Professional Documents
Culture Documents
ABSTRACT
In this paper we have compared the OLGA HD stratified flow model against transient
ramp-up and pseudo transient experiments from a Statoil funded experimental campaign
performed at the SINTEF Multiphase Flow Laboratory. The experiments were designed
to specifically target the transition point from low to high liquid holdup; the
accumulation point. We have explained how the removal of the liquid, when the gas flow
is increased to inside the multiple holdup solution region, can be described by a steady
state fully developed flow approximation. OLGA HD showed excellent predictions
compared to the accumulation point experiments. We have also compared the time
needed to remove the liquid for a fully transient simulation of experiments with ramp-up
into the multiple holdup solution region. A retuned version of OLGA HD, presented in
[1], performed well also on the experiments where OLGA 2014.1 HD overpredicted the
time needed to remove the liquid.
1 INTRODUCTION
The work described in this paper originates from a project funded by Statoil where the
uncertainty of OLGA for a gas-condensate field offshore Tanzania has been evaluated.
The Tanzania project is described in more details in [2]. The project consists of two
parts: the “Large Scale Liquid Loading Two-phase Flow Tests” campaign carried out at
the SINTEF Tiller large scale test facility, and the “Core Model Evaluation and Flow
Assurance Risk Study” done by Schlumberger. The experiments are described in more
detail in [3].
The OLGA HD stratified flow model is the next generation flow model for stratified
flows, the predominant flow regime in a gas condensate pipeline. In this paper we will
validate the OLGA HD 2014.1 flow model against some of the transient and pseudo-
transient experiments done in the Large Scale Liquid Loading Two-phase Flow Tests
campaign.
The new data from the SINTEF Multiphase Flow Laboratory consist of measurements
for both two and three phase flow. For simplicity we will in this paper only consider the
two-phase experiments.
+ = 0, (1)
+ = 0, (2)
where , are the gas/liquid densities, , are the gas/liquid (bulk) velocities and
, are the gas/liquid volume fractions.
+
(3)
ℎ
=− − − − sin − cos ,
+
(4)
ℎ
=− − + − sin − cos ,
where , , are the gas-wall, liquid-wall and gas-liquid interfacial shear stresses,
, , are the gas-wall, liquid-wall and gas-liquid interfacial wetted lengths, ℎ is the
height of the gas-liquid interface, is the pipe inclination relative to the horizontal plane,
is the acceleration due to gravity and A is the pipe cross sectional area.
Assuming steady state fully developed flow the mass balance equations (1)-(2) are
trivially true and the momentum balance (3)-(4) reduces to
− = + − sin , (5)
− = − − sin . (6)
(5)-(6) are also approximately valid as long as the neglected dynamic terms in (3)-(4) are
small compared to the gravity and friction terms.
Eliminating the pressure drop between (5) and (6) gives the corresponding holdup
equation:
1− − g sin = − 1− + (7)
The OLGA HD development was initiated by the HORIZON Joint Industry Project
sponsored by Chevron, Conoco, ENI, Exxon, Shell, Statoil and Total, see Biberg et al.
[4].
2.3 Multiple solutions for the steady state fully developed two-phase model
The holdup equation for fully developed steady state flow (7) may exhibit multiple
solutions depending on the input parameters. It is argued, e.g. by Barnea and Taitel in
[5], that when multiple solutions exist, the low solution is always stable, the middle
solution is linearly unstable and the high solution is non-linearly unstable for large
perturbations. These results are supported by the results in Section 4.2.
In OLGA-S, the steady state fully developed flow version of OLGA, it is always
assumed that the lowest holdup solution is the correct solution. In this paper, OLGA-S
will not refer to the commercial product OLGA-S, but the physical model inside it.
Figure 1 shows the liquid holdup for different superficial gas velocities. For this case we
notice a large region with multiple solutions. In the following, the abbreviations USG and
USL are the superficial gas and liquid velocity respectively.
USG1 exhibits only a single solution α1 which is a relatively high solution for the liquid
holdup. If the gas rate is ramped up from USG1 to USG3, still only a single solution
exists; however, α3 is a low solution for the liquid holdup. If the gas rate is ramped up
from USG1 to USG2, the liquid holdup will, according to the theory that the low solution
is the stable one, end up at the low solution α2low.
