Full-depth reservoir physical and mechanical parameter characterization method, device, equipment and medium

Abstract

The application discloses a full-depth reservoir physical and mechanical parameter characterization method, device, equipment and medium, relates to oil and gas development, and comprises the following steps: acquiring and processing multi-source data in a wellbore range of a reservoir, and constructing and generating an artificial intelligence model; the prediction parameter results of the artificial intelligence model are constrained and corrected; the preprocessed multi-source data is jointly input into the artificial intelligence model, the constrained and corrected reservoir physical and mechanical parameter results are fused and completed, and a full-depth parameter field is formed; a numerical simulation model is constructed based on the parameter field, a reservoir mechanical response process is simulated, and simulation results are obtained; the parameter field is inverted and iteratively corrected by using the simulation results, and target full-depth reservoir physical and mechanical parameter characterization results are obtained. The application realizes multi-source data fusion, AI driving and simulation inversion closed loop, significantly improves the integrity and reliability of reservoir physical and mechanical parameter characterization, and is suitable for different types of reservoir engineering.

Description

Technical Field

[0001] This invention relates to the field of oil and gas development technology, specifically to methods, devices, equipment, and media for characterizing the physical and mechanical parameters of reservoirs at all depths. Background Technology

[0002] Reservoir physical and mechanical parameters are crucial for fracturing design, wellbore stability evaluation, and optimization of oil and gas reservoir development strategies. These parameters include porosity, density, Young's modulus, Poisson's ratio, and fracturing pressure. Current technologies primarily rely on well logging interpretation methods, core and cuttings experiments, and numerical simulation to obtain these parameters.

[0003] Well logging interpretation methods are often used to estimate reservoir physical parameters and have the advantage of describing wellbore continuity. However, these methods are mostly based on empirical formulas or statistical regression, which makes it difficult to reveal the heterogeneous characteristics of complex reservoirs and have limited ability to extend reservoirs between wells.

[0004] Core and cuttings testing methods can directly obtain pore structure and mechanical properties through microscopic testing or 3D CT scanning, yielding highly accurate results. However, limited by sampling intervals and experimental conditions, the data are only locally representative and cannot cover the entire well depth range.

[0005] Numerical simulation methods can utilize geomechanical models and employ techniques such as finite element method and finite difference method to simulate and invert reservoir response. However, these methods are highly dependent on the initial input data, and simulation results are prone to deviation when the input is incomplete or lacks precision.

[0006] In recent years, artificial intelligence (AI) methods have begun to be applied to reservoir parameter prediction and modeling, which can fit complex nonlinear relationships and improve data utilization efficiency. However, existing studies mostly use discriminative or regression models, which lack the ability to complete parameters under conditions of missing data. At the same time, due to insufficient physical constraints, the prediction results may deviate from the actual reservoir patterns.

[0007] In summary, existing methods have the following prominent problems:

[0008] 1. Insufficient data fusion: Well logging, core, and simulation data are difficult to couple effectively, resulting in large scale differences;

[0009] 2. Limited coverage: Experimental data is limited to local well sections and cannot achieve continuous characterization across the entire depth;

[0010] 3. Lack of physical consistency: AI predictions fail to fully incorporate mechanical mechanisms, posing a risk of bias.

[0011] 4. Lack of closed-loop optimization: The lack of an iterative mechanism of "data-AI-simulation-inversion" makes it difficult to dynamically improve the accuracy of results. Summary of the Invention

[0012] In view of the above problems, the present invention is proposed to provide a method, apparatus, device and medium for characterizing the physical and mechanical parameters of a full-depth reservoir to overcome or at least partially solve the above problems.

[0013] According to one aspect of the embodiments of this application, a method for characterizing the physical and mechanical parameters of a full-depth reservoir is provided, the method comprising:

[0014] Acquire multi-source data within the reservoir wellbore area, including core three-dimensional voxel structure data, cuttings microphysical property data, and logging curve parameters;

[0015] Multi-source data is preprocessed and feature extracted to obtain a standardized input feature vector for modeling.

[0016] Based on the standardized input feature vector, the reservoir physical and mechanical parameters are initially modeled to obtain a generative artificial intelligence model.

[0017] The preprocessed logging curve parameters are used to constrain and correct the predicted parameter results of the artificial intelligence model, resulting in constrained and corrected reservoir physical and mechanical parameter results.

[0018] The preprocessed multi-source data is jointly input into the artificial intelligence model, which drives the artificial intelligence model to fuse and complete the results of the reservoir physical and mechanical parameters after constraint correction, forming a parameter field of the whole depth within the reservoir wellbore range;

[0019] A numerical simulation model is constructed based on the parameter field, and the reservoir mechanical response process is simulated using the numerical simulation model to obtain simulation results;

[0020] The simulation results are used to invert and iteratively correct the parameter field, and the physical and mechanical parameters of the target reservoir at all depths are characterized.

[0021] Furthermore, the preprocessing and feature extraction of multi-source data to obtain a standardized input feature vector for modeling further includes:

[0022] A linear interval mapping method is used to normalize multi-source data, so as to transform different physical quantities in multi-source data into a unified interval;

[0023] Denoising the multi-source data;

[0024] For well logging parameters in multi-source data, missing data in the well logging parameters are completed based on the interpolation of the well logging parameters and the confidence weighting factor output by the artificial intelligence model.

[0025] By combining principal component analysis with a deep learning-based autoencoder, dimensionality reduction and deep feature extraction are performed on the preprocessed multi-source data to obtain a standardized input feature vector.

[0026] Furthermore, the completion formula used to complete the missing data in the well logging curve parameters is as follows:

[0027]

[0028] in, This represents the parameters of the completed logging curve at a well depth of z; This represents the preprocessed logging curve parameters at a well depth of z. This represents the confidence weighting factor output by the artificial intelligence model based on the data's credibility. This represents the correction amount corresponding to well depth z, obtained by interpolation based on adjacent well sections.

[0029] Furthermore, the constrained reservoir physical and mechanical parameters are calculated using the following formula:

[0030]

[0031]

[0032] in, This represents the corrected reservoir physical and mechanical parameters at a well depth of z. This represents the predicted parameters at well depth z obtained by the artificial intelligence model; Indicates constraint parameters; This represents the set of constraints corresponding to the logging curve parameters at a well depth of z. This represents the comparable logging parameters obtained by mapping the predicted parameter results; This represents the constrained corrected reservoir physical and mechanical parameters at a well depth of z. This represents the confidence factor at a well depth of z.

[0033] Furthermore, the preprocessed multi-source data is jointly input into the artificial intelligence model, driving the model to fuse and complete the constrained and corrected reservoir physical and mechanical parameter results, forming a parameter field covering the entire depth within the reservoir wellbore. This further includes:

[0034] The preprocessed multi-source data and the constrained corrected reservoir physical and mechanical parameters are uniformly mapped to the depth coordinate system to construct a multi-source parameter set.

[0035] The multi-source parameter set is weighted and fused using a fusion function to obtain the fused and completed reservoir physical and mechanical parameters. A parameter field is then formed based on these fused and completed parameters. The fusion function is:

[0036]

[0037]

[0038] in, This represents the reservoir physical and mechanical parameters after fusion and completion at a well depth of z; n represents the total number of data components in the multi-source parameter set; This represents the normalized weight of the i-th data component at well depth z in the multi-source parameter set; This represents the i-th data component at well depth z in the multi-source parameter set; This represents the basic weighting factor of the i-th data component at a well depth of z; This represents the weight correction factor of the AI ​​model for the i-th data component at a well depth of z during the fusion process; This represents the basic weighting factor of the j-th data component at a well depth of z; This represents the weight correction factor of the AI ​​model for the j-th data component at well depth z during the fusion process.

[0039] Furthermore, the simulation results are used to invert and iteratively correct the parameter field, and the resulting physical and mechanical parameter characterization results of the target full-depth reservoir further include:

[0040] Based on the differences between simulation results and measured data, an inversion optimization framework is constructed.

