Vous êtes sur la page 1sur 10

IPTC 10382

Experience With Simulation of Condensate Banking Effects in Various


Gas/Condensate Reservoirs
J.W. Barker, Total S.A.

Copyright 2005, International Petroleum Technology Conference


Many gas condensate reservoirs are close to their dew
This paper was prepared for presentation at the International Petroleum Technology
Conference held in Doha, Qatar, 2123 November 2005.
point pressure at initial conditions, so there may be no
reduction of productivity because the initial productivity
This paper was selected for presentation by an IPTC Programme Committee following review
of information contained in an proposal submitted by the author(s). Contents of the paper, as is already a reduced one due to condensate deposition
presented, have not been reviewed by the International Petroleum Technology Conference
and are subject to correction by the author(s). The material, as presented, does not necessarily
right from the start of production.
reflect any position of the International Petroleum Technology Conference, its officers, or Velocity-dependent relative permeability (VDRP) effects
members. Papers presented at IPTC are subject to publication review by Sponsor Society
Committees of IPTC. Electronic reproduction, distribution, or storage of any part of this paper may mean that there are many gas condensate reservoirs
for commercial purposes without the written consent of the International Petroleum Technology where condensate banking either does not occur or has no
Conference 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 effect on productivity (i.e. the gas relative permeability
acknowledgment of where and by whom the paper was presented. Write Librarian, IPTC, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435. near the well is no lower than that far from the well,
whether or not the condensate saturation near the well is
Abstract greater than that in the bulk of the reservoir).
In this paper, we describe our experience of using the velocity- Extensive experimental research into the dependence of
dependent relative permeability model developed by Heriot- gas and condensate relative permeability on flow velocity and
Watt University and the generalized pseudo-pressure model of gas-oil interfacial tension has been conducted at Heriot-Watt
Fevang and Whitson to simulate condensate banking effects. University over the past decade1-5, sponsored by a consortium
For cases where the generalized pseudo-pressure model fails, of oil companies. Results of the experiments have clearly
we discuss alternative approaches to simulating condensate demonstrated that relative permeability increases as flow
banking effects in full field models, together with their velocity increases and as gas-oil interfacial tension decreases.
limitations. Our methodology is illustrated on a number of A mathematical model for relative permeability has been
different examples, drawn from reservoirs in Indonesia, the derived from the results3 and has been implemented in some
Middle East and the North Sea where Total is either the commercial reservoir simulators. The model includes non-
operator or an active partner. Darcy flow (turbulence) effects. This model can be used to
study the condensate banking phenomenon in detail, using
Introduction grids fine enough to resolve the near-well region.
Condensate banking is a well-known phenomenon in gas Other research, notably by Fevang and Whitson6, has led to
condensate reservoirs. It occurs when the bottom-hole pressure a generalized pseudo-pressure (GPP) model for well in-flow
of a production well falls below the dew-point pressure. performance in gas condensate reservoirs, which has also been
Condensate drops out in the reservoir around the well and implemented in some commercial simulators. This model
rapidly accumulates as each cubic metre of fresh gas produced allows condensate banking effects to be modeled in a much
deposits further condensate due to the pressure near the well coarser grid of the type typically used for full field modeling
being lower than the pressure further away from the well. The studies.
accumulation of condensate near the well reduces the gas In this paper, we describe our experience of using these
saturation there, in turn reducing the gas relative permeability models to simulate condensate banking effects in a number of
and hence reducing the the well productivity. different reservoirs. First however, we briefly recall the main
Although the potential for the occurrence of condensate points of each of the models.
banking is widely recognized, there is no consensus within the
industry as to when it occurs and whether or not it will have an Effect of flow velocity and interfacial tension on
effect on well productivity. There are several possible reasons relative permeability
for this: The mathematical model derived from the results of the
Detecting the presence of a condensate bank from well experiments at Heriot-Watt University is described in
test data is not easy because it may be masked by well references 3 and 7. The effects of flow velocity and (gas-oil)
bore storage effects, because the data may not be of good interfacial tension are not independent and are expressed via a
enough quality, or because velocity-dependent relative dimensionless capillary number defined by:
permeability effects complicate the analysis. vg g
Decline in productivity with time may be caused by Nc = (1)

effects other than condensate banking.
Note that the same value of Nc is used for both phases.
2 IPTC 10382

