presented in performance of the well. There is always some uncertainty
Figure 9 and Figure 10. An excellent production history associated to the model used to arrive at a satisfactory historical
match is obtained. The reservoir pressure history match is also performance match. Based upon the history match results, the
very good. engineer can obtain a range of future performances, and hence
can make a decision in light of uncertainties.
6 ARSENE B. BITSINDOU, M.G. KELKAR SPE 52170
The computer program presented in this paper is capable of 6. Carvalho, R. Thompson, L.G., Redner, R. and Reynolds, A.C.:
history matching the production data as well as predicting the Simple Procedure for Imposing Constraints for Nonlinear Least
future performance under different scenarios. The program has Square Optimization, paper SPE 29582.
been validated with the help of both synthetic and field data. The 7. Carvalho, R.: Nonlinear Regression: Application to Well Test
program provides a logical improvement to conventional nodal Analysis, PhD Dissertation, University of Tulsa, Tulsa, OK
analysis. (1993).
Nomenclature
b = decline exponent for hyperbolic decline behavior. Table 1-SYNTHETIC DATA: INPUT PARARMETERS
Bg = gas formation volume factor, cf/scf
= reservoir porosity Type of decline = exponential
Pressure decrement [psia] = 50
G = gas in place, Mscf Optimization tolerance = 0.000001
G = gas in place decrease, Mscf
P = pressure, psia. Reservoir
Initial pressure [psia] = 5011
P = pressure drop, psia Initial temperature [F] = 212
Q = gas flow rate, Mscf/D Pay [ft] = 64.5
Qobs = observed flow rate, Mscf/D Skin = 116.5
Sg = gas saturation, fraction Drainage radius [ft] = 9108
Permeability [md] = 11
T = temperature, R Porosity [fraction] = 0.06
Tobs = Observed production Time, days Water saturation [fraction] = 0.533
T = elapsed time, days
Fluid properties
Vb = reservoir bulk volume, Mscf Specific gravity of produced gas = 0.646
set of 3 independent regression variables Oil density [API] = 51.1
Specific gravity of produced water = 1.0
Subscript
Completion
g = gas Hole diameter [in] = 8.496
R = reservoir Casing diameter [in] =5
WF = at bottom hole in well flowing conditions Perforated interval [ft] = 17
Perforation diameter [in] = 0.36
Perforation tunnel length [in] = 12.33
Acknowledgement Perforation density [SPF] =4
The authors thank the University of Tulsa for provinding the Mode of perforation = overbalance
computer facilities used to conduct this study. They also express Tubing inside diameter [in] = 1.945
Tubing roughness [ft] = 0.00015
their gratitude to Dr Leslie G. Thompson of the University of Tubing length [ft] = 8688.0
Tulsa, and Stuart Cox of Marathon Oil Co. for their comments Hole inclination angle [degree] = 90
and suggestions. They are also grateful to Marathon Oil Co. for Pressure drop correlation: Beggs and Brill
providing the field data used during the test of the computer
Production
program. Oil/Gas ratio, [SBBLO/MMscf] = 145.0
Water/Gas ratio, [SBBLW/MMscf] = 0.0
References Well head pressure, [psia] = 2250.0
Well head temperature, [F] = 111.0
1. Brown, K.E. et al.: The technology of artificial lift methods, Reference separator pressure, [psia] = 14.7
Pennwell Publishing Company, Tulsa, OK (1984), volume 4. Reference separator temperature, [deg F] = 60.0
2. Perez, G. and Kelkar, B.G.: A Simplified Method to Predict
Over-all Production Performance, Journal of Canadian Petroleum Limits of regression parameters
Technology, January-February, 1990, Volume 29, No. 1. KMIN [md] = 0.0 KMAX [md] = 100.0
SMIN = -5.0 SMAX = 175.0
3. Bitsindou, A.: Gas Well Production Optimization Using Dynamic reMIN [ft] = 2500.0 reMAX [ft] = 10000.0
Nodal Analysis, M.S. thesis, University of Tulsa, Tulsa, OK
(1998).