[0041] Based on the inversion optimization framework and iterative update formula, the parameter field is iteratively updated until the preset convergence condition is met, thus obtaining the physical and mechanical parameter characterization results of the target full-depth reservoir; wherein, the iterative update formula is:

[0042]

[0043] in, This represents the parameter set at the (k+1)th iteration; This represents the set of parameters at the k-th iteration. Indicates the learning rate or iteration step size; This represents the gradient of the objective function in the inversion optimization framework with respect to the parameters of the parameter set at the k-th iteration; Indicates the correction factor; This indicates an AI-driven correction operator used to dynamically adjust the inversion results using the fused and completed reservoir physical and mechanical parameters. This represents the reservoir physical and mechanical parameters after fusion and completion at a well depth of z.

[0044] According to another aspect of the embodiments of this application, a full-depth reservoir physical and mechanical parameter characterization device is provided, the device comprising:

[0045] The data acquisition module is suitable for acquiring multi-source data within the reservoir wellbore area, including core three-dimensional voxel structure data, cuttings microphysical property data, and logging curve parameters.

[0046] The data processing module is suitable for preprocessing and feature extraction of multi-source data to obtain standardized input feature vectors for modeling.

[0047] The model building module is suitable for performing preliminary modeling of reservoir physical and mechanical parameters based on standardized input feature vectors to obtain generative artificial intelligence models;

[0048] The parameter constraint module is suitable for using preprocessed logging curve parameters to constrain and correct the predicted parameter results of the artificial intelligence model, forming constrained and corrected reservoir physical and mechanical parameter results.

[0049] The fusion and completion module is suitable for inputting preprocessed multi-source data into the artificial intelligence model, driving the artificial intelligence model to fuse and complete the reservoir physical and mechanical parameter results after constraint correction, forming a parameter field of the whole depth within the reservoir wellbore range;

[0050] The simulation solution module is suitable for building numerical simulation models based on parameter fields, simulating reservoir mechanical response processes using numerical simulation models, and obtaining simulation results.

[0051] The inversion optimization module is suitable for using simulation results to invert and iteratively correct the parameter field, and obtain the characterization results of the physical and mechanical parameters of the target reservoir at all depths.

[0052] According to another aspect of the embodiments of this application, a computing device is provided, including: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other through the communication bus;

[0053] The memory is used to store at least one executable instruction, which causes the processor to perform the operation corresponding to the above-mentioned full-depth reservoir physical and mechanical parameter characterization method.

[0054] According to another aspect of the embodiments of this application, a computer storage medium is provided, which stores at least one executable instruction that causes a processor to perform operations corresponding to the above-described full-depth reservoir physical and mechanical parameter characterization method.

[0055] According to another aspect of the embodiments of this application, a computer program product is provided, including at least one executable instruction that causes a processor to perform operations corresponding to the above-described full-depth reservoir physical and mechanical parameter characterization method.

[0056] According to the technical solution provided by this invention, a closed loop of multi-source data fusion, AI-driven simulation inversion is realized, significantly improving the integrity and reliability of reservoir physical and mechanical parameter characterization, and applicable to different types of reservoir engineering. Specifically, at the single-well scale, a multi-scale, multi-dimensional data system is established by combining core three-dimensional voxel structure data, cuttings microphysical property data, and logging curve parameters. Fusion and completion of multi-source data are achieved through artificial intelligence models, greatly improving parameter integrity and effectively solving the problem of discontinuous and incomplete parameters caused by traditional methods relying on single logging data. Driven by generative artificial intelligence structures such as diffusion models, variational autoencoders, or generative adversarial networks, the standardized input feature vectors corresponding to multi-source data are mapped to the latent space. Combined with latent distribution constraints and optimized loss functions, the predictive ability is enhanced, realizing generative prediction of reservoir physical and mechanical parameters, overcoming the limitations of existing methods that can only perform simple regression fitting and are difficult to capture complex nonlinear relationships. Utilizing density, sonic transit time, and resistivity... Using logging parameters such as natural gamma as constraints, and combining them with predicted key mechanical parameters such as porosity, a dynamic fusion of predicted parameter results and actual logging data is achieved through parameter correction formulas and confidence weighting mechanisms. This ensures the physical consistency, rationality, and reliability of parameters across the entire depth range. In the simulation and inversion optimization stages, the fused and completed reservoir physical and mechanical parameter results are introduced into the constitutive relation, and then the parameters are dynamically corrected through iterative update formulas, forming a closed-loop process of "prediction-simulation-inversion." Compared with existing technologies, this significantly improves the accuracy and adaptability of reservoir physical and mechanical parameter characterization. This scheme has outstanding engineering application value, enabling high-precision characterization of reservoir physical and mechanical parameters across the entire well depth range under complex geological conditions. It can directly serve engineering stages such as drilling design, well completion engineering, and fracturing operations, providing more reliable inputs of key mechanical parameters such as porosity, elastic modulus, and Poisson's ratio for oil and gas development. This effectively reduces the risk of wellbore instability and fracture propagation, demonstrating good applicability and promising prospects for widespread application.

[0057] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0058] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0059] Figure 1 A flowchart illustrating a method for characterizing the physical and mechanical parameters of a full-depth reservoir according to an embodiment of this application is shown.

[0060] Figure 2 A schematic diagram of multi-source data acquisition is shown;

[0061] Figure 3 A schematic diagram of the fusion completion and simulation framework is shown;

[0062] Figure 4 A schematic diagram illustrating the principles of inversion and iterative correction is shown.

[0063] Figure 5 A comparison chart showing the porosity prediction and measured core data of well L1 is presented.

[0064] Figure 6 A bar chart comparing the prediction errors of different methods is shown;

[0065] Figure 7 Box plots comparing the prediction errors of different methods are shown;

[0066] Figure 8 A structural block diagram of a full-depth reservoir physical and mechanical parameter characterization device according to an embodiment of this application is shown;

[0067] Figure 9 A schematic diagram of the structure of a computing device according to an embodiment of this application is shown. Detailed Implementation

[0068] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0069] Figure 1 A flowchart illustrating a method for characterizing the physical and mechanical parameters of a full-depth reservoir according to an embodiment of this application is shown, as follows. Figure 1 As shown, the method includes the following steps:

[0070] Step S101: Obtain multi-source data within the reservoir wellbore area.

[0071] This application proposes a method for characterizing full-depth reservoir physical and mechanical parameters by combining generative artificial intelligence, well logging data, and numerical simulation. By fusing multi-source data such as core 3D voxel structure data, rock cuttings microphysical property data, and well logging curve parameters, and utilizing the data completion capabilities of generative artificial intelligence and simulation-driven inversion optimization, the method efficiently achieves the fusion and completion of multi-source data. Under physical constraints, it obtains continuous, reliable, and high-precision full-depth reservoir physical and mechanical parameter characterization results, realizing a closed loop of multi-source data fusion, AI-driven, and simulation inversion. This significantly improves the completeness and reliability of reservoir physical and mechanical parameter characterization, and is applicable to different types of reservoir engineering, providing reliable data support for engineering decisions such as fracturing design, wellbore stability evaluation, and oil and gas reservoir development.

[0072] To achieve high-precision characterization of reservoir physical and mechanical parameters across the entire depth, multi-source, multi-scale data acquisition was performed on the target wellbore within its reservoir wellbore range, resulting in multi-source data within the reservoir wellbore range. This multi-source data included core three-dimensional voxel structure data, cuttings microphysical property data, and logging curve parameters.

[0073] Figure 2 A schematic diagram of multi-source data acquisition is shown, such as... Figure 2 As shown, core CT scanning, such as CT slicing of three-dimensional core samples, is used to obtain three-dimensional voxel structure data of the core; rock cuttings microscopic experiments, such as observing two-dimensional images under a microscope, are used to obtain rock cuttings microphysical property data; and logging curve parameters, including density curve parameters, sonic curve parameters, resistivity curve parameters, etc., are obtained through conventional logging methods.

