A method and system for predicting equivalent parameters of fluid-solid coupling in fracture-cavity media

Abstract

The present invention discloses a method and system for predicting equivalent parameters of fluid-solid coupling of fracture-cavity medium, and relates to the field of numerical simulation of oil reservoirs. The method comprises: inputting basic information parameters into an equivalent parameter prediction model to obtain predicted values ​​of equivalent parameters of the fluid-solid coupling model of the target fracture-cavity medium; the equivalent parameters include effective permeability, effective Biot coefficient, effective elastic coefficient and effective storage coefficient; the predicted values ​​of the equivalent parameters are used to construct the fluid-solid coupling model of the target fracture-cavity medium; the fluid-solid coupling model of the target fracture-cavity medium is used to perform numerical simulation of the fracture-cavity medium oil reservoir; wherein, the method for determining the equivalent parameter prediction model is: training a neural network model with equivalent parameters in the macro fluid fluid-solid coupling model determined by homogenization theory. The present invention can improve the calculation efficiency of equivalent parameters, thereby realizing efficient numerical simulation of the fracture-cavity medium oil reservoir.

Description

Technical Field

[0001] The invention relates to the field of reservoir numerical simulation, and in particular to a method and system for predicting equivalent parameters of fluid-solid coupling of fracture-cavity medium. Background Art

[0002] Fracture-cavity media are common in carbonate rocks. The presence of fractures and cavities has an important influence on the hydraulic behavior of rocks. How to effectively quantify and analyze this effect remains a challenging problem in earth science. The fluid-solid coupling model of fracture-cavity media is used to numerically simulate fracture-cavity media reservoirs. The numerical simulation results are of great significance for the exploration and development of fracture-cavity media reservoirs. The calculation method of equivalent parameters in the existing fracture-cavity media fluid-solid coupling model is usually: only through theoretical calculation, the corresponding mathematical model is obtained, and then the effective parameters are further calculated. However, due to the limitation of computing power, the calculation is still very cumbersome, the calculation efficiency of equivalent parameters is very low, and the reuse rate is very low. Summary of the invention

[0003] Based on this, an embodiment of the present invention provides a method and system for predicting equivalent parameters of fluid-solid coupling in fracture-cavity media, so as to improve the calculation efficiency of equivalent parameters, thereby realizing efficient numerical simulation of oil reservoirs in fracture-cavity media.

[0004] To achieve the above object, the present invention provides the following solutions:

[0005] A method for predicting equivalent parameters of fluid-solid coupling in fracture-cavity media, comprising:

[0006] Obtaining basic parameter information of the target fracture-cavity medium; the basic parameter information includes Young's modulus, Poisson's ratio, porosity, Biot coefficient and storage coefficient;

[0007] Inputting the basic information parameters into an equivalent parameter prediction model, obtaining predicted values ​​of equivalent parameters of the fluid-solid coupling model of the target fracture-cavity medium; the equivalent parameters include effective permeability, effective Biot coefficient, effective elastic coefficient and effective storage coefficient; the predicted values ​​of the equivalent parameters are used to construct the fluid-solid coupling model of the target fracture-cavity medium; the fluid-solid coupling model of the target fracture-cavity medium is used to perform numerical simulation on the fracture-cavity medium reservoir;

[0008] Wherein, the method for determining the equivalent parameter prediction model is:

[0009] The fluid-solid coupling model of the training fracture-hole medium is constructed at the microscopic scale by using the Biot equation of the poroelastic zone in the training fracture-hole medium, the Stokes equation of the free fluid zone in the training fracture-hole medium and the extended boundary conditions on the porous-fluid pseudo interface.

[0010] The macro fluid-solid coupling model is obtained by performing a size analysis on the fluid-solid coupling model of the training fracture-cavity medium through homogenization theory;

[0011] The equivalent parameters in the macro fluid-solid coupling model are used as inputs of a neural network model for training, and the trained neural network model is determined as the equivalent parameter prediction model.

[0012] Optionally, the using the equivalent parameters in the macro fluid-solid coupling model as input of a neural network model for training, and determining the trained neural network model as the equivalent parameter prediction model specifically includes:

[0013] The equivalent parameters in the macro fluid-solid coupling model are divided into training data, test data and verification data according to a set ratio;

[0014] Using the training data as input of the neural network model, and training with mean absolute percentage error as the loss function to obtain a preliminarily trained neural network model;

[0015] Using the test data to test the preliminarily trained neural network model to obtain a tested neural network model;

[0016] The tested neural network model is verified using the verification data to obtain a verified neural network model, and the verified neural network model is determined as a trained neural network model.

[0017] Optionally, the step of performing a dimensional analysis on the training fluid-solid coupling model of the fracture-hole medium by homogenization theory to obtain a macro fluid fluid-solid coupling model specifically includes:

[0018] Performing a dual-scale asymptotic analysis on the fluid-solid coupling model of the training fracture-cavity medium, obtaining the Biot equation in the poroelastic zone after the asymptotic analysis, the Stokes equation after the asymptotic analysis, and the extended boundary conditions on the porous-fluid pseudo-interface after the asymptotic analysis;

[0019] A mechanical property analysis is performed based on the Biot equation in the poroelastic zone after the asymptotic analysis, the Stokes equation after the asymptotic analysis, and the extended boundary conditions on the porous-fluid pseudo-interface after the asymptotic analysis to obtain a macroscopic Navier equation of deformable fractured-cavity rock;

[0020] Perform fluid flow characteristic analysis based on the Biot equation in the poroelastic zone after the asymptotic analysis and the Stokes equation after the asymptotic analysis to obtain a macroscopic Darcy equation;

[0021] Perform fluid mechanics coupling equation analysis based on the Biot equation in the poroelastic zone after the asymptotic analysis and the Stokes equation after the asymptotic analysis to obtain a fluid mechanics equation;

[0022] A macroscopic fluid-solid coupling model is constructed according to the macroscopic Navier equation of deformable fracture-cavity rock, the macroscopic Darcy equation and the fluid mechanics equation.

