Fracture-vuggy carbonate rock hydraulic fracture propagation prediction method and system based on mechanism data dual-drive model

Abstract

The invention relates to a fracture-vuggy carbonate rock hydraulic fracture propagation prediction method and system based on a mechanism data dual-drive model. The method comprises the following steps: constructing a fracture-vuggy carbonate rock reservoir model; constructing a hydraulic fracturing model of a natural fracture and hole distribution form in the fracture-vug type carbonate reservoir model; constructing a fracture-vug type carbonate reservoir geomechanical model; determining crack propagation of the hydrofracture model based on the fracture-vug type carbonate reservoir geomechanical model and an extended finite element method; on-site fracture-cavity type reservoir fracturing construction data are collected, numerical simulation is conducted on the hydraulic fracturing model, and a fracture expansion data set is obtained; constructing a crack propagation prediction model, and training the crack propagation prediction model by using the crack propagation data set to obtain a trained crack propagation prediction model; and using the trained crack propagation prediction model to predict and discriminate the crack propagation form.

Description

Technical Field

[0001] This invention relates to a method and system for predicting the propagation of hydraulic fractures in fractured carbonate rocks based on a dual-drive model of mechanistic data, belonging to the field of oil and gas field development engineering. Background Technology

[0002] Deep carbonate reservoirs contain abundant oil and gas resources and are currently a key type of oil and gas reservoir being developed globally. Of the world's proven oil and gas reserves, 60% are marine carbonate reservoirs. In China, carbonate rocks are mainly distributed in the Lower Paleozoic and Precambrian strata. Currently, large-scale carbonate oil and gas fields have been discovered in the Sichuan Basin, the Tarim Basin and Shunbei area of ​​Xinjiang, the Qaidam Basin of Qinghai, and the Ordos Basin of North China, representing major growth points for future reserve and production increases. Carbonate reservoirs have poor porosity and permeability, and due to their complex internal structure and strong heterogeneity, reservoir stimulation and production enhancement are extremely difficult. Currently, carbonate reservoir development mainly employs two modes: acid fracturing and proppant fracturing. Propant fracturing is more suitable for deep carbonate reservoirs with high closure stress. Because of the development of caves and fractures inside carbonate rocks, hydraulic fracturing or acid fracturing fractures can be affected by caves, fractures, and pores in the carbonate reservoir during their propagation, causing the fractures to deflect. This results in significant differences in the morphology of the fracture propagation system. Therefore, there is great uncertainty in predicting the propagation of fracture-cavity carbonate fractures, and the effect of fracturing the fracture-cavity body is not ideal. As a result, the design of fracturing and production enhancement for fracture-cavity carbonates is extremely difficult.

[0003] High-precision fracture propagation simulation methods have always been a research hotspot in oil and gas engineering. In recent years, scholars from China and abroad have conducted extensive theoretical research on fracture propagation problems in different types of reservoirs, establishing fracturing models that can characterize the geomechanical features of geological bodies and simulate fracture propagation. These models mainly include the finite element method, boundary element method, extended finite element method, and discrete element method. Simultaneously, many researchers have attempted to use well test analysis, fracturing construction methods, and mathematical statistics to invert fracturing fracture morphology and then further predict fracture propagation effects. Compared to traditional fracture propagation methods, fracture-vuggy fracture propagation is not only affected by horizontal principal stresses but also requires consideration of the influence of vuggy structures on local stresses. Furthermore, complex mechanical behaviors such as crossing and communicating occur after encountering karst caves during fracture propagation. Fracture propagation in vuggy structures is itself a multi-field, multi-factor coupled problem. Therefore, numerical simulation methods for simulating fracture-vuggy fracture propagation typically involve a large computational workload and a long computation period.

[0004] Currently, scholars have applied artificial intelligence concepts to fracturing fracture propagation prediction models. By combining mathematical and logical methods with the fracturing process, they have achieved intelligent prediction and assessment of fracturing fracture propagation. Machine learning can train models on large amounts of data, and the training data can be supplemented at any time, making the learned models more suitable for field needs and overcoming the application limitations of traditional numerical models. However, most current models mainly judge the relationship between fractures and natural fractures. Furthermore, there is limited research on fracture propagation prediction methods that combine machine learning and numerical simulation methods to consider the characteristics of fracture and cavity development in carbonate reservoirs. Therefore, based on the mutual advantages of machine learning and numerical simulation methods, constructing a prediction model for hydraulic fracture propagation in carbonate fracture and cavity structures can not only quantitatively study fracture propagation but also improve the calculation speed of fracturing propagation simulation, providing guidance and rapid feedback for optimizing field fracturing processes. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for predicting hydraulic fracture propagation in fractured-vuggy carbonate rocks based on a dual-drive model of mechanistic data. It is mainly used to solve the problems of low computational efficiency and limited prediction accuracy of fracture propagation in conventional fractured-vuggy carbonate reservoirs. By drawing on the advantages of machine learning in fracture propagation simulation, this invention fills the gap in existing technologies and obtains a simulation process and method with high computational efficiency and solution accuracy. Terminology Explanation: 1. Solid Deformation Field: Also known as the continuous solid deformation field, it is a core concept in continuum mechanics. It describes the state of deformation of each point inside a solid with time and spatial position under the action of external forces.

[0006] 2. Extended Finite Element Method (XFEM): This is an improved finite element method specifically designed to solve problems involving interface discontinuities (such as cracks and material interfaces). It introduces special functions to represent these discontinuities without modifying the mesh structure, making it a powerful tool for simulating crack initiation and propagation.

[0007] 3. Partition of Unity Method (PUM): This is a weighted approximation method based on the partition definition function (i.e., partition function). It expands the solution space by multiplying the local enhancement function with the traditional finite element shape function, and can effectively represent local singularities, discontinuities and multi-scale features.