[0074] Specifically, high-resolution X-ray three-dimensional CT scanning was performed on the core samples retrieved from the wellbore to obtain three-dimensional voxel structure data of the core. The three-dimensional voxel structure data of the core can be represented as follows: To enhance the robustness of the data, this embodiment also introduces a weighted correction to the three-dimensional porosity in the three-dimensional voxel structure data of the core using the voxel segmentation confidence score output by a generative artificial intelligence model. The formula is as follows:

[0075]

[0076] in, This represents the weighted three-dimensional porosity in the three-dimensional voxel structure data of the core; N represents the total number of voxels. This represents the confidence level of the artificial intelligence model in determining that the i-th voxel belongs to the pore, and is simply referred to as the voxel segmentation confidence level. Its value ranges from 0 to 1. This represents the volume of the i-th voxel, i.e., the volume of the single voxel; Indicates the total volume of the CT scan;

[0077] Two-dimensional microscopic imaging and physical property testing were performed on rock cuttings samples recovered during the drilling process to calculate two-dimensional porosity. Furthermore, microscopic mechanical parameters can be obtained, such as the microscopic elastic modulus of rock fragments, which can be obtained through nanoindentation experiments. And hardness H. Therefore, the microscopic physical property data of rock cuttings obtained in the embodiments of this application may include two-dimensional porosity. Microelastic modulus of rock fragments The hardness H of the rock fragments.

[0078] The physical property curve parameters, i.e., logging curve parameters, are obtained at the full depth using conventional logging methods. These logging curve parameters may include density. Sound wave time difference resistivity and natural gamma etc., its expression is:

[0079]

[0080] in, This represents the set of logging curve parameters at a well depth of z. This represents the density at a well depth of z; This represents the acoustic transit time at a well depth of z; This represents the resistivity at a well depth of z; This represents the natural gamma at a well depth of z; z represents the well depth position.

[0081] Optionally, to achieve joint utilization of two-dimensional and three-dimensional data, multi-source data fusion correction can be further performed in this embodiment. The weighted three-dimensional porosity and two-dimensional porosity are fused and corrected using a depth-related weighting factor, and the fusion correction formula is as follows:

[0082]

[0083] in, This indicates the porosity after fusion correction; The depth-related weighting factor, determined based on well logging curve parameters, can be derived from density. Sound wave time difference Wait until it is determined; This represents the weighted three-dimensional porosity in the three-dimensional voxel structure data of the core. This represents the two-dimensional porosity in rock cuttings microphysical property data.

[0084] Through the multi-source acquisition and fusion correction of the above-mentioned core CT scan, cuttings microscopy experiments and conventional logging, a complete input dataset can be formed, which provides a foundation for subsequent data processing and modeling.

[0085] Step S102 involves preprocessing and feature extraction of the multi-source data to obtain a standardized input feature vector for modeling.

[0086] Preprocessing includes normalization, denoising, and missing data completion. By unifying and structuring multi-source data, the dimensional differences and noise interference between data from different sources and at different scales can be effectively eliminated, ensuring that the data can be used as standardized input for artificial intelligence models.

[0087] Specifically, a linear interval mapping method is used to normalize multi-source data, transforming different physical quantities in the multi-source data into a unified interval. Its expression is:

[0088]

[0089] in, This represents the normalized value of a physical parameter. The original data representing this physical parameter; This indicates the maximum value of the physical parameter; This represents the minimum value of the physical parameter. Normalization effectively avoids bias caused by data of different dimensions and magnitudes when inputting into artificial intelligence models.

[0090] Furthermore, denoising processing is required for multi-source data to filter out noise. For example, this is necessary for core three-dimensional voxel structure data. Isolated noise can be eliminated by three-dimensional median filtering, and its formal expression can be:

[0091]

[0092] in, This represents the filtered three-dimensional voxel structure data of the core; median indicates median filtering. Indicates coordinates as Three-dimensional voxel structure data of the core; Indicated by The set of neighborhood voxels centered on the center.

[0093] In addition, missing data in the multi-source data needs to be completed. Considering that wellbore depth sampling often has missing segments, the missing data in the well logging parameters of the multi-source data can be completed based on the interpolation of the well logging parameters and the confidence weighting factor output by the artificial intelligence model. The completion formula used is as follows:

[0094]

[0095] in, This represents the parameters of the completed logging curve at a well depth of z; This represents the preprocessed logging curve parameters at a well depth of z. This represents the confidence weighting factor output by the artificial intelligence model based on the data's credibility. This represents the correction amount corresponding to well depth z, obtained by interpolation based on adjacent well sections. By employing a completion method that combines interpolation based on well logging curve parameters with correction based on artificial intelligence model weights, the completed data can be made closer to the real data, effectively improving the usability and accuracy of the data.

[0096] After normalization, denoising, and completion processing of the multi-source data, preprocessed multi-source data is formed. Key feature parameters can then be extracted from this preprocessed multi-source data to form a standardized input feature vector. Feature extraction is achieved jointly using Principal Component Analysis (PCA) and a deep learning-based autoencoder. Specifically, PCA and the deep learning-based autoencoder are combined to perform dimensionality reduction and deep feature extraction on the preprocessed multi-source data, resulting in a standardized input feature vector, expressed as:

[0097]

[0098] in, This represents a standardized input feature vector used by artificial intelligence models. Indicates the feature extraction operator; This represents preprocessed multi-source data, including weighted three-dimensional porosity. Two-dimensional porosity Microelastic modulus of rock fragments The hardness H of the rock cuttings and the parameters of the completed logging curves. .

[0099] Through the above preprocessing and feature extraction processes, data from different sources and at different scales can be transformed into standardized, noise-free, missing-complete, and representative input vectors, providing a data foundation for the subsequent construction of artificial intelligence models.

[0100] Step S103: Based on the standardized input feature vector, the reservoir physical and mechanical parameters are initially modeled to obtain a generative artificial intelligence model.

[0101] Generative artificial intelligence models can be at least one of diffusion models, variational autoencoders, or generative adversarial networks. First, the input feature vectors are standardized. Mapped to latent space vectors ,in, Let represent the m-th eigenvector. Its mapping transformation relationship can be expressed as:

[0102]

[0103] in, Represents a potential space vector, used to represent an abstract representation of reservoir physical and mechanical parameters; Indicates the encoding function; Represents the standardized input feature vector; This represents the set of parameters for the encoder.

[0104] Then, the latent space vector is decoded using the decoding function. Predicted parameters converted into reservoir physical and mechanical parameters:

[0105]

[0106] in, This represents the predicted parameters at a well depth of z obtained by the artificial intelligence model, including porosity, Young's modulus, Poisson's ratio, and fracture pressure. Indicates the decoding function; Represents the latent space vector; This represents the set of parameters for the decoder.

[0107] Furthermore, to improve the consistency between the predicted parameters and the actual data, latent distribution constraints are introduced during the training process of generative artificial intelligence models. Taking a variational autoencoder as an example, its latent distribution satisfies:

[0108]

[0109] in, This represents the approximate posterior distribution obtained by the inference network; This represents the target latent distribution, which is a real but not directly obtainable latent distribution.

[0110] Furthermore, during model training, a loss function is defined as the optimization objective:

[0111]

[0112] in, Represents the loss function; Represents the actual parameters obtained from real well logging or experiments; This represents the prediction parameter results obtained from the artificial intelligence model. This represents the mean square error term; Represents the regularization constraint term; This represents the balance factor, used to adjust the weights between the mean squared error term and the regularization constraint term.

[0113] Through the above process, a model with a generative artificial intelligence model as its core was constructed, realizing the mapping from multi-source standardized input feature vectors to preliminary predictions of full-depth reservoir physical and mechanical parameters, providing an input basis for subsequent parameter constraints and fusion completion.

[0114] Step S104: Use the preprocessed logging curve parameters to constrain and correct the predicted parameter results of the artificial intelligence model, and form constrained and corrected reservoir physical and mechanical parameter results.

[0115] The preprocessed logging curve parameters are depth-continuous parameters, which may include one or more of the following: density, sonic transit time, resistivity, and natural gamma curve parameters. These preprocessed logging curve parameters are used as constraints to constrain and correct the predicted parameters of the artificial intelligence model, ensuring the vertical continuity and physical rationality of the reservoir's physical and mechanical parameters.

[0116] Specifically, the preprocessed logging curve parameters are extracted into a constraint set:

[0117]

[0118] in, This represents the constraint set corresponding to the logging curve parameters at a well depth of z, including density. Sound wave time difference resistivity Natural Gamma etc.; z represents the well depth position.

