You are on page 1of 17

Presented at the IBC Technical Services Conference on Optimisation of Gas Condensate Fields, Aberdeen, 26-27 June 1997

CALCULATING WELL PRODUCTIVITY IN GAS CONDENSATE RESERVOIRS

Robert Mott

AEA Technology plc Winfrith, Dorchester, Dorset DT2 8DH

1. Introduction
Well productivity is a critical issue in the development of many gas condensate reservoirs. Liquid build-up around the well can cause a significant reduction in productivity, even in reservoirs where the fluid is very lean. This paper reviews recent developments in the understanding of near-well behaviour in condensate reservoirs, and in estimating well productivity through numerical simulation.

2. Mechanism of Liquid Build-Up


Figures 1 and 2 shows the pressure and liquid saturation around a producing well, as a function of the distance from the well, in a lean gas condensate reservoir below its dew point. Away from the well, only the gas phase will be mobile. As gas flows towards the well and its pressure falls, the capacity of the gas to vaporise oil will reduce, so that liquid will condense from the flowing gas phase. This liquid condensate will initially be immobile and will accumulate in the near-well region. As the flow rate of gas is greater near the well, liquid will accumulate faster in the region near to the well. Liquid condensate will continue to accumulate until its saturation exceeds the critical value, when liquid will start to flow. Eventually a semi steady state will be reached where the combined flowing fluid composition (in both liquid and vapour phases) is constant in the region around the well. In the region within about 10 feet of the well, the semi steady state will be reached quite quickly, because of the small volume of this region compared with the throughput of gas. The liquid saturation in this region must be large enough for the liquid to be mobile, so that a liquid saturation of 20% or more can be achieved even if the liquid saturation in the deep reservoir is only 1 or 2 %. The loss in productivity due to liquid build-up is determined by the value of krg near to the well, compared with the value of krg in the deep reservoir. The loss in productivity is more sensitive to the relative permeability curves than to fluid PVT properties.

To demonstrate this point, 1D single well compositional simulations were carried out for 5 different fluids ranging from lean to near-critical, with oil-gas ratios ranging from 50 to 280 stb/MMscf. Figure 3 shows the liquid saturations for these fluids in constant composition expansion tests. The simulations were initialised at a pressure just above the dew point. To assess the effect of liquid build-up, we use the normalised productivity index which is a ratio of the actual well productivity to the well productivity if no liquid build-up occurs. Figure 4 shows the normalised productivity index versus time, and indicates a similar effect for the five fluids, confirming that it is the relative permeabilities which are most important in determining productivity loss.

3. Field Examples
3.1 NORTH SEA LEAN GAS CONDENSATE RESERVOIR
We have analysed well test results in a lean North Sea gas condensate reservoir where the maximum liquid drop out is 2%. This is Fluid A in Figure 3. Single well radial simulations were used to calculate well head pressures at different gas flow rates, which were compared with field measurements after about 1 year of production. Simulations with the measured gas-oil relative permeability curves gave well head pressures which were much lower than the field measurements (see Figure 5), suggesting that the simulations were underestimating well productivity. In order to match the field data, the oil relative permeability curve was modified as shown in Figure 6. This gave a much better match to the measured well head pressures. Figure 7 shows the normalised productivity indices for the two simulation runs with the original and modified relative permeabilities. This result shows how simulations with measured relative permeability curves can overestimate the loss in well productivity due to liquid blockage. In this case the calculated productivity was about 50% of the value indicated by well test results. The relative permeability data for this reservoir were measured in a steady state experiment using mineral oil, brine and nitrogen at low pressure. It would appear that data from this type of experiment may not be appropriate for reservoir conditions where the liquid phase is produced by retrograde condensation.

3.2 OTHER EXAMPLES


Mobils results for the Arun field [1] in Indonesia are similar to those for the North Sea reservoir. Arun is a large gas condensate field with maximum liquid drop out of 1%. Well productivity fell by about 50% as the reservoir pressure dropped below the dew point. Single well simulations were used to model well tests, and the measured relative permeability curves had to be increased before the simulation results fitted field data. In this case the gas curve was adjusted as shown in Figure 8. A recent paper on the Britannia gas condensate field [2] also reports that measured relative permeability curves needed to be increased to match well test performance. In this case a variation of relative permeability with capillary number was used in the simulations.