[0008] 4. Level Set Method (LSM): This is a numerical method for capturing and tracking the evolution of interfaces (such as cracks, phase boundaries, and free surfaces). It represents the interface through a zero level set of a high-dimensional implicit function, which allows topological changes of the interface (such as splitting and merging) to be handled naturally.

[0009] 5. XGBoost Machine Learning Algorithm Model (eXtreme Gradient Boosting): This is an ensemble learning algorithm based on gradient boosting. It constructs a classifier by integrating multiple weak classifiers (usually CART trees). It is characterized by high efficiency, accuracy, and scalability, and is one of the most widely used machine learning methods today.

[0010] 6. ABAQUS Finite Element Software: This is a powerful finite element software for engineering simulation, capable of solving problems ranging from relatively simple linear analyses to many complex nonlinear problems. ABAQUS includes a rich library of elements that can simulate arbitrary geometries. It also has a library of various types of material models that can simulate the properties of typical engineering materials, including metals, rubber, polymers, composite materials, reinforced concrete, compressible hyperelastic foams, and geological materials such as soil and rock. As a general-purpose simulation tool, ABAQUS can solve a large number of structural (stress / displacement) problems and simulate many other problems in other engineering fields, such as heat conduction, mass diffusion, thermoelectric coupling analysis, acoustic analysis, geotechnical analysis (fluid permeability / stress coupling analysis), and piezoelectric analysis.

[0011] The technical solution of the present invention is as follows: The first aspect of this invention provides a method for predicting the propagation of hydraulic fractures in fractured-vuggy carbonate rocks based on a dual-driven model of mechanistic data, comprising: Step 1: Construct a fractured-vuggy carbonate reservoir model; Step 2: Construct a hydraulic fracturing model of the distribution morphology of natural fractures and pores in the fractured-vuggy carbonate reservoir model; Step 3: Construct a geomechanical model of fractured-vuggy carbonate reservoirs; Step 4: Based on the geomechanical model of fractured-vuggy carbonate reservoirs and the extended finite element method, determine the crack propagation of the hydraulic fracturing model; Step 5: Collect on-site fracturing data of fractured-vuggy reservoirs and perform numerical simulation on the hydraulic fracturing model to obtain fracture propagation dataset; Step 6: Construct a crack propagation prediction model. Train the crack propagation prediction model using the crack propagation dataset to obtain a trained crack propagation prediction model. Step 7: Use the trained crack propagation prediction model to predict and identify crack propagation patterns.

[0012] According to a preferred embodiment of the present invention, constructing a fractured-vuggy carbonate reservoir model includes: Collect target reservoir parameters, including elastic modulus, Poisson's ratio, tensile strength, maximum horizontal principal stress, minimum horizontal principal stress, and porosity; Geometric modeling was performed using ABAQUS finite element software. Modeling materials were defined and the collected target reservoir parameters were assigned values ​​to construct and generate a fractured-vuggy carbonate reservoir model.

[0013] According to a preferred embodiment of the present invention, a hydraulic fracturing model is constructed in a fractured-vuggy carbonate reservoir model to demonstrate the distribution morphology of natural fractures and pores; comprising: Hydraulic fracturing models for the distribution patterns of natural fractures and pores include: single pore model, continuous pore model, single fracture model, and fracture-cavity combination model; The single hole model is represented as: (1); in, The diameter of the hole, For stress difference, For displacement, Where H is the viscosity of the fracturing fluid, and H{} represents the path offset or deflection characteristics of the fracture as it passes through the cavity or fracture. It is a function that represents the calculation process of the simulation software and describes the influence of the input variables on the target output H; The continuous hole model is represented as: (2); in, The diameter of the hole, The spacing between holes, For the internal pressure of the hole, For stress difference, For displacement, This refers to the viscosity of the fracturing fluid. This represents the calculation process of the simulation software; The single crack model is represented as: (3); in, Cohesion within natural cracks, The internal friction angle of a natural crack. For the approximation angle, For stress difference, For displacement, This refers to the viscosity of the fracturing fluid. This represents the calculation process of the simulation software; The combined model of the seam and hole is represented as follows: (4); in, The diameter of the hole, For the internal pressure of the hole, Cohesion within natural cracks, The internal friction angle of a natural crack. For the approximation angle, For stress difference, For displacement, This refers to the viscosity of the fracturing fluid. This represents the calculation process of the simulation software.

[0014] According to a preferred embodiment of the present invention, a geomechanical model for fractured-vuggy carbonate reservoirs is constructed, comprising: The geomechanical model of fractured-vuggy carbonate reservoirs includes the solid deformation field equation, the fluid pressure field control equation in the rock matrix, and the fluid pressure field control equation in the rock fractures: The equations for solid deformation fields include stress equilibrium equations, displacement-strain equations, and constitutive equations, among which the stress equilibrium equations are expressed as: (5); in, yes x Normal stress in the direction; It is the front side y Direction pointing x Shear stress in the direction; yes x Volumetric stress in the direction; Front is x Direction pointing y Shear stress in the direction; yes y Normal stress in the direction; yes y Volumetric stress in the direction; The displacement-strain equation is expressed as: (6); in, , They represent , Displacement in direction; , They are respectively , The positive strain in the direction; Shear strain; The constitutive equation is expressed as: (7); in , They are respectively , Normal stress in the direction; Represents shear stress; Indicates the shear modulus of rock; Indicates the bulk modulus of rock; Represents the Biot coefficient; Pore ​​pressure; Represents the shear stress components in the x and y directions; The governing equations for the fluid pressure field in the rock matrix are: (8); in, For fluid density; Porosity; Bulk modulus of the fluid; Represents the Biot coefficient; The bulk modulus of the rock matrix; Permeability of the rock matrix; For fluid viscosity; Pore ​​pressure; For the source and sink of fluids in the matrix; For rock volumetric strain; For time; The governing equation for the fluid pressure field in rock fractures is: (9); in, For fluid density; Bulk modulus of the fluid; Pore ​​pressure; Permeability of rock fractures; For fluid viscosity; It serves as both the source and sink of fluid within the crack.