If the capillary number is less than a threshold value, Ncb, sp = * (1 S w,irr ) c k d (8)
the base relative permeability curve (krb) is used. Low
capillary number corresponds to the low velocity conditions assuming that the end-point relative permeability to gas is
encountered in the bulk of the reservoir and to the low equal to unity.
velocity/high interfacial conditions under which relative
permeability curves are conventionally measured. Generalised Pseudo-Pressure Model
Once the capillary number exceeds the threshold, an The usual inflow performance relationship for a gas in a
interpolation between the base relative permeability curve and reservoir simulator is:
a miscible curve (krm) is used: Q = CF .k r (S b ).( b w ) (9)
1 1
k r = r n k rb + (1 r n )k rm (2) where CF is the connection transmissibility, subscript b
denotes a grid block value and is the real gas pseudo-
Here the interpolation parameter, r, is given by: pressure defined by:
N cb
r= (3) s RT P
1
Nc ( P) = 2( ) dP (10)
M Pref
B
and the miscible relative permeability curve is a straight line
with a residual saturation that reduces from its base value (Srb) This relationship assumes constant values of saturation and
to zero as the capillary number increases: relative permeability within the grid block. Non-Darcy flow
effects are accounted for by a rate-dependent skin.
S * Sr
k rm = (4) The generalized pseudo-pressure (GPP) model of Fevang
1 S r
and Whitson6 accounts for the variations of saturation and

( )
mr
relative permeability within the grid block that are caused by
Sr = 1 e S rb (5) condensate banking. The relative permeability term in
S equation (9) is no longer treated as a constant, but is included
S* = (6) in a pseudo-pressure integral similar to that of equation (10).
1 Sw Details are given in references 6 and 7. Evaluation of the
The model thus has three parameters, Ncb, m and n, which may integral now requires knowledge of the variation of saturation
be different for each phase (gas and condensate). More with pressure. This can be obtained by assuming that the
generally, n may vary with saturation, but we do not use this flowing composition is constant over the pressure range of
feature. interest.
This assumption requires that the so-called region 1
Non-Darcy Flow extends over most of the grid block and that conditions in the
Reservoir simulation models are generally based on Darcys grid block are changing relatively slowly (i.e. that a pseudo
Law, which states that the pressure gradient is proportional to steady state exists). Region 1 is the region around the well in
the flow rate. This is valid at sufficiently low rate, but at which both gas and condensate are mobile6. The assumption is
slightly higher rates, the pressure gradient is observed to therefore more likely to be valid for smaller grid blocks than
increase more rapidly than the flow rate. This can be described for larger ones.
by adding a quadratic term to obtain the well-known VDRP and the non-Darcy flow effect can be incorporated
Forchheimer equation: in the GPP model7. However, this requires additional
dp g v g assumptions in order to determine the fluid velocity at each
= + g v g2 (7) pressure in the pseudo-pressure integral. In the software we
dx k use7, the flow is assumed to be radial and the grid block
Non-Darcy flow effects are usually negligible, except in the pressure is assumed to occur at the Peaceman pressure
vicinity of wells in some gas reservoirs. Where such effects equivalent radius (which isnt necessarily the case when
are seen, they are often referred to as turbulence, though in condensate is present).
fact the flow remains laminar and the extra pressure drop is
due to inertial effects. In full field simulation models, non- Methodology
Darcy flow effects are often modeled simply as a rate- The first question in any condensate banking study is whether
dependent skin. or not a significant reduction in productivity is likely to occur.
The Forchheimer or non-Darcy flow coefficient is To answer this we use a finely-gridded single well model. For
usually measured for single-phase flow (in the presence of vertical or near-vertical wells this is usually a radial model,
irreducible water saturation), or inferred from multi-rate well either one- (single layer) or two-dimensional (layered model).
tests. When condensate is deposited in the reservoir, it is For horizontal or deviated wells, or for fractured wells, a
necessary to account for the variation of with gas saturation. Cartesian local grid refinement is used. Ideally, this model is
In the VDRP model developed by Heriot-Watt University3, compositional and accounts for VDRP and non-Darcy flow. If
this dependence is expressed as: appropriate, it may be necessary to model explicitly the
( S g ) = * S gc (kk rg ) d (8) perforations and any formation damage or stimulation.
If productivity impairment does seem to be a possibility,
Note that the value of for single phase flow is given by: the second question is how to simulate it in a full field model.
Ideally, the GPP method would always be used. Indeed, if we
IPTC 10382 3