[0023] Optionally, the step of using the training data as an input of a neural network model and training with mean absolute percentage error as a loss function to obtain a preliminarily trained neural network model specifically includes:

[0024] Constructing a fully connected neural network model; the fully connected neural network model includes an input layer, four hidden layers and an output layer connected in sequence;

[0025] The training data is used as the input of the fully connected neural network model, and the mean absolute percentage error is used as the loss function for training to obtain a preliminarily trained neural network model.

[0026] Optionally, the step of using the training data as input of the fully connected neural network model and training with mean absolute percentage error as a loss function to obtain a preliminarily trained neural network model specifically includes:

[0027] The Relu function is used as the activation function of the hidden layer in the fully connected neural network model, the Sigmoid function is used as the activation function of the output layer in the fully connected neural network model, and the training data is used as the input of the fully connected neural network model. The random steepest descent method is adopted, and the mean absolute percentage error is used as the loss function for training to obtain a preliminarily trained neural network model.

[0028] The present invention also provides a fracture-hole medium fluid-solid coupling equivalent parameter prediction system, comprising:

[0029] A basic parameter information acquisition module is used to acquire basic parameter information of the target fracture-cavity medium; the basic parameter information includes Young's modulus, Poisson's ratio, porosity, Biot coefficient and storage coefficient;

[0030] An equivalent parameter prediction module is used to input the basic information parameters into an equivalent parameter prediction model to obtain the predicted values ​​of the equivalent parameters of the fluid-solid coupling model of the target fracture-cavity medium; the equivalent parameters include effective permeability, effective Biot coefficient, effective elastic coefficient and effective storage coefficient; the predicted values ​​of the equivalent parameters are used to construct the fluid-solid coupling model of the target fracture-cavity medium; the fluid-solid coupling model of the target fracture-cavity medium is used to perform numerical simulation on the fracture-cavity medium reservoir;

[0031] An equivalent parameter prediction model determination module is used to determine the equivalent parameter prediction model; the equivalent parameter prediction model determination module specifically includes:

[0032] The training model construction submodule is used to construct a fluid-solid coupling model of the training fracture-hole medium at a microscopic scale by using the Biot equation of the poroelastic zone in the training fracture-hole medium, the Stokes equation of the free fluid zone in the training fracture-hole medium, and the extended boundary conditions on the porous-fluid pseudo interface;

[0033] A macro model construction submodule is used to perform a dimensional analysis on the fluid-solid coupling model of the training fracture-hole medium through homogenization theory to obtain a macro fluid fluid-solid coupling model;

[0034] The model training submodule is used to train the equivalent parameters in the macro fluid-solid coupling model as input of the neural network model, and determine the trained neural network model as the equivalent parameter prediction model.

[0035] Optionally, the model training submodule specifically includes:

[0036] A data division unit, used for dividing the equivalent parameters in the macro fluid-solid coupling model into training data, test data and verification data according to a set ratio;

[0037] A preliminary training unit, used to use the training data as input of the neural network model and perform training using mean absolute percentage error as a loss function to obtain a preliminary trained neural network model;

[0038] A model testing unit, used to test the preliminarily trained neural network model using the test data to obtain a tested neural network model;

[0039] The model verification unit is used to verify the tested neural network model using the verification data to obtain a verified neural network model, and determine the verified neural network model as a trained neural network model.

[0040] Optionally, the macro model construction submodule specifically includes:

[0041] A dual-scale progressive analysis unit is used to perform dual-scale progressive analysis on the fluid-solid coupling model of the training fracture-cavity medium to obtain the Biot equation in the poroelastic zone after the progressive analysis, the Stokes equation after the progressive analysis, and the extended boundary conditions on the porous-fluid pseudo interface after the progressive analysis;

[0042] A macroscopic Navier equation construction unit is used to perform mechanical property analysis according to the Biot equation in the poroelastic zone after the asymptotic analysis, the Stokes equation after the asymptotic analysis and the extended boundary conditions on the porous-fluid pseudo interface after the asymptotic analysis, so as to obtain the macroscopic Navier equation of deformable fracture-cavity rock;

[0043] A macroscopic Darcy equation construction unit is used to perform fluid flow characteristic analysis based on the Biot equation in the poroelastic zone after the asymptotic analysis and the Stokes equation after the asymptotic analysis to obtain the macroscopic Darcy equation;

[0044] A fluid mechanics equation construction unit is used to perform fluid mechanics coupling equation analysis based on the Biot equation in the poroelastic zone after the asymptotic analysis and the Stokes equation after the asymptotic analysis to obtain a fluid mechanics equation;

[0045] A macroscopic fluid-solid coupling model is constructed according to the macroscopic Navier equation of deformable fracture-cavity rock, the macroscopic Darcy equation and the fluid mechanics equation.