[0119] Next, the preliminary predicted parameters output by the artificial intelligence model are compared with the constraint set to construct a parameter correction formula:

[0120]

[0121] in, This represents the corrected reservoir physical and mechanical parameters at a well depth of z. This represents the predicted parameters at well depth z obtained by the artificial intelligence model; These represent constraint parameters used to control the correction strength. This represents the set of constraints corresponding to the logging curve parameters at a well depth of z. This represents the comparable logging parameters obtained by mapping the predicted parameter results.

[0122] Furthermore, to avoid the excessive influence of abnormal logging points on the confinement effect, a weighted formula based on confidence factors is introduced. The predicted parameter results are fused with the corrected reservoir physical and mechanical parameter results using the confidence factors to obtain the final confinement-corrected reservoir physical and mechanical parameter results.

[0123]

[0124] in, This represents the constrained corrected reservoir physical and mechanical parameters at a well depth of z. The confidence factor at well depth z can be determined based on the weighting mechanism of well logging curve quality evaluation or artificial intelligence model, and its value range is [value range missing]. .

[0125] Through the above-mentioned constraint and weighting process, dynamic correction between the predicted parameters of the generative artificial intelligence model and the actual logging data is achieved, so that the final parameter results maintain the trend of AI prediction at the whole well depth scale, and also have the physical constraints of logging curves.

[0126] Step S105: The preprocessed multi-source data is jointly input into the artificial intelligence model, driving the artificial intelligence model to fuse and complete the results of the reservoir physical and mechanical parameters after constraint correction, forming a parameter field of the entire depth within the reservoir wellbore.

[0127] Based on the constrained and corrected reservoir physical and mechanical parameters, the preprocessed core three-dimensional voxel structure data (such as weighted three-dimensional porosity, pore structure, fracture morphology, etc.), rock cutting microphysical property data (such as two-dimensional porosity obtained by microscopic imaging, microscopic elastic modulus of rock cuttings, etc.) and logging curve parameters are fused and completed to drive the artificial intelligence model to form a parameter field of the entire depth within the reservoir wellbore, that is, a continuous parameter distribution of the entire depth.

[0128] Specifically, the preprocessed multi-source data and the constrained corrected reservoir physical and mechanical parameters are uniformly mapped to the depth coordinate system to construct a multi-source parameter set:

[0129]

[0130] in, Represents a set of multi-source parameters; This represents the weighted three-dimensional porosity at a well depth of z; This represents the two-dimensional porosity at a well depth of z; H represents the microscopic elastic modulus of the rock cuttings; H represents the hardness of the rock cuttings. This indicates the parameters of the completed well logging curve; This represents the constrained corrected reservoir physical and mechanical parameters at a well depth of z.

[0131] Next, the multi-source parameter set is weighted and fused using a fusion function to obtain the fused and completed reservoir physical and mechanical parameter results. A parameter field is formed based on the fused and completed reservoir physical and mechanical parameter results. In other words, this parameter field is the fused and completed parameter field, which records the fused and completed reservoir physical and mechanical parameter results.

[0132] The fusion function is:

[0133]

[0134]

[0135] in, This represents the reservoir physical and mechanical parameters after fusion and completion at a well depth of z; n represents the total number of data components in the multi-source parameter set; Represents a multi-source parameter set The normalized weights corresponding to the i-th data component at a well depth of z satisfy the following conditions: ; Represents a multi-source parameter set The i-th data component at a well depth of z; This represents the basic weighting factor of the i-th data component at a well depth of z; This represents the weight correction factor of the AI ​​model for the i-th data component at a well depth of z during the fusion process, which is used to dynamically adjust the weight based on the consistency or difference between the data. This represents the basic weighting factor of the j-th data component at a well depth of z; This represents the weight correction factor of the AI ​​model for the j-th data component at well depth z during the fusion process.

[0136] The aforementioned fusion function achieves a weighted combination of different data sources. Furthermore, a weight correction factor from an artificial intelligence model is introduced into the weights, followed by normalization processing. This enables the fusion, completion, and optimization of the reservoir physical and mechanical parameters after constraint correction. The resulting parameter field can be used as input for simulation solutions, realizing the transition from data-driven to physical modeling.

[0137] Step S106: Construct a numerical simulation model based on the parameter field, and use the numerical simulation model to simulate the reservoir mechanical response process to obtain simulation results.

[0138] A numerical simulation model is constructed based on the reservoir physical and mechanical parameters obtained by fusion and completion in the parameter field, and the stress, strain and stability distributions are solved across the entire depth range. Figure 3 A schematic diagram of the fusion completion and simulation framework is shown.

[0139] Using the merged and completed reservoir physical and mechanical parameters from the parameter field as input, including porosity, density, microelastic modulus, Poisson's ratio, and fracture development coefficient, a three-dimensional numerical simulation model is established. This numerical simulation model needs to satisfy the static equilibrium equations of continuum mechanics.

[0140]

[0141] in, Represents the stress tensor; Represents the divergence of the stress tensor; This represents the volume force vector, which includes gravity and pore pressure gradient.

[0142] Furthermore, the parameter field obtained after fusion and completion in step S105 is introduced into the constitutive relation, and its formula is:

[0143]

[0144] in, Represents the stress tensor; Represents the strain tensor; This represents the stiffness matrix, whose components are determined by the fused and completed parameters, including the microelastic modulus, Poisson's ratio, and crack correction factor.

[0145] Apply wellbore boundary conditions to reflect actual stress conditions:

[0146]

[0147] in, represents the wellbore radius; z represents the well depth position; This represents the fluid pressure inside the wellbore at a depth of z, typically in the range of 40 to 60 MPa. This represents the radial stress at the well wall at a depth of z. This represents the circumferential stress at the wellbore wall at a depth of z. This represents the pore pressure value obtained from well logging or well testing, with a typical gradient of 1.55 to 1.65 MPa / 100m. This boundary condition can reflect the actual stress state relatively well.

[0148] Wellbore pressure and pore pressure constraints are applied to the boundary conditions, and the above equations are solved by simulation using finite element or finite difference numerical methods. The simulation results include the stress field, displacement field and fracture pressure distribution in the entire depth range of the reservoir, which provides a physical constraint basis for subsequent inversion optimization.

[0149] Step S107: The simulation results are used to invert and iteratively correct the parameter field to obtain the physical and mechanical parameter characterization results of the target full-depth reservoir.

[0150] Based on the differences between simulation results and actual logging and well testing data, an inversion optimization framework is constructed to iteratively correct the parameter field in order to improve prediction accuracy and physical consistency.

[0151] Define the objective function in the inversion optimization framework to measure the difference between simulation results and measured data. The formula is:

[0152]

[0153] in, Represent the objective function; This represents the set of parameters to be optimized; z represents the well depth. Indicates in the parameter set The reservoir physical and mechanical parameters obtained from the simulation; This represents the parameter corresponding to the measured data at well depth z.

[0154] To avoid overfitting or inconsistencies, physical constraints and AI-driven correction operators are introduced to iteratively update the parameters. Specifically, based on the inversion optimization framework and iterative update formula, the parameter field is iteratively updated until a preset convergence condition is met, yielding the physical and mechanical parameter characterization results of the target full-depth reservoir; the iterative update formula is as follows:

[0155]

[0156] in, This represents the parameter set at the (k+1)th iteration; This represents the set of parameters at the k-th iteration. Indicates the learning rate or iteration step size; This represents the gradient of the objective function in the inversion optimization framework with respect to the parameters of the parameter set at the k-th iteration; Indicates the correction factor; This refers to an AI-driven correction operator used to utilize the fused and completed reservoir physical and mechanical parameters. The inversion results are dynamically adjusted to reflect data consistency, physical priors, and constraint information. This represents the reservoir physical and mechanical parameters after fusion and completion at a well depth of z.

[0157] To ensure the stability of the solution, a convergence condition is also set: when the objective function... Less than a preset threshold (e.g., the preset threshold could be 10) -3 The inversion process is considered convergent when the number of iterations reaches the maximum limit (e.g., 1000). Based on actual computational stability criteria, dynamic threshold adjustment or early stopping strategies can also be used to avoid overfitting and ineffective iterations.