had complete confidence in the GPP method, it would not be reduction in productivity below the dew point. Even though
necessary to use the finely gridded single well model. the condensate saturation still reaches 30% near the well, the
However, in our experience, the GPP is not always valid. gas relative permeability remains above 0.2 due to the positive
Additionally, if the full field model is black oil, it is not effect of the high velocity. The third simulation includes the
currently possible to take into account the VDRP effects with non-Darcy flow effect as well as VDRP. This leads to a slight
the software available to us. Thus, alternative methods are reduction in productivity before the dew-point is reached
needed. (when the Forchheimer coefficient is equal to its single-
The most obvious alternative is to use an LGR around each phase value), and to a greater reduction afterwards, when
well. This is feasible in certain cases, but the LGR cells near increases as the gas saturation decreases. Overall, however,
the well usually have to be very small to resolve the the positive effect of the velocity-dependence outweighs the
condensate bank explicitly and the computing time may negative effect of the non-Darcy flow in this case.
therefore be prohibitive, especially if there are many wells. A
possible compromise is to use both an LGR and the GPP Cartesian Model Results. Figure 4 shows results of the
model. In this case, the LGR does not have to be fine enough same three simulations made with the Cartesian model. In the
to resolve the condensate bank, so computing time remains Cartesian model, the non-Darcy effect is represented by a rate-
reasonable, while the GPP method may work better on the dependent skin (D-factor) rather than via the Forchheimer
smaller cells of the LGR. coefficient . Also, because of software limitations, the
We shall discuss some other alternatives later in the paper. velocity-dependent relative permeability and the D-factor
Whichever method is used, a key point is to ensure that the could not both be used in the same simulation. However,
results of the fine grid single well model are matched using a Figure 4 shows that the non-Darcy flow effect is negligible in
coarse grid model of the same well the coarse grid model the Cartesian model because the coefficient remains at its
having the same grid size as the full field model. single-phase value throughout. The effect of velocity-
In the following sections, we illustrate aspects of this dependent relative permeability is also small, because the
methodology on a number of different cases. region of high velocity around the well is not resolved, due to
the large grid block size.
Case Study Reservoir A
Reservoir A is located in the North Sea and is relatively rich in GPP Model Results. Figure 5 compares the radial and
condensate (~2000 sm3/Msm3) with a maximum condensate Cartesian models for the case with no VDRP and no non-
saturation of 0.24 in a constant volume depletion experiment Darcy flow. It also shows the result of using the GPP inflow
(Figure 1). Initial pressure is well above the dew point. A relationship in the Cartesian model. The Cartesian model with
study was made to estimate the potential impact of condensate the conventional inflow relationship under-estimates the
deposition on well productivity once the dew point is reached. productivity reduction, compared to the radial model. The
A single well, one-dimensional, radial model was set up reduction also occurs later in the Cartesian model than in the
using average reservoir properties in the vicinity of a radial one. When the GPP inflow relationship is used, the
particular well (see Table 1). Very small grid blocks are used Cartesian model results are quite similar to the radial model
near the well in order to accurately represent the deposition of ones, especially just after the dew point is reached. The GPP
condensate there. Fluid properties were described by a 7- model captures the earlier impact of condensate deposition
component equation of state model. Base relative permeability seen in the radial model. However, later on, the GPP model is
curves for gas and condensate are shown in Figure 2. A small somewhat optimistic compared to the radial model.
amount of mobile water is present in the model. A gas Figure 6 makes the same comparison for the case with
production rate is set for the well, subject to a minimum tubing VDRP, but still with no non-Darcy flow. The Cartesian model
head pressure (THP). Vertical flow performance tables are with the conventional inflow relationship is now pessimistic
used to relate THP to flowing bottom-hole pressure (BHP). compared to the radial model. However, when the GPP inflow
A Cartesian model was also set up with grid blocks the relationship is used, the Cartesian model becomes overly
size of those used in the full field model. This model covers optimistic, predicting essentially no loss of productivity at all.
the same area as the radial model, and uses the same reservoir Some numerical instability is also evident in the GPP case,
properties, EOS model and base relative permeability curves. despite damping of the changes in the blocking factor from
one time step to the next.
Radial Model Results. Figure 3 shows the predicted gas Finally, Figure 7 makes the same comparison for the case
production rate for three simulations made with the radial with both velocity-dependent relative permeability and non-
model. The first simulation uses the base relative permeability Darcy flow. The GPP model is again overly optimistic, in fact
curves throughout, neglecting the velocity-dependence. It also more so. The radial model is the only one to see any
neglects the non-Darcy flow effect. When the pressure at the significant impact of the non-Darcy flow.
well reaches the dew point, there is a sharp reduction in To summarise, in this case the GPP model seems to work
productivity due to the formation of a bank of condensate with reasonably well without VDRP, but not with VDRP. Nor does
saturation up to 35%. At its maximum, the condensate bank it capture the non-Darcy flow effect.
extends to a distance of between 10 and 20 metres from the
well. The gas relative permeability falls from 0.8 to about
0.04. The second simulation includes VDRP with the base
parameter values shown in Table 2: there is now only a slight
4 IPTC 10382