[0046] Optionally, the preliminary training unit specifically includes:

[0047] A model building subunit, used to build a fully connected neural network model; the fully connected neural network model includes an input layer, four hidden layers and an output layer connected in sequence;

[0048] The model training subunit is used to use the training data as the input of the fully connected neural network model and perform training with the mean absolute percentage error as the loss function to obtain a preliminarily trained neural network model.

[0049] Optionally, the model training subunit specifically includes:

[0050] The training subunit is used to use the Relu function as the activation function of the hidden layer in the fully connected neural network model, use the Sigmoid function as the activation function of the output layer in the fully connected neural network model, and use the training data as the input of the fully connected neural network model, adopt the random steepest descent method, and use the mean absolute percentage error as the loss function for training to obtain a preliminarily trained neural network model.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] The embodiment of the present invention proposes a method and system for predicting equivalent parameters of fluid-solid coupling of fracture-cavity media, inputting basic information parameters into an equivalent parameter prediction model to obtain predicted values ​​of equivalent parameters of the fluid-solid coupling model of the target fracture-cavity media; the equivalent parameters include effective permeability, effective Biot coefficient, effective elastic coefficient and effective storage coefficient; wherein, the method for determining the equivalent parameter prediction model is: training a neural network model with equivalent parameters in the macro fluid fluid-solid coupling model determined by homogenization theory. The present invention combines theoretical calculation with neural network prediction methods, reduces computational complexity, and can improve the computational efficiency and reuse rate of equivalent parameters, thereby realizing efficient numerical simulation of fracture-cavity media reservoirs, and further improving the efficiency of exploration and development of fracture-cavity media reservoirs. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0054] Figure 1 A flow chart of a method for predicting equivalent parameters of fluid-solid coupling in a crack-hole medium provided in an embodiment of the present invention;

[0055] Figure 2 A schematic diagram of a slit-hole type medium provided in an embodiment of the present invention;

[0056] Figure 3 A schematic diagram of a specific periodic slit-hole medium provided by an embodiment of the present invention;

[0057] Figure 4 A structural diagram of a fluid-solid coupling equivalent parameter prediction system for fracture-hole type media provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0060] Under the combined action of geological tectonic movement and karstification, a large number of macroscopic cracks and caves are often generated in the strata. Their spatial scale is much larger than the intergranular pores, which has an important impact on the fluid flow in the strata. Such media are widely present in nature and are closely related to human life. However, due to the multi-scale nature of cracks and caves in fracture-cavity media in space, there are many problems in experimental measurement, and the measurement results often cannot reflect the true condition of the strata. Therefore, theoretical calculation has gradually received attention and has become a powerful means to predict the equivalent parameters of fracture-cavity complex media. Therefore, it is necessary to combine theoretical calculation with neural network prediction methods to improve the prediction accuracy of equivalent parameters of fracture-cavity media.

[0061] The present invention upgrades the theory through the homogenization theory, upgrades the fluid-solid coupling model established by the Biot equation, the Stokes equation and the extended boundary condition (BJS boundary condition) on the porous-fluid pseudo interface to a macro equivalent model, and calculates the equivalent parameters. Then, the equivalent parameters are trained by using a neural network, and finally the effective prediction of the equivalent parameters of the fluid-solid coupling model of the fracture-hole type medium is realized.

[0062] Figure 1 This is a flow chart of the method for predicting equivalent parameters of fluid-solid coupling in crack-hole medium provided by an embodiment of the present invention. Figure 1 The method for predicting equivalent parameters of fluid-solid coupling of crack-hole medium of this embodiment includes:

[0063] Step 101: Obtain basic parameter information of a target fracture-cavity medium; the basic parameter information includes Young's modulus, Poisson's ratio, porosity, Biot coefficient and storage coefficient.

[0064] Step 102: Input the basic information parameters into an equivalent parameter prediction model to obtain predicted values ​​of equivalent parameters of the fluid-solid coupling model of the target fracture-cavity medium; the equivalent parameters include effective permeability, effective Biot coefficient, effective elastic coefficient and effective storage coefficient; the predicted values ​​of the equivalent parameters are used to construct the fluid-solid coupling model of the target fracture-cavity medium; the fluid-solid coupling model of the target fracture-cavity medium is used to perform numerical simulation on the oil reservoir of the fracture-cavity medium.

[0065] Wherein, in step 102, the method for determining the equivalent parameter prediction model is:

[0066] 1) The fluid-solid coupling model (microscale fluid-solid coupling model) of the training fracture-hole medium is constructed at the microscale by using the Biot equation of the poroelastic zone in the training fracture-hole medium, the Stokes equation of the free fluid zone in the training fracture-hole medium and the extended boundary conditions on the porous-fluid pseudo interface.

[0067] The flow of the fluid is considered to be isothermal, single-phase, incompressible, and the viscosity of the fluid is constant. Under small deformation conditions, the rock is poroelastic.

[0068] Slit-type media such as Figure 2 As shown in the figure, Ω is a bounded fractured rock mass, i.e., a fracture-cavity medium. Ω is the poroelastic zone in the non-overlapping region. p and free fluid area Ω v The union of Γ pv Ω p and Ω v The interface between these two regions, Ω f is the fracture area, Ω m is the matrix area, and Vug is the cave area.

[0069] Ω p Biot equation and Ω in v The Stokes equations in , and Γ pv The corresponding interface conditions are as follows.

