Complexity of Two-Phase Flow in Porous Media

Transcription

1 1 Complexity of Two-Phase Flow in Porous Media Rennes September 16, 2009 Eyvind Aker Morten Grøva Henning Arendt Knudsen Thomas Ramstad Bo-Sture Skagerstam Glenn Tørå Alex Hansen

2 2 Declining oil production, increasing energy consumption: Buying time for developing of alternative energy sources. Paul Roberts: Arguably the most serious crisis ever to face industrial society The End of Oil: On the Edge of a Perilous New World (2004) Low-permeability carbonate rock: All fluid transport through fractures -But where are the fractures, and are there enough of them? Recovering stuck oil: EOR Tar sands of Alberta, Canada

3 3 Lecture Plan: 1. Transient vs. Steady-State Flow 2. Numerical Network Model 3. Steady-State Flow in Numerical Model 4. Parallel Flow Instabilies 5. Statistical Mechanics of Steady-State Flow 6. Some Thermodynamics 7. Film Flow in Crevices 8. Summary

4 4 1. Transient vs. Steady-State Flow

5 5 Experimental Background K.J. Måløy s lab. at the Univ. of Oslo Inlet Q P L Defending Fluid: 90% by weight Glycerol - water solution Viscosity: Pa s Invading Fluid: Air Viscosity ratio: M = Interfacial tension: N/m W Outlet Q

6 6 Flooding Experiments Dimensionless numbers: Instabilities: Transients Ca = capillary number = viscous forces/ capillary forces M = Ca = M = viscosity ratio Ca = Ca = 0.22 (Måløy, Univ. Oslo)

7 7 Ca = Ca = Ca = 0.120

8 8 Steady State Flow What enters a RVE is statistically the same as what leaves it. An RVE inside a reservoir will typically experience steady-state conditions, and not the transients in a flooding experiment.

9 9 Steady state signifies that apart from statistical fluctuations, the system remains constant or changes slowly on the scale of the fluctuations. Hence, even if both fluids move and fluid clusters break up and merge, we will be dealing with a steady state as long as the averaged macroscopic parameters describing the flow remain constant (or change slowly on the time scale of the fluctuations). A setup for studying steady-state flow in the laboratory: Region of spatially homogeneous steady-state flow. Tallakstad, Måløy, Univ. Oslo

10 10 Clusters in spatially homogeneous steady-state flow Clusters have been artificially colored. Tallakstad, Måløy

11 11 Steady state Old wisdom: Equilibrium phenomena are easy Non-equilibrium phenomena are difficult Flooding Equilibrium thermodynamics is 150 years old Non-equilibrium thermodynamics is still being developed Porous media: Flooding has been studied for a long time Steady-state flow has been largely neglected

12 12 2. Numerical Network Model

13 13 Numerical Modeling Ca = M = 100 Ca = M = (Aker et al., 1998) Ca = M=1

14 14 Network of connected Pores Sharp corners: Wetting fluid forms films in the crevices. For now: no crevices, no films

15 15 Dynamics at the pore level: How do the interfaces move?

16 16 Coalescence and Instabilities: Snap-off

17 17 Coalescence and Instabilities: Coalescence

18 18

19 19 3. Steady State Flow in Numerical Model

20 20 Steady-state flow implemented on a torus Largest non-wetting cluster. Flow Direction (Ramstad, 2006)

21 21 Starting from a band of red fluid, the interfaces initially develop capillary fingering and stable displacement. Capillary fingering Wetting fluid Flow direction Non-wetting fluid Ca = M=1 Stable displacement

22 22

23 23 Steady-state configuration is independent of initial preparation of system: It is a state. (It is independent of initial conditions given the flow is fast enough for none of the fluids to get stuck due to capillary forces.) Flow direction Drainage and imbibition are now microscopic concepts.