[0015] According to a preferred embodiment of the present invention, the crack propagation in a hydraulic fracturing model is determined based on a geomechanical model of fractured-vuggy carbonate reservoirs and an extended finite element method; including: The change in shape function of a hydraulic fracture, generated in hydraulic fracturing simulation, as it passes through an element is described by a step expansion shape function; the expression for the step expansion shape function is... As shown below: (10); in, l Indicates the signed distance from a point to the crack surface or the extension line of the crack tip; The location of the crack is determined by the level set function, i.e., the signed distance function, as shown below: (11); in, Points representing cracks, yes Orthogonal projection on the crack surface; It is a cracked surface The projection of the outer normal at the point; This indicates a point on the crack. [.] is a sign function that determines whether a point lies on the crack normal. Indicates time, This represents the crack distance function calculated in the level set method; The location of the crack tip is determined by constructing a level set function, which is shown below: (12); in, , These are the polar coordinates of the crack tip. The horizontal set function representing the crack tip; Based on the type function, i.e., the extension function referenced in the extended finite element, a continuous displacement field function describing the crack is formed, as shown below: (13); in, This represents the total displacement field function constructed in XFEM, that is, the displacement solution approximated by the XFEM method in a region containing cracks or discontinuities. These are nodes set at the tips of Type I and Type II cracks; These are standard degrees of freedom; It is the number of nodes through which the crack passes; It is the step expansion shape function; These represent the degrees of freedom at the crack surface, the tip of a type I crack, and the tip of a type II crack, respectively. Indicates the first i Finite element shape function of nodes, Indicates the number of nodes. Indicates the first j Finite element shape function of nodes, This indicates the node corresponding to the crack tip reinforcement node. k shape function, and The corresponding singular function at the crack tip represents the stress singularity; The maximum principal stress criterion is chosen as the criterion for determining crack initiation in the final two-dimensional numerical model, i.e., the hydraulic pressure model. The expression is: (14); in, Represents the stress criterion function. This indicates the current maximum principal stress or the current actual stress at a certain point. It is the critical maximum principal stress that reservoir rocks can withstand, measured in MPa; when the critical stress value is exceeded, the rocks begin to deteriorate. Based on the maximum principal stress criterion, the energy release rate is introduced to determine crack propagation, as shown below: (15); in, Normal fracture energy release rate, N / mm; It is the tangential energy release rate, N / mm; It is the energy release rate of a composite material crack, in N / mm; This indicates the energy release rate during slip fracture.

[0016] According to a preferred embodiment of the present invention, data from on-site fracturing operations in fractured-vuggy reservoirs are collected, and numerical simulations are performed on a hydraulic fracturing model to obtain a fracture propagation dataset; including: Field fracturing data for fractured reservoirs includes wellbore azimuth, fractured cavity dimensions, two-dimensional horizontal stress difference, fracturing fluid discharge rate, and fracturing fluid viscosity. The hydraulic fracturing model was numerically simulated using on-site fractured-vuggy reservoir fracturing construction data, and the numerical simulation results were obtained. The field fracturing data and numerical simulation results of fractured-vuggy reservoirs were used as the fracture propagation dataset.

[0017] According to a preferred embodiment of the present invention, a crack propagation prediction model is constructed, and the crack propagation prediction model is trained using a crack propagation dataset to obtain a trained crack propagation prediction model; including: The crack propagation prediction model is the XGBoost machine learning algorithm model; The wellbore azimuth, fracture cavity size, two-way horizontal stress difference, fracturing fluid discharge rate, fracturing fluid viscosity, and numerical simulation results of the hydraulic fracturing model from the fracture propagation dataset are used as input parameters. The input parameters are processed using range standardization, as shown below: (16); in, The input parameters are the standardized sample data. It is the minimum value among all similar values ​​in the input parameters. The maximum value among all similar values ​​in the input parameters; Indicates the input parameters; The i-th sample, i.e., the standardized sample data, is input into the XGBoost machine learning algorithm model and processed... tThe optimized predicted value is expressed as follows: (17); in, Indicates the first i One sample in t The predicted value after optimization. This represents the predicted value from the previous round. For the XGBoost machine learning algorithm model, the first i Sample The output is the fitted value of the residual; The objective function is approximated using Taylor expansion, i.e., the objective function is optimized, as shown below: (20); in, This indicates the output value of the new tree. a quadratic function, and These are the first and second derivatives of the loss function, respectively. Represents the loss function; Represents the regularization term; Through continuous iterative training, a well-trained crack propagation prediction model is obtained, which predicts two types of cracks: deflection and penetration through holes.

[0018] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of a method for predicting the propagation of hydraulic fractures in fractured carbonate rocks based on a dual-drive model of mechanistic data.

[0019] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of a method for predicting the propagation of hydraulic fractures in fractured carbonate rocks based on a dual-drive model of mechanistic data.

[0020] A second aspect of the present invention provides a system for predicting the propagation of hydraulic fractures in fractured carbonate rocks based on a dual-drive model of mechanistic data, comprising: The fractured-vuggy carbonate reservoir model building module is configured to: build a fractured-vuggy carbonate reservoir model; The hydraulic fracturing model construction module is configured to: construct a hydraulic fracturing model of the distribution morphology of natural fractures and pores in a fractured-vuggy carbonate reservoir model; The geomechanical model building module is configured to: build a geomechanical model of fractured-vuggy carbonate reservoirs; The crack propagation module of the hydraulic fracturing model is configured to determine the crack propagation of the hydraulic fracturing model based on the geomechanical model of fractured-vuggy carbonate reservoirs and the extended finite element method. The numerical simulation module is configured to: collect on-site fractured-vuggy reservoir fracturing construction data and perform numerical simulation on the hydraulic fracturing model to obtain fracture propagation dataset; The crack propagation prediction model training module is configured to: build a crack propagation prediction model, train the crack propagation prediction model using a crack propagation dataset, and obtain a trained crack propagation prediction model. The prediction module is configured to predict and discriminate crack propagation morphology using a trained crack propagation prediction model.