[0070] ① Biot equation in the poroelastic zone:

[0071]

[0072] in, is the divergence symbol, σ p and σ s denote the total stress tensor of the porous medium and the stress tensor of the solid skeleton, respectively. The density of the saturated porous medium is ρ p =ρ f +(1-φ)ρ s , where φ is the porosity of the porous medium, ρ f and ρ s are the densities of the fluid skeleton and the solid skeleton respectively. p is the matrix pore pressure; α is the Biot coefficient, which is a second-order tensor; a is the fourth-order elastic tensor; e(u s ) is u s The strain tensor of s is the displacement of the solid skeleton; for u s Differentiate with respect to time. v p is the Darcy velocity of the fluid in the porous region relative to the solid skeleton; w p is the average velocity of the fluid in the porous zone; k is the permeability tensor of the porous medium; μ is the fluid viscosity; g is the gravitational acceleration; γ is the storage coefficient; Indicates p p Differentiation of time; is the gradient of matrix pore pressure.

[0073]

[0074] Represents the gradient of the solid skeleton displacement.

[0075] ② Stokes equation in free fluid zone:

[0076]

[0077] Among them, σ f is the fluid stress tensor, p f is the fluid pressure, v f is the free fluid velocity, I represents the unit tensor, e(v f ) is v f The strain tensor of .

[0078]

[0079] represents the free fluid velocity gradient.

[0080] ③Γ pv Interface conditions on:

[0081]

[0082] where n is the unit normal vector outward from Ωv to Ωp; τ is the unit tangent vector along the interface. The first line in equation (5) is a mass balance, which requires that the normal fluid flux must be continuous across the interface. The second line represents momentum conservation, which requires that the stress in the porous matrix be balanced with the stress in the fluid. The third line is a balance of the normal component of the stress in the fluid phase across the interface. The last line simulates the BJS condition with frictional slip, α BJ is the BJ slip coefficient, Γ pf Ω p and Ω f The interface between these two areas.

[0083] Now, consider the homogenization process of the above Biot-Stokes system. An important feature of this method is that there are two different length scales: the mesoscale l, which is typically the size of the representative unit volume (REV); and the macroscale L, which characterizes the global changes in external forces and boundary data. Let ε = l / L and ε < < 1.

[0084] Consider a crack-shaped matrix Ω with a characteristic length of L, such as Figure 3 As shown, the slit-hole matrix Ω has a microstructure with a period of Y, where Y = Y p ∪Y f , where Y p Ω p The cycle, Yf Ω f The period of the periodic microscale cell is l, and the ratio of different characteristic lengths ε=l / L is introduced.

[0085] According to the conventions of homogenization theory, it is assumed that any point can be described by two coordinates: x describes the global position of the point, and y describes the position of the point in the periodic unit. Therefore, x and y are determined by the scaling constant ε: y ~ ε -1 x. According to the chain rule, the following relationship holds:

[0086]

[0087] in and denote the gradient operators with respect to x and y respectively. Due to the separation of scales, each physical quantity All of them are functions of these two spatial variables. The variables in the fluid-solid coupling model, stress, pressure, displacement and velocity, are expanded according to the following asymptotic expansion formula (7):

[0088]

[0089] in, for The first-level expansion result is: for The secondary expansion result is: for The three-level expansion result.

[0090] Next, based on the dual-scale asymptotic homogenization theory, the scale problems in different regions can be obtained.

[0091] 2) The fluid-solid coupling model of the training fracture-hole medium is analyzed in size by homogenization theory to obtain a macroscopic fluid-solid coupling model.

[0092] 3) Using the equivalent parameters in the macro fluid-solid coupling model as input of a neural network model for training, and determining the trained neural network model as the equivalent parameter prediction model.

[0093] Wherein, step 2) specifically includes:

[0094] 21) Perform dual-scale asymptotic analysis on the fluid-solid coupling model of the training fracture-cavity medium to obtain the Biot equation in the poroelastic zone after asymptotic analysis, the Stokes equation after asymptotic analysis, and the extended boundary conditions on the porous-fluid pseudo interface after asymptotic analysis. The specific process of dual-scale asymptotic analysis is:

[0095] Substitute the expanded equation into equations (1), (3), and (5), and transform using equation (6) to combine ε -1 and ε 0 The power term can be obtained:

[0096] Biot equation for elastic pore region:

[0097]

[0098] σ m represents the stress tensor of the matrix; v m is the Darcy velocity of the matrix relative to the solid skeleton; the superscript (0) indicates the first term on the right side of the equal sign after expansion according to formula (7), such as, is m After expanding according to formula (7), the first term on the right side of the equal sign, i.e., σ m The first-level expansion result of ; the superscript (1) indicates the second term on the right side of the equal sign after expansion according to formula (7); Y m represents one period within the elastic pore region.

[0099] Stokes equations in caves:

[0100]

[0101] Interface conditions:

[0102]

[0103] Y fm represents a period of the interface between the crack region and the matrix region. From the second and fifth equations of (8), the first equation of (9), and the third equation of (10), we can obtain:

[0104] From this we can see that and It only changes on the macroscopic scale x and has nothing to do with the microscopic scale y.

