Theory and Application of Numerical Simulation of Chemical Flooding in High Temperature and High Salt Reservoirs

Applications, theoretical analysis and numerical methods are introduced for the simulation of mechanical models and principles of the porous flow in high temperature, high salt, complicated geology and large-scale reservoirs in this paper. Considering petroleum geology, geochemistry, computational permeation fluid mechanics and computer technology, we state the models of permeation fluid mechanics and put forward a sequence of implicit upwind difference iteration schemes based on refined fractional steps of the upstream, which can compute the pressures, the saturation and the concentrations of different chemistry components. A type of software applicable in major industries has been completed and carried out in numerical analysis and simulations of oil extraction in Shengli Oil-field, which brings huge economic benefits and social benefits. This software gives many characters: spatial steps are taken as ten meters, the number of nodes is up to hundreds of thousands and simulation time period can be tens of years and the high-order accuracy can be promised in numerical data. Precise analysis is present for simplified models of this type and that provides a tool to solve the international famous problem.


Introduction
At present an effective method, water-flooding, to keep the pressure of reservoirs is popular in the world, and the recovery efficiency is more outstanding than any other natural exploring forms.It gives more benefits and helps Chinese oil fields keep high quantity production.It continues to be more important and a strategic job to develop the exploiting efficiency of crude oil in the way water-flooding driving.
Much crude oil remains in the reservoir after water-flooding exploiting, which stays underground due to the constraint of capillary force, or doesn't move due to slight influenced region and the fluidity ratio between displacement phase and driven phase.How to develop the displacement efficiency?A popular method considered is that the injected mixture includes chemical addition agents such as polymer, surface active agent and alkali.Polymer can optimize the fluidity of displacement phase, modify the ratio with respect to driven phases, balance the leading edges well, weaken the inner porous layer, and increase the efficiency of displacement and the pressure gradient.Surface active agent and alkali can decrease interfacial tensions of different phases, then make the bounded oil move and gather.Some hypotheses should be made for the mathematical models.Local thermodynamic equilibrium holds in the reservoir, solid phase has no motion, and the rock and mixture fluid are slightcompressible, of Fick dispersion, ideal and suitable for Darcy Law.
The equilibrium equation of multi-phase, multi-components and slight compressible mixture is formulated by a nonlinear coupled system of partial differential equations.It is hard to solve this system because many modern numerical methods such as mixed element, finite element, finite difference and numerical algebra, will be involved in the simulation.In general speaking, based on physical means the pressure function is solved by an implicit scheme and the concentration values are obtained by an explicit solver or an implicit solver.The scholars try to find good ways analyzing the data and numerical results and doing some research work in simulation, which describe the whole process of chemistry displacements very well and help the engineers control the rules and process of displacement and forecast the recovery efficiency of natural oil and compute the oil percentage of output liquid and the percent of polymer and surface active agent.By numerical research the curves describing different components motion are shown, and some plans are made about the beginning and end of injected liquid and some related parameters of natural oil efficiency are derived.These conclusions, important techniques in chemistry displacements, can be used in forecasting the characters of fields, choosing different optimization plans, establishing the models of chemical displacements of reservoir, completing computational software and carrying out the numerical simulation.Petroleum engineers and mathematicians pay more attention to modern new techniques of exploiting natural oil.
Yuan visited United States and accomplished some work cooperate with Prof. R. E. Ewing during 1985 to 1988, and kept a series of research in theoretical analysis and applications of numerical simulation.With the leading of Yuan several research groups undertook some important projects from 1991 to 1995 such as "Eighth-Five" national key science and technology program (the Program for Tackling Key Programs) (85-203-01-087) entitled "research and application of the polymer displacement software" [1]- [6].The software was applied in designing plan and research work of polymer displacements in industrial production region of Daqing Oilfield.Many conclusions from actual numerical results are illustrated such as effects of fragments, fragments setting of rinsing protection, quantity of polymer, and used in actual simulations which give rise to outstanding economic and social benefits [7]- [9].Later the authors undertook a key tackling program of oil administration of Daqing Oilfield (DQYJ-1201002-2006-JS-9565)-solving development of mathematical models and completing explain of reservoir [10].This software system is also applied in three compound combination flooding of Gudong Little Well experimental region of Shengli Oilfield, polymer flooding of Gudong Middle One experimental region, optimization of combination flooding expanded experimental region of Gudong West region and feasibility of active water flooding of Gudong eighth region, and many interesting results are obtained [11].
Theory, method and application of numerical simulation are studied for high temperature, high salt, complicated geology and large-scale reservoirs and the principle of chemical flooding in this paper.Based on the former research, the conclusions and more discussion of the national major special project on science and technology (2008ZX05011-004) "Study on key technology of chemical flooding numerical simulation in high temperature and high salt reservoirs (on numerical simulation)" are given, which consists of permeation fluid mechanical models of numerical simulation of high temperature and high salt polymer flooding and compound combination flooding, numerical methods, applicable software, theoretical analysis and applications in oilfields.