[0021] The beneficial effects of this invention are as follows: 1. The propagation law of hydraulic fractures in carbonate rocks containing pores and natural fractures, established based on machine learning methods, can provide strong theoretical support and optimization tools for fracturing and stimulation of carbonate reservoirs.

[0022] 2. During construction, input parameters are updated in real time to dynamically reflect the crack propagation status, facilitating on-site construction optimization decisions.

[0023] 3. By identifying abnormal crack propagation trends in advance, safety risks such as unexpected layer penetration and wellbore instability can be effectively avoided.

[0024] 4. Provides key support for intelligent fracturing operations and digital twin reservoirs. Attached Figure Description

[0025] Figure 1 This is a flowchart of the method for predicting the propagation of hydraulic fractures in fractured carbonate rocks based on a dual-drive model of mechanistic data, as described in this invention. Figure 2 This is a single-hole hydraulic fracturing model for the present invention; Figure 3 This is a hydraulic fracturing model with continuous holes for the present invention; Figure 3 (a) is a hydraulic fracturing model with a single cavity; Figure 3 (b) is a hydraulic fracturing model with two cavities; Figure 3 (c) is a hydraulic fracturing model with three cavities; Figure 3 The middle (d) is a hydraulic fracturing model with four holes. Figure 4 This invention provides a hydraulic fracturing model for a single natural fracture. Figure 5 This is a hydraulic fracturing model of the fracture-cavity assembly of the present invention; Figure 6 The simulation results of changing the diameter of the hole under the single hole crack model of the present invention; Figure 6 (a) shows the simulation results for a single-hole crack model with a hole diameter of 1m; Figure 6 (b) shows the simulation results for a single-hole crack model with a hole diameter of 2m; Figure 6 (c) shows the simulation results for a single-hole crack model with a hole diameter of 3m; Figure 6 In the middle (d), the simulation results are for a hole with a diameter of 4m; Figure 7 The simulation results of changing the ground stress difference under the single-hole crack model of the present invention; Figure 7 In the middle (a), the simulation results are shown for a stress difference of 5 MPa under a single-hole fracture model. Figure 7 (b) shows the simulation results of a single-hole fracture model with a ground stress difference of 10 MPa; Figure 7 (c) shows the simulation results of a single-hole fracture model with a ground stress difference of 15 MPa; Figure 7 In the middle (d), the simulation results are shown for a stress difference of 20 MPa under a single-hole fracture model. Figure 8 The simulation results of changing the geostress difference under the continuous hole-crack model of the present invention are shown. Figure 8 (a) shows the simulation results of the next hole in the continuous hole crack model; Figure 8 (b) shows the simulation results of two holes under the continuous hole crack model; Figure 8 (c) shows the simulation results of three holes under the continuous hole crack model; Figure 8 (d) shows the simulation results of four holes under the continuous hole crack model; Figure 9 The simulation results of pore fracturing under different fracture-cavity combinations are presented in this invention. Figure 9 (a) shows the simulation results of pore pressure of 55 MPa under different fracture-cavity combinations; Figure 9 (b) shows the simulation results of pore pressure of 50 MPa under different fracture-cavity combinations; Figure 9 (c) shows the simulation results of pore pressure of 45 MPa under different fracture-cavity combinations; Figure 9 (d) shows the simulation results of pore pressure of 40 MPa under different fracture-cavity combinations; Figure 10 This is a schematic diagram comparing the prediction model and simulation results under 10 parameter combinations of the present invention. Detailed Implementation

[0026] The present invention will be further described below with reference to the embodiments and accompanying drawings, but is not limited thereto.

[0027] Example 1 A method for predicting hydraulic fracture propagation in fractured-vuggy carbonate rocks based on a dual-driven model of mechanistic data, such as... Figure 1 As shown, it includes: Step 1: Construct a fractured-vuggy carbonate reservoir model; Step 2: Construct a hydraulic fracturing model of the distribution morphology of natural fractures and pores in the fractured-vuggy carbonate reservoir model; Step 3: Construct a geomechanical model of fractured-vuggy carbonate reservoirs; Step 4: Based on the geomechanical model of fractured-vuggy carbonate reservoirs and the extended finite element method, determine the crack propagation of the hydraulic fracturing model; Step 5: Collect on-site fracturing data of fractured-vuggy reservoirs and perform numerical simulation on the hydraulic fracturing model to obtain fracture propagation dataset; Step 6: Construct a crack propagation prediction model. Train the crack propagation prediction model using the crack propagation dataset to obtain a trained crack propagation prediction model. Step 7: Use the trained crack propagation prediction model to predict and identify crack propagation patterns.

[0028] Example 2 The difference between the method for predicting hydraulic fracture propagation in fractured-vuggy carbonate rocks based on a dual-driven model of mechanistic data described in Example 1 and the method described in Example 1 is as follows: Constructing a fractured-vuggy carbonate reservoir model; including: Collect target reservoir parameters, including elastic modulus, Poisson's ratio, tensile strength, maximum horizontal principal stress, minimum horizontal principal stress, and porosity; Geometric modeling was performed using ABAQUS finite element software. Modeling materials were defined and the collected target reservoir parameters were assigned values ​​to construct and generate a fractured-vuggy carbonate reservoir model.

