Analysis of Multiphase Flow Instabilities in the Girassol Deep Offshore Production System

Analysis of Multiphase Flow Instabilities in the Girassol Deep Offshore Production System Erich Zakarian, Dominique Larrey Total E&P, Process Department, France ABSTRACT After over 6 years as operator of Girassol, a deepwater oil field development in Angola, Total has acquired a large amount of data and strong experience in deep offshore production. The subsea production system of Girassol includes several conventional subsea loops connected to a FPSO by 135m water depth. Gas can be injected at the base of the production risers for activation and flow stabilization. To ensure efficient topside processing, hydraulic stability of the multiphase feed stream is one of the key issues for operators. However, when flowing conditions are modified, undesirable hydraulic instabilities may occur in a production line and cause significant flow rate fluctuations at the riser outlet. We describe these instabilities from field measurements and propose their classification between riser-induced slugging and hydrodynamic/terrain-induced slugging. Their analysis is performed with the support of dynamic multiphase simulation, including slug tracking. 1 INTRODUCTION Our analysis of the Girassol field operational data started in 6. A first paper, published in December 7, presented the preliminary validation of the simulator OLGA against production data history, essentially through a large number of well tests and partly through hydraulic stability tests (1). This second paper exposes a deeper analysis of the hydraulic instabilities observed in the subsea production system of Girassol. Some of these instabilities were purposely generated during stability tests to obtain accurate data and relevant flowing conditions for further analysis. The rest of them was observed during normal operation and was carefully reported by control room operators for further investigations. 2 THE GIRASSOL FIELD Girassol is a deepwater oil field development, located 15 km off the coast of Angola in the Block 17 (2). Initial reservoir pressure and temperature are about 25 bar and 7 C respectively. Oil API gravity is approximately 32 ; its density is about 86 kg/m 3 at storage conditions; its viscosity is 1 cp at reservoir conditions and ranges from 6.5 cp to

35 cp at surface conditions. The gas-oil ratio of the production fluid (GOR) is in order of 11- Sm 3 /Sm 3, while the water cut ranges from to 7%. The field is tied-back to a FPSO (Floating Production Storage & Offloading) through five 8'' (.32m I.D.) piggable production loops, namely P1, P, P3, P4, P5, in approximately 135 meters of water: cf. Figure 1. Oil production started on December 1. Two years later, the nearby and similar Jasmim field was tied-back to the Girassol FPSO with an identical loop, namely P6. The latter came on stream in November 3. FPSO Offloading Buoy Riser Towers Injection Line 2 wellheads manifold Production Loop Production well Injection Well Umbilical Figure 1. Girassol subsea production system The 24 production wells of Girassol and Jasmim are connected to the production loops through 2-slot manifolds at a maximum 6km distance from the FPSO. They are equipped for chemical injection, continuously at downhole and possibly at Christmas tree for batch treatments. Each well can be routed to either left or right line of a loop. Gas can be injected at the base of the production risers for activation and flow stabilization. Flow assurance issues such as wax deposition or hydrate formation are primarily covered during normal production operations by an extensive thermal insulation of the subsea production system (3). Thermal performance is achieved with the gathering of the production flowlines (right and left lines of a production loop) within seabed bundles and riser towers (two production loops per riser tower). During shut-down conditions, the full system is designed to be preserved from hydrate formation (after a no-touch time) through: o methanol injection at wellheads, jumpers and manifolds from 2 service lines; o displacement of live oil with dehydrated and stabilized dead oil circulation from the FPSO into the production loops.