[0158] Figure 4 A schematic diagram illustrating the principles of inversion and iterative correction is shown, as follows: Figure 4 As shown, error calculation (i.e., difference calculation) is performed on the simulation results and measured data. Then, the parameters are iteratively updated, and convergence is determined according to preset convergence conditions. When the inversion process is determined to be converged, the final optimized parameters are obtained, which are the physical and mechanical parameter characterization results of the target reservoir at the full depth. Through the inversion optimization process of error calculation—parameter update—convergence determination, the corrected distribution of reservoir physical and mechanical parameters across the entire well depth is obtained, making the simulation results consistent with the measured data and satisfying the predetermined physical constraints.

[0159] The effectiveness of the full-depth reservoir physical and mechanical parameter characterization method proposed in this application will be verified through two specific examples below:

[0160] Example 1:

[0161] Well L1 in Block M was selected as the research object. This well has a depth of 4850m, a four-section wellbore structure, and a reservoir type of sandstone interbedded with carbonate rocks, with relatively well-developed fractures. To achieve high-precision characterization of the full-depth reservoir physical and mechanical parameters, the integrated full-depth reservoir physical and mechanical parameter characterization method driven by generative AI, logging, and simulation proposed in this application was applied. Multi-source data from this well were acquired, processed, modeled, constrained, fused, simulated, and optimized to verify the effectiveness of this application.

[0162] (1) During the data acquisition phase: In well L1, core samples were taken every 300m, and high-resolution X-ray CT scans were used to obtain three-dimensional voxel structure data of the cores. The resolution is 2.5 μm / voxel. To enhance the reliability of the three-dimensional porosity results, the voxel segmentation confidence score output by the generative artificial intelligence model was introduced to weight and correct the three-dimensional porosity in the core three-dimensional voxel structure data. After correction, the three-dimensional porosity of well L1 ranges from 6.8% to 13.2%.

[0163] Rock cuttings samples obtained during drilling were collected, and two-dimensional microscopic imaging and physical property testing were performed to obtain two-dimensional porosity. The microscopic elastic modulus of the rock fragments was obtained through nanoindentation experiments. (range 12.5–21.3 GPa) and hardness H (range 0.62–0.95 GPa).

[0164] Obtain logging parameters for well L1, including density. Sound wave time difference resistivity and natural gamma etc., represented as ：

[0165] in, This represents the set of logging curve parameters at a well depth of z. This represents the density at a well depth of z; This represents the acoustic transit time at a well depth of z; This represents the resistivity at a well depth of z; The natural gamma at well depth z represents the depth of the well. Finally, the weighted three-dimensional porosity and two-dimensional porosity are fused and corrected using a depth-related weighting factor to achieve joint utilization of two-dimensional and three-dimensional data. Through the above multi-source, multi-scale data acquisition and fusion correction, the initial input parameter set for well L1 is formed.

[0166] (2) During the data processing stage:

[0167] The initial input parameter set is normalized, denoised, imputed, and feature extracted to obtain the standardized input feature vector of well L1. Specifically, missing data in the logging curve parameters are imputed, and the imputation formula uses a confidence weighting factor output by the artificial intelligence model based on data reliability. The value ranges from 0.75 to 0.92.

[0168] (3) In the model building phase:

[0169] A generative artificial intelligence model is used to model the standardized input feature vector of the L1 well. A variational autoencoder (VAE) structure is selected, and a balance factor is used in the loss function during model training. The value is 0.15.

[0170] (4) During the parameter constraint phase:

[0171] The predicted parameters of the artificial intelligence model were constrained and corrected using preprocessed well logging curve parameters, involving porosity (7%–14%), density (2.35–2.65 g / cm³), sonic transit time (190–260 μs / m), resistivity (8–35 Ω·m), and natural gamma (75–110 API). The constrained parameters in the parameter correction formula are... The value is 0.65. Further analysis is performed using the confidence factor. (Values ​​range from 0.6 to 0.8) The predicted parameter results are fused with the corrected reservoir physical and mechanical parameter results.

[0172] (5) During the integration and completion phase:

[0173] Constructing a multi-source parameter set for well L1 By using a fusion function to perform weighted fusion of the multi-source parameter set, the reservoir physical and mechanical parameters after fusion and completion are obtained.

[0174] (6) During the simulation solution stage:

[0175] A numerical simulation model was constructed based on the fused and completed reservoir physical and mechanical parameters to solve for the stress, strain, and stability distributions across the entire depth range. The components of the stiffness matrix in the constitutive relation were determined by the fused and completed parameters, including the microelastic modulus (18–28 GPa), Poisson's ratio (0.21–0.29), and fracture correction factor.

[0176] Fluid pressure inside the wellbore in wellbore boundary conditions The pore pressure value is 42 MPa, obtained from well logging or well testing. The average gradient is 1.58 MPa / 100 m. The finite element method was used for simulation to obtain simulation results, which include the stress field, displacement field, and fracture pressure distribution throughout the entire depth of the reservoir.

[0177] (7) In the inversion optimization stage:

[0178] When discrepancies exist between simulation results and measured data, an inversion optimization framework is constructed. Based on this framework and the iterative update formula, the parameter field is iteratively updated until a preset convergence condition is met, yielding the characterization results of the physical and mechanical parameters of the target full-depth reservoir. Specifically, the iterative update formula... =0.05, The preset convergence condition is: when the objective function... The optimization is considered to have converged when the number of iterations reaches 1000.

[0179] (8) Analysis of implementation effect:

[0180] The porosity prediction curves for the entire depth were obtained based on fusion completion and simulation solutions, and compared with measured core data. The results show that the porosity prediction curves generally match the measured values ​​well, and the vast majority of measured values ​​fall within the 95% confidence interval. For easier visualization, Table 1 shows the comparison results of porosity prediction and measurement for well L1, recording the predicted values, measured values, and their upper and lower confidence intervals for some well depth ranges of well L1.

[0181] Table 1. Comparison of predicted and measured porosity of Well L1

[0182]

[0183] As shown in Table 1, the deviation between the predicted porosity and the measured porosity is controlled within ±7%, and the trend remains consistent, verifying the reliability of the fusion completion and simulation closed-loop method across the entire well depth range. Figure 5 This shows a comparison chart between the predicted porosity of well L1 and the measured core data. Figure 5 The figure shows the porosity prediction curve, measured core data points, and uncertainty band of the 95% confidence interval. This figure can intuitively reflect the degree of agreement between the porosity prediction results and the measured core data, as well as the range of uncertainty.

[0184] Based on the statistical results of the full-depth prediction curve, the final range of reservoir physical and mechanical parameters for well L1 is as follows: porosity ranges from 6.5% to 14.1%; microelastic modulus ranges from 17.8 to 28.6 GPa; Poisson's ratio ranges from 0.22 to 0.28; and the deviation between the predicted fracturing pressure and the measured fracturing data is controlled within ±7%.

[0185] Therefore, the full-depth reservoir physical and mechanical parameter characterization method proposed in this application can effectively improve the prediction accuracy of key mechanical parameters such as porosity, microelastic modulus and Poisson's ratio while ensuring the consistency of basic logging parameters such as density.

[0186] Example 2:

[0187] Well T2 in Block T was selected as the research object. This well has a depth of 6200m, a three-section wellbore structure, and is located in a deep carbonate reservoir with low porosity and a well-developed fracture system. To achieve high-precision characterization of the full-depth reservoir physical and mechanical parameters, the integrated full-depth reservoir physical and mechanical parameter characterization method proposed in this application, which combines generative AI, logging, and simulation-driven methods, was systematically analyzed to verify the applicability of this application under different reservoir types.

[0188] (1) During the data acquisition phase: Core samples were taken from well T2 at depths of 2200–2400m, 3800–4000m, and 5200–5400m. High-resolution X-ray CT scanning was used to obtain three-dimensional voxel structure data of the cores. The resolution was 3.0 μm / voxel. To enhance the reliability of the three-dimensional porosity results, the voxel segmentation confidence score output by a generative artificial intelligence model was introduced to weight and correct the three-dimensional porosity in the core three-dimensional voxel structure data. After correction, the three-dimensional porosity of well T2 ranged from 3.5% to 8.9%. Meanwhile, two-dimensional microscopic testing of the rock cuttings samples showed that the two-dimensional porosity... The range is 4.1%-9.3%; microelastic modulus The range of hardness is 25.6–34.2 GPa; the range of hardness H is 1.10–1.42 GPa.