[0105] 22) According to the Biot equation of the poroelastic zone after the asymptotic analysis, the Stokes equation after the asymptotic analysis and the extended boundary conditions on the porous-fluid pseudo interface after the asymptotic analysis, a mechanical performance analysis is performed to obtain the macroscopic Navier equation of deformable fractured rock. The specific process of mechanical performance analysis is as follows:

[0106] Using the first three equations of (8), the second equation of (10), and The following questions can be obtained

[0107]

[0108] Represents the interface, where

[0109]

[0110] is a third-order tensor, ζ(y) is a vector. Substituting (12) into (11), we can get the following two local primitive problems, defining ξ kh (y) is the solution of (13).

[0111]

[0112] in δ ij is the Kronecker symbol. Similarly, the vector ζ(y) satisfies:

[0113]

[0114] Then, combining the second and third equations of equation (9) and equation (10), we get

[0115]

[0116] Define the total pressure as:

[0117]

[0118] Define the average operator:

[0119]

[0120] Using (16), the total stress on Y is On average, using the first equation (9), and (16), we get:

[0121]

[0122] in:

[0123]

[0124] Then integrate the seventh equation of (8) with respect to Yp, and use the divergence theorem and periodicity conditions to obtain:

[0125]

[0126] Similarly, integrating the second equation of (9) yields:

[0127]

[0128] Adding equations (19) and (20) together we get:

[0129]

[0130] Then, divided by the volume |Y|, (20) becomes:

[0131] Formula (21) is the macroscopic Navier equation of deformable fractured rock.

[0132] 23) Perform fluid flow characteristic analysis based on the Biot equation in the poroelastic zone after the asymptotic analysis and the Stokes equation after the asymptotic analysis to obtain the macroscopic Darcy equation. The specific process of fluid flow characteristic analysis is:

[0133] Combining the 6th and 8th equations of (9) and the 2nd, 3rd and 4th equations of (9), we can obtain:

[0134]

[0135] Combination and (22) can be converted to:

[0136]

[0137] in, Indicates the true flow rate,

[0138]

[0139] The first and third equations of formula (23) can be written as follows:

[0140]

[0141] Among them, e i is the standard Cartesian basis vector in the ith direction, let and is the periodic solution of the following Darcy-Stokes problem in the basic unit cell, p is the poroelastic region, v is the cave region

[0142]

[0143] Applying the average operator to combine the second equation of (8) and (24) yields

[0144]

[0145] In the formula, is the velocity of fluid passing through the total pores in fractured porous rocks; φ T is the total porosity of fractured cave rock, including matrix, fractures and caves. The macro equivalent permeability k of fractured cave rock can be defined as eff as follows:

[0146]

[0147] The subscripts pm and pf represent porous matrix and porous fractures, respectively. f is the crack aperture.

[0148] Combining (27) and (28), we can obtain the macroscopic Darcy equation (29):

[0149]

[0150] 24) Perform fluid mechanics coupling equation analysis based on the Biot equation of the poroelastic zone after the asymptotic analysis and the Stokes equation after the asymptotic analysis to obtain the fluid mechanics equation. The specific process of the fluid mechanics coupling equation analysis is:

[0151] Combining the 9th equation of (8) and the 5th equation of (9),

[0152]

[0153] According to the definition of Darcy's total velocity in equation (27) and equation (24), we get equation (31):

[0154]

[0155] Integrating the above equation over Y and applying the average operator yields:

[0156]

[0157] Then, we further apply the divergence theorem and the periodic condition of the second term on the left, and use equation (14) to obtain:

[0158]

[0159] 25) A macroscopic fluid-solid coupling model is constructed based on the macroscopic Navier equation of deformable fracture-cavity rock, the macroscopic Darcy equation and the fluid mechanics equation. The construction process of the macroscopic fluid-solid coupling model is as follows:

[0160] The macroscopic equations of deformable fractured rock obtained from (14), (21), (29), and (33) in the previous sections can be summarized as:

[0161]

[0162] Solving (34), we can obtain the corresponding equivalent parameter k eff , a eff , α eff , γ eff , k effrepresents the effective permeability; a eff Represents the effective Biot coefficient; α eff represents the effective elastic coefficient; γ eff Represents the effective storage coefficient.

[0163] Wherein, step 3) specifically includes:

[0164] 31) The equivalent parameters in the macro fluid-solid coupling model are divided into training data, test data and verification data according to a set ratio.

[0165] 32) Using the training data as the input of the neural network model, and training with the mean absolute percentage error as the loss function, to obtain a preliminarily trained neural network model. Specifically: constructing a fully connected neural network model; the fully connected neural network model includes an input layer, four hidden layers and an output layer connected in sequence; using the training data as the input of the fully connected neural network model, and training with the mean absolute percentage error as the loss function, to obtain a preliminarily trained neural network model.

[0166] Among them, using the training data as the input of the fully connected neural network model, and training with the mean absolute percentage error as the loss function to obtain a preliminarily trained neural network model, specifically includes: using the Relu function as the activation function of the hidden layer in the fully connected neural network model, using the Sigmoid function as the activation function of the output layer in the fully connected neural network model, and using the training data as the input of the fully connected neural network model, using the stochastic steepest descent method, and training with the mean absolute percentage error as the loss function to obtain a preliminarily trained neural network model.

