1 B015 SIMULATION OF AIR INJECTION IN LIGHT-OIL FRACTURED RESERVOIRS – FROM THE MATRIX BLOCK SCALE TO THE CROSSSECTION SCALE BY USING A DUALPOROSITY MODEL Philippe DELAPLACE, Sébastien LACROIX, Bernard BOURBIAUX (IFP) and Yann LAGALAYE (TOTAL) Institut Français du Pétrole, 1&4 avenue de Bois Préau, 92852 RUEIL-MALMAISON Cedex, FRANCE Abstract Air injection can be an economical alternative for pressure maintenance of fractured reservoirs as it avoids re-injecting a valuable associated gas and/or generating or importing a make-up gas. In addition, the oil recovery can be enhanced thanks to the thermal effects associated with oil oxidation. However, such an improved recovery method requires a careful assessment of the involved reservoir displacement mechanisms, in particular the magnitude and kinetics of matrix-fracture transfers. Considering the situation of a light-oil fractured reservoir, compositional thermal simulations of matrix-fracture transfers were carried out on a finegrid single-porosity model of a matrix block or a stack of blocks surrounded by air-invaded fractures then on the equivalent (up-scaled) dual-porosity model, respectively (a) to identify the main physical mechanisms controlling matrix-fracture transfers during air injection (b) to dispose of a reliable simulation tool usable for field-scale predictions. Firstly, the reference fine-grid simulations show that gas diffusion and thermodynamic transfers are the major physical mechanisms controlling the global kinetics of matrix-fracture transfers and the resulting oxidation of oil. The chronology of extraction of oil components from the matrix blocks is clearly interpreted in relation with phase transfers. Then, the predictions of the dual-porosity model are shown to be in very good agreement with those of the reference model, thanks to a specific numerical formulation which ensures a proper up-scaling of diffusion and inter-phase transfers at the overall scale of matrix blocks. 1 - Physical Data Petrophysical Data The petrophysical and thermodynamic properties used in our simulations are largely inspired from the Ekofisk field (Thomas et al., 1983, 1991, Agarwal et al., 1999, Jensen et al. 2000). The matrix medium has a permeability K of 1mD, a porosity Φ equal to 30%. The calorific capacity of the unsaturated rock is equal to 2.35 Jg-1°C-1 and the overall thermal conductivity of the fluid-saturated rock equals 1.8 Wm-1°C-1. The working pressure is 5600 Psi, very close to the bubble point pressure and temperature is 266°F. The irreducible water saturation, Swi, is 0.15, the critical gas saturation, Sgc, is 0.01 and the residual oil saturations, Sorw and Sorg, are both equal to 0.25. Capillary pressures and relative permeability curves are shown figure 1. 9th European Conference on the Mathematics of Oil Recovery — Cannes, France, 30 August - 2 September 2004 2 Kr WATER-OIL Pc WATER-OIL 1 2 kr WATER kr OIL 0.9 1.5 0.8 0.7 1 Capillary Pressure (Bar) 0.6 0.5 0.4 0.3 0.2 0.5 0 -0.5 -1 0.1 0 0 0.1 0.2 0.3 0.4 0.5 0.6 Water Saturation 0.7 0.8 0.9 1 -1.5 0 0.1 0.2 0.3 0.4 0.5 0.6 Water Saturation 0.7 0.8 1 Kr OIL-GAS Pc GAS-OIL 1 0.03 kr GAS kr OIL 0.9 0.025 Capillary Pressure (Bar) 0.9 0.8 0.7 0.02 0.6 0.5 0.015 0.4 0.01 0.3 0.2 0.005 0.1 0 0 0 0.1 0.2 0.3 0.4 Gas Saturation 0.5 0.6 0.7 0 0.1 0.2 0.3 0.4 0.5 0.6 Gas Saturation 0.7 0.8 0.9 1 Figure 1: Relative Permeabilities and Capillary Pressure Curves Multi-component Oil A light oil representative of Ekofisk field was adopted as the hydrocarbon phase for most simulations. In order to take into account the specific volatility of the light fractions and the possibilities of oxidation of the heavy fractions of reservoir oil, the following optimal set of pseudo-components (molar concentrations in parenthesis) is considered to model accurately the oil phase : - the pure component methane (C1 = 60%), - two intermediate pseudo-components (C2C3 = 12% and C4C9 = 14%) - and two heavy pseudo-components (C10-C17 = 8% and C18-C30 = 6%). To complete this thermodynamic system, we need extra components namely nitrogen (N2 = 80%) and oxygen (O2 = 20%) for air and carbon dioxide resulting from the oxidation of heavy components. The Peng-Robinson Equation Of State (EOS) is used to model the fluid behaviour: gas-liquid equilibrium, density of each component for each phase in the presence of air and/or CO2. Air, water and carbon dioxide Air consists of 20% oxygen and 80% nitrogen which are declared as components of the EOS and may be partially dissolved in the oil phase. In the same way, carbon dioxide is also declared as a component of the EOS but the water component is considered separately because we don’t yet dispose of a complete three-phase thermodynamical flash