[0189] Obtain logging parameters for well T2, including density. (2.50–2.78 g / cm³), sound wave transit time (170–205 μs / m), resistivity (30–95 Ω·m) and natural gamma (45–70 API), etc. Furthermore, the weighted three-dimensional porosity and two-dimensional porosity are fused and corrected using depth-related weighting factors.

[0190] (2) During the data processing stage:

[0191] The initial input parameter set is normalized, denoised, imputed, and feature extracted to obtain the standardized input feature vector of well T2. Specifically, missing data in the logging curve parameters are imputed, and the imputation formula uses a confidence weighting factor output by the artificial intelligence model based on data reliability. The value ranges from 0.70 to 0.88. The completion results show that the missing segments of the deep resistivity and porosity curves have been reasonably restored.

[0192] (3) In the model building phase:

[0193] A generative artificial intelligence model is used to model the standardized input feature vector of well T2. A diffusion model is selected, and a balance factor is used in the loss function during model training. The value is 0.20.

[0194] (4) During the parameter constraint phase:

[0195] The predicted parameters of the artificial intelligence model are constrained and corrected using preprocessed well logging curve parameters, involving porosity (3%–10%), density (2.55–2.75 g / cm³), sonic transit time (170–210 μs / m), resistivity (25–100 Ω·m), and natural gamma (40–75 API). The constraint parameters in the parameter correction formula are... The value is 0.60. Further analysis is performed using the confidence factor. (Values ​​range from 0.65 to 0.80) The predicted parameter results are fused with the corrected reservoir physical and mechanical parameter results.

[0196] (5) During the integration and completion phase:

[0197] Constructing a multi-source parameter set for well T2 The reservoir physical and mechanical parameters are obtained by weighted fusion of multiple source parameter sets using a fusion function. The normalized weights in the fusion function are used for this purpose. The values ​​range from 0.1 to 0.25. The results show that the continuity of the parameter curve is significantly enhanced after fusion and completion.

[0198] (6) During the simulation solution stage:

[0199] A numerical simulation model was constructed based on the fused and completed reservoir physical and mechanical parameters to solve for the stress, strain, and stability distributions across the entire depth range. The components of the stiffness matrix in the constitutive relation were determined by the fused and completed parameters, including the microelastic modulus (26–34 GPa), Poisson's ratio (0.20–0.26), and fracture correction factor.

[0200] Fluid pressure inside the wellbore in wellbore boundary conditions The pore pressure value is 56 MPa, obtained from well logging or well testing. The average gradient is 1.65 MPa / 100 m. The finite element method was used for simulation to obtain simulation results, which include the stress field, displacement field, and fracture pressure distribution throughout the entire depth of the reservoir.

[0201] (7) In the inversion optimization stage:

[0202] When discrepancies exist between simulation results and measured data, an inversion optimization framework is constructed. Based on this framework and the iterative update formula, the parameter field is iteratively updated until a preset convergence condition is met, yielding the characterization results of the physical and mechanical parameters of the target full-depth reservoir. Specifically, the iterative update formula... =0.04, 2; The preset convergence condition is: when the objective function... The optimization is considered to have converged when the number of iterations reaches 1200.

[0203] (8) Analysis of implementation effect:

[0204] To further verify the effectiveness of the method proposed in this application, ablation experiments were conducted for comparison. The comparison methods included "well logging regression only", "without generative completion", and "without simulation inversion loop closure", and error comparison analysis was performed with the method proposed in this application. The statistical results are shown in Table 2.

[0205] Table 2. Comparison of prediction errors using different methods

[0206]

[0207] As shown in Table 2, the method proposed in this application significantly outperforms the comparative methods in terms of both mean error and standard deviation. Specifically, compared with "well logging regression only", the mean error of the method proposed in this application is reduced by approximately 53.6%; compared with "without generative completion", the mean error of the method proposed in this application is reduced by approximately 41.9%; and compared with "without simulation inversion loop closure", the mean error of the method proposed in this application is reduced by approximately 31.6%.

[0208] Based on the statistical results of the full-depth prediction curve, the final range of reservoir physical and mechanical parameters for well T2 is as follows: porosity ranges from 3.2% to 9.7%; microelastic modulus ranges from 25.8 to 33.9 GPa; Poisson's ratio ranges from 0.21 to 0.25; and the deviation between the predicted fracturing pressure and the measured fracturing data is controlled within ±6%.

[0209] The results show that the method proposed in this application is applicable not only to sandstone interbedded with carbonate rock reservoirs (i.e., Example 1) but also to deep carbonate rock reservoirs (i.e., Example 2), and maintains high consistency and generalizability under different geological backgrounds, demonstrating good adaptability.

[0210] Figure 6 In the form of bar charts, Figure 7 The prediction errors of the different methods described above are compared using box plots. Figure 6 The bar chart visually compares the average error levels of different methods. Figure 7 The box plot shows the error distribution range and the median. Figure 7 The red line in the diagram represents the median of this set of analyzed data; the circles represent outliers, individual samples with particularly large or small errors. The results show that the method proposed in this application not only has the smallest average error but also the highest error distribution concentration and the lowest dispersion, verifying the crucial role of generative AI fusion and simulation closed-loop in characterizing reservoir physical and mechanical parameters.

[0211] The full-depth reservoir physical and mechanical parameter characterization method provided in this application realizes multi-source data fusion, AI-driven simulation inversion closed loop, significantly improving the integrity and reliability of reservoir physical and mechanical parameter characterization, and is applicable to different types of reservoir engineering. Specifically, at the single-well scale, a multi-scale, multi-dimensional data system is established by combining core three-dimensional voxel structure data, cuttings microphysical property data, and logging curve parameters. This system is then fused and completed using an artificial intelligence model, achieving multi-source data fusion and completion, greatly improving parameter integrity, and effectively solving the problem of discontinuous and incomplete parameters caused by traditional methods relying on single logging data. Driven by generative artificial intelligence structures such as diffusion models, variational autoencoders, or generative adversarial networks, the standardized input feature vectors corresponding to multi-source data are mapped to the latent space. Combined with latent distribution constraints and optimized loss functions, the predictive ability is enhanced, realizing generative prediction of reservoir physical and mechanical parameters, overcoming the limitations of existing methods that can only perform simple regression fitting and are difficult to capture complex nonlinear relationships. Utilizing density, acoustic... Using logging parameters such as wave transit time, resistivity, and natural gamma as constraints, and combining them with predicted key mechanical parameters such as porosity, a dynamic fusion of predicted parameter results and actual logging data is achieved through parameter correction formulas and a confidence weighting mechanism. This ensures the physical consistency, rationality, and reliability of the parameters across the entire depth range. In the simulation and inversion optimization stages, the fused and completed reservoir physical and mechanical parameter results are introduced into the constitutive relation, and then the parameters are dynamically corrected through iterative update formulas, forming a closed-loop process of "prediction-simulation-inversion." Compared with existing technologies, this significantly improves the accuracy and adaptability of reservoir physical and mechanical parameter characterization. This scheme has outstanding engineering application value, enabling high-precision characterization of reservoir physical and mechanical parameters across the entire well depth range under complex geological conditions. It can directly serve engineering stages such as drilling design, well completion engineering, and fracturing operations, providing more reliable inputs of key mechanical parameters such as porosity, elastic modulus, and Poisson's ratio for oil and gas development. This effectively reduces the risk of wellbore instability and fracture propagation, demonstrating good applicability and promising prospects for widespread application.

[0212] Figure 8 A structural block diagram of a full-depth reservoir physical and mechanical parameter characterization device according to an embodiment of this application is shown, as follows: Figure 8 As shown, the device includes: a data acquisition module 810, a data processing module 820, a model building module 830, a parameter constraint module 840, a fusion completion module 850, a simulation solution module 860, and an inversion optimization module 870.

[0213] The data acquisition module 810 is suitable for acquiring multi-source data within the reservoir wellbore area. This multi-source data includes core three-dimensional voxel structure data, cuttings microstructure data, and logging parameters.

[0214] The data processing module 820 is suitable for: preprocessing and feature extraction of multi-source data to obtain standardized input feature vectors for modeling.

[0215] The model building module 830 is suitable for: performing preliminary modeling of reservoir physical and mechanical parameters based on standardized input feature vectors to obtain a generative artificial intelligence model.