Case Study Reservoir B wells B-1 (40) and B-2 (40), suggesting a stronger
Reservoir B is a gas reservoir in Indonesia. The reservoir dependence on permeability than initially estimated
consists of several producing layers, and within each layer It is encouraging that the behaviour of the wells can be
there are many independent sand bodies. Gas composition matched with parameters not too far from their initial values.
varies with depth and between layers. The condensate content The lack of complete consistency between the matches serves
is generally low (100-400 sm3/Msm3) but the saturation as a reminder that uncertainty remains on the parameter values
pressure is usually close to the initial pressure. Most of the for the velocity-dependent relative permeability model, as it
wells have commingled production from several layers. does on any reservoir property. It should also be noted that, in
Three wells were selected for detailed analysis. Each had the simulation model, condensate banking is the only
produced for a lengthy period from a single sand body, and a mechanism for reducing productivity so although the observed
back pressure test (BPT) had been performed at the beginning reduction in PI can be explained by condensate banking, other
and at the end of this period. explanations are not excluded.
All three wells showed a lower productivity during the
second BPT than during the first. Conventional analysis of the Predictive Simulations. The matched models were re-run
BPTs revealed no indication of the presence of a condensate with a fixed THP, starting from the initial conditions. A
bank, except for an increased skin for wells B-1 and B-3 (11.7 classical pseudo-pressure based productivity index was
and 0 instead of 4.2 and -2.2 respectively). calculated from the gas production rate, bottom hole and
A study was made to determine whether condensate reservoir pressures at each time step (i.e. making no allowance
banking could explain the productivity reduction and if so, to for the presence of condensate).
examine the implications for future well productivity. This PI is plotted against reservoir pressure for well B-2 in
A one-dimensional, radial model was set up for each well. Figure 10 (a number of cases are shown with different THP
A total of 36 grid blocks were used, with size varying limit, but all show the same behaviour). The corresponding
gradually from 0.5 metres near the well to over 100 metres far values of condensate saturation in the first grid block are
from the well. Parameters used in the models (after history shown in Figure 11, along with the liquid drop-out curve from
matching) are shown in Table 3. A 7-component EOS model a CVD experiment (scaled to account for the presence of
was used for the fluid properties. Well B-1 produces a richer water). There is a drop in productivity corresponding to the
gas than the other two wells. Base relative permeability curves formation of a condensate bank around the well. Condensate
are shown in Figure 8 (the same curves are used for each well saturation near the well reaches about 30% even though the
but scaled to the appropriate irreducible water saturation). maximum liquid drop-out is only 2%. Productivity later
Parameters for the VDRP model and for the non-Darcy flow recovers as the condensate saturation near the well decreases.
effect were estimated from the Heriot-Watt University This is a sort of reverse banking effect as the bottom hole
experimental data. pressure becomes less than the pressure corresponding to the
maximum liquid drop-out.
History Matching. The actual historical production rates The behaviour of the other two wells is similar, except that
were imposed on each well, and the model parameters were for the lower permeability well B-3, productivity does not
adjusted to obtain a match to the pressures recorded during the recover. At the minimum productivity, one third to one half of
different flow periods of the BPTs. The parameters adjusted the initial productivity is lost in all three wells. At its
were the mechanical skin S, the non-Darcy flow coefficient , maximum, the condensate bank extends to over 50 metres
and the parameters m and Ncb in the velocity-dependent from well B-1, between 10 and 20 metres for the other wells.
relative permeability model. For well B-1, the dew-point
pressure was also adjusted. The values shown bold in Table 3 GPP Model. Cartesian models with 99 grid cells were built
are those that were adjusted during the matching. Figure 9 for each well and the predictive simulations re-run using the
shows the match obtained for well B-1. The quality of the GPP option. Results are shown in Figures 12-14. For well B-1
matches for the other wells is similar. (Figure 12), the GPP model is pessimistic, predicting an initial
The matches to the three wells are in many ways rate that is lower than that given by the radial model. For well
consistent, but not entirely so: B-2 (Figure 13), the GPP model is initially slightly optimistic
The matched values for mechanical skin are very but in general works very well. For well B-3 (Figure 14), the
different. GPP model works very well throughout.
The non-Darcy coefficient is divided by 10 for wells B- Exactly why the GPP model should work well in two cases
1 and B-3, but multiplied by 20 for well B-2. This is at the but not in the third is not completely clear. The distinguishing
limit of the range of uncertainty on this parameter, and feature of the B-1 well is that it produces the richer gas, but it
although the value still increases with increasing is not particularly rich in absolute terms. Further investigation
permeability, the variation is much bigger than in the appears to show that the hypothesis of a constant flowing
original estimation. composition throughout the cell containing the well is not
The velocity-dependent relative permeability parameter valid for well B-1.
Ncb reduced from 2x10-6 to 2x10-7 for wells B-2 and B-3
but left unchanged for well B-1. Case Study Reservoir C
The velocity-dependent relative permeability parameter m Reservoir C is a gas reservoir in the Middle East with a
reduced from 30 to 3 for well B-3, but unchanged for condensate content of 482106 m3/m3. Initial pressure is about
IPTC 10382 5