In this paper we will study ramp-up scenarios similar to increasing the gas rate from
USG1 to USG2 and how the liquid is removed during the transition from one stable
solution α1 to the other stable solution α2low. We will call this situation a ramp-up into the
multiple solution region.
Figure 1. Liquid holdup for different superficial gas velocities. Ramp-up from
USG1 to USG2 takes the solution into the multiple holdup solution region.
Figure 2. Ramp-up from outside the multiple solution region into the multiple
solution region. There exist three distinct propagating hydraulic gradients.
H1 is the liquid holdup for the steady state solution prior to ramp-up.
H4 is the liquid holdup for the steady state solution after the ramp-up.
There exist two quasi-stable liquid holdup solutions H2 and H3.
By using a simple mass balance it can be shown that the average propagating speed of
the hydraulic gradient is given by = . Nossen et al argued in [6] that
in the SINTEF experiments studied here, the liquid rate at the outlet during the second
and third sweeps is given by the liquid rate at the accumulation point for the current gas
rate. For the example in Figure 2 the difference in the liquid holdups H2 and H3 (second
sweep) is larger than the difference in the liquid holdups H3 and H4 (third sweep),
explaining the difference in propagation speed of the hydraulic gradient for the second
and third sweep. The discussion of sweep 1 is outside the scope of this paper.
Figure 3. Multiple solution curve for the inlet superficial liquid velocity and the
superficial liquid velocity at the accumulation point for superficial gas velocity
USG2. Liquid holdup for USG1 is given by α1. Liquid holdup for USG2 after steady
state is reached is given by α2low. The two intermediate holdup solutions during the
liquid removal; H2 and H3 in Figure 2, are given by the high holdup solution α2high
for (USG2,USLinlet) and the holdup solution α2acc at the accumulation for
(USG2,USLacc).
Through the Statoil funded experimental campaign at the SINTEF Multiphase Flow
Laboratory Large Scale Loop at Tiller, several steady state, transient and pseudo transient
experiments were conducted. Most transient experiments were ramp-up experiments into
and beyond the multiple solution region. During the experiments it was noticed that the
liquid outlet rate during sweep two and three (see Figure 2) was not influenced by the
inlet rate. We argued in Section 2.4 that this corresponds to the accumulation point, see
also Nossen et al. [6]. As a result of this discovery, experiments were performed where
the liquid inlet rate was adjusted to keep the experiment in the second sweep situation
(middle plot Figure 2). When the gas rate was increased, the liquid outlet rate quickly
stabilized on the liquid rate determined by the accumulation point. This allowed for
frequent increments of the gas rate and hence the measurement of the corresponding
liquid outlet rate at the accumulation point at several gas rates. These experiments will be
referred to as pseudo-transient experiments, or as screening experiments. The liquid
holdups on both sides of the hydraulic gradient were measured in addition to the liquid
outlet rate.
= (8)
− cos /
As described in [3], 5 different gamma densitometers located along the test section
allowed the holdup before and after the hydraulic gradient to be measured, equivalent to
high acc
α2 and α2 from Figure 3. Early in the measurement campaign, the liquid holdup
values were not considered important and no special care was taken to keep the hydraulic
gradient in the middle of the pipe. Consequently, for some of the measurements the
hydraulic gradient influenced the holdup at the inlet or the outlet depending on the
position of the hydraulic gradient. The liquid holdup measurements with severe influence
by the hydraulic gradient have been removed from the comparisons.
−
= . (9)
The error is presented as mean error, Root Mean Square error (RMS) and the percentage
of comparisons within ±5,10 and 20%. We will show the errors both in superficial gas
and liquid velocity. The error in the superficial gas velocity is considered to be the most
relevant error since the gas volume rate is very close to the total volume flow rate in a
typical gas-condensate field. Consequently, the uncertainty in the flow rate for a specific
behaviour of the field is best represented by the error in the superficial gas rate. However,
the error in the superficial liquid velocity will be important when comparisons against the
liquid removal from the full ramp-up experiments are discussed in Section 4.2.