[0216] The parameter constraint module 840 is suitable for: using the preprocessed logging curve parameters to constrain and correct the predicted parameter results of the artificial intelligence model, forming constrained and corrected reservoir physical and mechanical parameter results.

[0217] The fusion completion module 850 is suitable for: inputting preprocessed multi-source data into an artificial intelligence model, driving the artificial intelligence model to fuse and complete the results of reservoir physical and mechanical parameters after constraint correction, and forming a parameter field of the whole depth within the reservoir wellbore range.

[0218] The simulation solution module 860 is suitable for: constructing a numerical simulation model based on the parameter field, simulating the reservoir mechanical response process using the numerical simulation model, and obtaining simulation results.

[0219] The inversion optimization module 870 is suitable for: using simulation results to invert and iteratively correct the parameter field, and obtaining the characterization results of the physical and mechanical parameters of the target full-depth reservoir.

[0220] Furthermore, the data processing module 820 is further adapted to: normalize the multi-source data using a linear interval mapping method to convert different physical quantities in the multi-source data into a unified interval; denoise the multi-source data; complete the missing data in the well logging curve parameters based on the interpolation of the well logging curve parameters and the confidence weighting factor output by the artificial intelligence model; and perform dimensionality reduction and deep feature extraction on the preprocessed multi-source data by combining principal component analysis and a deep learning-based autoencoder to obtain a standardized input feature vector.

[0221] Furthermore, the completion formula used to complete the missing data in the well logging curve parameters is as follows:

[0222]

[0223] in, This represents the parameters of the completed logging curve at a well depth of z; This represents the preprocessed logging curve parameters at a well depth of z. This represents the confidence weighting factor output by the artificial intelligence model based on the data's credibility. This represents the correction amount corresponding to well depth z, obtained by interpolation based on adjacent well sections.

[0224] Furthermore, the constrained reservoir physical and mechanical parameters are calculated using the following formula:

[0225]

[0226]

[0227] in, This represents the corrected reservoir physical and mechanical parameters at a well depth of z. This represents the predicted parameters at well depth z obtained by the artificial intelligence model; Indicates constraint parameters; This represents the set of constraints corresponding to the logging curve parameters at a well depth of z. This represents the comparable logging parameters obtained by mapping the predicted parameter results; This represents the constrained corrected reservoir physical and mechanical parameters at a well depth of z. This represents the confidence factor at a well depth of z.

[0228] Furthermore, the fusion completion module 850 is further adapted to: uniformly map the preprocessed multi-source data and the constrained corrected reservoir physical and mechanical parameter results to the depth coordinate system to construct a multi-source parameter set; perform weighted fusion of the multi-source parameter set through a fusion function to obtain the fused and completed reservoir physical and mechanical parameter results, and form a parameter field based on the fused and completed reservoir physical and mechanical parameter results; wherein, the fusion function is:

[0229]

[0230]

[0231] in, This represents the reservoir physical and mechanical parameters after fusion and completion at a well depth of z; n represents the total number of data components in the multi-source parameter set; This represents the normalized weight of the i-th data component at well depth z in the multi-source parameter set; This represents the i-th data component at well depth z in the multi-source parameter set; This represents the basic weighting factor of the i-th data component at a well depth of z; This represents the weight correction factor of the AI ​​model for the i-th data component at a well depth of z during the fusion process; This represents the basic weighting factor of the j-th data component at a well depth of z; This represents the weight correction factor of the AI ​​model for the j-th data component at well depth z during the fusion process.

[0232] Furthermore, the inversion optimization module 870 is further adapted to: construct an inversion optimization framework based on the difference between simulation results and measured data; and iteratively update the parameter field according to the inversion optimization framework and the iterative update formula until the preset convergence condition is met, thereby obtaining the physical and mechanical parameter characterization results of the target full-depth reservoir; wherein, the iterative update formula is:

[0233]

[0234] in, This represents the parameter set at the (k+1)th iteration; This represents the set of parameters at the k-th iteration. Indicates the learning rate or iteration step size; This represents the gradient of the objective function in the inversion optimization framework with respect to the parameters of the parameter set at the k-th iteration; Indicates the correction factor; This indicates an AI-driven correction operator used to dynamically adjust the inversion results using the fused and completed reservoir physical and mechanical parameters. This represents the reservoir physical and mechanical parameters after fusion and completion at a well depth of z.

[0235] The full-depth reservoir physical and mechanical parameter characterization device provided in this application embodiment realizes multi-source data fusion, AI-driven simulation inversion closed loop, significantly improving the integrity and reliability of reservoir physical and mechanical parameter characterization, and is applicable to different types of reservoir engineering. Specifically, at the single-well scale, a multi-scale, multi-dimensional data system is established by combining core three-dimensional voxel structure data, cuttings microphysical property data, and logging curve parameters. Through fusion and completion using an artificial intelligence model, multi-source data fusion and completion are achieved, greatly improving parameter integrity and effectively solving the problem of discontinuous and incomplete parameters caused by traditional methods relying on single logging data. Driven by generative artificial intelligence structures such as diffusion models, variational autoencoders, or generative adversarial networks, the standardized input feature vectors corresponding to multi-source data are mapped to the latent space. Combined with latent distribution constraints and optimized loss functions, the predictive ability is enhanced, realizing generative prediction of reservoir physical and mechanical parameters, overcoming the limitations of existing methods that can only perform simple regression fitting and are difficult to capture complex nonlinear relationships. Utilizing density, acoustic... Using logging parameters such as wave transit time, resistivity, and natural gamma as constraints, and combining them with predicted key mechanical parameters such as porosity, a dynamic fusion of predicted parameter results and actual logging data is achieved through parameter correction formulas and a confidence weighting mechanism. This ensures the physical consistency, rationality, and reliability of the parameters across the entire depth range. In the simulation and inversion optimization stages, the fused and completed reservoir physical and mechanical parameter results are introduced into the constitutive relation, and then the parameters are dynamically corrected through iterative update formulas, forming a closed-loop process of "prediction-simulation-inversion." Compared with existing technologies, this significantly improves the accuracy and adaptability of reservoir physical and mechanical parameter characterization. This scheme has outstanding engineering application value, enabling high-precision characterization of reservoir physical and mechanical parameters across the entire well depth range under complex geological conditions. It can directly serve engineering stages such as drilling design, well completion engineering, and fracturing operations, providing more reliable inputs of key mechanical parameters such as porosity, elastic modulus, and Poisson's ratio for oil and gas development. This effectively reduces the risk of wellbore instability and fracture propagation, demonstrating good applicability and promising prospects for widespread application.

[0236] The present invention also provides a non-volatile computer storage medium storing at least one executable instruction that can execute the full-depth reservoir physical and mechanical parameter characterization method in any of the above method embodiments.

[0237] This invention provides a computer program product comprising at least one executable instruction or computer program that enables a processor to perform operations corresponding to the full-depth reservoir physical and mechanical parameter characterization method in any of the above method embodiments.

[0238] Figure 9The diagram shows a structural schematic of a computing device according to one embodiment of the present application. The specific embodiments of the present application do not limit the specific implementation of the computing device.

[0239] like Figure 9 As shown, the computing device may include: a processor 902, a communication interface 904, a memory 906, and a communication bus 908.

[0240] The processor 902, communication interface 904, and memory 906 communicate with each other via communication bus 908. Communication interface 904 is used to communicate with other network elements, such as clients or other servers. Processor 902 executes program 910, specifically performing the relevant steps in the embodiment of the full-depth reservoir physical and mechanical parameter characterization method for computing devices described above.

[0241] Specifically, program 910 may include program code that includes computer operation instructions.

[0242] The processor 902 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application. The computing device includes one or more processors, which may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0243] Memory 906 is used to store program 910. Memory 906 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0244] Specifically, program 910 can be used to cause processor 902 to execute the full-depth reservoir physical and mechanical parameter characterization method in any of the above method embodiments. The specific implementation of each step in program 910 can be found in the corresponding descriptions of the steps and units in the above full-depth reservoir physical and mechanical parameter characterization embodiments, and will not be repeated here. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described equipment and modules can be referred to the corresponding process descriptions in the foregoing method embodiments, and will not be repeated here.