[0167] In practical applications, the training data must be standardized before entering the neural network model. The formula is:

[0168]

[0169] Eighty percent of the data is used as training data, ten percent of the data is used as test data, and ten percent of the data is used as validation data. The hidden layer is activated by the Relu function, and the output layer is activated by the Sigmoid function. The constructed neural network is trained by the SGD method. The mean absolute percentage error (MAPE) is used as the loss function in the training to calculate the difference between the model prediction value and the true value as shown in formula (29). The input layer is the basic characteristic information of the formation and Young's modulus, Poisson's ratio, porosity, Biot coefficient, storage coefficient, etc. The output layer result is the numerical value k of the equivalent parameter of the fracture-cavity medium fluid-solid coupling model eff (effective permeability), a eff (effective Biot coefficient), αeff (effective elastic coefficient), γ eff (Storage coefficient).

[0170]

[0171] 33) Using the test data to test the preliminarily trained neural network model to obtain a tested neural network model.

[0172] 34) Using the verification data to verify the tested neural network model, a verified neural network model is obtained, and the verified neural network model is determined as a trained neural network model. The trained neural network model is saved for subsequent prediction of equivalent parameters of the fracture-hole medium.

[0173] The method for predicting equivalent parameters of fluid-solid coupling in crack-hole medium of this embodiment has the following advantages:

[0174] This embodiment establishes a macro-equivalent fracture-cavity medium fluid-solid coupling model through the theory of homogenization scale upgrade, and on this basis, obtains macro-effective elastic coefficients such as effective permeability, effective Biot coefficient, effective elastic coefficient and effective storage coefficient, and forms a model for predicting fracture-cavity medium fluid-solid coupling equivalent parameters through learning and processing of a large amount of obtained data by neural network. Through this invention, the equivalent parameters of the fracture-cavity medium fluid-solid coupling model can be effectively predicted, and it is more accurate and faster than traditional methods. The trained model can be reused, thereby realizing efficient numerical simulation of fracture-cavity medium reservoirs, and thus improving the efficiency of exploration and development of fracture-cavity medium reservoirs.

[0175] The fluid-solid coupling model of fracture-cavity medium is mainly used to simulate the formation of fracture-cavity reservoirs and calculate the fluid coefficient. It can be applied to fracture-cavity reservoirs such as Tarim Oilfield.

[0176] The present invention also provides a fracture-hole medium fluid-solid coupling equivalent parameter prediction system. Figure 4 This is a structural diagram of the equivalent parameter prediction system for fluid-solid coupling of crack-hole medium provided by an embodiment of the present invention. Figure 4 , the system comprising:

[0177] The basic parameter information acquisition module 201 is used to acquire the basic parameter information of the target fracture-hole type medium; the basic parameter information includes Young's modulus, Poisson's ratio, porosity, Biot coefficient and storage coefficient.

[0178] The equivalent parameter prediction model determination module 202 is used to determine the equivalent parameter prediction model.

[0179] The equivalent parameter prediction module 203 is used to input the basic information parameters into the equivalent parameter prediction model to obtain the predicted values ​​of the equivalent parameters of the fluid-solid coupling model of the target fracture-cavity medium; the equivalent parameters include effective permeability, effective Biot coefficient, effective elastic coefficient and effective storage coefficient; the predicted values ​​of the equivalent parameters are used to construct the fluid-solid coupling model of the target fracture-cavity medium; the fluid-solid coupling model of the target fracture-cavity medium is used to perform numerical simulation on the oil reservoir of the fracture-cavity medium.

[0180] The equivalent parameter prediction model determination module 202 specifically includes:

[0181] The training model construction submodule is used to construct a fluid-solid coupling model of the training fracture-hole medium at the microscopic scale by using the Biot equation of the poroelastic zone in the training fracture-hole medium, the Stokes equation of the free fluid zone in the training fracture-hole medium, and the extended boundary conditions on the porous-fluid pseudo interface.

[0182] The macro model construction submodule is used to perform dimensional analysis on the fluid-solid coupling model of the training fracture-hole medium through homogenization theory to obtain a macro fluid fluid-solid coupling model.

[0183] The model training submodule is used to train the equivalent parameters in the macro fluid-solid coupling model as input of the neural network model, and determine the trained neural network model as the equivalent parameter prediction model.

[0184] In one example, the model training submodule specifically includes:

[0185] The data division unit is used to divide the equivalent parameters in the macro fluid-solid coupling model into training data, test data and verification data according to a set ratio.

[0186] The preliminary training unit is used to use the training data as the input of the neural network model and to perform training with the mean absolute percentage error as the loss function to obtain a preliminary trained neural network model.

[0187] The model testing unit is used to test the preliminarily trained neural network model using the test data to obtain a tested neural network model.

[0188] The model verification unit is used to verify the tested neural network model using the verification data to obtain a verified neural network model, and determine the verified neural network model as a trained neural network model.

[0189] In one example, the macro model construction submodule specifically includes:

[0190] The dual-scale progressive analysis unit is used to perform dual-scale progressive analysis on the fluid-solid coupling model of the training fracture-cavity medium to obtain the Biot equation in the pore elastic zone after the progressive analysis, the Stokes equation after the progressive analysis, and the extended boundary conditions on the porous-fluid pseudo-interface after the progressive analysis.

[0191] The macroscopic Navier equation construction unit is used to perform mechanical property analysis based on the Biot equation in the pore elastic zone after the asymptotic analysis, the Stokes equation after the asymptotic analysis and the extended boundary conditions on the porous-fluid pseudo-interface after the asymptotic analysis, so as to obtain the macroscopic Navier equation of deformable fractured rock.

[0192] The macroscopic Darcy equation construction unit is used to perform fluid flow characteristic analysis according to the Biot equation of the poroelastic zone after the asymptotic analysis and the Stokes equation after the asymptotic analysis to obtain the macroscopic Darcy equation.