Overall the predictions from OLGA HD are good compared to the measured superficial
gas velocity. The error in the superficial liquid velocity is relatively larger due to the
small values of the superficial liquid velocities and the non-linear increase in the
superficial liquid velocity at the accumulation point as a function of the superficial gas
velocity.
The error statistics for the superficial velocities can be found in Table 1 and the
comparisons are shown in Figure 5.
Table 1 Relative error statistics for superficial gas (USG) and liquid (USL) velocity.
#cases = 95 Mean RMS ±5% ±10% ±20%
USG 0.059 0.082 35.8 78.9 98.9
USL -0.26 0.49 3.2 7.4 16.8
Figure 5. Computed vs Measured for USG and USL. The dashed lines are
±20% error.
The measurements have been sorted for different pipe diameters and inclinations in order
to better show how they affect the predictions. The results can be found in Figure 6-
Figure 11. The continuous lines are the results from the OLGA-S HD simulations, the
small dots are the results from the transient OLGA HD simulations while the large dots
are the measured values. The left plot contains the liquid Froude number at the
accumulation point for different gas Froude numbers. The right plot contains the liquid
holdup upstream and downstream the hydraulic gradient, corresponding to α2high and α2acc
from Figure 3. α2high is the upper line while α2acc is the lower line in the right plots.
Notice that for increasing gas Froude numbers the low holdup solution, α2acc, increases.
The agreement between the quasi steady state results from the transient OLGA HD
simulation and the steady state fully developed approximation from OLGA-S HD is very
good. This supports the claim that the steady state fully developed model (7) is a good
approximation and hence a good explanation model. However, the deviation in the high
liquid holdup solution predicted by OLGA HD compared to OLGA-S HD increases for
lower gas rates. This is because a lower gas rate gives a lower liquid rate and hence a
larger difference in the holdup across the hydraulic gradient, α2high and α2acc respectively,
and consequently the influence of the hydraulic gradient increases for this relatively short
pipeline. When the length of the pipeline is increased, the deviation between OLGA HD
and OLGA-S HD is reduced.
The predictions of the liquid holdup are good overall, considering the uncertainty in the
holdup measurements due to the influence of the hydraulic gradient. The deviation
compared to the measurements is larger for the 8'' 5 degrees inclined pipe and the 12'' 1
degree inclined pipe. We have no good explanation of why the holdups for these two
configurations are predicted less accurately than the remaining configurations. Note also
that for the very low holdups, the measurements are in the order of the measurement
uncertainty which in [3] has been estimated to 0.001 volume fraction.
Figure 10. Comparisons against the 12 inch 2.5◦ inclined pipeline experiments.
We will compare the computed liquid holdup against the liquid holdup measurements
from the 5 different gamma densitometers. The instruments were positioned at 10, 28, 47,
66 and 85 m measured from the start of the inclined pipeline. For more details on the
experimental setup we refer to [3].
By looking at the change in liquid holdup as a function of time we are measuring the time
it takes to remove the liquid, given by
−
= , (10)
−
where L is the length of the pipe from which the liquid is removed, and are
the liquid holdups before and after the sweep, and , are the superficial
liquid velocities into and out of the pipeline. Consequently, the three computed values
contributing to the difference in predictions compared to the measurements are
, and . From these three computed values, is normally the most
sensitive value and the sensitivity increases with the reduction of − . In
Figure 13 the error in the time required to remove the liquid is shown as a function of the
error in the prediction of the superficial liquid velocity at the accumulation point. The left
plot shows an example where the gas is increased from USG1=3 m/s to USG2=3.09 m/s.
For this example the inlet superficial liquid velocity is set to the superficial liquid
velocity at the accumulation point of USG1=3 m/s. For a predicted close to
, the error in the liquid removal time goes to infinity as shown in the right plot in
Figure 13. If is predicted to be lower than , the pipe will accumulate liquid
instead of removing the liquid. For this example we have assume no error in the
predicted liquid holdup values.
To make the results more readable, both the measurements and the OLGA HD results
have been averaged using a moving average algorithm to remove noise and high
frequency waves.
We will show comparisons for three different ramp-up cases using two OLGA HD
versions; the 2014.1 HD version and a retuned HD version described in [1]. In Figure 12
the three different cases are illustrated. Case 1 is outside the multiple holdup solution
region while both case 2 and 3 are ramped up into the multiple holdup solution. Case 2
has the lowest inlet liquid rate while case 1 has the highest.