[0245] The algorithms and displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The required structure for constructing such systems is apparent from the above description. Furthermore, this invention is not directed to any particular programming language. It should be understood that the contents of the invention described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing the best mode of implementation of the invention.

[0246] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0247] Similarly, it should be understood that, in order to streamline this disclosure and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof. However, this method of disclosure should not be interpreted as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as reflected in the claims, inventive aspects lie in fewer than all features of a single foregoing disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of the invention.

[0248] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0249] Furthermore, those skilled in the art will understand that although some embodiments described herein include certain features but not others included in other embodiments, combinations of features from different embodiments are intended to be within the scope of the invention and form different embodiments. For example, in the claims, any of the claimed embodiments can be used in any combination.

[0250] The various component embodiments of the present invention can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some or all of the components according to the embodiments of the present invention. The present invention can also be implemented as a device or apparatus program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such programs implementing the present invention can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0251] It should be noted that the above embodiments are illustrative of the invention and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

Claims

1. A method for characterizing the physical and mechanical parameters of a reservoir at all depths, characterized in that, The method includes: Acquire multi-source data within the reservoir wellbore area, including core three-dimensional voxel structure data, cuttings microphysical property data, and logging curve parameters; Multi-source data is preprocessed and feature extracted to obtain a standardized input feature vector for modeling. Based on the standardized input feature vector, the reservoir physical and mechanical parameters are initially modeled to obtain a generative artificial intelligence model. The predicted parameter results of the artificial intelligence model are constrained and corrected using the preprocessed logging curve parameters to form constrained and corrected reservoir physical and mechanical parameter results. The preprocessed multi-source data is jointly input into the artificial intelligence model, which drives the artificial intelligence model to fuse and complete the results of the reservoir physical and mechanical parameters after constraint correction, forming a parameter field of the whole depth within the reservoir wellbore range; A numerical simulation model is constructed based on the parameter field, and the reservoir mechanical response process is simulated using the numerical simulation model to obtain simulation results; The simulation results are used to invert and iteratively correct the parameter field to obtain the physical and mechanical parameter characterization results of the target full-depth reservoir.

2. The method for characterizing the physical and mechanical parameters of a full-depth reservoir according to claim 1, characterized in that, The preprocessing and feature extraction of multi-source data to obtain a standardized input feature vector for modeling further includes: A linear interval mapping method is used to normalize multi-source data, so as to transform different physical quantities in multi-source data into a unified interval; Denoising the multi-source data; For well logging curve parameters in multi-source data, missing data in the well logging curve parameters are filled in based on the interpolation of the well logging curve parameters and the confidence weighting factor output by the artificial intelligence model; By combining principal component analysis with a deep learning-based autoencoder, dimensionality reduction and deep feature extraction are performed on the preprocessed multi-source data to obtain the standardized input feature vector.

3. The method for characterizing the physical and mechanical parameters of a full-depth reservoir according to claim 2, characterized in that, The completion formula used to complete the missing data in the well logging curve parameters is as follows: in, This represents the parameters of the completed logging curve at a well depth of z; This represents the preprocessed logging curve parameters at a well depth of z. This represents the confidence weighting factor output by the artificial intelligence model based on the data's credibility. This represents the correction amount corresponding to well depth z, obtained by interpolation based on adjacent well sections.

4. The method for characterizing the physical and mechanical parameters of a full-depth reservoir according to claim 1, characterized in that, The constrained reservoir physical and mechanical parameters are calculated using the following formula: in, This represents the corrected reservoir physical and mechanical parameters at a well depth of z. This represents the predicted parameter result at well depth z obtained by the artificial intelligence model; Indicates constraint parameters; This represents the set of constraints corresponding to the logging curve parameters at a well depth of z. This represents the comparable logging parameters obtained by mapping the predicted parameter results; This represents the constrained corrected reservoir physical and mechanical parameters at a well depth of z. This represents the confidence factor at a well depth of z.

5. The method for characterizing the physical and mechanical parameters of a full-depth reservoir according to claim 1, characterized in that, The step of jointly inputting the preprocessed multi-source data into the artificial intelligence model to drive the artificial intelligence model to fuse and complete the reservoir physical and mechanical parameter results after constraint correction, forming a parameter field of the entire depth within the reservoir wellbore range, further includes: The preprocessed multi-source data and the constrained corrected reservoir physical and mechanical parameters are uniformly mapped to the depth coordinate system to construct a multi-source parameter set. The multi-source parameter set is weighted and fused using a fusion function to obtain the fused and completed reservoir physical and mechanical parameters. The parameter field is then formed based on these fused and completed reservoir physical and mechanical parameters. The fusion function is: in, This represents the reservoir physical and mechanical parameters after fusion and completion at a well depth of z; n represents the total number of data components in the multi-source parameter set. This represents the normalized weight of the i-th data component at a well depth of z in the multi-source parameter set; This represents the i-th data component at well depth z in the multi-source parameter set; This represents the basic weighting factor of the i-th data component at a well depth of z; This represents the weight correction factor of the AI ​​model for the i-th data component at a well depth of z during the fusion process; This represents the basic weighting factor of the j-th data component at a well depth of z; This represents the weight correction factor of the AI ​​model for the j-th data component at a well depth of z during the fusion process.

6. The method for characterizing the physical and mechanical parameters of a full-depth reservoir according to any one of claims 1-5, characterized in that, The step of using the simulation results to invert and iteratively correct the parameter field to obtain the characterization results of the physical and mechanical parameters of the target full-depth reservoir further includes: Based on the differences between the simulation results and the measured data, an inversion optimization framework is constructed. Based on the aforementioned inversion optimization framework and iterative update formula, the parameter field is iteratively updated until a preset convergence condition is met, thereby obtaining the physical and mechanical parameter characterization results of the target full-depth reservoir; wherein, the iterative update formula is: in, This represents the parameter set at the (k+1)th iteration; This represents the set of parameters at the k-th iteration. Indicates the learning rate or iteration step size; This represents the gradient of the objective function in the inversion optimization framework with respect to the parameters of the parameter set at the k-th iteration; Indicates the correction factor; This indicates an AI-driven correction operator used to dynamically adjust the inversion results using the fused and completed reservoir physical and mechanical parameters. This represents the reservoir physical and mechanical parameters after fusion and completion at a well depth of z.

7. A device for characterizing the physical and mechanical parameters of a reservoir at full depth, characterized in that, The device includes: The data acquisition module is suitable for acquiring multi-source data within the reservoir wellbore area, including core three-dimensional voxel structure data, cuttings microphysical property data, and logging curve parameters. The data processing module is suitable for preprocessing and feature extraction of multi-source data to obtain standardized input feature vectors for modeling. The model building module is adapted to perform preliminary modeling of reservoir physical and mechanical parameters based on the standardized input feature vector to obtain a generative artificial intelligence model; The parameter constraint module is suitable for constraining and correcting the predicted parameter results of the artificial intelligence model using the preprocessed logging curve parameters, so as to form constrained and corrected reservoir physical and mechanical parameter results. The fusion and completion module is suitable for jointly inputting preprocessed multi-source data into the artificial intelligence model, driving the artificial intelligence model to fuse and complete the reservoir physical and mechanical parameter results after constraint correction, forming a parameter field of the whole depth within the reservoir wellbore range; The simulation solution module is suitable for constructing a numerical simulation model based on the parameter field, using the numerical simulation model to simulate the reservoir mechanical response process, and obtaining simulation results. The inversion optimization module is suitable for using the simulation results to invert and iteratively correct the parameter field, so as to obtain the physical and mechanical parameter characterization results of the target full-depth reservoir.

8. A computing device, characterized in that, include: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction, which causes the processor to perform the operation corresponding to the full-depth reservoir physical and mechanical parameter characterization method as described in any one of claims 1-6.

9. A computer storage medium, characterized in that, The computer storage medium stores at least one executable instruction, which causes the processor to perform the operation corresponding to the full-depth reservoir physical and mechanical parameter characterization method as described in any one of claims 1-6.

10. A computer program product, characterized in that, It includes at least one executable instruction that causes the processor to perform the operation corresponding to the full-depth reservoir physical and mechanical parameter characterization method as described in any one of claims 1-6.

Images