Prediction of high viscosity two

Prediction of high viscosity twophase flows with LedaFlow
Jørn Kjølaas, SINTEF
Contents





Heavy oil - background
Experimental data with viscous oils
LedaFlow 1D predictions
LedaFlow Q3D predictions
Conclusions
2
Heavy oil - background
 Easily accessible light crude oil is slowly running out
 Off-shore production is moving to much deeper waters (expensive)
 Heavy oil is available at much lower depth (relatively cheap)
 There is much more heavy than light oil (i.e. Venezuela)
 Difficult to produce
 High viscosity (up to 100 Pa.s)
 Generally high sulfur and heavy metal content
 Refining process is expensive
 Accurate flow predictions are needed for proper design of
viscous oil transport facilities.
3
LedaFlow
 A next generation tool for prediction of multiphase pipeline
flows – from well to processing facility.
 Project vision: develop step change technology for
pipeline design, operation, training and trouble-shooting.
 Key model features:
WHP
1D
1D
 Point model
 1D pipeline model
 Special models (wells, valves, etc.)
 Profile model (cross-section)
Q3D
Q3D
Down
Comer
Q3D
Riser
1D
Q3D
 Q3D multidimensional model
 Networks of 1D and Q3D pipe segments
Q3D
 Thermal and compositional modeling
4
LedaFlow
 Validation data obtained at SINTEF’s lab at Tiller (Trondheim)
 Until recently, the main focus has been on light oils (condensate)
 The focus has now shifted towards viscous oils
5
Experimental data with viscous
oils (gas-oil)
6
Experimental data (gas-oil)
 Gockal (2008)
 173 data points
 Horizontal pipe
 50mm diameter
 Air + viscous oil (o=200-600 cP)
 Atmospheric pressure
Courtesy of TULSA University
 SINTEF Medium Scale data (FACE++)
 421 data points
 Horizontal pipe
 69mm diameter
 NexBase + SF6 (o=100 cP)
 NexBase/Exxsol mixture + SF6 (o=35 cP)
 4 and 8 bar pressure (g=26, 46 kg/m3)
 LDA measurements
7
Experimental data (gas-oil)
 IFE FACE in-kind data set (Statoil)
 250 data points
 =0º, 1º, 5º
 100mm diameter
 o=70-160 cP
 8 bar pressure (g=45 kg/m3)
 SINTEF Large Scale Campaign (2012)
 788 data points
 =0º, 0.5º, 90º
Top view
 189mm diameter
 NexBase + N2 (o=100 cP)
 20, 45, 85 bar pressure (g=25, 50, 95 kg/m3)
 LDA measurements
 Proprietary data (not available)
Side view
8
LedaFlow 1D
9
LedaFlow 1D
 SINTEF MS single phase data: The transition from laminar to
turbulent flow sets in earlier than what is reported by other authors.
This can probably be attributed to disturbances in the process that
generate turbulence.
 The laminar-turbulent
transition is often important
in high-viscosity systems,
and this transition has a
significant impact on the gasliquid flow regime.
 A particular challenge is to
correctly predict this
transition in stratified flow.
10
LedaFlow 1D
 SINTEF MS data (100 cP): The flow regime predictions are
incorrect in some cases.
LedaFlow SLUG
LedaFlow STRAT
11
LedaFlow 1D
 SINTEF MS data: The flow regime predictions are very
important with viscous oils!
12
LedaFlow 1D
 SINTEF MS data (35 cP): The flow regime predictions are
perhaps a little better for 35 cP than at 100 cP.
LedaFlow SLUG
LedaFlow STRAT
13
LedaFlow 1D
 SINTEF Medium Scale 35 cP data: The overall predictions
are quite reasonable.
 The flow regimes are better predicted than at 100 cP.
 The penalty for erroneous flow regime is less severe.
14
LedaFlow Q3D
15
LedaFlow Q3D
 All equations are derived in 3-dimensions and then
averaged over horizontal slices
 Three fluid and 9-field model
 Solves turbulence transport equations for all phases.
 Calculates transition from separated to dispersed flow and
vice versa.
 Compressible N-field flow solver
 Solves for both straight and curved pipes
16
LedaFlow Q3D (3)
 Pipe geometry can be curved
 Adds terms in momentum balance (centrifugal force, etc.)
a
b
c
d
17
Viscous oil simulations
 Simulation parameters
 Based on experiment by Gokcal et al. (2008)2
 g = 1.13/1.2 kg.m-3, l = 889 kg.m-3
 g = 1.87E-05 Pa.s, l = 0.181/0.587 Pa.s
  = 0.02 N.m-1
 Wall roughness = 10 m
 L = 40/60 m, D = 0.0508 m, = 0o
 Grid: 200 x 20 cells (length x height)
 Inlet: g = 0.5, l = 0.5, Us,g = 0.1/0.5/1.0/2.0 m/s, Us,l = 0.5 m/s
 tmax=0.01 sec; CFL=0.5
 Constant particle size (3 mm droplets, 1 mm bubbles)
2
B. Gokcal et al. (2008), SPE projects facilities and construction
18
Viscous oil (2)
 Usg=0.5
(pipe diameter is magnified x50)

1.0
(x50)

5.0
(x75)

10.0
(x75)
19
Viscous oil (3)
 High viscosity (0.587 Pa.s)
Pressure gradient
7000
dPdX (Pa/m)
6000
5000
4000
Experiment
3000
LedaFlow
2000
1000
0
0
1
2
3
Usg (m/s)
4
5
6
Viscous oil (4)
 "Low" viscosity
(0.181 Pa.s)
3000
Pressure gradient
dPdX (Pa/m)
2500
2000
1500
Experiment
1000
LedaFlow‐Lvisc
500
0
0
1
2
3
Usg (m/s)
4
5
6
Conclusions
 In recent years, due to the increasing interest in heavy oil production,
a lot of viscous-oil experimental data has been produced.
 We have compared the predictions of LedaFlow 1D (which has been
developed using low viscosity data) to some viscous oil data sets.
 Some challenges have been identified:
 The flow regime predictions are critical, because slugs have a huge
impact on the pressure drop (for low-viscosity oils, the penalty for getting
the flow regime wrong is much less severe).
 There are also challenges with regards to the determination of
laminar/turbulent flow. This is however not straight-forward in stratified
flow.
 These and other challenges will soon be addressed in a model
development activity targeted at improving the predictions in viscous
flows. This is now possible since we have a lot of experimental data.
 LedaFlow Q3D provides good results for viscous oil flows, and will be
used to obtain improved insight into these situations.
22
Thank you for your attention!
23