A large number of pressure and temperature sensors are available and operational since start-up, providing measurements at high frequency (from half an hour to less than one minute). The following pieces of information were specially recovered for this study: pressure and temperature at wellheads, production jumpers, manifolds and chokes; opening of chokes at wellhead and topsides (riser outlet and gas-lift injection inlet); injected gas-lift volume flow rates; well routing indicators to the left or right line of a production loop. The fluid produced from the Girassol and Jasmim loops is routed to a single first-stage separator, via a common production manifold. In case of production testing, each riser can be individually connected to a test separator, via a test manifold. The operating pressure of these two separators is approximately the same: 25 barg. 3 FLOW INSTABILITIES The nature of flow instabilities occurring in a pipeline-riser system is strongly dependent on the geometrical profile of the flowline laid on the seabed: o A downward inclined flowline towards the base of a riser is potentially prone to riser-induced slugging and terrain slugging: at low flow rates, when pressure drop is dominated by gravity, liquid accumulates at low points, initiating the formation of slugs (4)(5)(6). o An upward inclined flowline is potentially prone to terrain slugging and hydrodynamic slugging: at low flow rates, liquid accumulation in dips will initiate the formation of slugs. At high flow rates, short slugs can form in upward sloping sections and grow along the flowline (7). Figure 2 gives a schematic view of the geometry of Girassol/Jasmim flowlines. From above observations, we can predict that the following loops may produce under riserinduced slugging and/or terrain slugging conditions: P, P3, P4, P5. The two remaining loops, namely P1 and P6, may produce under terrain slugging or hydrodynamic slugging conditions. Note that to some extent, riser-induced slugging may occur in P6 since the last flowline section of this loop is slightly downward inclined towards the riser base. In the Girassol FPSO control room, hydraulic instabilities are detected when the pressure fluctuation within a flowline is significant. Practically, the maximum amplitude of a pressure instability measured at the closest subsea manifold should not exceed 5 bar. This arbitrary but practical criterion allows operators to detect significant instabilities compared to the usual noisy signal of a slug flow whose maximum pressure fluctuation is about 1 bar. Mitigation of hydraulic instabilities is achieved from the following actions: relative opening of the wellhead choke(s) to boost the production, partial closure of the topside choke to increase the back-pressure, ramp-up of the gas-lift rate to enhance the liquid entrainment in the riser.

Distance [m] -128 3 4 5 6 7 8-129 Subsea manifold M51 P5 loop - -131 P3 loop Water depth [m] -13-133 -134-135 P loop Subsea manifold M31 P6 loop P4 loop Riser base -136-137 Subsea manifolds M12 & M13 Subsea manifold M11 P1 loop -138 Figure 2. Girassol/Jasmim flowline geometry 3.1 First example of instability: riser-induced slugging This section gives an example of a riser-induced instability reported by control room operators on Oct 28 th, 5. It occurred in the right line of the P3 loop; a single well, namely P312, was allocated to the line through the subsea manifold M31: o From 7: pm to :4 am, the well production flow rate was progressively reduced with the partial closure of the wellhead choke from 54% to 44%. The flow became unstable, generating large pressure fluctuations in the flowline: cf. Figure 3 and Figure 4. o At :4 am, the amplitude of the pressure oscillation was about 35 bar with a time period of 5 min. An attempt to mitigate this severe instability was launched at :45 am with the closure of the topside production choke from 37% to 35%. o Finally, stronger mitigation means were implemented to kill the instability: the wellhead choke was reopened from 44% to 52% to boost the production; the topside choke was closed from 35% to 18% to increase the back-pressure; the gas-lift rate was ramped-up from 7 to 8 ksm 3 /d to enhance the liquid entrainment in the riser. At 3: am, the instability was killed. Note that the fluid pressure in the flowline went up from 87 barg to 97 barg despite the increase of the gas-lift flow rate. Topside choking is the main reason to this significant pressure change: indeed, from the trend of the pressure drop across the topside riser head choke (see Figure 4), we can easily deduce that the topside back-pressure has roughly increased by 1 bar since the pressure at the riser choke outlet is approximately constant (controlled first-stage separator pressure).