100 bars above the dew point. The objective of the study was ACTIONR and WPIMULT keywords of the Eclipse
to determine the optimal operating strategy to maximize simulator). Further multipliers of 0.7, 0.6 and 0.6 are applied
condensate production. A three-dimensional, Cartesian, layer- at region pressures of 500, 450 and 420 bars respectively.
cake model was built, covering the drainage volume of a Figure 16 shows that a good match to cumulative production
single well. The model contains 13 reservoir layers, separated has indeed been achieved by using these multiplier values.
by impermeable shales. Layer thickness varies from 3 to 30 Note that multipliers are applied for region pressures well
metres (the thickest layers are represented by two numerical above the dew point pressure of 390 bars: this is because the
layers); layer permeability varies from 1 to 150 mD. bottom-hole pressure reaches the dew point and a condensate
The grid size used in the Cartesian models was varied: bank starts to form while the cell pressures are still well above
models 1, 2 and 3 used cells of size 300m300m, 100m100m the dew-point. The low permeability in this case results in a
and 42.9m42.9m respectively (each model covers the same large drawdown. However, the drawdown obviously depends
area). Further, each model was run three times, first with no on the rate, illustrating one disadvantage of this approach
special treatment for condensate banking, second with a radial which is that it may not be robust to changes in rate or other
LGR in the column of cells containing the well, and third operating conditions. The other disadvantage is that the
without the LGR but using the GPP option. The radial LGR matching process may be relatively time-consuming.
contains 10 cells between the well radius and an outer radius Finally note that in this case, the overall multiplying factor
of 30m. A fluctuating gas production rate was imposed, is 0.70.70.60.6=0.18, i.e. a 5-fold reduction in connection
corresponding to the historical production rate of a particular transmissibility. However, VDRP has been neglected, so the
well in the field. Initially, velocity-dependent relative result may be pessimistic.
permeability and non-Darcy flow are neglected.
Figure 15 shows the predicted bottom-hole pressure for Upside and downside cases. It should be remembered
each model and each option (radial LGR, GPP or neither). The that there is always uncertainty associated with the parameters
Cartesian models with no LGR see no condensate effect. All needed to model the condensate banking effect accurately. In
three models with LGR give the same result, with an extra particular, the parameters in the velocity dependent relative
drawdown, compared to the previous cases, that increases permeability model are relatively poorly constrained at
gradually one the dew point is reached and attains 60 bars by present, due to the limited amount of experimental data
the end of the simulation. All three models with the GPP available. There is also uncertainty over the effect of the
option are pessimistic, particularly just after the dew point is perforations and formation damage in the near-well region.
reached when an extra drawdown of over 100 bars is Figures 3-7 presented results for Reservoir A using the
predicted. The error decreases gradually at later times. Greater base case parameters for the VDRP model given in Table 2.
errors are seen on the coarser grids than on the finer ones. Figures 17-18 show results from radial model simulations,
The likely explanation for the failure of the GPP model in with VDRP and non-Darcy flow effects, for a range of
this case is that the zone of condensate deposition is parameters. In these simulations, the parameters were varied
significantly smaller than the grid block size. If the simulation one at a time between the minimum and maximum values
is run for a longer period, the model may become more given in Table 2. Also shown are the results of two Cartesian
accurate as the size of the zone increases. model simulations, one in which no special action is taken to
model condensate banking, the other in which the connection
Alternatives to the GPP Approach transmissibility is halved when the average pressure in the cell
We have seen that the GPP model may not always give good containing the well falls to 6 bars above the dew point (this
results, for a number of possible reasons. Alternative second case was in fact obtained by matching against a radial
approaches for accounting for condensate banking effects in a model neglecting VDRP effects, which was one of the more
full field model are therefore sometimes needed. As already pessimistic radial models). It can be seen that there is a
mentioned, the simplest would be to use an LGR around each considerable spread of results from the radial models, and that
well, but this may be prohibitive in terms of computing time. the two Cartesian models more or less bracket the radial model
We therefore discuss three other possible alternatives. responses, at least in terms of cumulative production (Figure
18). These two cases, neglect of condensate banking and
Manual adjustment of connection factors. In this reduction of the connection transmissibility by a factor of two,
approach, the idea is to manually adjust the connection factors could thus be retained as up- and down-side cases in the full
during the coarse grid simulation so as to match the results of field model study.
the fine grid single well model. The coarse grid simulation We note in passing that the parameter with the most impact
model here is the single well model, having the same grid size on the results is Ncb. The two extreme radial runs in Figures
as the full field model. 17-18 correspond to the minimum and maximum values of
This is illustrated in Figure 16 for another case study, this parameter. The range of values used for Ncb in this case is
Reservoir D, a low permeability (~1 mD), relatively rich possibly a bit wide, however.
(~900 sm3/Msm3) gas condensate reservoir in the North Sea.
In this case, a region is defined consisting of the column of Sensitivity case. This approach is similar to the previous
cells in which the well is completed. A multiplicative factor of one but it dispenses with the radial models. It consists of
0.7 is applied to the transmissibility of all the connections neglecting condensate banking in the base case model, and
when the region pressure falls below 580 bars (using the testing a reduction of the connection factors as a sensitivity
case. The reduction factor is chosen arbitrarily, usually
6 IPTC 10382

