A three-dimensional dynamic inversion method of ground stress field based on multi-scale adaptive algorithm

Abstract

This invention relates to the field of geostress field data processing technology, and particularly to a three-dimensional dynamic inversion method for geostress fields based on a multi-scale adaptive algorithm. The method includes the following steps: acquiring a geological dataset of the target region; constructing a three-dimensional initial model of the geostress field based on the geological dataset, and performing spatial partitioning and meshing of the geological bodies to obtain three-dimensional mesh model data of the geostress field; performing multi-scale regional partitioning on the three-dimensional mesh model data of the geostress field, and establishing a multi-scale weighting function to obtain multi-scale partition mapping information; constructing a cross-scale boundary adaptive transfer mechanism, establishing a stress tensor continuity constraint model at the multi-scale partition boundaries, and obtaining cross-scale stress boundary coupling data. This invention achieves high-precision, dynamic, and continuous inversion of geostress fields in complex geological structures through a multi-scale adaptive algorithm and dynamic closed-loop optimization.

Description

Technical Field

[0001] This invention relates to the field of geostress field data processing technology, and in particular to a three-dimensional dynamic inversion method for geostress fields based on a multi-scale adaptive algorithm. Background Technology

[0002] In fields such as geological engineering, geotechnical mechanics, and energy development, accurate inversion of the geostress field is crucial for engineering safety assessment, fracture prediction, and fracturing optimization. Traditional geostress inversion methods are mainly based on single-scale theory and homogeneity assumptions, relying on limited measurement data and simplified models for inversion. While achieving quantitative calculation of the stress field to some extent, they suffer from poor adaptability to complex geological structures, local distortion of inversion results, and limited spatial resolution. Currently, multi-scale geostress modeling methods show certain advantages in dealing with formation heterogeneity and abrupt stress gradient changes, but significant technical bottlenecks remain in areas such as scale boundary continuity processing, initial parameter setting, and inversion convergence efficiency. On the one hand, traditional methods often use fixed-scale grids or manual region partitioning, leading to significant human intervention in the inversion results and prominent boundary discontinuities. On the other hand, existing methods lack dynamic adjustment mechanisms, making it difficult to optimize the model in real time based on residual feedback, resulting in a lack of adaptability and low iterative efficiency in the inversion process. Summary of the Invention

[0003] Therefore, it is necessary for the present invention to provide a three-dimensional dynamic inversion method for geostress field based on a multi-scale adaptive algorithm to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a three-dimensional dynamic inversion method for geostress field based on a multi-scale adaptive algorithm is proposed, comprising the following steps:

[0005] Step S1: Obtain the geological dataset of the target area; construct a three-dimensional initial model of the geostress field based on the geological dataset, and perform spatial division and meshing of the geological body to obtain the three-dimensional mesh model data of the geostress field;

[0006] Step S2: Perform multi-scale regional division on the three-dimensional mesh model data of the geostress field and establish a multi-scale weight function to obtain multi-scale partition mapping information; construct a cross-scale boundary adaptive transfer mechanism, establish a stress tensor continuity constraint model at the multi-scale partition boundary, and obtain cross-scale stress boundary coupling data;

[0007] Step S3: Based on multi-scale partitioning mapping information and cross-scale stress boundary coupling data, set adaptive initial inversion parameters for each scale region, construct a multi-scale adaptive inversion objective function, and perform joint inversion to obtain preliminary three-dimensional distribution results of the geostress field;

[0008] Step S4: Perform dynamic error assessment and residual distribution analysis on the preliminary three-dimensional distribution results of the geostress field, identify local anomalous areas, and perform high-resolution densification inversion on local areas based on the multi-scale adaptive inversion objective function to obtain optimized three-dimensional dynamic distribution results of the geostress field;

[0009] Step S5: Obtain actual microseismic response data; perform coupled comparative analysis between the optimized three-dimensional dynamic distribution results of the geostress field and the actual microseismic response data to generate a dynamic fit evaluation index; if the fit is lower than the preset threshold, update the multi-scale weight function and inversion parameters, and iteratively execute steps S2 to S5 until the convergence condition is met to obtain the final three-dimensional dynamic inversion data of the geostress field.

[0010] This invention improves the model's adaptability and resolution to complex geological structures by introducing a multi-scale region partitioning and weight adjustment mechanism; it constructs a cross-scale boundary stress tensor continuity constraint model to achieve physical consistency of multi-scale inversion boundaries and stability of parameter transfer; it adopts an adaptive initial parameter setting and dynamic error feedback mechanism to enhance the convergence efficiency and model adjustment capability of the inversion process; combined with a residual-driven local densification inversion strategy, it can perform high-resolution optimization of anomalous areas, significantly improving the overall inversion accuracy; finally, by introducing actual microseismic response data for fitting degree verification and closed-loop iterative optimization, it ensures that the inversion results have high geological consistency and physical credibility, thereby achieving refined, dynamic, and highly reliable inversion of the geostress field in complex tectonic areas. Attached Figure Description

[0011] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0012] Figure 1 This is a schematic diagram of the steps of a three-dimensional dynamic inversion method for geostress field based on a multi-scale adaptive algorithm according to the present invention.

