Professional Documents
Culture Documents
Comparisons Between Engineering Software Predictions and West of Africa Deepwater Field Data Including Thermal Performances
Comparisons Between Engineering Software Predictions and West of Africa Deepwater Field Data Including Thermal Performances
ABSTRACT
This paper presents a comparison between field data from a West Africa deep offshore
field operated by TOTAL and software available for design engineering. Multiphase flow
simulations have been performed for several production loops to evaluate the software
ability to reproduce steady state operations (slugging, pressure and temperature profiles,
etc.) and the full preservation sequence including shutdown and fluid displacement by
dead oil circulation.
Results show that software predictions are generally in good agreement with field
observation especially with regards to pressure variation and timing. However the
structure of the interface (i.e. mixing zone) between the cold dead oil and the live oil is
not well reproduced for the loop operated in hybrid mode (i.e. only one branch in
production, the other one being already preserved with dead oil).
1 INTRODUCTION
In oil dominated systems, one of the most common means of preservation against hydrate
formation in deep water subsea production networks is to displace the live oil by
circulation of stabilized dead oil. In a conventional design, the dead oil is pumped from
the cargo tanks and the production lines and risers are looped to allow circulation from
both ends.
However ensuring the displacement of a multiphase production fluid sitting in deep water
risers and flowlines involve complex transient phenomena such as thermal effect (cool-
down and warm-up, cold spots effects as manifolds, subsea connectors), occurrence of
important pressure variations (depressurization, packing, unbalanced liquid head in
risers) and multiphase flow effects (phase segregation, fluid displacement, liquid slugs).
With increasing demand on Flow Assurance tools, developments works are performed
continuously by code suppliers to improve the simulation accuracy and operators have to
undertake systematic internal validation tests before qualifying released versions.
Nevertheless, despite the fact that experimental databases are widely used for the
assessment of the accuracy of the different mechanistic models, the final proof remains
the results of comparisons between multiphase flow assurance software and actual field
production data. Note that the use and access to field temperature data for the validation
of thermal models is relatively limited compared to the hydrodynamic models for which
development is supported using important experimental database and field production
data.
Figure 2 – Existing flow loops for multiphase flow tests (photos from website) –
Sintef (Left) & IFE (Right)
Nevertheless, if some of these flow loops offer the possibility to use large diameter pipes
and high pressure, most of them are equipped with reduced diameter pipes (4 inches is
the most common pipe diameter) at barely above atmospheric pressure and custom fluids
are used to approach the actual production fluids behavior. Semi empirical correlations
validated in small diameter and low pressure conditions are then implemented in
commercial codes developed to describe behavior in actual field situations where larger
diameters and higher pressures are observed. In this case, e.g. operating conditions in
West Africa fields, a comprehensive assessment of the model uncertainties has to be
carried out.
Extensive validation tests are performed by both multiphase flow simulation tool
developers and Oil and Gas operators. The objective of these validations is to assess
model’s accuracy and limitations according to the type of field to be developed (oil or
gas dominated, ultra-deep water, etc.) and operating conditions. Moreover, internal
databases gathering actual field data are commonly used for the validation of the
different releases as it is presented on the following figures.
The systematic evaluation program of the multiphase flow assurance software has to
emphasize on:
The accuracy of the model for simulation of gas dominated system and oil
dominated system
The conservatism of the different models: does the model over-predict or
under-predict the pressure drop, liquid content, slug flow pattern, etc.?
By considering the accuracy of the models, the definition of the design margins
to be accounted for by Flow Assurance engineers during project
The relevance of the various models with regard to the complexity of the field
development scheme, depending on the outcome of the validation tests.
20 %
50 %
External External
Figure 4 – Identification of over-prediction or under-prediction of horizontal liquid
hold-up – Comparisons with external database and Boussens internal database
Experimental data are nevertheless not sufficient enough to fully qualify the multiphase
flow simulators and operators have to build upon field data comparisons.