24 24 Reconstructed pore networks Pore Network from Berea Sandstone (3mm)3 Each pore is described by a number of geometric parameters. Reconstruction by e.g. merging thin slices

25 25 Setting up steady-state flow in reconstructed network: Must make the non-periodic structure periodic

26 26 Making the three-dimensional network periodic in the flow direction:

27 27 Evolution towards steady-state flow Steady state Ca = 0.015, M = 1, S = 0.5 non-wetting saturation

28 28

29 29 Largest non-wetting clusters at different saturation levels in steady state S=0.65 Ca=0.015, M=1 S=0.67 S=0.59 S=0.71 Critical saturation

30 30 Cluster size distribution corresponding to the saturations just shown. At criticality N(s) s-2.27±0.08 Cluster size Average over 20 samples

31 31 Relative density of largest cluster P vs. saturation S. Critical saturation vs. capillary number: Sc = A + B log Ca Interesting consequences for radial flow

32 32 4. Parallel Flow Instabilities

33 33

34 34 Instabilities at the boundary of fluids moving in parallel. Wetting fluid Non-wetting fluid Wetting fluid What happens at this boundary when the capillary number is increased? Flow direction Periodic boundary conditions in the flow direction.

35 35 Kelvin-Helmholtz Shear Instability?

36 36 Flow direction. Flow periodic in the flow direction.

37 37

38 38 Mechanism: The non-wetting fluid produces viscous fingers that penetrate into the wetting fluid. The fingers experience side flow. They break off. This produces clusters of non-wetting fluid that diffuse into the wetting fluid and away from the non-wetting layer. There are never any clusters of wetting fluid within the non-wetting layer.

39 39 Non-wetting saturation profile normal to flow direction as function of time.

40 40 The derivative of the non-wetting saturation profile. This defines a mixing length scale λ.

41 41 Profile is stable and moves towards the non-wetting fluid: effectively an imbibition process.

42 42 Wetting fluid Constant current of non-wetting bubbles Non-wetting fluid Mixing zone moving with constant speed Steady state in a reference frame moving with constant speed to the right.

43 43 Starting out with the wetting fluid in the middle: an equilibrium is set up. The profile survives!

44 44 This layer is stable! Non-wetting fluid Wetting fluid Non-wetting fluid

45 45 5. Statistical Mechanics of Steady-State Flow

46 46 Returning to the concept of a state. A state is described by a small number of macroscopic parameters When the system is in such a state, it evolves through following a path in the space of all possible microscopic configurations such that the macroscopic averages remain constant. The equations that governs the path are the microscopic flow equations. These equations are very hard to solve (numerically). But, they provide far too much information! This information is needed to understand transients, but not to describe the dynamical steady state. Is there a way to get the needed information to describe macroscopically the steady state without having to resort to complete overkill?

47 47 Sequence of configurations through time integration: t t 4 t t The order of the configurations has been randomized: Does this randomization change the statistics? 2

48 48 If order plays no role: All steady-state properties will be completely described by the configurational probability distribution W{cf} where {cf} signifies the positions of all interfaces between the immiscible fluids in the porous medium. A configuration is fully described by the position of all interfaces. This leads to the possibility to exchange time integration by more efficient methods.

49 49 Metropolis Monte Carlo Sampling Configurational probability W{cf} Old configuration Test configuration {cfold} {cftest} Chosen by random change of old configuration. Draw a random number r [0,1]. If W{cfold}/W{cftest} > r: Reject test configuration. If W{cfold}/W{cftest} r: Accept test configuration.

50 50 We need to know what W{cf} is? Assumption: W{cf}=W(E{cf}) Probability depends on single macroscopic function E=E{cf}. Energy

51 51 Statistical mechanics of steady-state flow Pressure field solves the Kirchhoff equations. Microscopic energy function: E{cf} = (1/2) min Σ <ij> (pi-pj-pcij)2/µ eff ij Pressure field solves the Kirchhoff equations. Temperature 1/T=dΣ /de Effective viscosity Capillary pressure W{cf} = e-e{cf}/t W{cf} = W(E{cf}) (All configurations with same energy are equally probable.) Boltzmann distribution Configurational entropy Σ = log N(E) Basic assumption