[0013] Figure 2 for Figure 1 A detailed flowchart of step S1;

[0014] Figure 3 for Figure 1 A detailed flowchart of step S2;

[0015] Figure 4 This is a navigation diagram illustrating the implementation process of an embodiment of the present invention. Detailed Implementation

[0016] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0017] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0018] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0019] To achieve the above objectives, please refer to Figures 1 to 4 This invention provides a three-dimensional dynamic inversion method for geostress field based on a multi-scale adaptive algorithm, the method comprising the following steps:

[0020] Step S1: Obtain the geological dataset of the target area; construct a three-dimensional initial model of the geostress field based on the geological dataset, and perform spatial division and meshing of the geological body to obtain the three-dimensional mesh model data of the geostress field;

[0021] Step S2: Perform multi-scale regional division on the three-dimensional mesh model data of the geostress field and establish a multi-scale weight function to obtain multi-scale partition mapping information; construct a cross-scale boundary adaptive transfer mechanism, establish a stress tensor continuity constraint model at the multi-scale partition boundary, and obtain cross-scale stress boundary coupling data;

[0022] Step S3: Based on multi-scale partitioning mapping information and cross-scale stress boundary coupling data, set adaptive initial inversion parameters for each scale region, construct a multi-scale adaptive inversion objective function, and perform joint inversion to obtain preliminary three-dimensional distribution results of the geostress field;

[0023] Step S4: Perform dynamic error assessment and residual distribution analysis on the preliminary three-dimensional distribution results of the geostress field, identify local anomalous areas, and perform high-resolution densification inversion on local areas based on the multi-scale adaptive inversion objective function to obtain optimized three-dimensional dynamic distribution results of the geostress field;

[0024] Step S5: Obtain actual microseismic response data; perform coupled comparative analysis between the optimized three-dimensional dynamic distribution results of the geostress field and the actual microseismic response data to generate a dynamic fit evaluation index; if the fit is lower than the preset threshold, update the multi-scale weight function and inversion parameters, and iteratively execute steps S2 to S5 until the convergence condition is met to obtain the final three-dimensional dynamic inversion data of the geostress field.

[0025] Preferably, step S1 includes the following steps:

[0026] Step S11: Obtain drilling data, well logging data, seismic profiles, geological mapping and structural interpretation data of the target area to obtain the geological dataset of the target area;

[0027] Step S12: Extract information on major faults, stratigraphic interfaces and lithological boundaries based on the geological dataset, establish a tectonic boundary framework, and obtain a geological structure framework model;

[0028] Step S13: Spatial division of different lithological units in the geological structure framework model to obtain spatial partitioning data of geological bodies;

[0029] Step S14: Perform irregular spatial grid partitioning based on the geological body spatial partitioning data to obtain a three-dimensional initial grid partitioning model of the geostress field;

[0030] Step S15: Apply boundary conditions and initial stress state to the three-dimensional initial mesh model to establish a three-dimensional initial model of the geostress field;

[0031] Step S16: Bind and integrate the three-dimensional initial model of the geostress field with the three-dimensional initial mesh partitioning model to generate the three-dimensional mesh model data of the geostress field.

[0032] In this embodiment of the invention, firstly, standardized drilling positions are set up in the target area, using boreholes with a diameter not less than [missing information]. A 91mm diamond coring method was used to obtain continuous core samples within a depth range of 0–2000m, along with corresponding drilling columnar sections and lithological comparison data. Simultaneously, logging data obtained through a combined direct wave and refracted wave processing method, including formation P-wave velocity, density curves, and elastic modulus data, was acquired. The logging depth was no less than 1000m, and the well spacing was less than 200m to ensure geological information resolution. Furthermore, using two-dimensional seismic reflection profile data, a structural boundary enhancement technique based on amplitude attributes was employed to extract the main fault location, stratigraphic unconformity, and sedimentary bedding characteristics. Integrating 1:5000 geological mapping results and structural interpretation data, a structural boundary framework was established with fault strike, dip angle, and extension depth accuracy controlled within ±5°, generating a three-dimensional geological structural framework model. Within this framework, based on the differences in the physical and mechanical parameters of lithological units, the model was divided into no fewer than 10 lithological units. The method first generates spatial partitioning data of geological bodies using a spatial partitioning method based on the principle of geometric consistency. Then, the Delaunay tetrahedral partitioning method for irregular three-dimensional structures is used to partition the spatial partitioning data of geological bodies, generating a three-dimensional initial mesh partitioning model of the geostress field with the side length of the partitioning unit controlled within the range of 10m to 50m. After comparing the mesh structure with the actual geological boundary, linear stress gradient boundary conditions are applied according to the principal direction of regional tectonic stress, with the boundary tensor value set to the range of 0 to 20MPa, and the initial stress state is constructed according to the geological time scale. On this basis, the structured data integration interface is used to bind the initial stress state results with the three-dimensional mesh partitioning data at the unit level, and the node number consistency verification method is used to ensure the model coupling accuracy. Finally, a three-dimensional mesh model data of the geostress field with a one-to-one correspondence between nodes, units, and stress tensors is generated.

[0033] This invention fully integrates heterogeneous data from multiple sources, including drilling, logging, seismic, and geological interpretation, comprehensively improving the integrity and accuracy of geological information. It enables precise extraction and structural modeling of key structural elements and lithological boundaries, helping to realistically reflect the complexity of underground geological structures. Through fine spatial partitioning of different lithological units, it enhances the model's ability to distinguish heterogeneous geological bodies. The use of irregular spatial grid partitioning allows the model to maintain the true geometric shape of the geological structure while improving the adaptability and computational efficiency of numerical simulation. Physical constraints on initial stress states and boundary conditions are established to ensure a reasonable initial mechanical basis for subsequent inversion processes. Finally, through model integration operations, it achieves the organic unity of structural information and stress fields, providing solid data support and a physical foundation for multi-scale, high-precision geostress inversion.

[0034] Preferably, step S2 includes the following steps:

[0035] Step S21: Perform cluster analysis on the spatial characteristic scales in the three-dimensional mesh model data of the geostress field to obtain the scale sensitivity classification results;

[0036] Step S22: Based on the scale-sensitive classification results, the three-dimensional grid model data of the geostress field is divided into regions to obtain preliminary results of multi-scale partitioning;

[0037] Step S23: Based on the preliminary results of multi-scale partitioning, construct the spatial mapping relationship between multi-scale partitioning and the original model to obtain multi-scale partitioning mapping information;

[0038] Step S24: Construct a cross-scale boundary adaptive transfer mechanism based on multi-scale partition mapping information, establish a stress tensor continuity constraint model at the multi-scale partition boundary, and obtain cross-scale stress boundary coupling data.