1 115 11 Gas-lift rate 9 8 15 95 9 85 8 7 6 5 4 3 Gas-lift rate [ksm3/d] 75 7 28-Oct-5 16:48 Flowline pressure at subsea manifold M31 28-Oct-5 28-Oct-5 29-Oct-5 : 29-Oct-5 2:24 29-Oct-5 4:48 29-Oct-5 7:12 1 29-Oct-5 9:36 Figure 3. Hydraulic instability in the P3 loop: flowline pressure and gas-lift rate Choke opening [%] 6 55 5 45 4 35 3 25 P312 wellhead choke opening Topside riser head choke opening Pressure drop across topside riser head choke 15 1 28-oct-5 28-oct-5 28-oct-5 29-oct-5 16:48 : Pressure drop across P312 wellhead choke 29-oct-5 29-oct-5 29-oct-5 2:24 4:48 7:12 4 35 3 25 15 1 5-5 29-oct-5 9:36 Pressure drop across choke [bar] Figure 4. Hydraulic instability in the P3 loop: choke opening 3.2 Second example of instability: hydrodynamic/terrain slugging This section gives a second example of an instability reported by control room operators on Feb. 27 th, 5. It occurred in the left flowline of the P1 loop; two wells, namely P122 and P131, were allocated to the line through the manifolds M12 and M13: o At 2: am, pressure instability appeared on the screens of the control room, approximately 11 hours after the partial closure of the topside choke from 37.6% to 34.6% and 9 hours after an increase of the gas-lift rate from 17 to 2 ksm 3 /d: cf. Figure 5 and Figure 6. The mean pressure in the flowline remained approximately the same (~123.7 barg) due to the significant increase of the gaslift rate. Compared to the instability previously described, the maximum amplitude of the pressure oscillation was rather small (~4 bar). However, mitigation of even small instabilities is always recommended to improve the efficiency of topside processing.

o From 6: am to 8:45 am, the progressive reopening of the topside choke failed to mitigate the flow instability, even with a larger opening (41%) than initially set (37.6%). Therefore, the choke was closed again but back to its initial position (38%). As no significant improvement was obtained, the gas-lift rate was reduced at 9:25 am, from 2 to 145 ksm 3 /d. o The instability vanished at 9:3 am but seemed to restart one hour later. The topside choke was further closed to 36% with no apparent stabilization effect. At 1:55 am, it was reopened from 36% to 37%. Again no improvement was observed. At 11:1 am, the topside choke opening was finally set back to 36%. The instability vanished definitely at 11:3 am. 129 128 Gas-lift rate 129 128 127 126 125 124 123 122 121 3 1 4:48 5:16 5:45 6:14 6:43 7:12 25 127 126 125 124 123 15 Gas-lift rate [ksm3/d] 122 121 1 26-Feb-5 14:24 26-Feb-5 16:48 Flowline pressure at subsea manifold M11 26-Feb-5 26-Feb-5 : 2:24 4:48 7:12 9:36 12: 5 14:24 Figure 5. Hydraulic instability in the P1 loop: flowline pressure and gas-lift rate Unlike the previous example, no attempt was made to mitigate this relatively small instability by increasing the production flow rate. Wellhead chokes were maintained at constant opening (8%). We see on this example that manual control of flow stability in the P1 production loop can be more delicate. Particularly, the reduction of the gas-lift rate helped to kill the instability whereas the reverse would be usually anticipated. This unexpected behavior has been observed in other deep-water oil field developments; Na Kika in the Gulf of Mexico is one of them (8).

Riser choke opening [%] 9 8 7 6 5 4 P122 & P131 wellhead choke opening Pressure drop across topside riser head choke Topside riser head choke opening 18 16 14 12 1 8 6 4 2 Pressure drop across choke [bar] 3 26-Feb-5 14:24 26-Feb-5 26-Feb-5 16:48 26-Feb-5 : 2:24 4:48 7:12 9:36 12: 14:24 Figure 6. Hydraulic instability in the P1 loop: choke opening Moreover, the response of the system to a change of operating conditions is not instantaneous; the instability occurred several hours after the partial closure of the topside choke and the increase of the gas-lift rate. The existence of a time delay between action and effect, in addition to the upward profile of the loop, suggests a different type of instability compared riser-induced slugging. The determination of the nature of the instability may help to understand the behavior of the system and possibly its operation. A close look at the pressure trend in Figure 5 gives some indication: the instability is rather periodic with a time period close to 15 minutes, suggesting a terrain-slugging phenomenon. Though, the presence of small pressure peaks over the main signal indicates the presence of hydrodynamic slugs. From the simple observation of this curve, some authors would classify this instability as terrain slugging of type II (9): liquid slugs are more or less aerated with gas bubbles at a varying rate; no full blockage of the line is apparent. The use of dynamic multiphase simulation could also bring useful information, provided that the fluid composition and the production flow rate of each well is known. While phase distribution (i.e. gas-oil ratio and water cut) can be reasonably derived from up-to-date well test data, the determination of the liquid production flow rate at a given time requires specific multiphase equipment (e.g. multiphase meter, test separator). Unfortunately, Girassol Christmas trees are not equipped with multiphase meter and the left line of the P1 loop was not under test on Feb 27 th 5. However, a very similar instability was observed during a field test performed in May/June 4 on the same line with the same production wells (P122 and P131). The test was performed with the test separator for an accurate measurement of the phase flow rates. In the second part of this paper, we propose to reproduce this test with dynamic simulation in order to characterize the hydraulic instabilities occurring in the P1 loop.