between 1.5 and 3, and applied at a pressure just above the In some cases it is reasonably clear that productivity
dew point. reductions due to condensate banking will be negligible, or of
This approach is justified if one is confident that the second order compared to other factors. In these cases, a
reduction in productivity due to banking will be negligible. productivity reduction may be retained as a sensitivity case,
This is often the case in high permeability (100 mD or greater) but no fine grid simulation would be performed. Ideally, the
reservoirs. Even where some reduction does occur, it may be conditions under which this approach can be taken with
of second order compared to other uncertainties on well confidence could be defined by a systematic study to
productivity such as formation permeability or non-Darcy determine the conditions under which productivity reductions
flow effects. Productivity reduction may also be a second occurs. Although we have made a start on this, the parameter
order effect if drawdown in the reservoir is much smaller than space is very large and no definitive conclusions can be drawn
the pressure losses in the tubing (often the case for deeply- as yet.
buried reservoirs of reasonable permeability).
We have in fact already carried out some investigations of Acknowledgements
a generic nature in an attempt to determine when condensate The author is very grateful for the contributions to this paper
banking leads to significant productivity reduction and when it made directly or indirectly by many colleagues at Total
does not. A systematic series of one-dimensional radial including Grald Hamon, Bruno Decroux, Olivier Bernet-
simulations was carried out for a range of permeability from Rollande, Batrice Glachant, Jean-Luc Boutaud-de-la-Combe,
0.001-1000mD and for four different values of CGR (744, Chantal Ducap, Roger Brousse, Guy Debray and Marie-Laure
1010, 1130 and 1400 sm3/Msm3). For each case, simulations Baratin; by Mohamed Al-Nuaimi of Dolpin Energy Ltd,
were run with and without the VDRP model. during his secondment to Total; and by Tu Loan Tran and
The results showed that productivity reduction (expressed Antoine Rech during their student placements at Total.
as a skin see Tables 4 and 5) occurred essentially for the
richer fluids in the lower permeability rocks. VDRP was References
significant for intermediate permeabilities (0.1-10mD): at 1. Danesh, A., Khazam, M., Henderson, G.D., Tehrani, D.H.,
lower permeability the velocity was not high enough while at and Peden, J.M., "Gas Condensate Recovery Studies",
higher permeability there was no reduction in productivity Proceedings of the UK DTI EOR Seminar, London (June
even without VDRP. 1994).
2. Henderson, G.D., Danesh, A., Tehrani, D.H., and Peden,
However, caution should be exercised in generalizing these J.M., "Measurement and Correlation of Gas Condensate
results. All were obtained using a single EOS model, the same Relative Permeability by the Steady-State Method", SPE
parameters in the VDRP and non-Darcy flow models, the Journal, June 1996.
same initial pressure with respect to the dew point, the same 3. Danesh, A., Tehrani, D.H., Henderson, G.D., Al-Kharusi, B.,
base relative permeability curves, etc. Further work is needed Jamiolahmady, M., Ireland, S. and Thomson, G., Gas
to complete an exhaustive search of the parameter space, even Condensate Recovery Studies: Near Wellbore Relative
for a single-layer, homogeneous model. Thus, we are not yet Permeability Including The Inertial Effect, Proceedings of
safe from a surprise if this method is adopted too often. the UK DTI IOR Seminar, London, June 2000.
4. Jamiolahmady, M., Danesh, A., Tehrani, D.H., Duncan,
D.B., A Mechanistic Model of Gas Condensate Flow in
Conclusions Pores", Transport In Porous Media, Vol 41, No. 1, pp. 11-46
We have outlined our methodology for simulating the effects (October 2000).
of condensate banking. Finely-gridded, compositional, single 5. Jamiolahmady, M., Danesh, A., Tehrani, D. H. and Duncan,
well models (radial or otherwise) are used, with velocity- D. B., Positive Effect of Flow Velocity on Gas-Condensate
dependent relative permeability and non-Darcy flow. These Relative Permeability: Network Modelling and Comparison
are considered to be the most accurate models available today, with Experimental Results, Transport in Porous Media Vol.
even though uncertainty remains on the input parameters. 52, No. 2, pp. 159-183, 2003.
If productivity impairment does seem to be a possibility, it 6. Fevang, O, and Whitson, C.H., Modeling Gas Condensate
can sometimes be simulated in a full field model using the Well Deliverability, SPE 30714, SPE Reservoir
Engineering, November 1996, pp. 221-230.
GPP method. However, this method should always be 7. Eclipse Technical Description, version 2004a_1,
compared against the finely-gridded model as we have seen Schlumberger.
several cases where it is not accurate. In cases where the GPP
model fails, introduction of a very fine LGR around each well Nomenclature
is the simplest alternative but is often prohibitively expensive. B = Formation volume factor
Combining a less fine LGR with the GPP model is a possible c,d = Parameters in VDRP model
solution in some cases. CF = Connection transmissibility (m3.cP/day/bar)
Another option is manual adjustment of the connection k = Permeability (mD)
transmissibilities, but this approach may be time-consuming kr = Relative permeability
and is not necessarily robust. A simpler alternative is to retain M = Molecular weight (kg/mol)
two cases, neglect of condensate banking and reduction of the m,n = Parameters in VDRP model
connection transmissibilities below the dew point by a given Nc = Capillary number
factor, as up- and down-side cases in the full field model. The p = Pressure (bar)
reduction factor can be calibrated against the finally-gridded Q = Injection/production rate (m3/day)
model neglecting VDRP, which is usually pessimistic.
IPTC 10382 7