Mathematical Model of Polymer Flooding and Compound Combination Flooding
This section includes five subsections: Section 2.1, basic hypothesis, Section 2.2, conservation equation of matter, Section 2.3, the pressure equation, Section 2.4, the concentration equation, Section 2.5, the concentration equation of chemical components.

Basic Hypothesis
Mathematical model of polymer flooding and compound combination flooding is derived under the following hypothesis.Local thermodynamic equilibrium holds in the reservoir, solid phase doesn't motion, rock and mixture fluid are slight-compressible, of Fick dispersion, ideal and suitable for Darcy Law [1]- [3] [7]- [10].

Conservation Equation of Matter
Under the primary hypothesis, a conservation equation of the i -th component is stated as follows dependent of the i -th concentration i C  : where il C represents the concentration the concentration of the i-th component in the l -phase, i Q means source sink term, p n means the number of phases, and subscripts l denotes the order of phases.The symbol i C  represents the total concentration of the i -th component, i.e. the summation of concentrations of the i -th component in different phases (including adsorbed phase): where cv n , the number of component whose volume couldn't be ignored, and ˆk C denotes the adsorption concentration of the component k .
The density of the i -th component is dependent of the pressure i ρ under slight compressible case： ( ) where o i ρ is the density of the i -th component under considering reference pressure r p , and p means the pressure and o i C is the coefficient of compressibility of the i -th component.
Suppose that the rock is compressible, then the function of the porosity φ and the pressure is ( ) where r C means the coefficient of compressibility of the rock.Darcy velocity, l u , is described by Darcy Law, ( ), where l p means the pressure of phases, K is the permeability tensor, D is the depth, rl K is the relative permeability, l µ is the viscosity and l γ is the proportion.
The dispersion flux is expressed in the following Fick formation： , , , .
The dispersion tensor il

The Pressure Equation
Considering all the conservative equations of each matter with positive volume together, using Darcy Law and capillary pressure to express the flux and the pressure relations of different phases respectively, and combining with the following constraints We can get the pressure equation of referenced phase: ( ) where and the total relative fluidity T λ is The total coefficient of compressibility t C , is a function dependent of the compressibility of rocks r C and components of the mixture

The Concentration Equation
Let w S and o S be concentrations of water phase and oil phase denoted by subscripts w and o denote and the relation 1 S S + = holds obviously.The equations on concentrations of water and oil phases are expressed as follows by conservation of the mass (1) By Darcy Law the velocities of different phases are derived by , where are the fluidities of two phases, and K , rw K and ro K are the absolute permeability tensor, relative permeabilities.K is the absolute permeability of medium, and w µ , o µ are the viscosities of water-oil phases dependent on their concentrations and the satura- tions of polymer and the two opposing principles in nature. 1 P , 2 P and 1 γ , 2 γ respectively denote the pres- sures and the densities of water phase and oil phase and D is the depth function, ( )

The Concentrations of Chemical Substance Components
Note that all the components (polymer and different principles) are mixed in the water phase and no transmission takes place, then , 0, . where is dependent of the absorptions.