[0039] In this embodiment of the invention, firstly, based on the obtained three-dimensional grid model data of the geostress field, a joint feature description method based on the center point position and side length of the grid cell is used to extract spatial scale feature vectors. Scale sensitivity analysis is then performed using hierarchical clustering based on Euclidean distance, with a clustering threshold of 0.15 and a maximum number of classes limited to 6. This yields the scale sensitivity classification label for each grid cell, i.e., the scale sensitivity classification result. Subsequently, grid cells with the same label are merged according to their spatial adjacency. A region extraction method based on spatial connectivity is used to remove areas smaller than 100. For discrete regions with fewer than 20 elements, preliminary results of multi-scale partitioning are generated. Based on this, for each multi-scale region element, its corresponding original 3D mesh model number, node coordinates, and topological relationships are recorded to construct a mapping index table from the multi-scale region to the original model. This index table includes the region number, starting element number, node number mapping relationship, and boundary element identifier field, forming multi-scale partitioning mapping information. At the boundary of the multi-scale partition, the set of node pairs on the shared mesh surface between adjacent regions is extracted, and the validity is screened based on the criteria of surface normal consistency and element volume difference ratio less than 0.2 to obtain the cross-scale contact boundary node set. Furthermore, based on this set and mapping information, the tensor projection direction and adjacent path length between each pair of boundary nodes are defined. The stress weight adjustment coefficient of each pair of nodes is calculated using a normalized weighting method, and a stress weight adjustment parameter set is constructed. Based on this parameter set and the relationship between the tensor principal axis directions at the boundary nodes, the tensor transition function expression between nodes is derived using a tensor interpolation function, thereby establishing a stress tensor continuity constraint formula. After discretizing the constraint formula, a stress tensor coordination matrix is ​​formed, and this matrix is ​​projected onto the boundary data structure of the original three-dimensional mesh model of the geostress field to realize the encoding and embedding of cross-scale coupled tensor data, and finally generate cross-scale stress boundary coupled data.

[0040] This invention automatically identifies scale differences in the geostress field by performing cluster analysis on spatial characteristic scales, achieving objectivity and intelligence in the regional division process and avoiding errors and subjectivity caused by manual division. Combining scale sensitivity results with multi-scale regional division enhances the model's adaptability and expressive ability to tectonic units of different scales. By constructing a spatial mapping relationship between the regional divisions and the original model, consistency and coordination between regional division and the overall geological structure are ensured. Furthermore, the introduction of a cross-scale boundary adaptive transfer mechanism and a stress tensor continuity constraint model not only effectively eliminates physical discontinuities at multi-scale boundaries but also ensures coordinated transmission of stress information between units of different scales, thus providing key support and physical consistency guarantees for high-precision, multi-scale linked geostress field inversion.

[0041] Preferably, step S24 includes the following steps:

[0042] Step S241: Construct a scale response weight function based on multi-scale partitioning mapping information to obtain multi-scale stress weight distribution data;

[0043] Step S242: Pair up and identify the boundary grid nodes of each region in the multi-scale partitioning mapping information to obtain a cross-scale contact boundary node set;

[0044] Step S243: Construct a cross-scale boundary adaptive transfer mechanism based on the cross-scale contact boundary node set and multi-scale stress weight distribution data to obtain the stress weight adjustment parameter set;

[0045] Step S244: Construct a stress tensor continuity constraint model based on the stress weight adjustment parameter group and the preset boundary node tensor state to obtain the multi-scale boundary stress tensor coordination matrix.

[0046] Step S245: Project the multi-scale boundary stress tensor coordination matrix onto the boundary data structure of the three-dimensional mesh model data of the geostress field to generate cross-scale stress boundary coupling data.

[0047] In this embodiment of the invention, firstly, based on the established multi-scale partitioning mapping information, the spatial coordinate range, grid density, cell volume, and regional structural complexity indices corresponding to each scale region are extracted. An exponential function combination is then used to define the scale response weight function, where the average cell volume within the region is denoted as... Node density is The function is in the form of ,in =1.0、 =0.02、 =1.5, the weight result is assigned to each grid node in the region to generate multi-scale stress weight distribution data; then, the set of grid nodes identified as region boundaries in the multi-scale mapping information is paired according to adjacency relationship and surface normal consistency. A face-by-face screening method is used to identify node pairs located at the boundary of different scale regions, where the Euclidean distance between nodes is less than 1.5 times the average element side length, to construct a cross-scale contact boundary node set; further, based on this set and the stress weight distribution data, the weight value of the element in each paired node is extracted, and the initial stress transfer coefficient is calculated using the two-node weighted average formula to form an unnormalized transfer parameter matrix, which is then combined with the boundary node adjacency... Based on the relational topology information, a summation normalization process is performed to generate a stress weight adjustment parameter set. Then, the preset tensor states of each boundary node are referenced. These states are given by static boundary load conditions and principal stress direction constraints. A linear tensor interpolation formula is used to construct the tensor continuity expression between nodes, and the expressions of all paired nodes are matrixed to assemble a multi-scale boundary stress tensor coordination matrix. Finally, the coordination matrix is ​​projected onto the boundary node subset of the geostress field 3D mesh model data structure according to the region index and node index rules. A sparse matrix interpolation slot structure is used to complete the tensor data embedding operation, and a unified numbering and encoding standard is adopted to generate cross-scale stress boundary coupling data.

[0048] This invention enhances the model's adaptability to geological structural complexity by constructing a scale response weight function to adjust the weights of stress information and model its response characteristics across different scales. It employs a boundary node pairing and identification mechanism to accurately extract cross-scale contact interfaces, improving the completeness and accuracy of boundary processing. During boundary information integration, a stress weight adjustment parameter set is introduced to achieve dynamic balance and coordinated adjustment of stress information across multiple scales. By constructing a stress tensor continuity constraint model, the numerical and physical continuity of the stress tensor at the boundary is ensured, eliminating the common boundary tensor abrupt change problem in traditional methods. Finally, the coordinated tensor information is effectively embedded into the original 3D mesh data structure, guaranteeing the physical consistency and numerical stability of the entire model under multi-scale coupling conditions, providing key boundary condition support for subsequent high-precision inversion.