R = Gas constant
r = Interpolation parameter Well B-1 B-2 B-3
S = Saturation Top depth (mSS) 3345.6 4099.7 4153.9
Sr = Residual saturation Outer radius (m) 1420 1020 467
t = Time (days) Well bore radius (m) 0.18 0.18 0.18
T = Absolute temperature (K) Net thickness (m) 3.9 18.7 11
v = Darcy velocity (m/day) Porosity 0.131 0.139 0.168
x = Distance (m) Permeability (mD) 81.4 135 7.1
= Non-Darcy flow coefficient (F) Initial water saturation 0.352 0.263 0.235
= Constant in Non-Darcy flow equation (F.mD-d) Initial pressure (bar) 343.4 388.7 389.3
= Viscosity (cP) Dew point pressure (bar) 317.8 344.9 344.9
= Density (kg/m3) Initial CGR (sm3/Msm3) 432 102 102
= Interfacial tension (dynes/cm) Skin 5.5 18 -3
= Real gas pseudo-pressure mg 40 40 3
Subscripts mc 40 40 3
b = base value or grid block value 1/ng 0.245 0.245 0.418
c = condensate 1/nc 0.017 0.017 0.018
g = gas Ncb 2x10-6 2x10-7 2x10-7
irr = irreducible c -0.21 -0.21 -0.21
m = miscible d -2.15 -2.15 -2.15
ref = reference value
* 4664 1732600 372
sp = single phase value
s = value at surface conditions
w = water or well Table 3. Parameters for radial models, Reservoir B

Parameter Radial Cartesian CGR Permeability (mD)


Model Value Model Value (m3/Mm3) .001 .01 .1 1 10 100 1000
Number of grid cells 2211 27271 1400 25 34 49 77 81 5 0
Grid cell size From 0.25 m 81 m 81 m 1130 14 24 32 43 38 2 0
to 362 m 1010 8 17 23 28 14 2 0
Reservoir thickness 181 m 744 0 4 4 4 3 2 0
Porosity 0.176
Net-to-gross 0.78 Table 4. Condensate skin for simulations without VDRP
Horizontal permeability 18.8 mD
Irreducible water saturation 0.26 CGR Permeability (mD)
(m3/Mm3) .001 .01 .1 1 10 100 1000
Critical condensate saturation 0.104
Well bore diameter 0.22 m 1400 21 34 39 8 5 1 0
Skin (mechanical) 1.0 1130 14 24 14 4 3 1 0
Target gas production rate 2 000 000 sm3/day 1010 9 17 8 3 2 2 0
Minimum THP 30 bars 744 0 4 3 3 1 0 0