Figure 13. (Left) The superficial liquid velocity at the accumulation point as a
function of the superficial gas velocity. (Right) The relative error in the time needed
to remove the liquid, as a function of the error in the superficial liquid velocity.
We assume no error in the predicted liquid holdup.
4.2.1.1 Case 1: Ramp-up for a case outside multiple solution region, 8'' 2.5 degrees
inclined pipe
The gas rate is ramped up to a high rate as shown in the left plot in Figure 14. Case 1 has
a higher liquid loading and does not exhibit multiple solutions as seen from the right plot
in Figure 14. In Section 2.4 we argued that when the case is not ramping up to inside the
multiple solution region, all the liquid is removed in a single sweep. This is confirmed
from Figure 15 where both the OLGA HD simulation and the experiments show only a
single sweep. We notice that OLGA HD underpredicts the liquid holdup.
The measurements show different liquid holdup values for the different gamma
densitometers, however, it is believed that for this case some of the difference is due to
measurement uncertainty since there is a non-consistent behaviour of the liquid holdup
measurements along the pipeline.
Figure 15. Case 1, 8'' 2.5 degrees inclined. Liquid holdup values for 5 different
positions in the pipeline. The surplus liquid is removed in a single sweep which is in
agreement with the right plot in Figure 14 showing no multiple solution region.
(Top) OLGA HD simulation. (Bottom) Measured values.
4.2.1.2 Case 2: Ramp-up to medium high gas rate, 8'' 2.5 degrees inclined pipe
The gas rate is ramped up to a medium rate inside the multiple holdup solution region as
shown in the left plot in Figure 16. The right plot in Figure 16 shows the multiple liquid
holdup solution curve for the liquid Froude number at the inlet and the liquid Froude
number at the accumulation point for the ramp-up gas Froude number, Frg2. From Figure
17 we notice that OLGA HD severely overpredicts the time needed to remove the liquid.
The two main contributing factors are; OLGA HD underpredicts the liquid outlet rate and
Figure 16. Case 2, 8'' 2.5 degrees inclined. (Left) Superficial gas/liquid rate and
outlet pressure taken from the experiment. (Right) The multiple holdup solution
curve for the inlet liquid Froude number (Frlin) and the outlet liquid Froude
number (Frlout) determined by the accumulation point for the ramp- up gas
Froude number (Frg2). The dash-dot line is the liquid Froude number at
the accumulation point for the different gas Froude numbers.
Figure 17. Case 2, 8'' 2.5 degrees inclined. Liquid holdup values for 5 different
positions in the pipeline. (Top) OLGA HD simulations. (Bottom) Measured values.
4.2.2.1 Case 2: Ramp-up to medium high gas rate, 8'' 2.5 degrees inclined pipe
We re-ran Case 2 using the retuned HD version from [1]. The gas rate is ramped up to a
medium rate as shown in left plot in Figure 18. The right plot in Figure 18 shows an
increase in prediction of the superficial liquid velocity at the accumulation point
compared to the results from OLGA 2014.1 HD (ref Figure 16) which suggests that the
time to remove the liquid is reduced. This is confirmed by the results in Figure 19. The
simulations are qualitatively very similar to the measurements and the time needed to
remove the liquid is less overpredicted.
4.2.2.2 Case 3: Ramp-up to low gas rate, 12'' 5 degrees inclined pipe
The gas rate is ramped up to a rate just inside the multiple solution region. The right plot
in Figure 20 shows the multiple liquid holdup solution curve for the liquid Froude
number at the inlet and the liquid Froude number at the accumulation point for the ramp-
up gas Froude number, Frg2. The small difference between the liquid rate in and out of
the pipeline makes it challenging to correctly predict the time needed to remove the
liquid. From Figure 21 we see that retuned OLGA HD very accurately predicts the time
needed to remove the liquid. The liquid holdup is slightly underpredicted.