Numerical Methods
This section consists of three subsections: Section 3.1 solving the pressure equation, Section 3.2 solving the concentration equation, and Section 3.3 structuring a numerical algorithm of components.

Solving the Pressure
Let the parameters with subscripts w , o be related with the water and the oil respectively, such as w S , o S and w P , o P denote the saturations and the pressures of water and oil phases.Note that the mixture fluid is only made by oil and water two phases in the model of polymer flooding and compound combination flooding and we can describe the pressure in a simple formula ( ) Using Darcy Law and the formula of capillary force, we rewrite the above equation as follows with respect to an unknown variable ( ) The initial values of the saturations are known at the beginning of simulations while the pressure values should be initialized in the following process.When the pressure n P at the n -th time step are known, then the flow velocities are obtained and the values of the saturation and components at the next step ( ) The pressure, denoted by a parabolic equation, is obtained by a seven-point central difference method.Considering the physical features of two-phase we assign rw λ and ro λ , the values of the left-side term, in ac- cordance with upstream principles.At the injected wells and produced wells with fixed quantities, the right source term can be assigned directly and the values of quantity are determined by the difference between the pressure of local regions and the pressure of bottom holes at injected wells and produced wells with fixed pressures.The production quantities of different phases are distributed by the relative fluidity of oil-water phases.In addition, the pressure equation is degraded into an equation of elliptic type, and the matrix is not strictly diagonal-dominated under an impressible assumption (the coefficients of compressibility are assumed to be zero).In the way of taking the diagonal unit be 1 and non-diagonal units be zero, the equation of feature edges is consistent with the equation of normal oil deposits.The data in feature edges don't need to be replaced by the values of solutions, which makes the quantity of physical data as a constant in the computation.If the program (necessary for the design) runs at feature edges virtual data will be used in the whole computation.
Given n P , using upstream seven-point central difference algorithm to compute ( ) .
The subscript denotes the upstream position in the first direction between the i -th point and the ( ) An assumption is given that there is no flow moving through the boundary.That is to say that its boundary condition is homogeneous of Neumann type.The quantities at the injected wells known and those at the produced wells , , , , = are given implicitly by the flowing bottom hole pressure f P , the pressures of phases and the relative mobility ratio.Distributed quantities are computed by an allocation program after the pressure values are obtained.It is easy to solve the saturation equation when the values of source and sink terms.

Solving the Saturation
The pressure of water phase w P at 1 n t + and the quantity w Q are known, then the saturation of water w S at 1 n t + are computed by the following discrete algorithm, , where the subscript i + denotes the upstream position according to the moving trend between i x and it is either i or 1 i + .It is hard to solve the nonlinear equations directly. , , , The program runs based on upstream sequence rule, and the iteration value at the ( ) l + -th step is computed until the relative error meets a designed requirement or the iteration reaches the steps when the values at upstream points are known.The final iterative value is denoted by 1 n w S + .

Numerical Algorithm of Concentration Components
The components of water phase keep conservation of the mass of anions, cations and molecules and other particles, whose equation is of diffusion-convection and convection dominated.It has more strengths such as high order of accuracy and high efficiency of simulation applying decomposition of operators into the nonlinear system and solving two subproblems: a hyperbolic equation of convection type and a diffusion equation.The former is solved implicitly by an upwind method, which can be carried out explicitly by an upstream technique.The latter is solved by alternating directions finite difference method, which can improve the computational speed.The concentration equation of k -component, a typical convection-diffusion equation, is simplified as follows where the dispersion tensor is a diagonal form.Given the saturation w S and the flow field w u , the concen- tration k C is to be computed by using an implicit upwind method to solve a convection problem, where the subscript k is ignored and C denotes a component concentration. , The values 1,0 n C + is obtained, then the diffusion equation is discretized alternatively in three directions.In x-direction, Then in y -direction, ( At last in z -direction, Then the discrete solutions C + are obtained and the computation runs in the next step.