calculation in our dual porosity simulator. In consequence, water vaporisation and dissolution of carbon dioxide in the water phase are neglected in this work. Oxidation process To formulate oxidation reactions, each pseudo-component k is modelled as the equivalent alkane, CnkH2nk+2, with nk the number of carbons determined from the molecular weight of the pseudo-component: < Mw > k = Mw [Cnk H2nk+2] = (12 nk + 2 nk + 2) = 14 nk +2 3 The reaction of oxidation of pseudo-component k is then approximated as: [CH 2 ]nk + 3nk O2 → nk [CO2 ] + nkH 2 O 2 (+ ∆Hk of Reaction ) . . . . . . . . . . . . . (1) This reaction of complete oxidation (with CO2 and H2O productions) is applied to each of the three heaviest pseudo-components with the following reaction rate: Qr = VK r e − Ea / RT (φρo So xk )α ( PyO2 ) β . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . (2) where: - the reaction rate, Qr, in moles per time unit, is given for a reference component (the oil pseudocomponent in our case) in the reaction formula, - V is the cell volume (if Qr is expressed at cell scale), - Kr is the exponential pre-factor, i.e. a constant characteristic of the reaction considered, with a dimension consistent with α and β values, - ρo the molar density of the oil phase, - xk the molecular concentration of the hydrocarbon pseudo-component, - yO2 the partial pressure of oxygen, - Ea, the energy of activation, - α and β are the reaction orders with respect to the hydrocarbon component concentration and to the oxygen partial pressure (α = 2 and β = 1 in this study). Actually, only the heaviest pseudo-component, C18-C30, is oxidised which provides the necessary heat enhancing the diffusion/vaporisation of the intermediate pseudo-components. Whereas reaction enthalpies are rather well known (from the reaction enthalpy of a methyl radical for instance), the pre-exponential factor and energy of activation require to be calibrated from representative experimental data. In this study, we adopted estimated values. The energy of activation, Ea, closely controls the kinetics of reaction, as the parameter of an exponential function. The same value (61250) was assumed for all three reactions, which is close to existing values found in Burger (Burger et al., 1985). Diffusion process The molecular diffusion flux in the “p” phase for the component “k” is expressed as: ( p) k (F ⎡ φ ⎤ ) = ⎢Tk ∆(Ck )⎥ ⎣ τ ⎦ ( p) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (3) where: - Tk = Dk.(A/∆X)eff is the diffusion transmissivity with (A/∆X)eff the effective surface to length ratio and Dk the effective diffusion coefficient described below; τ the matrix tortuosity; and ∆(Ck) the component concentration difference. The diffusion of components within the liquid phase can be neglected. For the vapor phase, effective diffusion coefficients Dk are computed at each time step as a function of pressure, temperature and the gas composition (using the Wilke’s Method described in Da Silva SPE 19672) because temperature and composition may vary within a large range of values during air injection. 2 - Up-Scaling Strategy This work is focused on the modelling of physical processes at different scales, from the 1D-X matrix block to the 2D-XZ stack of blocks and from constant boundary conditions to complex boundary conditions (variable in space and time). At each scale, single porosity simulations and equivalent dual porosity simulations were performed. The single porosity simulations performed on fine grids give insight of the physics of matrix-fracture transfers and provides reference solutions for the dual porosity model. We will use four physical models corresponding respectively to four levels of increasing complexity : 9th European Conference on the Mathematics of Oil Recovery — Cannes, France, 30 August - 2 September 2004 4 The “1D-X Block” model will be used for studying the pure matrix-fracture transfers, without any gravity effect. Without limitation of meshing, this model allows us to accurately follow saturation, oil or gas composition profiles along the X-axis, orthogonal to the matrix-fracture exchange surface. The “2D-XZ Block” model will be used for studying at the matrix block scale, a complete and realistic physical case including gas-oil gravity drainage. The “2D-XZ Continuous Column” model will be used for studying a realistic air injection in a fractured reservoir where gas contact moves down slowly in the fracture network inducing a complex boundary condition in terms of available oxygen for matrix oil oxidation. The “2D-XZ Stack of Blocks” model is a composite model consisting in a stack of continuous matrix blocks without capillary continuity to validate the new dual-porosity formulation at a