4. Special Phenomena Affecting Near-well Behaviour


The results for the field cases discussed above suggest that standard simulations can overestimate the effects of liquid blockage and therefore underpredict well productivity. Similar results are believed to have been found in other condensate fields. This suggests that there are special phenomena affecting near well flow, which are not represented in standard simulation models. Possible effects include High capillary number flow Non-Darcy flow Low IFT flow Water vaporisation Compositional effects Non-equilibrium PVT We have used generic simulation models to examine the possible impact of some of these effects on well productivity.

4.1 HIGH FLOW RATE EFFECTS


A number of experimental measurements have shown that relative permeability is increased at high capillary number [3, 4, 5]. The capillary number N c is a dimensionless quantity which measures the ratio of viscous to capillary forces, and is defined by Nc = ( flow rate). (viscosity ) IFT

As capillary number is proportional to flow rate, high values can occur in the region close to a gas well. Experiments show that significant improvements in mobility occur above a threshold capillary number which appears to be around 10-5 for gas and 10-3 to 10-5 for oil. Capillary numbers of this magnitude occur within about 10 feet of the well, and the effect is to improve well productivity. On the other hand, Non-Darcy or inertial flow effects will reduce well productivity. The magnitude of this effect will depend on the value of the Non-Darcy flow coefficient . There are problems in estimating , as there is a wide variation between results of the different published correlations for as a function of permeability and porosity. Single well simulation studies have been used to assess the impact of high capillary number and Non-Darcy flow effects on well productivity. For example, Figure 9 shows the calculated well bottom hole pressure (BHP) for a 10 mD reservoir containing rich gas condensate. Including the effects of high capillary number flow leads to a significant increase in BHP, and the well productivity is increased by up to a factor of 2. The effect of non-Darcy flow is a small reduction in productivity.

1.2 LOW IFT EFFECTS


A number of experiments have shown that a significant improvement in relative permeability occurs at very low interfacial tension. The threshold IFT for these changes is below 0.1 mN/m. Low IFT effects can be important in miscible gas flooding, and can be modelled in most compositional simulators. Figure 10 shows the IFT as a function of pressure for a rich gas condensate with dew point pressure of 6000 psi. The data are taken from the 3 smallest grid cells in a single well

simulation, and the overlaps between the curves are caused by compositional changes in the near-well region. A threshold IFT of 0.1 mN/m is also indicated, and IFTs below the threshold only occur at pressures above 5000 psi. Calculations on other fluids confirm the conclusion that very low gas-oil IFTs will occur only for rich condensates when the pressure is within about 1000 psi of the dew point. When a gas condensate well is pressure limited and productivity is a concern, the reservoir pressure will usually be too low for very low IFTs to occur. Changes in near-well mobility are unlikely to be due to low IFTs alone, but to a combination of high flow rates and moderate IFTs leading to high capillary number.

1.3 WATER VAPORISATION


Water vaporisation may also affect near-well behaviour, although it is not modelled in most simulation studies. The solubility of water in the gas phase increases as pressure decreases. As gas flows towards a production well and its pressure falls, it will be able to vaporise additional water, which may reduce the water saturation around the well, and thereby increase productivity. This effect is more important in high temperature reservoirs. Some simple simulations have indicated that all of the water within a few feet of the well could be removed due to vaporisation, and that the well productivity could be increased by around 30%.

1.4 NON-EQUILIBRIUM EFFECTS


Reservoir simulators assume instantaneous thermodynamic equilibrium between the fluids in a grid block. In the region near to the well, the very high flow rates may not allow time for equilibrium to occur. Liquid may be transported in the form of a mist within the gas phase. If this were the case, the amount of liquid condensed in the near-well region would be lower than predicted by the simulator, so that simulation results would underestimate well productivity. At present, these phenomena are not well understood and cannot be modelled with existing reservoir simulators.

1.5 DISCUSSION
The improvement in mobility at high capillary number is the most significant of the special phenomena affecting near-well flow, and is the most likely explanation for the observation that productivity losses are less severe than predicted from simulation. Although high capillary number effects have been demonstrated in a number of experiments, future work is needed to develop and validate models which can be used in reservoir simulators, and to understand how the high capillary number effects change between different rock types. The effects of water vaporisation, non-equilibrium behaviour and mist flow are less well understood, but could also cause significant improvements in mobility.