[0029] Constructing a hydraulic fracturing model of natural fracture and void distribution patterns within a fractured-vuggy carbonate reservoir model; including: Hydraulic fracturing models for the distribution patterns of natural fractures and pores include: single pore model, continuous pore model, single fracture model, and fracture-cavity combination model; The single-hole model (characteristic of hydraulic fracture propagation upon encountering a hole) is represented as follows: (1); in, The diameter of the hole, For stress difference, For displacement, Where H is the viscosity of the fracturing fluid, and H{} represents the path offset or deflection characteristics of the fracture as it passes through the cavity or fracture. It is a function that represents the calculation process of the simulation software and describes the influence of the input variables on the target output H. Its specific form needs to be determined through theoretical derivation, numerical simulation or experimental data fitting. The continuous void model (the characteristic of hydraulic fractures expanding upon encountering voids) is represented as follows: (2); in, The diameter of the hole, The spacing between holes, For the internal pressure of the hole, For stress difference, For displacement, This refers to the viscosity of the fracturing fluid. This represents the calculation process of the simulation software; The single fracture model (the propagation characteristics of a hydraulic fracture encountering a natural fracture) is represented as follows: (3); in, Cohesion within natural cracks, The internal friction angle of a natural crack. For the approximation angle, For stress difference, For displacement, This refers to the viscosity of the fracturing fluid. The fracture-cavity combination model (characteristics of hydraulic fractures encountering natural fractures and pore expansion) is represented as follows: (4); in, The diameter of the hole, For the internal pressure of the hole, Cohesion within natural cracks, The internal friction angle of a natural crack. For the approximation angle, For stress difference, For displacement, This refers to the viscosity of the fracturing fluid. This represents the calculation process of the simulation software.

[0030] Constructing a geomechanical model of fractured-vuggy carbonate reservoirs; including: The geomechanical model of fractured-vuggy carbonate reservoirs includes the solid deformation field equation, the fluid pressure field control equation in the rock matrix, and the fluid pressure field control equation in the rock fractures: The equations for solid deformation fields include stress equilibrium equations, displacement-strain equations, and constitutive equations, among which the stress equilibrium equations are expressed as: (5); in, yes x Normal stress in the direction (x-direction of the XOY coordinate system of the two-dimensional model); It is the front side y The direction (y-direction of the XOY coordinate system of the two-dimensional model) points to x Shear stress in the direction; yes x Volumetric stress in the direction; Front is x Direction pointing y Shear stress in the direction; yes y Normal stress in the direction; yes y Volumetric stress in the direction; The displacement-strain equation is expressed as: (6); in, , They represent , Displacement in direction; , They are respectively , The positive strain in the direction; Shear strain; The constitutive equation is expressed as: (7); in, , They are respectively , Normal stress in the direction; Represents shear stress; Indicates the shear modulus of rock; Indicates the bulk modulus of rock; Represents the Biot coefficient; Pore ​​pressure; This represents the shear stress components in the x and y directions, also known as in-plane shear stress. The governing equations for the fluid pressure field in the rock matrix are: (8); in, For fluid density; Porosity; Bulk modulus of the fluid; Represents the Biot coefficient; The bulk modulus of the rock matrix; Permeability of the rock matrix; For fluid viscosity; Pore ​​pressure; For the source and sink of fluids in the matrix; For rock volumetric strain; For time; The governing equation for the fluid pressure field in rock fractures is: (9); in, For fluid density; Bulk modulus of the fluid; Pore ​​pressure; Permeability of rock fractures; For fluid viscosity; It serves as both the source and sink of fluid within the crack.

[0031] Based on the geomechanical model of fractured-vuggy carbonate reservoirs and the extended finite element method (XFEM), crack propagation in a hydraulic fracturing model is determined; including: The basic principle of extended finite element is based on the unified partitioning method. The unified partitioning method introduces some strengthening functions into the traditional finite element displacement mode to reflect the discontinuity on both sides of the crack, and uses the level set method (LSM) to track the movement of the crack interface. The nodes through which the crack passes are given additional degrees of freedom. The change in the shape function of the crack (i.e., the hydraulic fracture generated in the hydraulic fracturing simulation) as it passes through the element (a mesh element in the numerical model) is described by a step expansion shape function; the expression for the step expansion shape function is... As shown below: (10); in, l H(l) represents the signed distance from a point to the crack surface or the extension line of the crack tip. It acts as an indicator or step function, dividing the space into two regions based on a point's position relative to the crack surface (i.e., which side of the crack it's on), and assigning these two regions different values ​​(+1 or -1). This distinction allows the numerical model to differentiate the physical fields (e.g., displacement fields) on either side of the crack within the same finite element element, effectively simulating crack opening or slippage without requiring precise meshing along the crack line. This is one of the core ideas behind XFEM's handling of discontinuities.

[0032] The location of the crack is determined by the level set function, i.e., the signed distance function, as shown below: (11); in, Points representing cracks, yes Orthogonal projection on the crack surface; It is a cracked surface The projection of the outer normal at the point; This indicates a point on the crack, usually x i It itself or a specially selected reference point, [.] is a sign function that determines whether a point lies on the crack normal and returns +1 or -1. Indicates time, This represents the crack distance function (position changing over time) calculated in the level set method. The location of the crack tip is determined by constructing a level set function, which is shown below: (12); in, , These are the polar coordinates of the crack tip. The horizontal set function representing the crack tip; Based on the type function, i.e., the extension function referenced in the extended finite element, a continuous displacement field function describing the crack is formed, as shown below: (13); in, This represents the total displacement field function constructed in XFEM, that is, the displacement solution approximated by the XFEM method in a region containing cracks or discontinuities. These are nodes set at the tips of Type I and Type II cracks; These are standard degrees of freedom; It is the number of nodes through which the crack passes; It is the step expansion shape function; These represent the degrees of freedom at the crack surface, the tip of a type I crack, and the tip of a type II crack, respectively. Indicates the first i Finite element shape functions (usually linear or higher-order interpolation functions) for each node. Indicates the number of nodes. Indicates the first j Finite element shape functions (usually linear or higher-order interpolation functions) for each node. This indicates the node corresponding to the crack tip reinforcement node. k shape function, and The corresponding singularity function at the crack tip is often taken from the Williams expansion, representing stress singularity; The maximum principal stress criterion is chosen as the criterion for determining crack initiation in the final two-dimensional numerical model, i.e., the hydraulic pressure model. The expression is: (14); in, This represents the stress criterion function (which reflects whether the current stress state has reached the failure condition; in the XFEM model, it can be used as a crack propagation criterion). This indicates the current maximum principal stress or the current actual stress at a certain point. It is the critical maximum principal stress that reservoir rocks can withstand, measured in MPa; when the critical stress value is exceeded, the rocks begin to deteriorate. Based on the maximum principal stress criterion, the energy release rate is introduced to determine crack propagation, as shown below: (15); in, Normal fracture energy release rate, N / mm; It is the tangential energy release rate, N / mm; It is the energy release rate of a composite material crack, in N / mm; This indicates the energy release rate during slip fracture.