[0049] Preferably, step S243 includes the following steps:

[0050] Step S2431: Perform topological adjacency analysis on the cross-scale contact boundary node set to obtain boundary node adjacency structure data;

[0051] Step S2432: Extract cross-scale mapping path sequences based on boundary node adjacency structure data and multi-scale partitioning mapping information to obtain a path index dataset;

[0052] Step S2433: Calculate the initial stress transfer coefficient based on the path index dataset and multi-scale stress weight distribution data to obtain the path transfer coefficient matrix;

[0053] Step S2434: Perform normalization adjustment on the path transfer coefficient matrix to obtain normalized weight distribution data;

[0054] Step S2435: Generate a stress weight adjustment parameter set based on the normalized weight distribution data and the cross-scale contact boundary node set.

[0055] In this embodiment of the invention, firstly, a grid connection table retrieval method based on node index numbers is used to perform topological adjacency analysis on the cross-scale contact boundary node set. Shared nodes on the same grid surface are identified through a node-by-node traversal, and an adjacency dictionary structure indexed by nodes is constructed. The adjacency criterion is that the Euclidean distance is less than 0.75 times the average side length of the cell. Boundary node adjacency structure data is output, recording the first-order adjacency information between nodes. Subsequently, based on the obtained adjacency structure data and multi-scale partition mapping information, all continuous mapping paths composed of boundary nodes are retrieved along the boundary direction using a breadth-first traversal method. Node sequence numbers are recorded in the order of path start, end, and intermediate jump points. The multi-scale region sequence traversed by each path is extracted, generating a cross-scale mapping path sequence and constructing a path index dataset. Next, based on the path index dataset and the multi-scale stress weight distribution data associated with each node in the corresponding path, the initial stress transfer coefficient is calculated using the weighted geometric average method. Let the first stress transfer coefficient be... There are in the path There are nodes, with corresponding weights of . The path transfer coefficient is defined as follows: The path coefficients are then used to construct a path transfer coefficient matrix. This matrix is ​​then normalized using a row-wise normalization method, where the transfer coefficient of each path is divided by the sum of all path transfer coefficients to ensure that the normalized value of each path coefficient falls within the specified interval. The data is then processed to generate normalized weight distribution data, with the sum set to 1. Finally, a one-to-one correspondence is established between the normalized weight distribution data and the cross-scale contact boundary node set. The proportion of each pair of nodes in the path is extracted, and its normalized weight value is combined to form a stress influence proportion factor, which is defined as the stress weight adjustment coefficient. The stress weight adjustment coefficients of all node pairs are then summarized to generate a stress weight adjustment parameter set.

[0056] This invention analyzes the topological adjacency relationships of contact boundary nodes, accurately reflecting the structural connectivity and spatial adjacency characteristics between boundary nodes, providing a topological foundation for constructing continuous stress transfer paths. By combining multi-scale mapping relationships to extract mapping path sequences, it achieves a systematic expression of stress transfer paths across different scale units. The path transfer coefficient matrix is ​​used to quantify the stress transfer capacity of each path, enhancing the model's ability to describe the stress transfer behavior of complex boundaries. Normalization processing improves the numerical stability and physical interpretability of the weight distribution data, effectively avoiding the transfer imbalance problem caused by excessively large weight ranges between paths. The final generated stress weight adjustment parameter set can serve as the core basis for fine-tuning the cross-scale boundary stress distribution, providing crucial support for achieving boundary continuity, adaptive adjustment, and multi-scale coordination in geostress field inversion.

[0057] Preferably, step S3 includes the following steps:

[0058] Step S31: Perform regional index parsing on the multi-scale partitioning mapping information to obtain the structural configuration data of each scale region;

[0059] Step S32: Based on the structural configuration data, set adaptive initial inversion parameters for each scale region to obtain inversion parameter setting data;

[0060] Step S33: Construct a multi-scale adaptive inversion objective function based on the inversion parameter setting data and the cross-scale stress boundary coupling data to obtain the inversion objective function model;

[0061] Step S34: Perform joint inversion solution on the inversion objective function model to obtain preliminary three-dimensional distribution results of the geostress field.

[0062] In this embodiment of the invention, firstly, region index parsing processing is performed on the multi-scale partitioning mapping information. The region number, grid node set, cell number set, and their corresponding topological relationships recorded in the multi-scale mapping index table are called. By traversing the index structure, structural configuration data corresponding to each scale region is generated. This configuration data includes the region's spatial location, number of cells, average cell volume, number of boundary nodes, scale level, and boundary connection information with adjacent regions. Subsequently, adaptive initial inversion parameters are set for each scale region based on the structural configuration data, wherein the cell volume is greater than 200. The region is set with a low-resolution parameter group, and the inversion step size is set to [value]. The stress update threshold is MPa, unit volume less than 100 The region is set with a high-resolution parameter set, and the inversion step size is set to [value]. The stress update threshold is At the region boundary, cross-scale weight adjustment parameters are introduced as constraint factors based on adjacency information. Inversion parameter setting data, including step size, constraint coefficients, update criteria, and upper limit of iteration count, is constructed for each region. Then, the above inversion parameter setting data is fused with the cross-scale stress boundary coupling data generated in step S24. An energy functional expression minimizing tensor differences is established for each scale region. Tensor continuity constraint terms, boundary compatibility terms, and scale adaptation terms are introduced to form a multi-scale adaptive inversion objective function, which is uniformly written into the objective function database as the inversion objective function model. Finally, a joint inversion solution operation is performed on this objective function model. A block-based iterative solution process is adopted, splitting the objective function corresponding to each scale region by region. The tensor distribution values ​​of all regions are jointly solved using the residual update method. The maximum number of iterations is set to 100 or the overall tensor difference is less than [a certain value]. MPa is used as the stopping criterion. The set of tensor values ​​of all elements in the region is output, and tensor reconstruction operation is performed to generate preliminary three-dimensional distribution results of the geostress field.