scale close to a reservoir cell. 3 – 1D-X Block Results For this first study, the boundary conditions are controlled by a high vertical flux of air in the fracture that ensures a constant gas composition of 80% nitrogen and 20% oxygen and drains most of the heat generated within the matrix block, holding a quasi constant temperature two or three degrees above the initial temperature(131°C to 134°C). Only the single porosity model is gridded along the x-axis and enables us to see the evolution of the gas saturation or oil composition during diffusion/vaporisation exchange between matrix and fracture. Figs. 2 gives, for different times, the saturation and oil composition profiles along a large matrix block (half block length = 50 cm): the pure diffusion case (equivalent to a nitrogen injection) is shown on the left whereas Low Temperature Oxidation effects (about 270°F) are included on the right. The diffusion process seems to be the dominant process at the short time scale and it is determinant in the control of the oxidation process. We can observe two different profiles of gas invasion : a single “piston” displacement in the purely-diffusive case and a “double piston” when the oxidation process occurs. These observations are at the starting point of a new dynamic formulation of matrix-fracture gas diffusion transfers in the “dual medium” version. This new formulation takes into account variable diffusion transmissivity values instead of constant ones in the conventional formulation. 4 - 2D-XZ Block Results with Ideal Boundary Conditions This model was used to study, at the matrix block scale, a complete and realistic physical case including gravity drainage. The boundary conditions are controlled by a significant gas diffusion exchange between the fracture and a large external air tank that also maintains a constant gas composition of 80% nitrogen and 20% oxygen. This new boundary condition radically differs from the previous one from the point of view of the thermal behaviour. In this case, the heat is drained off the matrix only by thermal conduction to the surrounding infinite medium having the same thermal properties as the matrix block. This boundary condition revealed itself to be close to an adiabatic condition for the matrix block where the temperature strongly increased, up to 400°C. It is nevertheless closer to realistic conditions than the previous isothermal conditions because the latter required a fracture displacement rate much higher than practical reservoir values. Figure 3 shows the evolution with time of oil components (mass in place and mass produced). 5 OIL COMPOSITION PROFILES at time = 160 days OIL COMPOSITION PROFILES at time = 160 days 1 1 Water Phase 0.9 0.6 Gas Phase 0.5 0.4 0.3 C1 0.2 O2 0.1 and N2 dissolved in oil phase 0 0 5 10 15 20 25 30 X-Depth (cm) 35 40 45 0.7 0.6 C1 0.3 0.2 CO2 0.1 and N2 dissolved in oil phase 5 0.9 0.8 0.8 Oil Sat. Norm. Composition Oil Sat. Norm. Composition 0.9 0.7 0.6 0.5 0.4 0.3 45 0 0 50 5 0.9 0.9 0.8 0.8 0.7 0.6 0.5 0.4 C18-C30 C10-C17 10 15 20 25 30 X-Depth (cm) 35 40 45 50 45 50 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 50 OIL COMPOSITION PROFILES at time = 1800 days 1 Oil Sat. Norm. Composition Oil Sat. Norm. Composition OIL COMPOSITION PROFILES at time = 1800 days 0.2 45 0.3 1 0.3 40 0.4 0.1 40 35 0.5 0.1 35 20 25 30 X-Depth (cm) 0.6 0.2 20 25 30 X-Depth (cm) 15 0.7 0.2 15 10 OIL COMPOSITION PROFILES at time = 400 days 1 10 C4-C9 C2-C3 OIL COMPOSITION PROFILES at time = 400 days 5 C10-C17 0.4 0 0 50 Gas Phase 0.5 1 0 0 C18-C30 0.8 Oil Sat. Norm. Composition Oil Sat. Norm. Composition 0.8 0.7 Water Phase 0.9 0.1 5 10 15 20 25 30 X-Depth (cm) 35 40 45 50 0 0 5 10 15 20 25 30 X-Depth (cm) 35 40 Figure 2: Saturation and Oil Composition profiles versus time with only diffusion (on the left) and with Low Temperature Oxidation (on the right) 9th European Conference on the Mathematics of Oil Recovery — Cannes, France, 30 August - 2 September 2004 6 --- Figure 3: Oil components history for Drainage (….. ),Diffusion, ( ) and Reaction (______) . The simulation restricted to air-oil Drainage presents a trivial oil recovery without any compositional effect, contrary to Diffusion and Reaction cases: Figure 3 shows that vaporisation and mass transfer by diffusion control the recovery of the light components. Comparing the pure Diffusion case (close to a nitrogen injection) to Reaction case, we can see no sensitive effect on C1 produced but a sensitive effect on the C4-C9 produced and heavier fractions. If we focus on the Diffusion case about the two heaviest oil fractions, we can see the C10-C17 is produced very slowly whereas C18-C30 is not produced at all. Because the vaporisation/diffusion process is faster than drainage and firstly concerns the light components, the remaining oil becomes highly concentrated in heavy components and air-oil drainage slows down due to the oil viscosity increase which finally leads to the selective trapping of the C18-C30 fraction by capillary retention. In addition, comparing mass in place to mass produced for the Reaction case, we can see that half of the C18-C30 fraction is produced whereas all the C10-C17 fraction is rapidly produced. Actually, the thermal effect induced by the oxidation of part of the heaviest fraction strongly enhances the diffusion/vaporisation process. 