Table 1. Parameters for radial and Cartesian models, Table 5. Condensate skin for simulations with VDRP (bold
Reservoir A values are those significantly different from Table 4)

Parameter Minimum Base Value Maximum


mg 4 20 40
mc 4 20 40
1/ng 0.2 0.4 0.5
1/nc 0.01 0.015 0.03
Ncb 210-7 210-6 210-5
c -0.31 -0.21 -0.11
d -2.65 -2.15 -1.15
* - 227500 -

Table 2. Parameters for velocity-dependent relative


permeability model, Reservoir A
8 IPTC 10382

0.3
CONDENSATE SATURATION

0.2

CARTESIAN WITH VDRP

CARTESIAN

0.1

CARTESIAN WITH NON-


DARCY FLOW (D-FACTOR)
0.0
160 200 240 280 320
PRESSURE (BARSA)
Figure 1. Liquid dropout curve, Reservoir A. Figure 4. Gas production rate for Cartesian models, with and
without VDRP and non-Darcy flow, Reservoir A

1.0
RELATIVE PERMEABILITY

.75
GAS

.50
CARTESIAN

CARTESIAN
-GPP
.25

CONDENSATE
RADIAL
0.0
0 0.1 0.2 0.3 0.4 0.5
CONDENSATE SATURATION
Figure 2. Base relative permeability curves, Reservoir A.
Figure 5. Comparison of Cartesian and radial models without
VDRP or non-Darcy flow, Reservoir A

RADIAL WITH VDRP

CARTESIAN-GPP
RADIAL WITH VDRP AND
NON-DARCY FLOW (BETA) RADIAL

RADIAL
CARTESIAN

Figure 3. Gas production rate for radial models, with and without
VDRP and non-Darcy flow, Reservoir A
Figure 6. Comparison of Cartesian and radial models with VDRP
but without non-Darcy flow, Reservoir A
IPTC 10382 9

SM3.CP/DAY/BAR2

PI

CARTESIAN-GPP

CARTESIAN

RADIAL RESERVOIR PRESSURE

Figure 10. Productivity in predictive simulations, well B-2 (several


Figure 7. Comparison of Cartesian and radial models with both cases with different THP).
VDRP and non-Darcy flow, Reservoir A

1.0
VARIOUS SIMULATIONS
WITH DIFFERENT THP
RELATIVE PERMEABILITY

.75
GAS

CVD EXPERIMENT
.50

.25
RESERVOIR PRESSURE
CONDENSATE

0.0
Figure 11. Condensate saturation in first grid cell, predictive
0 0.1 0.2 0.3 0.4 0.5
CONDENSATE SATURATION simulations, B-2.
Figure 8. Base relative permeability curves, Reservoir B.

SIMULATED SIMULATED
MEASURED MEASURED

CARTESIAN

RADIAL

CARTESIAN
-GPP

BACK PRESSURE TEST 1 BACK PRESSURE TEST 2

Figure 9. Match to flowing bottom-hole pressures during back Figure 12. Gas production rate for radial and Cartesian models,
pressure tests, well B-1. Well B-1.
10 IPTC 10382

Cartesian
initial PI
CARTESIAN
Cartesian matched
with WPIMULT

CARTESIAN
-GPP

RADIAL

Radial Model

Figure 13. Gas production rate for radial and Cartesian models,
Well B-2. Figure 16. Cumulative gas production, Reservoir D

CARTESIAN

CARTESIAN

CARTESIAN
-GPP

RADIAL ALL RADIAL


MODELS CARTESIAN, 50%
CF REDUCTION

Figure 14. Gas production rate for radial and Cartesian models, Figure 17. Gas production rate for two Cartesian models and all
Well B-3 radial models with VDRP and non-Darcy flow, Reservoir A

BHP bars Gas Rate (m3/d)


600 2400000
42.9*42.9
42.9*42.9 LGR
42.9*42.9 PSEUPRES
CARTESIAN
100*100
500 100*100 LGR 2000000
100*100 PSEUPRES
300*300
300*300 LGR
300*300 PSEUPRES
SR87 Gas Rate
400 1600000
ALL RADIAL
MODELS
300 1200000

200 800000
CARTESIAN, 50%
CF REDUCTION

100 400000

0 0
juil 1998 dc 1998 juil 1999 janv 2000 juil 2000 janv 2001 juil 2001 janv 2002 juil 2002 janv 2003

Figure 15. Bottom-hole pressure, Reservoir C: radial LGR vs. GPP

Figure 18. Cumulative gas production for two Cartesian models


and all radial models with VDRP and non-Darcy flow, Reservoir A

Vous aimerez peut-être aussi