Viscosity Computation of High Temperature and High Salt Reservoir
The polymer hydrolysis can decrease the viscosity in the high temperature and high salt reservoirs, and the viscosity is different under different polymers.The viscosity is computed by ( ) ( ) ( ) where the values of parameters are defined as follows,

Computation Program Illustration
This section illustrates the computation program by Figure 1.

Actual Experimental Tests of Oil Fields
The adaptation efficiency of the software SLCHEM is tested in view of three aspects: dependability, universality and special applicability for large-scale oilfields.The numerical results are dependable by comparing with the actual results and popular business software computations.The software is applied successfully in different fields of Shengli Oilfield and the universality is tested.The software is used in large-scale oilfields and the special applicability is tested.

1) Experiments of the polymer flooding in small-scale oilfields
The rectangle computational domain (Tuo Block 28) is partitioned into 22 × 24 × 6 subdomains with uni-  1.
The comparison of moisture content of produced oil of SLCHEM, VIP, and actual results are shown in Figure 2.
The time cost of SLCHEM is about 0.44 hour, and the material balance error is satisfactory.The relative total error comparison with actual moisture content is about 7.8% before the polymer flooding is injected.All the results show that the computation of SLCHEM runs fast, the numerical results are reliable and this scheme can be applied into present oilfields production.

2) Experiments of surfactant-polymer flooding agents in middle-scale oilfields
The rectangle computational domain (Sheng Block 2) is partitioned into 82 × 74 × 7 subdomains with uniform steps and the spacial steps in x -direction and y -direction are taken as 33.60 m and 29.29 m, respectively.Formation reserve is 3.4938 × 10 8 m 3 .There are sixty three wells in this block.The simulation works about   3.
The comparison of moisture content of produced oil of SLCHEM and actual results are shown in Figure 4.
The time cost of SLCHEM is about 8.45 hours, and the material balance error is 10 −6 .The relative total error comparison with actual moisture content is about 7.0%.All the results show that the computation of SLCHEM runs fast, the numerical results are reliable and this scheme can be applied into present oilfields production., , ,  ( ) ( )  ( ) of transition layer is computed by using speedup method in x -direction by (34a), { } Using upstream order, implicit upwind Newton iteration to solve the saturation of water phase, then the values of oil saturation 1 Newton iteration under taking the value at the previous step as the initial condition

Figure 1 .
Figure 1.Computation program illustration.formsteps and the spacial steps in x-direction and y-direction are taken as 45.61 m and 47.35 m, respectively.The irreducible water saturation is 0.175, residual oil saturation is 0.25, and formation reserve is 9.9843 × 10 7 m 3 .There are fifty four wells in this block.The simulation works about 22,402 days and is considered in three periods.It works in November, 1966 and water is injected in March, 1967.The polymer, whose concentration is 1500 ppm, is injected from June, 2008 to April, 2015.Water is injected again from April, 2015 to January, 2030.The simulation results are illustrated in Table1.The comparison of moisture content of produced oil of SLCHEM, VIP, and actual results are shown in Figure2.The time cost of SLCHEM is about 0.44 hour, and the material balance error is satisfactory.The relative total error comparison with actual moisture content is about 7.8% before the polymer flooding is injected.All the results show that the computation of SLCHEM runs fast, the numerical results are reliable and this scheme can be applied into present oilfields production.2) Experiments of surfactant-polymer flooding agents in middle-scale oilfieldsThe rectangle computational domain (Sheng Block 2) is partitioned into 82 × 74 × 7 subdomains with uniform steps and the spacial steps in x -direction and y -direction are taken as 33.60 m and 29.29 m, respectively.Formation reserve is 3.4938 × 10 8 m 3 .There are sixty three wells in this block.The simulation works about