[0193] The fluid mechanics equation construction unit is used to perform fluid mechanics coupling equation analysis based on the Biot equation of the pore elastic zone after the asymptotic analysis and the Stokes equation after the asymptotic analysis to obtain the fluid mechanics equation.

[0194] A macroscopic fluid-solid coupling model is constructed according to the macroscopic Navier equation of deformable fracture-cavity rock, the macroscopic Darcy equation and the fluid mechanics equation.

[0195] In one example, the preliminary training unit specifically includes:

[0196] The model building subunit is used to build a fully connected neural network model; the fully connected neural network model includes an input layer, four hidden layers and an output layer connected in sequence.

[0197] The model training subunit is used to use the training data as the input of the fully connected neural network model and perform training with the mean absolute percentage error as the loss function to obtain a preliminarily trained neural network model.

[0198] In one example, the model training subunit specifically includes:

[0199] The training subunit is used to use the Relu function as the activation function of the hidden layer in the fully connected neural network model, use the Sigmoid function as the activation function of the output layer in the fully connected neural network model, and use the training data as the input of the fully connected neural network model, adopt the random steepest descent method, and use the mean absolute percentage error as the loss function for training to obtain a preliminarily trained neural network model.

[0200] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0201] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for predicting equivalent parameters of fluid-solid coupling in fracture-cavity media, characterized in that: include: Obtaining basic parameter information of the target fracture-cavity medium; the basic parameter information includes Young's modulus, Poisson's ratio, porosity, Biot coefficient and storage coefficient; Inputting the basic information parameters into an equivalent parameter prediction model, obtaining predicted values ​​of equivalent parameters of the fluid-solid coupling model of the target fracture-cavity medium; the equivalent parameters include effective permeability, effective Biot coefficient, effective elastic coefficient and effective storage coefficient; the predicted values ​​of the equivalent parameters are used to construct the fluid-solid coupling model of the target fracture-cavity medium; the fluid-solid coupling model of the target fracture-cavity medium is used to perform numerical simulation on the fracture-cavity medium reservoir; Wherein, the method for determining the equivalent parameter prediction model is: The fluid-solid coupling model of the training fracture-hole medium is constructed at the microscopic scale by using the Biot equation of the poroelastic zone in the training fracture-hole medium, the Stokes equation of the free fluid zone in the training fracture-hole medium and the extended boundary conditions on the porous-fluid pseudo interface. The macro fluid-solid coupling model is obtained by performing a size analysis on the fluid-solid coupling model of the training fracture-cavity medium through homogenization theory; The equivalent parameters in the macro fluid-solid coupling model are used as inputs of a neural network model for training, and the trained neural network model is determined as the equivalent parameter prediction model.

2. The method for predicting equivalent parameters of fluid-solid coupling in fracture-cavity medium according to claim 1 is characterized in that: The method of using the equivalent parameters in the macro fluid-solid coupling model as input of a neural network model for training, and determining the trained neural network model as the equivalent parameter prediction model, specifically includes: The equivalent parameters in the macro fluid-solid coupling model are divided into training data, test data and verification data according to a set ratio; Using the training data as input of the neural network model, and training with mean absolute percentage error as the loss function to obtain a preliminarily trained neural network model; Using the test data to test the preliminarily trained neural network model to obtain a tested neural network model; The tested neural network model is verified using the verification data to obtain a verified neural network model, and the verified neural network model is determined as a trained neural network model.

3. The method for predicting equivalent parameters of fluid-solid coupling in fracture-cavity medium according to claim 1 is characterized in that: The macro fluid-solid coupling model is obtained by performing a dimensional analysis on the fluid-solid coupling model of the training fracture-hole medium through the homogenization theory, specifically comprising: Performing a dual-scale asymptotic analysis on the fluid-solid coupling model of the training fracture-cavity medium, obtaining the Biot equation in the poroelastic zone after the asymptotic analysis, the Stokes equation after the asymptotic analysis, and the extended boundary conditions on the porous-fluid pseudo-interface after the asymptotic analysis; A mechanical property analysis is performed based on the Biot equation in the poroelastic zone after the asymptotic analysis, the Stokes equation after the asymptotic analysis, and the extended boundary conditions on the porous-fluid pseudo-interface after the asymptotic analysis to obtain a macroscopic Navier equation of deformable fractured-cavity rock; Perform fluid flow characteristic analysis based on the Biot equation in the poroelastic zone after the asymptotic analysis and the Stokes equation after the asymptotic analysis to obtain a macroscopic Darcy equation; Perform fluid mechanics coupling equation analysis based on the Biot equation in the poroelastic zone after the asymptotic analysis and the Stokes equation after the asymptotic analysis to obtain a fluid mechanics equation; A macroscopic fluid-solid coupling model is constructed according to the macroscopic Navier equation of deformable fracture-cavity rock, the macroscopic Darcy equation and the fluid mechanics equation.

4. The method for predicting equivalent parameters of fluid-solid coupling in crack-hole medium according to claim 2 is characterized in that: The training data is used as the input of the neural network model, and the mean absolute percentage error is used as the loss function for training to obtain a preliminarily trained neural network model, specifically including: Constructing a fully connected neural network model; the fully connected neural network model includes an input layer, four hidden layers and an output layer connected in sequence; The training data is used as the input of the fully connected neural network model, and the mean absolute percentage error is used as the loss function for training to obtain a preliminarily trained neural network model.