5. Experimental Data Requirements


The most important requirement is for gas-oil relative permeabilites at conditions which are representative of the near well region. The analysis of Fevang and Whitson [6] shows that the crucial parameter for productivity loss is krg as a function of the ratio krg / kro, and that the range of relevant values of krg / kro can be calculated from the fluid PVT data. Fevang and Whitson have proposed a technique for measuring the required gas-oil relative permeabilites, using reservoir fluid samples. If possible, relative permeabilities should be measured at a range of flow rates to understand the effect of high capillary number flow. PVT data are also required. Liquid viscosity data, which are not usually relevant for the deep reservoir region, are important for near-well effects. IFT data are also needed to allow capillary number to be calculated.

6. Calculating Productivity in Field Scale Simulation


Accurate calculation of near-well effects in condensate reservoirs usually requires simulation with very small grid cells near to the well. A large part of the pressure drawdown occurs within 10 feet of the well, so that radial models are needed with the inner grid cell having dimensions of about one foot. This presents problems in full field models, where typical grid cell dimensions are hundreds of feet. The traditional approach to this problem is to run independent single well simulations to estimate skin factors due to liquid build-up, and use these skin factors in the full field simulation. This is not ideal, as the skin factor may vary with pressure and flow rate, and there can be problems in ensuring consistent conditions between the two models. An alternative solution is to use Local Grid Refinement (LGR) around the well. A radial grid can be embedded within a single column of grid cells in the full field model. Local timestepping can be used to allow short timesteps on the radial grids and longer timesteps on the global grid. The use of LGRs results in a much more complex simulation model, can lead to a significant increase in run time, and may cause numerical problems in linking the solutions on the local and global grids. A third alternative is to use the pseudopressure method of Fevang and Whitson [6], which aims to give accurate simulation of near-well effects without the need for very small grid cells. The well inflow is calculated from a pseudopressure integral, which is analogous to the standard pseudopressure function used for dry gas reservoirs, but which also includes the gas relative permeability to take account of the reduced mobility due to liquid build-up near the well. The pseudopressure calculation has been implemented in AEA Technologys in-house compositional reservoir simulator, and tested on a range of problems. For example, a 3D compositional simulation model was set up using a rich condensate fluid on a 12x12x5 cartesian grid. The areal dimensions of each grid cell were 300 feet by 300 feet. A vertical production well was placed near to the centre of the model, and completed in all 5 layers. The

permeability was uniform in the vertical direction, but varied between layers. The well was produced at a plateau gas production rate of 10 MSCF/day with a limiting bottom hole pressure of 2000 psi. Three simulation runs were made 1. Regular grid with no local refinement or pseudopressure. 2. Radial local grid refinement (LGR) in the column of blocks where the well was completed. There were 4 cells in the radial direction, with the inner cell having dimensions of about 1 foot. Local timestepping was used on the radial grid. 3. Regular grid with pseudopressure. The results for gas production rate are shown in Figure 11. If we assume that the LGR calculation gives the most accurate result, the simulation with a regular grid overestimates gas production significantly. Using pseudopressure in the regular grid gives results which are much closer to the LGR run, but with the gas rate slightly lower after the end of the plateau production period. The computing times for these simulations showed that the pseudopressure calculation increased the run time by about 10%, whereas the radial LGR increased the run time by 140%. These results show that the pseudopressure method is a potentially useful tool for calculating well productivity in full field simulation models. It can also be used in simple spreadsheet calculations to estimate well productivity outside of a reservoir simulator. Current work is aimed at extending the method to allow for the changes in relative permeability due to high capillary number effects.

7. Conclusions
1. Liquid build-up in the near well region can cause a significant loss in productivity, even for very lean condensate fluids. 2. The most important parameters in determining productivity loss are the gas-oil relative permeability curves, expressed in terms of krg as a function of the ratio krg / kro. 3. Simulations using measured relative permeability curves often overestimate the losses in productivity due to liquid blockage. 4. Changes in relative permeability at high capillary number can have a significant effect on well productivity, and simulations should allow for this effect. 5. Relative permeabilities for near-well calculations need to be measured under conditions which are representative of the near-well region. 6. The pseudopressure technique provides a convenient and accurate way of modelling well productivity in full field simulation.

8. Acknowledgements
The work described in this report was carried out as part of a Joint Industry Project funded by the UK Department of Trade and Industry, Amoco UK Exploration Company, BP Exploration Operating Company Ltd, Esso Exploration and Production UK Ltd, Marathon Oil UK Ltd, Mobil North Sea Ltd, Phillips Petroleum Company UK Ltd, Saga Petroleum a.s and Texaco Britain Ltd.