Figure 18. Case 2, 8'' 2.5 degrees inclined using retuned HD. (Left) Superficial
gas/liquid rate and outlet pressure taken from the experiment. (Right) The multiple
holdup solution curve for the inlet liquid Froude number (Frlin) and the outlet
liquid Froude number (Frlout) determined by the accumulation point for the ramp-
up gas Froude number (Frg2). The dash-dot line is the liquid Froude number at
the accumulation point for the different gas Froude numbers.
Figure 20. Case 3, 12'' 5 degrees inclined using retuned HD. (Left) Superficial
gas/liquid rate and outlet pressure taken from the experiment. (Right) The multiple
holdup solution curve for the inlet liquid Froude number (Frlin) and the outlet
liquid Froude number (Frlout) determined by the accumulation point for the ramp-
up gas Froude number (Frg2). The dash-dot line is the liquid Froude number
at the accumulation point for all the gas Froude numbers.
5 CONCLUSIONS
The behaviour of the liquid removal during a ramp-up scenario into the multiple holdup
solution region can be approximated by a steady state fully developed model, indicating
that for this scenario the dynamic terms from the momentum equation (3)-(4) are
negligible compared to the friction and gravity terms. This is supported by the agreement
between the transient solver OLGA HD and the steady state fully developed model in
OLGA-S HD, and the agreement with the transient ramp-up and the pseudo-transient
screening experiments done by SINTEF, [3]. This shows that the friction model is crucial
for correctly predicting the arrival time and rate of the liquid at the receiving facility.
The results in this paper do not indicate that a ramp-up scenario for an entire pipeline can
be well predicted using a steady state fully developed approximation. In a real pipeline
the pressure and fluid properties will change with time and the gas and liquid rate will
not be constant, which will potentially move the system in and out of the multiple holdup
solution region, making the scenario impossible to predict except by using a full transient
solver. Still, the newly acquired understanding of the physics behind the liquid removal
should be useful.
The OLGA 2014.1 HD stratified model reproduces the SINTEF liquid accumulation
experiments well, both for the liquid holdup H2 upstream and H3 downstream of the
hydraulic gradient (Figure 2), and the liquid rate out of the pipeline. By looking at the
The prediction of the time needed to remove the liquid for a ramp-up scenario is
influenced by the computed liquid holdups, and in particular the liquid rate at the
accumulation point. OLGA 2014.1 HD predicts the liquid removal time well for some
cases but overpredicts for cases where the gas rate is just inside the multiple holdup
solution region. The retuned version of OLGA HD, presented in [1], predicts also these
challenging cases well.
6 ACKNOWLEDGMENTS
The authors would like to thank Statoil for funding parts of this work and for the
permission to publish the results. We would also like to Chris Lawrence from
Schlumberger for many useful discussions. In addition we would like to thank Jørn
Kjølaas and Marita Wolden from SINTEF for their valuable help in getting the details of
the experiments correct.
REFERENCES
[1] D. Biberg, G. Staff, N. Hoyer and H. Holm, "Accounting for flow model
uncertainties in gas-condensate field design using the OLGA High Definition
Stratified Flow Model," in 17th International Conference on Multiphase Production
Technology, 2015.
[2] H. Holm, "Tanzania Flow Assurance Challenges," in 17th International Conference
on Multiphase Production Technology, 2015.
[3] J. Kjølaas, M. Wolden, T. E. Unander, H. Holm and P. S. Johansson, "A
comprehensive study of low-liquid-loading two- and three-phase flows," in 17th
International Conference on Multiphase Production Technology, 2015.
[4] D. Biberg, H. Holmås, G. Staff, T. Sira, J. Nossen, P. Andersson, C. Lawrence, B.
Hu and K. Holmås, "Basic flow modelling for long distance transport of wellstream
fluids," in 14th International Conference on Multiphase Production Technology,
2009.
[5] D. Barnea and Y. Taitel, "Structural and interfacial stability of multiple solutions for
stratified flow," International journal of multiphase flow, pp. 821-830, 1992.
[6] J. Nossen, T. Sira, T. Vanvik and H. Holm, "Analysis of hydraulic gradients in large
scale experiments," in International Conference on Multiphase Production
Technology, 2015.
[7] D. Biberg, The HD Stratified Flow Model, in Design og drift flerfasesystemer for
olje og gass. Tekna - Teknisk-naturvitenskapelig forening, Oslo, 2012.