[0033] Data on fracturing operations in fractured-vuggy reservoirs were collected from the field, and numerical simulations were performed using a hydraulic fracturing model to obtain a fracture propagation dataset; including: Field fracturing data for fractured reservoirs includes wellbore azimuth, fractured cavity dimensions, two-dimensional horizontal stress difference, fracturing fluid discharge rate, and fracturing fluid viscosity. Numerical simulation of the constructed hydraulic fracturing model was performed using on-site fractured reservoir fracturing construction data, and the numerical simulation results were obtained (the intersecting and propagating morphology of hydraulic fractures with pores and natural fractures under different horizontal geostress differences and fracturing construction parameters). The field fracturing data and numerical simulation results of fractured-vuggy reservoirs were used as the fracture propagation dataset.

[0034] Construct a crack propagation prediction model, train the model using a crack propagation dataset, and obtain a trained crack propagation prediction model; including: The crack propagation prediction model is the XGBoost machine learning algorithm model; The wellbore azimuth, fracture cavity size, two-way horizontal stress difference, fracturing fluid discharge rate, fracturing fluid viscosity, and numerical simulation results of the hydraulic fracturing model from the fracture propagation dataset are used as input parameters. The input parameters are processed using range standardization, as shown below: (16); in, The input parameters are the standardized sample data. It is the minimum value among all similar values ​​in the input parameters. The maximum value among all similar values ​​in the input parameters; Indicates the input parameters; The XGBoost machine learning model builds a classifier by integrating multiple weak classifiers (usually CART trees), and is characterized by high efficiency, accuracy and scalability. It is one of the most widely used machine learning methods today. The i-th sample, i.e., the standardized sample data, is input into the XGBoost machine learning algorithm model and processed... tThe optimized predicted value is expressed as follows: (17); in, Indicates the first i One sample in t The predicted value after optimization. This represents the predicted value from the previous round. For the XGBoost machine learning algorithm model, the first i Sample The output is the fitted value of the residual; The goal of the XGBoost machine learning algorithm model is to minimize the objective function, as shown below: (18); in, This represents the total loss function, or objective function, of the XGBoost machine learning algorithm model; It is a loss function (usually the squared error or logistic loss function), representing the first... i Predicted values ​​for each sample Compared with the true value The error between them; n represents the number of samples, This indicates the newly added trees in this round. It is a regularization term, represented as: (19); in, It is the number of leaf nodes in the tree of the XGBoost machine learning algorithm model. Each leaf node corresponds to a numerical output, which represents the error correction amount of the sample in this round. The leaf outputs of all trees are accumulated to form the final prediction value. It is the weight of each leaf; This represents the structural complexity penalty term incurred for each newly added leaf node. This represents the L2 regularization coefficient that controls the size of the leaf output value; To transform the original complex nonlinear loss function into a quadratic form that can be efficiently computed and optimized, a Taylor expansion is used to approximate the objective function, i.e., the optimization objective function, and transform it into a quadratic form that can be efficiently computed and optimized, as shown below: (20); in, This indicates the output value of the new tree. a quadratic function, and These are the first and second derivatives (pseudo-residuals) of the loss function, respectively. Represents the regularization term; Through continuous iterative training, a well-trained crack propagation prediction model is obtained, which predicts two types of cracks: deflection and penetration through holes. The XGBoost machine learning model algorithm training process includes: Initialize the model and set the initial prediction value (the mean of the training data). In each iteration: (1) Calculate the pseudo residuals, i.e., the first and second derivatives of the loss function; (2) Construct a new tree to fit these pseudo residuals; (3) Update the model's predictions; Repeat the above process (processes 1, 2, and 3) until the stopping condition is met (the loss function no longer decreases significantly). Establish a predictive model: ; Where X is the input parameter, including wellbore azimuth, fracture cavity size, two-way horizontal stress difference, fracturing fluid discharge, fracturing fluid viscosity and hydraulic fracturing model, and Y is the output parameter (corresponding to two types: deflection and penetration, with 2 output parameters).

[0035] Example 3 A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the method for predicting the propagation of hydraulic fractures in fractured carbonate rocks based on a dual-drive model of mechanistic data as described in Embodiment 1 or 2.

[0036] Example 4 A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for predicting the propagation of hydraulic fractures in fractured carbonate rocks based on a dual-drive model of mechanistic data as described in Embodiment 1 or 2.

[0037] Example 5 A system for predicting the propagation of hydraulic fractures in fractured-vuggy carbonate rocks based on a dual-driven model using mechanistic data includes: The fractured-vuggy carbonate reservoir model building module is configured to: build a fractured-vuggy carbonate reservoir model; The hydraulic fracturing model construction module is configured to: construct a hydraulic fracturing model of the distribution morphology of natural fractures and pores in a fractured-vuggy carbonate reservoir model; The geomechanical model building module is configured to: build a geomechanical model of fractured-vuggy carbonate reservoirs; The crack propagation module of the hydraulic fracturing model is configured to determine the crack propagation of the hydraulic fracturing model based on the geomechanical model of fractured-vuggy carbonate reservoirs and the extended finite element method. The numerical simulation module is configured to: collect on-site fractured-vuggy reservoir fracturing construction data and perform numerical simulation on the hydraulic fracturing model to obtain fracture propagation dataset; The crack propagation prediction model training module is configured to: build a crack propagation prediction model, train the crack propagation prediction model using a crack propagation dataset, and obtain a trained crack propagation prediction model. The prediction module is configured to predict and discriminate crack propagation morphology using a trained crack propagation prediction model.