9. References

1 D. Afidick, N.J.Kaczorowski and S. Bette. Production Performance of a Retrograde Gas Reservoir: A Case Study of the Arun Field. SPE 28749, presented at the SPE Asia Pacific Oil and Gas Conference, Melbourne, November 1994. 2 P.H.Diamond et al. Probabilistic Prediction of Well Performance in a Gas Condensate Reservoir. SPE 36894, presented at the SPE European Petroleum Conference, Milan, 2224 October 1996. 3 G.D.Henderson et al. Measurement and Correlation of Gas Condensate Relative Permeability by the Steady-State Method, SPE 30770, presented at the 1995 SPE Annual Technical Conference and Exhibition, Dallas, 22-25 October 1995. 4 H.L.Chen et al. Determination of Relative Permeability and Recovery for North Sea Gas Condensate Reservoirs, SPE 30769, presented at the 1995 SPE Annual Technical Conference and Exhibition, Dallas, 22-25 October 1995. 5 W.Boom et al. On the Use of Model Experiments for Assessing Improved Gas-Condensate Mobility under Near-Wellbore Flow Conditions, SPE 36714, presented at the 1996 SPE Annual Technical Conference and Exhibition, Denver, 6-9 October 1996. 6 O.Fevang and C.H.Whitson. Modelling Gas Condensate Well Deliverability, SPE Res Eng, November 1996.

10

Figure 1. Pressure in Region near to Gas Condensate Well

3200

3100

3000

Pressure

2900

2800

2700

2600

2500

500 Distance from well (feet)

1000

Figure 2. Liquid (oil) Saturation in Region near to Gas Condensate Well

16 14 12 Liquid saturation (%) 10 8 6 4 2 0 0 500 Distance from well (feet) 1000

critical oil saturation

Figure 3. Liquid Saturations in Constant Composition Expansion for 5 Fluids

0.50

Liquid Fraction (V/Vtotal)

0.40

0.30

0.20

Fluid a Fluid b Fluid c Fluid d Fluid s

0.10

0.00 0 1000 2000 3000 4000 5000 6000 7000

Pressure (psia)

Figure 4. Normalised Productivity Index for 1D Radial Simulations with 5 Fluids


1.2
Fluid A Fluid B

1
Normalised Productivity Index

0.8

Fluid C Fluid D

0.6

Fluid S

0.4

0.2

0
0 500 1000 Time (days) 1500 2000

Figure 5. Measured and Calculated Well Head Pressures for North Sea Lean Condensate Reservoir

2500

2000 Wellhead pressure (psia)

1500 Field data 1000 Simulation (measured rel perms) 500 Simulation (adjusted rel perms) 0 0 10 20 30 40 50 60 Gas flow rate (MMSCF/day)

Figure 6. Relative Permeability Curves for North Sea Lean Condensate Reservoir
1

0.8

Relative permeability

0.6

Krg: measured Krog: measured

0.4

Krog: adjusted to match well test data


0.2

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Gas saturation

Figure 7 Normalised Well Productivity Indices forNorth Sea Lean Condensate Reservoir

1 Adjusted rel perms 0.8 Normalised PI Measured rel perms

0.6

0.4

0.2

0 0 50 100 150 Time (days) 200 250 300

Figure 8. Relative Permeabilities used in Arun Reservoir Simulation (from ref 1)

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Gas saturation Krg (adjusted to match well test data) Kro (measured) Krg (measured)

Figure 9. Well Bottom Hole Pressures for 1D Radial Simulation - effect of Capillary Number and non-Darcy Flow

8000 reservoir pressure 7000 6000 pressure (psia) 5000 4000 3000 2000 1000 0 2 4 6 8 time (years) 10 12 14 BHP - base case BHP - Krg varies with Nc BHP - non-Darcy flow

Figure 10. Gas-Oil Interfacial Tension for Rich Gas Condensate


2.0 1.8 1.6 1.4

IFT (mN/m)

1.2 1.0 0.8 0.6 0.4 0.2 0.0 2000

Threshold IFT for rel perm changes

3000

pressure (psia)

4000

5000

6000

Figure 11. Gas Production Rates Predicted from Different Simulation Models

3D model with rich condensate fluid. Gas production rate for BHP = 2000 psi.
10000 gas production rate (MSCF/day)

8000

6000 No grid refinement 4000 Radial LGR Pseudopressure

2000

0 10 15 20 25 cumulative gas production (BSCF)

You might also like