3 FIELD DESCRIPTION
Spools, manifolds have also been included in the different models in order to capture the
impact on these “cold spots” onto the overall thermal performances of the production
network and simulate potential gas pocket traps induced by their geometries. The Figure
7 shows the typical arrangement of the spool and manifold piping between PIP drop-
down points as used for cool-down simulations (example for M3011 manifold).
As manifolds co-ordinates were not available, their depths have been assumed to be the
average of the two PIP drop down points. Manifold and spool piping between PIP drop-
down points is modeled as follows:
15m of low-insulated manifold piping assumed horizontal on seabed (285mm
ID, U=2.81 W/m2K) surrounded by 275mm ID, U =1.7 W/m2K spools,
15m long, 1m deep "liquid traps" added on either side to allow liquid to drain
from the manifold during a shut-down (conservative for cool-down calculation
purpose)
Straight spool sections assumed between the end of the liquid traps and the PIP
drop down points
For the upstream manifolds (M3012, 22, 32, 43) only, the liquid traps and
spool piping upstream of the manifold have not been considered.
For slugging simulations, unless specified otherwise, default Delay Constant of 150 has
been used in OlgaTM. Fine mesh (8D length) has been considered in LedaflowTM to
enable the use of slug capturing simulations.
To validate this approach, a mass source approach including a single PVT for each of the
mixtures has been compared to the standard approach with a standard flow rate on one of
the production lines. Discrepancies on pressure and temperature drop predictions have
been observed within 1% on average. The only noticeable difference is a discrepancy of
0.5 kg/s on the gas lift injection rate (roughly equivalent to 50 kSm3/d), which is
considered negligible with regard to the overall gas flow rate in the riser (typically
around 1000 kSm3/d).
Thus the standard approach was considered acceptable for this comparison work. Note
that this approach has the added advantage to simplify the models (no need to define
additional nodes), to avoid loading multiple PVT files and the need for units conversion
(which is common known source of error). However it should be noted that this modeling
approach is not directly applicable to every field and should be checked thoroughly
before being used as it depends strongly on the fluid GOR and pseudo process
considered.
For preservation simulations, the choke opening is adjusted based on actual choke valve
opening values from field data.
The separator pressure is set at 10 Bara with a pressure boundary node. First separation
stage onboard FPSO is composed of 2 separators in parallel. These separators present
large surge volume (up to 50 m3 each) and receive the production from several other
risers. Thus the overall gas and liquid throughput feeding one separator is expected to be
relatively stable as not all of the risers are subject to slugging.
The benefit of modeling the separator instead of a pressure node has been assessed. The
results highlighted that the difference with regard to dynamic was negligible and the
pressure node approach has thus been retained for this study.
Static cool-downs have thus been simulated at specific locations along the production
network. As no temperature sensors are implemented onto the Pipe-In-Pipe production
line, only manifolds and Integrated Production Bundle (IPB) have been considered in
these preliminary comparisons. These cool-down simulations have been carried out after
a preservation sequence of the production network such that the fluid in place is well
known.
Field data showed that the thermal performances for IPB and manifold cross sections
filled with dead oil are in good agreement with both analytic calculations and dynamic
simulation using OlgaTM and LedaflowTM models. The simulations tend to always be
slightly conservative. Typical results are shown in the Figure 10 for a pipe section of 200
m filled with dead oil, ID = 274.3 mm and thermal properties from Table 2.
Based on the results obtained, no tuning of the thermal properties of the insulating
material has been deemed required for manifold and IPB sections. The thermal properties
for the PIP production line have been defined according to Factory Acceptance Tests.