5. The method for predicting equivalent parameters of fluid-solid coupling in fracture-cavity medium according to claim 4 is characterized in that: The method of using the training data as the input of the fully connected neural network model and training with the mean absolute percentage error as the loss function to obtain a preliminarily trained neural network model specifically includes: The Relu function is used as the activation function of the hidden layer in the fully connected neural network model, the Sigmoid function is used as the activation function of the output layer in the fully connected neural network model, and the training data is used as the input of the fully connected neural network model. The random steepest descent method is adopted, and the mean absolute percentage error is used as the loss function for training to obtain a preliminarily trained neural network model.

6. A fracture-cavity medium fluid-solid coupling equivalent parameter prediction system, characterized in that: include: A basic parameter information acquisition module is used to acquire basic parameter information of a target crack-hole medium; The basic parameter information includes Young's modulus, Poisson's ratio, porosity, Biot coefficient and storage coefficient; An equivalent parameter prediction module is used to input the basic information parameters into an equivalent parameter prediction model to obtain the predicted values ​​of the equivalent parameters of the fluid-solid coupling model of the target fracture-cavity medium; the equivalent parameters include effective permeability, effective Biot coefficient, effective elastic coefficient and effective storage coefficient; the predicted values ​​of the equivalent parameters are used to construct the fluid-solid coupling model of the target fracture-cavity medium; the fluid-solid coupling model of the target fracture-cavity medium is used to perform numerical simulation on the fracture-cavity medium reservoir; An equivalent parameter prediction model determination module is used to determine the equivalent parameter prediction model; the equivalent parameter prediction model determination module specifically includes: The training model construction submodule is used to construct a fluid-solid coupling model of the training fracture-hole medium at a microscopic scale by using the Biot equation of the poroelastic zone in the training fracture-hole medium, the Stokes equation of the free fluid zone in the training fracture-hole medium, and the extended boundary conditions on the porous-fluid pseudo interface; A macro model construction submodule is used to perform a dimensional analysis on the fluid-solid coupling model of the training fracture-hole medium through homogenization theory to obtain a macro fluid fluid-solid coupling model; The model training submodule is used to train the equivalent parameters in the macro fluid-solid coupling model as input of the neural network model, and determine the trained neural network model as the equivalent parameter prediction model.

7. The system for predicting equivalent parameters of fluid-solid coupling of fracture-cavity media according to claim 6, characterized in that: The model training submodule specifically includes: A data division unit, used for dividing the equivalent parameters in the macro fluid-solid coupling model into training data, test data and verification data according to a set ratio; A preliminary training unit, used to use the training data as input of the neural network model and perform training using mean absolute percentage error as a loss function to obtain a preliminary trained neural network model; A model testing unit, used to test the preliminarily trained neural network model using the test data to obtain a tested neural network model; The model verification unit is used to verify the tested neural network model using the verification data to obtain a verified neural network model, and determine the verified neural network model as a trained neural network model.

8. The system for predicting equivalent parameters of fluid-solid coupling of fracture-cavity media according to claim 6, characterized in that: The macro model construction submodule specifically includes: A dual-scale progressive analysis unit is used to perform dual-scale progressive analysis on the fluid-solid coupling model of the training fracture-cavity medium to obtain the Biot equation in the poroelastic zone after the progressive analysis, the Stokes equation after the progressive analysis, and the extended boundary conditions on the porous-fluid pseudo interface after the progressive analysis; A macroscopic Navier equation construction unit is used to perform mechanical property analysis according to the Biot equation in the poroelastic zone after the asymptotic analysis, the Stokes equation after the asymptotic analysis and the extended boundary conditions on the porous-fluid pseudo interface after the asymptotic analysis, so as to obtain the macroscopic Navier equation of deformable fracture-cavity rock; A macroscopic Darcy equation construction unit is used to perform fluid flow characteristic analysis based on the Biot equation in the poroelastic zone after the asymptotic analysis and the Stokes equation after the asymptotic analysis to obtain the macroscopic Darcy equation; A fluid mechanics equation construction unit is used to perform fluid mechanics coupling equation analysis based on the Biot equation in the poroelastic zone after the asymptotic analysis and the Stokes equation after the asymptotic analysis to obtain a fluid mechanics equation; A macroscopic fluid-solid coupling model is constructed according to the macroscopic Navier equation of deformable fracture-cavity rock, the macroscopic Darcy equation and the fluid mechanics equation.

9. The system for predicting equivalent parameters of fluid-solid coupling of fracture-cavity medium according to claim 7, characterized in that: The preliminary training unit specifically includes: A model building subunit, used to build a fully connected neural network model; the fully connected neural network model includes an input layer, four hidden layers and an output layer connected in sequence; The model training subunit is used to use the training data as the input of the fully connected neural network model and perform training with the mean absolute percentage error as the loss function to obtain a preliminarily trained neural network model.

10. The system for predicting equivalent parameters of fluid-solid coupling of fracture-cavity medium according to claim 9, characterized in that: The model training subunit specifically includes: The training subunit is used to use the Relu function as the activation function of the hidden layer in the fully connected neural network model, use the Sigmoid function as the activation function of the output layer in the fully connected neural network model, and use the training data as the input of the fully connected neural network model, adopt the random steepest descent method, and use the mean absolute percentage error as the loss function for training to obtain a preliminarily trained neural network model.

Images