[0063] This invention uses regional indexing to accurately grasp the structural characteristics of regions at various scales, providing a scientific basis for parameter setting. Based on the dynamic adjustment of adaptive initial inversion parameters, it improves the adaptability of the inversion model to regions at different scales, enhancing the flexibility and accuracy of the inversion process. The construction of the multi-scale adaptive inversion objective function integrates cross-scale boundary coupling information, ensuring the continuity and physical consistency of the model across spatial scales. The joint inversion solution method effectively integrates multi-scale information, achieving synergistic optimization of the inversion process, significantly improving the accuracy and stability of the three-dimensional distribution results of the geostress field, and providing reliable technical support for stress field inversion under complex geological conditions.

[0064] Preferably, step S34 includes the following steps:

[0065] Step S341: Perform model dimension expansion on the inversion objective function model to obtain inversion dimension configuration data for each scale region;

[0066] Step S342: Set the joint inversion parameter set based on the inversion dimension configuration data of each scale region to obtain the initial parameters of the joint inversion;

[0067] Step S343: Construct a joint optimization solution process based on the joint inversion initial parameters to obtain the joint inversion process configuration data;

[0068] Step S344: Perform multiple rounds of numerical optimization iterations on the configuration data of the joint inversion process to obtain a multi-scale stress response numerical result set;

[0069] Step S345: Reconstruct the three-dimensional geostress tensor distribution based on the multi-scale stress response numerical result set to obtain the preliminary three-dimensional distribution results of the geostress field.

[0070] In this embodiment of the invention, firstly, the inversion objective function model is subjected to dimensional expansion processing. The entire geostress field is divided into multiple independent scale regions using a region block numbering method. For each region, the number of nodes, the number of tensor components, the proportion of boundary nodes, and the number of stress coupling constraint terms are extracted. Inversion dimension configuration data is formed using the region number as an index. This configuration data explicitly specifies the stress tensor solution dimension, the number of constraint conditions, and the computation block size for each region. Based on this dimension configuration data, a joint inversion parameter set is set. The solution variables are set according to the independence of the tensor distribution characteristics in the three principal stress directions, and the tensor update step size is defined as... MPa, constraint penalty coefficient is 0.8, tensor boundary pass factor is 1.2, and upper limit of error tolerance is set to MPa, and uniformly written into the joint inversion initial parameter set; then construct the joint optimization solution process, and perform local iteration and boundary cooperative transfer processing in the order of region number. In each iteration, the stability of the local solution is judged by the least square residual convergence criterion, and the cross-scale contact node pairs are found according to the boundary node number. The boundary node tensor values ​​are corrected by weighted averaging, and the joint inversion process configuration data is generated, including the iteration sequence table, boundary cooperative transfer path, residual evaluation standard and intermediate result buffer configuration; based on the configuration data, multiple rounds of numerical optimization iteration are performed, using a combination of synchronous update per region and staggered update across regions. In each round of iteration, the stress response of each region is internally balanced and adjusted, with the boundary tensor consistency error being less than 0.5%. MPa is used as the termination standard for each cycle, and the stress response result set corresponding to each scale region is finally output. The stress response numerical result set is input into the three-dimensional space reconstruction module. The complete stress tensor data of each grid node is restored by node interpolation. The tensor reconstruction process adopts the tensor principal value direction coincidence interpolation method to ensure the physical consistency between the tensor components. The tensor is written into the three-dimensional data structure in sequence according to the grid number to complete the preliminary tensor reconstruction of the three-dimensional distribution of the entire geostress field and obtain the preliminary three-dimensional distribution result of the geostress field.

[0071] This invention expands the inversion objective function dimensionally, dividing the inversion tasks into regions of different scales, thus improving the targeting and rationality of parameter configuration. The joint setting and process construction of the inversion parameter set achieves system integration and collaborative optimization of multi-scale information, enhancing the overall consistency of the inversion process. Multiple rounds of numerical optimization iterations effectively improve the convergence speed and solution stability of the inversion model, reducing the impact of local extremum traps. Finally, the reconstruction of the three-dimensional geostress tensor based on the numerical response results ensures the spatial continuity and physical authenticity of the inversion results, significantly improving the accuracy and reliability of the three-dimensional distribution of the geostress field, and providing solid technical support for geostress assessment under complex geological conditions.

[0072] Preferably, step S4 includes the following steps:

[0073] Step S41: Calculate the global error field based on the preliminary three-dimensional distribution results of the geostress field to obtain error distribution data;

[0074] Step S42: Perform residual statistics and spatial fitting analysis on regions of each scale based on error distribution data to obtain residual distribution information;

[0075] Step S43: Perform anomaly detection and clustering on the residual distribution information to obtain local anomaly region identification data;

[0076] Step S44: Extract the corresponding local geostress field three-dimensional mesh model based on the local anomaly area identification data to obtain local area model data;

[0077] Step S45: Refine the mesh and reset the boundary conditions of the local region model data to obtain a high-resolution inversion input model;

[0078] Step S46: Perform local densification inversion based on the high-resolution inversion input model and the multi-scale adaptive inversion objective function to obtain optimized local distribution data of the geostress field;

[0079] Step S47: Perform weighted fusion of the optimized local distribution data of the geostress field and the preliminary three-dimensional distribution results of the geostress field to obtain the optimized three-dimensional dynamic distribution results of the geostress field.