4 DYNAMIC SIMULATION OF HYDRAULIC INSTABILITIES The dynamic simulator OLGA was originally used for the design of the subsea production system of Girassol (1). A study has been recently conducted to validate a model of each production loop against a large number of well tests (1). These models simulate each line, from the furthest manifold to the topside production choke. Interaction with the gas-lift system is also included through the simulation of the gas-lift line from the topside choke outlet to the riser base; the production wells are not represented. The figure below reminds the results obtained from the comparison between measured and calculated pressure drop, from the closest subsea manifold to the topside production choke inlet. 11 Calculated Pressure Drop [bara] 9 8 7 6 5 4 3 1 1 3 4 5 6 7 8 9 11 Measured Pressure Drop [bara] P1 P P3 P4 P5 P6 +/- 1% +/- % Figure 7. Girassol/Jasmim well tests performed from Jan. 5 to Aug. 6 - Measured pressure drop vs. OLGA v.4.17.3 - From the closest subsea manifold to the topside production choke inlet The pressure drop is estimated with an error close to 1% in average for each loop except P6: cf. Figure 8. This error is consistent with the margin usually recommended for the design of oil production systems when pressure drop is calculated with OLGA. Moreover, these results are quite acceptable since no tuning was performed to fit field data. 3% Average error [%] % 1% % 21% 11% 11% 9% 7% 7% P1 P P3 P4 P5 P6 Figure 8. Girassol/Jasmim well tests performed from Jan. 5 to Aug. 6 - Measured pressure drop vs. OLGA v.4.17.3 - Average error

4.1 Hydraulic stability field test of the P1 production loop May/June 4 A multi-well test was performed on the left line of the P1 loop from May 29 th to June 3 rd, 4. Two wells, namely P122 and P131, were allocated to the line through the manifolds M12 and M13. The objective was to collect useful information about production flow rates and flow instability in order to optimize the gas-lift flow rates injected at riser base. During the test, the gas-lift rate was decreased, and kept constant at least few hours at several main stages to allow proper well testing (19, 15,, 7 and 5 ksm 3 /d). The flow became unstable two hours after the gas-lift rate was set down to 5 ksm 3 /d and half-an-hour after the partial closure of topside riser choke. The latter was progressively closed to maintain the topside back-pressure as constant as possible: cf. Figure 9 and Figure 1. Note that the opening of the wellhead chokes was kept constant throughout the whole test. A time-average of the recorded measurements is given in Table 1. 14 139 138 14 139 138 137 136 135 134 133 132 131 19:4 :9 :38 21:7 3 25 137 136 135 134 133 Flowline pressure at subsea manifold M12 15 Gas-lift rate [ksm3/d] 132 5 131 Gas-lift rate : 2:24 4:48 7:12 9:36 12: 14:24 16:48 4-Jun-4 : Figure 9. P1 stability field test: measured flowline pressure and gas-lift rate vs. time