52 52 All of thermodynamics now follows. Partition function: Z=Σ {cf} W{cf} Average energy (dissipation): E=Σ {cf} Thermodynamic variables: Pressure P Total flow rate Q Saturation S Fractional flow f Energy E Capillary pressure Pc Entropy Σ Temperature T Viscosity ratio M Free energy G = -T log Z E{cf} W{cf} / Z E=E(Σ,P) ( E/ P)Σ = - Q ( E/ Σ )P = T ( T/ P)Σ = - ( Q/ Σ )P

53 53 Numerical model: Controlling P (or Q) and saturation S. Tallakstad, Måløy, Univ. Oslo: Controlling P (or Q) and fractional flow f. Monte Carlo: S=S(T,P) f=f(t,p)

54 54 6. Some Thermodynamics

55 55 Temperature: Boltzmann weight: W{cf} = exp[-(1/2t) min Σ <ij> (pi-pj-pcij)2/µ eff ij ] Viscous pressures are all proportional to pressure P W{cf}=W(P,pc,T) Rescaling: W(P, pc, T) = W(P/T1/2,pc/T1/2,1) Setting T P2 Equation of state W(P, pc, T) = W(1, 1/Ca,1) Only dynamical parameter is Capillary number (besides M).

56 56 1st Law of Thermodynamics de = δ Q - δ W Reversible work related to change in flow rate Q; P dq. Irreversible work on the microscale related to creating new interfaces in the pores. 2nd Law of Thermodynamics In an isolated system, a process can occur only if it increases the entropy of the system. The steady state is at maximum entropy

57 57 3rd Law of Thermodynamics T 0 Σ 0 The zero-temperature state, i.e., when P = 0, the state with the least number of interfaces is the optimal one. This is when the two fluids are fully separated. This state has zero entropy.

58 58 A thermodynamic equation (Not yet a complete story) Fractional flow as a function of saturation S. Pressure difference as a function of saturation S. Keeping total flow rate Q constant.

59 59 Valid for a wide range of M. P = A (df/ds)q + B At constant Q

60 60 One more observation df/ds has a maximum for some F. P = C F2 + D F + E

61 61 This is the analytical form of the fractional flow vs. saturation curve. This corresponds to having an analytical form for the relative permeability curves. Experimental results from the literature

62 62 More Phase Transitions in the Steady State (2D system) (Single-phase flow of wetting fluid) (Single-phase flow non-wetting fluid) Single vs. multiphase flow.

63 63 Three-dimensional parameter Space: Saturation S Viscosity Ratio M Capillary Number Ca Both fluids move Only one fluid moves

64 64 7. Film Flow in Crevices

65 65 P: Pressure drop in bulk fluid p: Pressure drop in film Q: Bulk flow rate q: Film flow rate A: Bulk area a: film area

66 66 The Wedge Problem in a Closed Geometry Darcy equations for bulk and film flow: Film conservation

67 67 Surface Tension Eliminating P, p and q: Young relation

68 68 A Diffusion-Convection-Reaction Equation:

69 69 Diffusion constant Convection speed Reaction constant Shock fronts Diffusion constant increases with increasing concentration.

70 70 The area along the tube is constant: When Q = 0: No bulk flow. Convection term is zero.

71 71 No bulk flow: Diffusion only y=x2/t

72 72 Solutions: x/t1/2

73 73 Physical solution x/t1/2 Shock front

74 74 Bulk flow: Non-linear diffusion-convection equation Solitary wave solution Integration constants Z=x-ut Solitary wave speed

75 75 Solitary wave speed u (a+ + a-) a+ a-

76 76 Solitary wave speed u (a+ + a-)

77 77 a Stationary solutions: a1 a2 x

78 78 a x

79 79 Solvable under steady-state conditions for a class of profiles

80 80 Asymptotic solution: There is a divergence for finite AT/Ai for which a diverges. Critical area at which film goes unstable.

81 81 This instability leads to bubble formation in the pores.

82 82 AT = (x + 0.5)2

83 83 8. Summary

84 84 Transient vs. Steady-State Flow Numerical Network Model Steady-State Flow in Numerical Model Parallel Flow Instabilies Statistical Mechanics of Steady-State Flow Some Thermodynamics Film Flow in Crevices