[0080] In this embodiment of the invention, firstly, a global error field calculation is performed on the preliminary three-dimensional distribution results of the geostress field. The tensor difference between the node-level stress tensor and the boundary continuity constraint is used as the basis for error estimation. By constructing the error tensor norm function and calculating the error modulus on all grid nodes, three-dimensional error distribution data is formed, with its value range limited to 0 to 5 MPa and a resolution of one level per 10m grid. Subsequently, based on the error distribution data, residual statistics are performed within the region according to the multi-scale regional index, extracting the maximum residual value, mean, standard deviation, and residual direction distribution characteristics. A spatial fitting method is used to construct residual isosurfaces for each region and analyze their distribution trends, forming residual distribution information containing numerical, location, and structural information. Further, anomaly detection and cluster identification are performed on this residual information. High residual dense regions are extracted using a density threshold-based discrimination method, with the clustering threshold set to 2 MPa and the minimum clustering volume to be 100. The process involves identifying regions of localized stress incompatibility, marking their spatial indices and boundary element nodes, and outputting local anomaly region identification data. Based on this identification data, relevant elements and their connection structures are extracted from the corresponding 3D geostress field mesh to construct local region model data. The extracted range is then expanded by 10% to ensure boundary continuity. After obtaining the local region model data, the mesh is refined using tetrahedral subdivision, reducing the element volume to 1 / 4 of its original volume while controlling the mesh subdivision error to within 1%. Simultaneously, tensor boundary constraints are reapplied at the region boundaries, with tensor boundary values ​​linearly interpolated based on the average stress values ​​of adjacent regions, generating a high-resolution inversion input model. Based on this model and the constructed multi-scale adaptive inversion objective function, the inversion iteration cycle is limited to 50 cycles. The local region inversion calculation is performed using a tensor residual update method, outputting optimized local geostress field distribution data with tensor accuracy controlled within a specified range. Within MPa; finally, the optimized local distribution data of the geostress field and the original preliminary three-dimensional distribution results of the geostress field are weighted and fused. The fusion weight is determined based on the residual standard deviation in the local area. The higher the standard deviation, the greater the weight of the optimization result. The fusion method adopts the form of tensor weighted superposition, and the node tensor components are weighted and averaged. The data is written into a unified three-dimensional spatial data structure to finally generate the optimized three-dimensional dynamic distribution results of the geostress field.

[0081] This invention achieves accurate identification and quantification of error distribution in inversion results by comprehensively calculating the error field and combining residual statistics and spatial fitting analysis. It employs anomaly detection and clustering identification methods to accurately locate local anomaly regions, providing a scientific basis for targeted optimization. Mesh refinement and boundary condition resetting of the anomaly region model significantly improve the spatial resolution and physical realism of the local model. Local densified inversion is performed based on a high-resolution input model, effectively correcting local inversion errors and improving the model's ability to depict details. Finally, weighted fusion optimization of local and global inversion results ensures the continuity and stability of the three-dimensional dynamic distribution of the geostress field, achieving high-precision, dynamically adaptable geostress inversion and providing a solid guarantee for engineering applications in complex geological environments.

[0082] Preferably, step S5 involves coupling and comparing the optimized three-dimensional dynamic distribution of the geostress field with the actual microseismic response data to generate a dynamic fit evaluation index, including the following steps:

[0083] The actual microseismic response data were filtered and denoised, and then resampled using time windows to obtain standardized microseismic response data.

[0084] A spatial distribution model of microseismic response is constructed based on standardized microseismic response data to obtain microseismic response field data;

[0085] The optimized three-dimensional dynamic distribution of the geostress field is spatially coupled and compared with the microseismic response field data to obtain the coupled residual distribution results.

[0086] Multi-scale weighted statistical analysis was performed on the coupled residual distribution results to obtain the dynamic fit evaluation index.

[0087] In this embodiment of the invention, firstly, raw microseismic waveform data acquired from the same three-component seismic detector array within the target area over the past three months are collected. The sampling frequency is 1000 Hz, and the duration is no less than 60 seconds. A bandpass filter is used to filter the microseismic waveform data in the 20-250 Hz frequency band. A smoothing window function is then used to divide the signal into time windows and resample the signal in 10-second increments. The processed data is then mean-normalized to obtain unitless standardized microseismic response data. Secondly, based on the maximum amplitude value and trigger time of each measuring point in the standardized microseismic response data, the spatial coordinates and response intensity information of the microseismic event are extracted. A three-dimensional regular grid is constructed by combining the spatial location of the detectors. At each grid node, the microseismic response intensity value is calculated using a three-dimensional inverse distance weighted interpolation method, generating a spatial resolution of 20... The microseismic response field data of m is then used. Subsequently, the optimized three-dimensional dynamic distribution results of the geostress field are interpolated using the same spatial grid to ensure that the two sets of data are consistent in spatial resolution and grid structure. The squared difference between the microseismic response intensity value and the principal stress modulus of the geostress tensor is calculated at each corresponding grid node. The squared difference results of all grid nodes are summarized to obtain the coupling residual distribution results in the three-dimensional spatial domain. Then, the coupling residual distribution results are spatially partitioned and statistically processed according to the geological structure partitioning and scale sensitivity classification results. The mean, standard deviation and coefficient of variation of the residual values ​​in each partition are calculated respectively. The statistical indicators of each partition are weighted and accumulated according to the scale response weight to finally generate a dynamic fit evaluation index with unified dimensions. This index is limited to [0,1]. The closer the value is to 1, the higher the coupling degree. If the evaluation index is less than 0.75, the initial geostress parameter configuration is updated according to the residual distribution characteristics and the geostress inversion process is iteratively executed until the evaluation index meets the threshold requirement.

[0088] This invention effectively improves the signal-to-noise ratio and temporal consistency of microseismic data through filtering and denoising and time window resampling, ensuring data reliability and standardization. It constructs a spatial distribution model of microseismic response, achieving precise spatial location and quantitative representation of microseismic events. Spatial coupling and comparison of geostress field data and microseismic response field accurately reveals the correlation between stress field changes and microseismic activity. Utilizing multi-scale weighted statistical analysis, it comprehensively reflects the fitting quality at different spatial scales, forming a dynamic fitting degree evaluation index. This provides a scientific basis for the dynamic correction and parameter optimization of the inversion model, effectively promoting the dynamic adaptive iteration and refinement of geostress field inversion results.

[0089] Preferably, if the fitting degree is lower than a preset threshold in step S5, the multi-scale weight function and inversion parameters are updated, and steps S2 to S5 are iteratively executed until the convergence condition is met, resulting in the final three-dimensional dynamic inversion data of the geostress field, including:

[0090] Based on the dynamic fit evaluation index, it is determined whether it is lower than the preset threshold, and the fitting result judgment label is obtained.