M304-1 52.8 51.0 (-1.7°) 51.1 (-1.7°) 71.8 72.5 (+1%) 69.0 (-4%)
RH 45.0 45.2 (+0.2°) 45.4 (+0.4°) 19.9 17.3 (-13%) 18.5 (-7%)
FPSO 45.0 44.7 (-0.3°) 44.9 (-0.1°) 10.0 10.3 (+3%) 10.7 (+7%)
M304-3 50.7 50.6 (-0.1°) 50.6 (-0.1°) 65.0 63.7 (-2%) 67.7 (+4%)
M304-2 49.6 49.6 ( 0.0°) 49.6 ( 0.0°) 64.0 63.2 (-1%) 67.2 (+5%)
LEFT
M304-1 51.3 50.8 (-0.5°) 50.8 (-0.5°) 67.0 62.7 (-6%) 66.7 (-0%)
RH 45.2 45.4 (+0.2°) 45.5 (+0.3°) 15.8 17.7 (+12%) 19.5 (+23%)
FPSO N/A 44.9 45.1 10.0 10.1 (+3%) 10.7 (+7%)
Figure 11 – Right branch steady state – Pressure (left) & temperature (right) results
Figure 12 – Left branch steady state – Pressure (left) & temperature (right) results
4.7 Slugging analysis
For slugging analysis comparisons, the field data slug frequency has been calculated
based on pressure variation recorded at topside. It is assumed that the pressure variations
are representative of the slug size due to the change in hydrostatic pressure in the riser. It
is nevertheless important to keep in mind that the sampling of the pressure transmitter,
although quite loose, is considered of sufficient quality to extract a rough evaluation of
the slug frequency.
According to the slugging analyses performed for the production loop 302 and
summarized in Figure 14 and Figure 15, the following conclusions can be observed:
The slug frequency distribution is well predicted by LedaflowTM and
OlgaTM 7.2 (model with no node at riser base). An example of pressure trend
results with OlgaTM compared with field data is also reported in Figure 14,
Whatever the OlgaTM releases, the default delay constant of 150 gives
reasonably good results. Although Shea 2004 correlation suggests a delay
constant of 297, it was not considered necessary to run another simulation.
In OlgaTM, the presence of a node at the riser base seems to slightly deteriorate
the prediction with regards to slug frequency. This is relevant when
considering the way the nodes are taken into account in the simulations,
OlgaTM 5 results are not as consistent as OlgaTM 7 with regards to slug
frequency distribution.
It is also important to note that the slug frequency predicted by the Shea correlation (155
slug/h) is in acceptable agreement with field observations (peak at 100 slug/h).
Figure 13 – P302 – Right – Pressure trends – Comparisons of field production data
with OlgaTM 5 results – Slug tracking starts at t=4hours (DC = 150)
Figure 14 – P302 – Right – Slug frequency distribution
Equation 2– Shea correlation (2004) & Delay Constant (OlgaTM user manual)
The maximum surge volume calculated from OLGATM and LedaflowTM results is about 1
m3 which is likely to be under evaluated in this case as the pressure variations are under
estimated by both simulators as show on Figure 15. Note that the liquid level variations
recorded in the first stage separator range from 66% to 68% which suggests a liquid
surge volume of roughly 3 m3 (assuming a working surge volume capacity of 50 m3
between 66% and 100%).
The maximum surge volume calculated from OlgaTM and LedaflowTM results is about 3
m3 which is likely to be in agreement with actual surge volume coming from this riser as
the pressure variations are well reproduced by both simulators as show in Figure 16. This
is consistent with 1st stage separator level readings with a recorded oil level variation of
2% (between 65% and 67%) for a working surge volume of 50 m3.
M304 DO Circulation 24‐January‐2013
Circulation from Left to Right Branches
160 300
WP DO SD
140
DP 250
Dead Oil Rate (m3/h)
120
Pressure (bara)
200
100
80 150
60
100
40
50
20
0 0
Time (h)
Upstream Riser‐L Choke (bara) Upstream Riser‐R Choke (bara) M304‐1 L (bara)
M304‐1 R (bara) M304‐2 L (bara) M304‐2 R (bara)
DO Rate (m3/h)
Figure 19 – P302 Dead Oil circulation – Temperatures trends vs. field data
5 CONCLUSION
Multiphase flow software present, in overall, a fair agreement with field data for steady
state (pressure drop prediction within the 10% and temperature discrepancies less than
2°C) and static conditions during shutdown. The discrepancies between the models and
field data, after a Cool Down of 20 hours ranges from 0.5°C to 4°C and the numerical
models results are usually slightly conservative. This has been made possible as the fluid
properties, materials properties from Factory Acceptance Tests, initial operating
conditions are well known. No additional tuning was required on the models
implemented in the two multiphase flow software.