55 31 5 45 4 35 3 25 Pressure at topside riser head choke inlet Topside riser head choke opening 3.5 3 29.5 29 28.5 28 27.5 27 Riser choke opening [%] : 2:24 4:48 7:12 9:36 12: 14:24 16:48 26.5 4-Jun-4 : Figure 1. P1 stability field test: Measured inlet pressure and opening of the topside riser choke vs. time Gas-lift rate [ksm 3 /d] 19 15 7 5 Total liquid flow rate [Sm 3 /d] 644 679 5762 5497 544 Associated GOR [Sm 3 /Sm 3 ] 15 12 11 13 Water cut [%] 43. 45.5 42.2 42.1 43.3 Pressure at manifold M11 [barg].5 131.6 133.2 133.6 133.8 Temperature at manifold M11 [ C] 63.6 63. 62.4 62.2 62.1 Pressure at riser choke inlet [barg] 36.1 35.8 36. 35.7 35. Temperature at riser choke inlet [ C] 58.6 58.5 58.3 58.2 58.3 Temperature at gas-lift choke outlet [ C] 46.9 44.6 39.8 39. 42.1 Table 1. P1 stability field test May-June 4 Measurements The five operating cases in Table 1 were run independently with OLGA. No instability was found for all gas-lift rates, including the lowest one: 5 ksm 3 /d. However when interactions between terrain and hydrodynamic slugging are suspected, it is recommended to run OLGA with the Slug Tracking option (1). A hydraulic instability was effectively predicted with Slug Tracking at a gas-lift rate between 4 and 3 ksm 3 /d: cf. Figure 11. This result is quite satisfactory since a recent on-site investigation showed that the uncertainty on the measurements of the gas-lift volume flow rates could reach 3% at 5 ksm 3 /d. Indeed, the design of the Girassol gas-lift meters is based on a vortex method, using an intrusive probe that is mainly sensitive to high gas flow rates (the uncertainty is about 5% in that case); the gas-lift system of Girassol is designed to inject up to 3 ksm 3 /d of gas in each riser. Note in Figure 11 that the mean pressure and the amplitude of the pressure oscillation are fairly well predicted at ksm 3 /d gas-lift rate despite a poor estimate of the time

period. The default settings of the Slug Tracking module were used in this work. Particularly, a default value of the slug initiation frequency (DELAYCONST=15) was considered; this parameter is defined as the number of pipe diameters a slug must travel before a new slug can be initiated. A sensitivity analysis to this parameter could improve the results. In addition, other possible improvements are currently investigated: extension of the simulator to the wellbore and/or to the topsides, sensitivity analysis to modeling parameters (void in slug, bubble point, etc). 138 137 136 Measured pressure at manifold M12 138 137 136 135 134 133 132 131 135 134 133 132 131 129 128 Calculated pressure GL rate = 4 ksm 3 /d 19:4 :9 :38 21:7 129 128 Calculated pressure GL rate = 3 ksm 3 /d 19:4 :9 :38 21:7 138 138 137 137 136 136 135 134 133 132 131 129 128 Calculated pressure GL rate = ksm 3 /d 19:4 :9 :38 21:7 135 134 133 132 131 129 128 Calculated pressure GL rate = 1 ksm 3 /d 19:4 :9 :38 21:7 Figure 11. P1 stability field test: Measured pressure at subsea manifold M12 vs. OLGA for different gas-lift rates From these results, we can suppose that the hydraulic instabilities observed in the P1 loop are generated from a combination of terrain and hydrodynamic slugging. To confirm this assumption, a series of simulations was performed to determine the contribution of the gravity to the pressure drop vs. friction: given the phase distribution and boundary conditions from the last column in Table 1, the pressure drop along the flowline (from manifold M12 to riser base) and along the riser (from riser base to topside choke inlet) was calculated as a function of the liquid flow rate (oil + water). The results of the simulations, without Slug Tracking, are given in Figure 12. Gravity seems to prevail for very low liquid flow rates: a slope change of the pressure drop curve in the flowline is apparent at Sm 3 /d. However, flow instability is predicted for a critical liquid flow rate between 3 and 35 Sm 3 /d: below this critical value, the pressure drop oscillates with time between a minimum and a maximum; the amplitude of the fluctuation is not higher than 4 bars.