4. Fetkovich, M.J., Fetkovich, E.J. and Fetkovich, M.D.: Useful
Concepts for Decline-Curve Forecasting, Reserve Estimation, and
Analysis, paper SPE 28628 presented at the 1994 SPE Annual
Technical Conference and exhibition, New Orleans, Sept. 25-28.
5. Fetkovich, M.J.: Decline Curve Analysis Using Type Curves,
JPT, June 1980, 1065-1077.
SPE 52170 GAS WELL PRODUCTION OPTIMIZATION USING DYNAMIC NODAL ANALYSIS 7
TABLE 3-HISTORY MATCH FOR CASE #1 Fig. 1-System description and pressure losses.
Regression Calculated Initial Comment
1850
Drainage radius [ft] 4532.7 1000.0 Estimated
1800
1750
1700
TABLE 4-HISTORY MATCH FOR CASE #2 0 500 1000 1500 2000
Time [days]
Regression Calculated Initial Comment
2500
Observed data [after adding errors between
18000 -10% and +10% to the original data] Observed reservoir pressure
16000 Predicted rate [No error added to the 2000 Predicted reservoir pressure
original data]
10000
1000
8000
6000
500
4000
2000 0
0 500 1000 1500 2000 2500 3000
0 Time [days]
0 500 1000 1500 2000 2500 3000
Time [days] Fig. 6-Case #1:reservoir pressure history match.
3500 20000
Observed rate Observed rate [Well head pressure = 870 psia]
History match [skin=116.5] 18000
Predicted rate [Well head pressure = 870 psia]]
3000 Skin=116.5
16000 Well head pressure= 870 psia
Skin=50.0
Skin=0.0 Well head pressure = 700 psia
2500 14000
Well head pressure = 500 psia
1500 8000
6000
1000
4000
500
2000
0 0
0 1000 2000 3000 4000 5000 6000 7000 8000 0 1000 2000 3000 4000 5000 6000
Time [days]
Time [days]
16000
16000
14000 Observed rate [SPF=4]
14000
Observed rate Predicted rate [SPF=4]
12000 SPF=4
12000 Predicted rate SPF=8
10000 SPF=12
10000
Rate [Mscf/D]
Rate {Mscf/D]
8000
8000
6000
6000
4000
4000
2000
2000
0
0
0 1000 2000 3000 4000 5000 6000
0 500 1000 1500 2000 2500 3000
Time [days]
Time [days]
Fig. 5-Case #1: production history match. Fig. 8-Case #1: sensitivity of rate with respect to perforation
density.
SPE 52170 GAS WELL PRODUCTION OPTIMIZATION USING DYNAMIC NODAL ANALYSIS 9
3000
Observed rate [Perforated interval= 17 ft]
2000
History match [Perforated interval= 17 ft]
Perforated interval= 17 ft
Observed rate 2500
1950 Perforated interval= 64 ft
Predicted rate
Rate [Mscf/D]
1900 2000
Rate [Mscf/D]
1850 1500
1800
1000
1750
500
0 500 1000 1500 2000 2500 3000 3500 4000
1700
Time [days]
0 500 1000 1500 2000
Time [days]
Fig. 11-Case #2: sensitivity of rate with respect to perforated
interval.
Fig. 9-Case #2: production history match.
5050 2000
Observed rate [tubing size = 1.995 in.]
5000 Predicted reservoir pressure 1800 Predicted rate [Tubing size = 1.995 in.]
4850
1000
4800 800
4750 600
400
4700
200
4650
0
0 500 1000 1500 2000
0 2000 4000 6000 8000 10000 12000 14000
Time [days] Time [days]
Fig. 10-Case #2: reservoir pressure history match. Fig. 12-Case #2: sensitivity of rate with respect to tubing inside
diameter.