The thermal behavior observed in the production line during transient operations such as
cool-down, preservation sequence, and warm-up is also well reproduced with the
different multiphase flow software (only cool-down and warm-up for LedaflowTM) .This
indicates that dynamic preservation sequences, although involving complex multiphase
flow phenomenon, can be modeled to evaluate the overall thermal performance of a
design with acceptable accuracy, reducing the need to excessive margins on thermal
specifications and ultimately reducing the cost of a design.
This comparison work gives also a better understanding of the multiphase flow behavior
in the subsea system and allows the flow assurance engineer consolidating the operating
guidelines of existing subsea facilities and envisaging more innovative solutions for the
debottlenecking and future development of both topsides and subsea facilities.
Nevertheless, further works are still required to fully understand and model the
preservation sequence, especially the potential mixing at the interface between live oil
and dead oil. These studies should rely on finer mesh discretization in 1D model to better
capture the front if any at the interface. With the boom of full 3D models in the Oil &
Gas industry, and the spread of the access to high CPU resources, CFD modeling might
be also considered. Whatever the solution retained, long CPU time should be expected.
As a matter of fact, TOTAL E&P has fostered an acknowledged expertise and technical
advances in Flow Assurance leading the Group to conduct bold developments especially
in deepwater fields.
ABBREVIATIONS
ACKNOWLEDGEMENTS
The authors gratefully acknowledge the support and encouragement they receive from
TOTAL E&P while preparing this manuscript.
REFERENCES
[1] Fabre J. a Liné A., 1992, Modeling of two phases slug flow, Annual review of fluid
mechanics, 24, 21-46
[2] Valluri P., Spelt P.D.M., Lawrence C.J., Hewitt G.F., 2008, Numerical simulation of
the onset of slug initiation in laminar horizontal channel flow, International Journal of
multiphase flow 34 206 – 225
[3] E. Al-Safran, 2009, Investigation and prediction of slug frequency in Gas / Liquid
horizontal pipe flow, Journal of petroleum science and engineering 69 (2009) 143 – 155
[4] G.J. Zabaras, 2000, Prediction of slug frequency for Gas / Liquid flows, SPE journal
5 (3), September 2000
[5] T. Danielson, K. Bansal, B. Djoric, D. Larrey, S.T. Johansen, A. de Leedbeeck, J.
Kjolaas, 2012, Simulation of slug flow in Oil and Gas pipelines using a new transient
simulator, OTC 23353
[6] R. Belt, B. Djoric, S. Kalali, E. Duret, D. Larrey, 2011, Comparison of commercial
multiphase flow simulators with experimental and field databases, BHRG conference.
Comparisons between engineering software
predictions and West of Africa deepwater
field data including thermal performances
Christophe CANDELIER & Fabien PAPOT
Flow Assurance specialists – TOTAL E&P
Introduction
• World’s largest and easiest to exploit deepwater
field already under production
Extensive sensitivity studies performed to: Extensive internal validation tests performed to
• Define operating envelope of the network assess the accuracy of software
• Ensure compliance with operating constraints • Comparisons to experimental database
• Comparisons to field production data
Comparisons with experimental data
• Unbiased evaluation - Different database to the ones used for software development
• 4 production lines
– Length ranging from 8.2 km to 13.5 km
– Riser topography similar for all production lines
– Upward and downward bathymetries
• 1 upward production line
• 1 with flat bathymetry
• 2 downward bathymetries
Production loops Length (km)
P301 13.52
P302 8.2
P303 10.56
P304 13.32
Field characteristics PIP
• Specificities for slugging analysis • Riser top chokes, spools and manifolds
– Default DC used in Olga (DC = 150) included
– 8D mesh used in Ledaflow
• Ledaflow slug capturing predicts slugging in agreement with site measures without
specific tuning
• Confidence in results obtained when simulating dynamic preservation sequences
– Confirmation of actual thermal performance of equipment
– Consolidation of operating guidelines
• Perspectives
– Optimization of thermal specifications of the subsea equipment
– Can allow unlocking reserves by proposing innovative architectures and preservation
procedures