As expected, when running OLGA with the Slug Tracking option, the flow is unstable for a wider range of liquid flow rates: cf. Figure 13. The critical liquid flow rate increases somewhere between 4 and 45 Sm 3 /d, which is closer to the value of 544 Sm 3 /d where the flow appeared to be unstable during the test: cf. last column in Table 1. Pressure drop in flowline [bar] 24 22 18 16 14 12 1 8 6 4 2 Minimum pressure drop in flowline [bar] Maximum pressure drop in flowline [bar] Minimum pressure drop in riser [bar] Maximum pressure drop in riser [bar] 3 4 5 6 7 8 9 Liquid flow rate [Sm 3 /d] Figure 12. OLGA simulation of the P1 stability field test: calculated pressure drop along the flowline and the riser without Slug Tracking 9 8 7 6 5 4 3 1 Pressure drop in riser [bar] Pressure drop in flowline [bar] 24 22 18 16 14 12 1 8 6 4 2 Minimum pressure drop in flowline [bar] Maximum pressure drop in flowline [bar] Minimum pressure drop in riser [bar] Maximum pressure drop in riser [bar] 3 4 5 6 7 8 9 Liquid flow rate [Sm 3 /d] 9 8 7 6 5 4 3 1 Pressure drop in riser [bar] Figure 13. OLGA simulation of the P1 stability field test: calculated pressure drop along the flowline and the riser with Slug Tracking From a close look at the simulation results, we see that the flow is also unstable for some liquid flow rates around 6 Sm 3 /d, which is above the critical value previously quoted (544 Sm 3 /d). Such result is possibly an indication of flow conditions where hydrodynamic slug growth along the flowline could be observed as mentioned earlier in this paper (7). In Figure 14, the minimum and maximum pressures at the manifold M12 are given with respect to the liquid flow rate; in Figure 15, the pressure trend calculated with OLGA is given for three representative liquid flow rates. Below 3 Sm 3 /d, flow instability is such that operators would probably classify it as severe.

Pressure at manifold at M12 [bara] 16 15 14 1 11 Minimum pressure at manifold M12 [bara] Maximum pressure at manifold M12 [bara] Hydrodynamic slug growth? Terrain/hydrodynamic slugging 3 4 5 6 7 8 9 Liquid flow rate [Sm3/d] Figure 14. OLGA simulation of the P1 stability field test: calculated pressure at subsea manifold M12 with Slug Tracking 15 145 14 Calculated pressure at liquid rate = 6 Sm 3 /d Calculated pressure at liquid rate = 5 Sm 3 /d Pressure [bara] 135 125 1 115 11 15 Calculated pressure at liquid rate = Sm 3 /d.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Time [h] Figure 15. OLGA simulation of the P1 stability field test: calculated pressure at subsea manifold M12 with Slug Tracking 4.2 Hydraulic stability field test of the P5 production loop A single-well test on the left line of the P5 loop was performed from May 12 th to 15 th, 4. One well, namely P511, was allocated to the line through the manifold M51. The wellhead choke opening was maintained constant as long as no instability was observed. Again, the objective was to collect useful information about flow instability in order to optimize the gas-lift flow rates injected at riser base. During the test, the gas-lift rate was decreased, and kept constant at least few hours at several main stages to allow proper well testing (25,, 15 ksm 3 /d). The pressure in the flowline started to oscillate at a gas-lift rate of ksm 3 /d. However, this case can be considered as stable since the amplitude of the pressure oscillation was decreasing with time: cf. Figure 16. At a gas-lift rate of 7 ksm 3 /d, the flow was severely unstable with the typical pressure trend of a riser-induced slug flow (5). We deduce that the flow stability limit is given for a gas-lift rate somewhere between 7 ksm 3 /d and ksm 3 /d.

125 Gas-lift rate 18 1 16 115 11 15 95 9 85 Flowline pressure at subsea manifold M51 14 1 8 6 4 Gas-lift rate [ksm3/d] 8 13-May-4 5:16 13-May-4 17:16 14-May-4 5:16 14-May-4 17:16 15-May-4 5:16 Figure 16. P5 stability field test: flowline pressure and gas-lift rate The simulation of this test with OLGA confirmed the transition to an unstable flow with decreasing gas-lift rate: given the phase distribution and boundary conditions from the field test results, the onset of instability was predicted for a gas-lift rate between 8 and 9 ksm 3 /d: cf. Figure 17. However, the amplitude and the period of the pressure fluctuation at the lowest gas-lift rate (7 ksm 3 /d) are much under-predicted: amplitude of 6 bar instead of 35 bar and time period of 8 minutes instead of 3 hours. In this particular case, no better result was obtained with the use of the Slug Tracking module. Other possible improvements are currently investigated: extension of the simulator to the wellbore and/or to the topsides, sensitivity analysis to modeling parameters (void in slug, bubble point, etc). Pressure at manifold M51 [bara] 124 122 1 118 116 114 Gas-lift rate = 7 ksm 3 /d Gas-lift rate = 8 ksm 3 /d 112 Gas-lift rate = 9 ksm 3 /d Gas-lift rate = ksm 3 /d 11 1 2 3 Time [h] Figure 17. P5 riser stability test: calculated pressure at manifold M51 vs. time 5 CONCLUSION Hydraulic instabilities recorded in the subsea production system of Girassol were analyzed with the support of dynamic multiphase simulation. They can be classified with respect the geometrical profile of the flowline laid on the seabed:

o In downward inclined flowlines towards the base of the risers, severe instabilities are mainly due to riser-induced slugging at low production flow rate. Their mitigation is systematically achieved from the following actions: relative opening of the wellhead chokes to boost the production, partial closure of the topside choke to increase the back-pressure, ramp-up of the gas-lift rate to enhance the liquid entrainment in the riser. o In upward inclined flowlines, instabilities combine both hydrodynamic slugging and terrain slugging. The latter prevails progressively at decreasing production flow rate, making instabilities more severe. A significant time delay between modification of operating conditions and flow response is observable, making the control of instabilities more difficult compared to riser-induced slugging. Particularly, an increase of the gas-lift rate may not have a positive impact. The onset of instabilities with decreasing gas-lift rates was well predicted with OLGA despite a mismatch between measured and calculated oscillations of the flowline pressure. The use of the Slug Tracking module was required to capture hydrodynamic/terrain-slugging instabilities in upward inclined flowlines. ACKNOWLEDGMENTS The Girassol field is operated by Total E&P Angola under a Production Sharing Agreement awarded by Sonangol to the Block 17 Contractor Group including Total, Esso Exploration Angola Limited, BP, Statoil and Hydro. Total would like to thank Sonangol and Partners for their support in the preparation of this paper and their permission to publish. REFERENCES (1) Zakarian, E. and Larrey, D.: A Systematic Investigation of Girassol Deepwater Field Operational Data to Increase Confidence in Multiphase Simulation, paper IPTC 11379, International Petroleum Technology Conference, Dubai, U.A.E., December 7 (2) Serceau, A., Girassol project presentation and challenges, Paper OTC 14166, 2 Offshore Technology, Conference and Exhibition, Houston, USA, May 2 (3) Saint-Pierre, T., Constant, A. and Vu, V.K.: Girassol: The management of Flow Assurance Constraints, Paper 14169, 2 Offshore Technology, Conference and Exhibition, Houston, USA, May 2 (4) Zakarian, E.: Analysis of Two-Phase Flow Instabilities in Pipe-Riser Systems, ASME Pressure Vessels and Piping conference, Seattle, USA, July (5) Taitel, Y., Vierkand, S., Shoham, O. and Brill, J.P.: Severe Slugging in a Riser System: Experiments and Modeling, Int. J. Multiphase Flow, vol. 16, pp. 57-68, 199 (6) Zheng, G., Brill, J.P and Taitel, Y.: "Slug flow behavior in a hilly terrain", Int. J. Multiphase Flow, vol., pp. 63-79, 1994 (7) Valle, A. and Utvik, O.H.: Field tests and analysis of hydrodynamic slugging in multiphase crude oil flow lines, 12 th Multiphase Technology Conference, BHR Group, Barcelona, Spain, May 5 (8) Lockett, T.J., Chupin, G., Vanstone, D., Ng, T.S., Saidi, F., Groseth, M.A., Mackay, D.: Understanding flow stability on the Na Kika deepwater flowline system, 13 th Multiphase Production Technology Conference, BHR Group, Edinburgh, UK, June 7 (9) Nydal, O.J., Audibert, M., Johansen, M.: Experiments and modelling of gas-liquid flow in an S-shaped riser, 1 th International Conference, Multiphase '1, BHR Group, Cannes, France, June 1 (1) OLGA, User manual, Transient multiphase flow simulator, Version 5, 6