Claims

1. A method for predicting the propagation of hydraulic fractures in fractured-vuggy carbonate rocks based on a dual-driven model of mechanistic data, characterized in that, include: Step 1: Construct a fractured-vuggy carbonate reservoir model; Step 2: Construct a hydraulic fracturing model of the distribution morphology of natural fractures and pores in the fractured-vuggy carbonate reservoir model; Step 3: Construct a geomechanical model of fractured-vuggy carbonate reservoirs; Step 4: Based on the geomechanical model of fractured-vuggy carbonate reservoirs and the extended finite element method, determine the crack propagation of the hydraulic fracturing model; Step 5: Collect on-site fracturing data of fractured-vuggy reservoirs and perform numerical simulation on the hydraulic fracturing model to obtain fracture propagation dataset; Step 6: Construct a crack propagation prediction model. Train the crack propagation prediction model using the crack propagation dataset to obtain a trained crack propagation prediction model. Step 7: Use the trained crack propagation prediction model to predict and identify crack propagation patterns.

2. The method for predicting the propagation of hydraulic fractures in fractured-vuggy carbonate rocks based on a dual-driven model of mechanistic data as described in claim 1, characterized in that, Constructing a fractured-vuggy carbonate reservoir model; including: Collect target reservoir parameters, including elastic modulus, Poisson's ratio, tensile strength, maximum horizontal principal stress, minimum horizontal principal stress, and porosity; Geometric modeling was performed using ABAQUS finite element software. Modeling materials were defined and the collected target reservoir parameters were assigned values ​​to construct and generate a fractured-vuggy carbonate reservoir model.

3. The method for predicting the propagation of hydraulic fractures in fractured-vuggy carbonate rocks based on a dual-driven model of mechanistic data as described in claim 2, characterized in that, A hydraulic fracturing model of natural fracture and void distribution patterns was constructed in a fractured-vuggy carbonate reservoir model. include: Hydraulic fracturing models for the distribution patterns of natural fractures and pores include: single pore model, continuous pore model, single fracture model, and fracture-cavity combination model; The single hole model is represented as: (1)； in, The diameter of the hole, For stress difference, For displacement, Where H is the viscosity of the fracturing fluid, and H{} represents the path offset or deflection characteristics of the fracture as it passes through the cavity or fracture. It is a function that represents the calculation process of the simulation software and describes the influence of the input variables on the target output H; The continuous hole model is represented as: (2)； in, The diameter of the hole, The spacing between holes, For the internal pressure of the hole, For stress difference, For displacement, This refers to the viscosity of the fracturing fluid. This represents the calculation process of the simulation software; The single crack model is represented as: (3)； in, Cohesion within natural cracks, The internal friction angle of a natural crack. For the approximation angle, For stress difference, For displacement, This refers to the viscosity of the fracturing fluid. This represents the calculation process of the simulation software; The combined model of the seam and hole is represented as follows: (4)； in, The diameter of the hole, For the internal pressure of the hole, Cohesion within natural cracks, The internal friction angle of a natural crack. For the approximation angle, For stress difference, For displacement, This refers to the viscosity of the fracturing fluid. This represents the calculation process of the simulation software.

4. The method for predicting the propagation of hydraulic fractures in fractured-vuggy carbonate rocks based on a dual-driven model of mechanistic data as described in claim 3, characterized in that, Constructing a geomechanical model of fractured-vuggy carbonate reservoirs; including: The geomechanical model of fractured-vuggy carbonate reservoirs includes the solid deformation field equation, the fluid pressure field control equation in the rock matrix, and the fluid pressure field control equation in the rock fractures: The equations for solid deformation fields include stress equilibrium equations, displacement-strain equations, and constitutive equations, among which the stress equilibrium equations are expressed as: (5)； in, yes x Normal stress in the direction; It is the front side y Direction pointing x Shear stress in the direction; yes x Volumetric stress in the direction; Front is x Direction pointing y Shear stress in the direction; yes y Normal stress in the direction; yes y Volumetric stress in the direction; The displacement-strain equation is expressed as: (6)； in, , They represent , Displacement in direction; , They are respectively , The positive strain in the direction; Shear strain; The constitutive equation is expressed as: (7)； in, , They are respectively , Normal stress in the direction; Represents shear stress; Indicates the shear modulus of rock; Indicates the bulk modulus of rock; Represents the Biot coefficient; Pore ​​pressure; Represents the shear stress components in the x and y directions; The governing equations for the fluid pressure field in the rock matrix are: (8)； in, For fluid density; Porosity; Bulk modulus of the fluid; Represents the Biot coefficient; The bulk modulus of the rock matrix; Permeability of the rock matrix; For fluid viscosity; Pore ​​pressure; For the source and sink of fluids in the matrix; For rock volumetric strain; For time; The governing equation for the fluid pressure field in rock fractures is: (9)； in, For fluid density; Bulk modulus of the fluid; Pore ​​pressure; Permeability of rock fractures; For fluid viscosity; It serves as both the source and sink of fluid within the crack.