[0091] If the fitting result determines that the label is not satisfied, then the multi-scale weight function is updated based on the coupled residual distribution result to obtain the updated multi-scale weight function data;

[0092] Based on the updated multi-scale weight function data, the geostress inversion constraint strategy and the optimized parameter set are adjusted to generate an updated inversion parameter configuration, and the updated inversion parameter configuration is used to iteratively execute steps S2 to S5.

[0093] If the fitting result determines that the label is satisfied, then the current optimized three-dimensional dynamic distribution result of the geostress field will be output as the final three-dimensional dynamic inversion data of the geostress field.

[0094] In this embodiment of the invention, firstly, a preset threshold of 0.75 is set for the dynamic fit evaluation index. Immediately after the evaluation index is calculated, a logical judgment is made. If the value is lower than 0.75, the current judgment result is assigned the "not satisfied" status label. In this state, based on the previously obtained coupling residual distribution results, the original multi-scale partition mapping information is reconstructed. The scale weight coefficients corresponding to regions with residual standard deviations greater than 0.3 MPa are selected for adjustment. A weighted correction factor is used to update the original scale response weight function, resulting in updated multi-scale weight function data. The updated weight function data is used to reset the regional priority in the geostress inversion process. Specifically, regions with weights greater than 0.7 are assigned higher partition index values, and the upper limit of the inversion iteration count is increased by 20%. Next, combined with the updated multi-scale weight function data, the constraint strategy and optimization parameter set in the geostress inversion process are adjusted. The boundary condition forcing coefficient is limited to the range [0.8, 1.2], and the initial principal stress value is subjected to a ±10% linear perturbation, forming the updated... Inversion parameter configuration: Input the updated inversion parameter configuration into the three-dimensional dynamic inversion process of the geostress field, and iterate in the order of steps S2 to S5. That is, starting from multi-scale regional division and boundary coupling data reconstruction, the inversion objective function is updated, joint inversion solution is performed, error identification and local densification inversion are performed, and the coupling and comparison with microseismic response data are performed again. The above process is repeated until the dynamic fit evaluation index is greater than or equal to 0.75 and lasts for two iteration cycles, and the mean change rate of the residuals is less than 1% for two consecutive iterations. After meeting the convergence criterion, the optimized three-dimensional dynamic distribution result of the geostress field is directly output as the final three-dimensional dynamic inversion data of the geostress field.

[0095] This invention can determine the degree of matching between the inversion results and the actual microseismic response data in real time, effectively identifying situations where the inversion accuracy is insufficient; based on the coupled residual distribution, it dynamically updates the multi-scale weight function to achieve adaptive adjustment to geological heterogeneity and local anomalies, enhancing the model's flexibility and responsiveness; by adjusting the inversion constraint strategy and optimizing the parameter set, the system optimizes the inversion process, improving the efficiency and convergence speed of numerical solutions; iteratively executing the multi-scale adaptive inversion process ensures that the inversion results continuously approach the true geostress distribution; the final output three-dimensional dynamic inversion data of the geostress field has high accuracy, high consistency, and dynamic adaptability, meeting the stringent requirements of engineering applications under complex geological conditions.

[0096] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is not limited by the foregoing description. Thus, all changes falling within the meaning and scope of the equivalents of the application are intended to be included within the scope of the invention.

[0097] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A three-dimensional dynamic inversion method for geostress field based on a multi-scale adaptive algorithm, characterized in that, Includes the following steps: Step S1: Obtain the geological dataset for the target area; A three-dimensional initial model of the geostress field was constructed based on the geological dataset, and the geological body space was divided and meshed to obtain the three-dimensional mesh model data of the geostress field. Step S2: Perform multi-scale regional division on the three-dimensional mesh model data of the geostress field and establish a multi-scale weight function to obtain multi-scale partitioning mapping information; A cross-scale boundary adaptive transfer mechanism is constructed, and a stress tensor continuity constraint model is established at the multi-scale partition boundary to obtain cross-scale stress boundary coupling data. Step S2 includes the following steps: Step S21: Perform cluster analysis on the spatial characteristic scales in the three-dimensional mesh model data of the geostress field to obtain the scale sensitivity classification results; Step S22: Based on the scale-sensitive classification results, the three-dimensional grid model data of the geostress field is divided into regions to obtain preliminary results of multi-scale partitioning; Step S23: Based on the preliminary results of multi-scale partitioning, construct the spatial mapping relationship between multi-scale partitioning and the original model to obtain multi-scale partitioning mapping information; Step S24: Construct a cross-scale boundary adaptive transfer mechanism based on multi-scale partition mapping information, establish a stress tensor continuity constraint model at the multi-scale partition boundary, and obtain cross-scale stress boundary coupling data. Step S24 includes the following steps: Step S241: Construct a scale response weight function based on multi-scale partitioning mapping information to obtain multi-scale stress weight distribution data; Step S242: Pair up and identify the boundary grid nodes of each region in the multi-scale partitioning mapping information to obtain a cross-scale contact boundary node set; Step S243: Construct a cross-scale boundary adaptive transfer mechanism based on the cross-scale contact boundary node set and multi-scale stress weight distribution data to obtain the stress weight adjustment parameter set. Step S243 includes the following steps: Step S2431: Perform topological adjacency analysis on the cross-scale contact boundary node set to obtain boundary node adjacency structure data; Step S2432: Extract cross-scale mapping path sequences based on boundary node adjacency structure data and multi-scale partitioning mapping information to obtain a path index dataset; Step S2433: Calculate the initial stress transfer coefficient based on the path index dataset and multi-scale stress weight distribution data to obtain the path transfer coefficient matrix; Step S2434: Perform normalization adjustment on the path transfer coefficient matrix to obtain normalized weight distribution data; Step S2435: Generate a stress weight adjustment parameter set based on the normalized weight distribution data and the cross-scale contact boundary node set; Step S244: Construct a stress tensor continuity constraint model based on the stress weight adjustment parameter group and the preset boundary node tensor state to obtain the multi-scale boundary stress tensor coordination matrix. Step S245: Project the multi-scale boundary stress tensor compatibility matrix onto the boundary data structure of the three-dimensional mesh model data of the geostress field to generate cross-scale stress boundary coupling data; Step S3: Based on multi-scale partitioning mapping information and cross-scale stress boundary coupling data, set adaptive initial inversion parameters for each scale region, construct a multi-scale adaptive inversion objective function, and perform joint inversion to obtain preliminary three-dimensional distribution results of the geostress field; Step S4: Perform dynamic error assessment and residual distribution analysis on the preliminary three-dimensional distribution results of the geostress field, identify local anomalous areas, and perform high-resolution densification inversion on local areas based on the multi-scale adaptive inversion objective function to obtain optimized three-dimensional dynamic distribution results of the geostress field; Step S5: Obtain actual microseismic response data; perform coupled comparative analysis between the optimized three-dimensional dynamic distribution results of the geostress field and the actual microseismic response data to generate a dynamic fit evaluation index; if the fit is lower than the preset threshold, update the multi-scale weight function and inversion parameters, and iteratively execute steps S2 to S5 until the convergence condition is met to obtain the final three-dimensional dynamic inversion data of the geostress field.