Figure 2 . 3 .
Figure 2. Curve of moisture content simulation of produced oil of polymer flooding in smallscale oilfields.$2362days and is considered in three periods.It works onApril 1, 1966  and water is injected on January 1, 1974.The polymer, whose concentration is 1600 ppm, is injected from August, 2010 to September, 2016.Surfactant, whose concentration is 0.4%, is injected from August, 2011 to October, 2011.The largest time step is less than five days and the simulation results are illustrated in Table2.The comparison of moisture content of produced oil of SLCHEM, ECLIPSE, and actual results are shown in Figure3.The time cost of SLCHEM is about 4.2 hours, and the material balance error is satisfactory.The relative total error comparison with actual moisture content is about 7.10%.All the results show that the computation of SLCHEM runs fast, the numerical results are reliable and this scheme can be applied into present oilfields production.3)Experiments of large-scale oilfieldsThe rectangle computational domain (Gudaoguan Block 3) is partitioned into 72 × 62 × 26 subdomains with uniform steps and the spacial steps in x -direction and y -direction are taken as 25 m.Formation reserve is 22,708,200.0m 3 .There are one hundred and ninety one wells in this block.The simulation works about 12,965 days and is considered in three periods.It works on September 1, 1971 and water is injected on September 1, 1974.The polymer, whose concentration is 1000 ppm, is injected from March 1, 1994 to November 30, 2003.Water is injected again from December 1, 2003 to March 1, 2007.The largest time step is less than ten days and the simulation results are illustrated in Table3.The comparison of moisture content of produced oil of SLCHEM and actual results are shown in Figure4.The time cost of SLCHEM is about 8.45 hours, and the material balance error is 10 −6 .The relative total error comparison with actual moisture content is about 7.0%.All the results show that the computation of SLCHEM runs fast, the numerical results are reliable and this scheme can be applied into present oilfields production.4) Numerical simulation of surfactant polymer flooding The polymer, whose concentration is 1000 ppm, is injected from March 1, 1994 to November 30, 2003.Surfactant, whose concentration is 0.5%, is injected from May 1, 1996 to December 1, 2003.Then water is injected again from December 1, 2003 to March 1, 2007.The largest time step is less than ten days and the simulation results are illustrated in Table4.Numerical results of SLCHEM of water flooding, polymer flooding and surfactant-polymer flooding are compared in Figure5.The time cost of SLCHEM is about 13.5 hours, and the material balance error is satis-

4 )
Numerical simulation of surfactant polymer flooding The polymer, whose concentration is 1000 ppm, is injected from March 1, 1994 to November 30, 2003.Surfactant, whose concentration is 0.5%, is injected from May 1, 1996 to December 1, 2003.Then water is injected again from December 1, 2003 to March 1, 2007.The largest time step is less than ten days and the simulation results are illustrated in Table4.Numerical results of SLCHEM of water flooding, polymer flooding and surfactant-polymer flooding are compared in Figure5.The time cost of SLCHEM is about 13.5 hours, and the material balance error is satis-

Figure 5 .
Figure 5. Curve of moisture content simulation of produced oil of surface active agent and combination flooding.factory.The software has three flooding functions and physical-chemical parameters can be processed effectively.The software simulates reliably and generally based on the above experiments of different scales oilfields and the simulation scale is up to hundreds of thousands nodes.

+
obtained by (34b), and the solution of pressure { } 1 n ijk P + is solved by (34c).The values of Darcy velocity { } 1 n ijk + U are computed by (35).Secondly, the values { }is computed by using speedup method in x 1 -direction by (36a), { } is obtained by (36b), and the solution )

Table 1 .
Numerical results of polymer flooding in small-scale oilfield.