5. The method for predicting the propagation of hydraulic fractures in fractured-vuggy carbonate rocks based on a dual-driven model of mechanistic data as described in claim 4, characterized in that, Based on the geomechanical model of fractured-vuggy carbonate reservoirs and the extended finite element method, crack propagation in a hydraulic fracturing model is determined; including: The change in shape function of a hydraulic fracture, generated in hydraulic fracturing simulation, as it passes through an element is described by a step expansion shape function; the expression for the step expansion shape function is... As shown below: (10)； in, l Indicates the signed distance from a point to the crack surface or the extension line of the crack tip; The location of the crack is determined by the level set function, i.e., the signed distance function, as shown below: (11)； in, Points representing cracks, yes Orthogonal projection on the crack surface; It is a cracked surface The projection of the outer normal at the point; This indicates a point on the crack. [.] is a sign function that determines whether a point lies on the crack normal. Indicates time, This represents the crack distance function calculated in the level set method; The location of the crack tip is determined by constructing a level set function, which is shown below: (12)； in, , These are the polar coordinates of the crack tip. The horizontal set function representing the crack tip; Based on the type function, i.e., the extension function referenced in the extended finite element, a continuous displacement field function describing the crack is formed, as shown below: (13)； in, This represents the total displacement field function constructed in XFEM, that is, the displacement solution approximated by the XFEM method in a region containing cracks or discontinuities. These are nodes set at the tips of Type I and Type II cracks; These are standard degrees of freedom; It is the number of nodes through which the crack passes; It is the step expansion shape function; These represent the degrees of freedom at the crack surface, the tip of a type I crack, and the tip of a type II crack, respectively. Indicates the first i Finite element shape function of nodes, Indicates the number of nodes. Indicates the first j Finite element shape function of nodes, This indicates the node corresponding to the crack tip reinforcement node. k shape function, and The corresponding singular function at the crack tip represents the stress singularity; The maximum principal stress criterion is chosen as the criterion for determining crack initiation in the final two-dimensional numerical model, i.e., the hydraulic pressure model. The expression is: (14)； in, Represents the stress criterion function. This indicates the current maximum principal stress or the current actual stress at a certain point. It is the critical maximum principal stress that reservoir rocks can withstand, measured in MPa; when the critical stress value is exceeded, the rocks begin to deteriorate. Based on the maximum principal stress criterion, the energy release rate is introduced to determine crack propagation, as shown below: (15)； in, Normal fracture energy release rate, N / mm; It is the tangential energy release rate, N / mm; It is the energy release rate of a composite material crack, in N / mm; This indicates the energy release rate during slip fracture.

6. The method for predicting the propagation of hydraulic fractures in fractured-vuggy carbonate rocks based on a dual-driven model of mechanistic data as described in claim 5, characterized in that, Data on fracturing operations in fractured-vuggy reservoirs were collected from the field, and numerical simulations were performed using a hydraulic fracturing model to obtain a fracture propagation dataset; including: Field fracturing data for fractured reservoirs includes wellbore azimuth, fractured cavity dimensions, two-dimensional horizontal stress difference, fracturing fluid discharge rate, and fracturing fluid viscosity. The hydraulic fracturing model was numerically simulated using on-site fractured-vuggy reservoir fracturing construction data, and the numerical simulation results were obtained. The field fracturing data and numerical simulation results of fractured-vuggy reservoirs were used as the fracture propagation dataset.

7. The method for predicting the propagation of hydraulic fractures in fractured-vuggy carbonate rocks based on a dual-driven model of mechanistic data as described in claim 1, characterized in that, Construct a crack propagation prediction model, train the model using a crack propagation dataset, and obtain a trained crack propagation prediction model; including: The crack propagation prediction model is the XGBoost machine learning algorithm model; The wellbore azimuth, fracture cavity size, two-way horizontal stress difference, fracturing fluid discharge rate, fracturing fluid viscosity, and numerical simulation results of the hydraulic fracturing model from the fracture propagation dataset are used as input parameters. The input parameters are processed using range standardization, as shown below: (16)； in, The input parameters are the standardized sample data. It is the minimum value among all similar values ​​in the input parameters. The maximum value among all similar values ​​in the input parameters; Indicates the input parameters; The i-th sample, i.e., the standardized sample data, is input into the XGBoost machine learning algorithm model and processed... t The optimized predicted value is expressed as follows: (17)； in, Indicates the first i One sample in t The predicted value after optimization. This represents the predicted value from the previous round. For the XGBoost machine learning algorithm model, the first i Sample The output is the fitted value of the residual; The objective function is approximated using Taylor expansion, i.e., the objective function is optimized, as shown below: (20)； in, This indicates the output value of the new tree. a quadratic function, and These are the first and second derivatives of the loss function, respectively. Represents the loss function; Represents the regularization term; Through continuous iterative training, a well-trained crack propagation prediction model is obtained, which predicts two types of cracks: deflection and penetration through holes.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method for predicting the propagation of hydraulic fractures in fractured carbonate rocks based on a dual-drive model of mechanistic data as described in any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for predicting the propagation of hydraulic fractures in fractured carbonate rocks based on a dual-drive model of mechanistic data as described in any one of claims 1-7.

10. A prediction system for hydraulic fracture propagation in fractured-vuggy carbonate rocks based on a dual-driven model using mechanistic data, characterized in that, include: The fractured-vuggy carbonate reservoir model building module is configured to: build a fractured-vuggy carbonate reservoir model; The hydraulic fracturing model construction module is configured to: construct a hydraulic fracturing model of the distribution morphology of natural fractures and pores in a fractured-vuggy carbonate reservoir model; The geomechanical model building module is configured to: build a geomechanical model of fractured-vuggy carbonate reservoirs; The crack propagation module of the hydraulic fracturing model is configured to determine the crack propagation of the hydraulic fracturing model based on the geomechanical model of fractured-vuggy carbonate reservoirs and the extended finite element method. The numerical simulation module is configured to: collect on-site fractured-vuggy reservoir fracturing construction data and perform numerical simulation on the hydraulic fracturing model to obtain fracture propagation dataset; The crack propagation prediction model training module is configured to: build a crack propagation prediction model, train the crack propagation prediction model using a crack propagation dataset, and obtain a trained crack propagation prediction model. The prediction module is configured to predict and discriminate crack propagation morphology using a trained crack propagation prediction model.

Images