7 5 - 2D-XZ Model of a Stack of Blocks with Realistic Boundary Conditions S in g le M e d iu m M a tr ix B lo c k D u a l M e d iu m 1 D u a l M e d iu m 2 m 2 0 .2 5 m F ra c tu re s 1 m m 2m 10 10 cells Dual Medium Dual Medium Cells for for 10 10 Blocks blocks 3 4 20 m 1 Dual 1 cell Medium Dual Medium Cell for for 10 Blocks 10 blocks Figure 4: Validation of a new Dual Porosity formulation in 3 upscaling steps New realistic boundary conditions are taken into account for these last numerical experiments simulating a slow descent of the air-oil contact in the fracture network. The first upscaling step consists in comparing the reference solution provided by a single-porosity model with the solution provided by a dual-porosity model gridded in the vertical direction in order to reliably predict gravity drainage. This step allowed us to validate the upscaling of lateral diffusion exchanges between fracture and matrix for a variable gas composition along the fracture. The oxygen is consumed in the upper part of the matrix rock while nitrogen, carbon dioxide and vaporised oil fractions are drained off by the fracture. The second upscaling step consists in comparing the results provided by this gridded dualporosity model with the results provided by a single-cell dual-porosity model where all the physics, including gravity, are up-scaled. We don’t show any result about these two first upscaling step since the upscaled results perfectly matched the references results. The third upscaling step consists in comparing a dual-porosity model made up of the superposition of N single-block cells as modelled in the previous step with a single-cell dual-porosity model of N blocks subjected to variable boundary conditions. A reference air injection flow rate was defined corresponding to a frontal velocity of one foot/day assuming a total displacement of the oil-saturated matrix pore volume (85% of pore volume). The results shown in Figure 5 correspond to the third up-scaling step (for an average frontal velocity of 1/5 foot/day). The single-cell model (broken line) and the ten-cell model (solid line) give very close results for all light and intermediate oil fractions. A time delay is however observed for the heaviest oil fraction, because of the illegitimate averaging of oxygen concentration along the stack for the single-cell model. 9th European Conference on the Mathematics of Oil Recovery — Cannes, France, 30 August - 2 September 2004 8 Figure 5: Mass of Oil Components recovered from the matrix rock for a stack of 10 matrix blocks (DZ = 2m , DX = DY = 50 cm) Conclusions Reference fine-grid simulations revealed the major physical mechanisms involved in matrix-fracture transfers during air injection in a fractured light-oil reservoir. Gas diffusion and thermodynamic phase equilibria control the kinetics of matrix-fracture transfers and the resulting oxidation of oil. We now dispose of a reliable field simulation tool disposing of all the required capabilities to predict air injection scenarios in fractured reservoirs, and other gas injection scenarios involving multiphase compositional matrix-fracture transfers. Acknowledgements This work was supported by Total and the authors would like to thank D. Foulon for fruitful discussions. References Agarwal B., Hermansen H., Sylte J. and Thomas L.K.. Reservoir Characterisation of Ekofisk Field: A Giant Fractured Chalk Reservoir in the Norwegian North-Sea – History Match. SPE 68096 presented at the Reservoir Simulation Symposium, Houston, TX, 1999. Burger J., Sourieau P. and Combarnous M.. Thermal Methods of Oil Recovery. Editions Technip, 1985. Da Silva F.V. and Belery P.. Molecular Diffusion in Naturally Fractured Reservoirs: A Decisive Recovery Mechanism. Paper SPE 19672 presented at the Ann. Tech. Conf. & Exh., San Antonio, TX, Oct. 8-11, 1989. Jensen T.B., Harpole K.J. and Osthus A.. EOR Screening for Ekofisk. Paper SPE 65124 the European Petroleum Conference in Paris, 2000. presented at Thomas L.K., Dixon T.N., Pierson R.G. and Hermansen H.. Ekofisk Nitrogen Injection. SPE 19839, SPE Formation Evaluation, June 1991.