2. The three-dimensional dynamic inversion method for geostress field based on a multi-scale adaptive algorithm according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Obtain drilling data, well logging data, seismic profiles, geological mapping and structural interpretation data of the target area to obtain the geological dataset of the target area; Step S12: Extract information on major faults, stratigraphic interfaces and lithological boundaries based on the geological dataset, establish a tectonic boundary framework, and obtain a geological structure framework model; Step S13: Spatial division of different lithological units in the geological structure framework model to obtain spatial partitioning data of geological bodies; Step S14: Perform irregular spatial grid partitioning based on the geological body spatial partitioning data to obtain a three-dimensional initial grid partitioning model of the geostress field; Step S15: Apply boundary conditions and initial stress state to the three-dimensional initial mesh model to establish a three-dimensional initial model of the geostress field; Step S16: Bind and integrate the three-dimensional initial model of the geostress field with the three-dimensional initial mesh partitioning model to generate the three-dimensional mesh model data of the geostress field.

3. The three-dimensional dynamic inversion method for geostress field based on a multi-scale adaptive algorithm according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Perform regional index parsing on the multi-scale partitioning mapping information to obtain the structural configuration data of each scale region; Step S32: Based on the structural configuration data, set adaptive initial inversion parameters for each scale region to obtain inversion parameter setting data; Step S33: Construct a multi-scale adaptive inversion objective function based on the inversion parameter setting data and the cross-scale stress boundary coupling data to obtain the inversion objective function model; Step S34: Perform joint inversion solution on the inversion objective function model to obtain preliminary three-dimensional distribution results of the geostress field.

4. The three-dimensional dynamic inversion method for geostress field based on a multi-scale adaptive algorithm according to claim 3, characterized in that, Step S34 includes the following steps: Step S341: Perform model dimension expansion on the inversion objective function model to obtain inversion dimension configuration data for each scale region; Step S342: Set the joint inversion parameter set based on the inversion dimension configuration data of each scale region to obtain the initial parameters of the joint inversion; Step S343: Construct a joint optimization solution process based on the joint inversion initial parameters to obtain the joint inversion process configuration data; Step S344: Perform multiple rounds of numerical optimization iterations on the configuration data of the joint inversion process to obtain a multi-scale stress response numerical result set; Step S345: Reconstruct the three-dimensional geostress tensor distribution based on the multi-scale stress response numerical result set to obtain the preliminary three-dimensional distribution results of the geostress field.

5. The three-dimensional dynamic inversion method for geostress field based on a multi-scale adaptive algorithm according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Calculate the global error field based on the preliminary three-dimensional distribution results of the geostress field to obtain error distribution data; Step S42: Perform residual statistics and spatial fitting analysis on regions of each scale based on error distribution data to obtain residual distribution information; Step S43: Perform anomaly detection and clustering on the residual distribution information to obtain local anomaly region identification data; Step S44: Extract the corresponding three-dimensional mesh model of local geostress field based on the local anomaly area identification data to obtain local area model data; Step S45: Refine the mesh and reset the boundary conditions of the local region model data to obtain a high-resolution inversion input model; Step S46: Perform local densification inversion based on the high-resolution inversion input model and the multi-scale adaptive inversion objective function to obtain optimized local distribution data of the geostress field; Step S47: Perform weighted fusion of the optimized local distribution data of the geostress field and the preliminary three-dimensional distribution results of the geostress field to obtain the optimized three-dimensional dynamic distribution results of the geostress field.

6. The three-dimensional dynamic inversion method for geostress field based on a multi-scale adaptive algorithm according to claim 1, characterized in that, Step S5 involves coupling and comparing the optimized three-dimensional dynamic distribution of the geostress field with the actual microseismic response data to generate a dynamic fit evaluation index, including the following steps: The actual microseismic response data were filtered and denoised, and then resampled using time windows to obtain standardized microseismic response data. A spatial distribution model of microseismic response is constructed based on standardized microseismic response data to obtain microseismic response field data; The optimized three-dimensional dynamic distribution of the geostress field is spatially coupled and compared with the microseismic response field data to obtain the coupled residual distribution results. Multi-scale weighted statistical analysis was performed on the coupled residual distribution results to obtain the dynamic fit evaluation index.

7. The three-dimensional dynamic inversion method for geostress field based on multi-scale adaptive algorithm according to claim 1, characterized in that, If the fit is lower than the preset threshold in step S5, the multi-scale weight function and inversion parameters are updated, and steps S2 to S5 are iteratively executed until the convergence condition is met, resulting in the final three-dimensional dynamic inversion data of the geostress field, including: Based on the dynamic fit evaluation index, it is determined whether it is lower than the preset threshold, and the fitting result judgment label is obtained. If the fitting result determines that the label is not satisfied, then the multi-scale weight function is updated based on the coupled residual distribution result to obtain the updated multi-scale weight function data; Based on the updated multi-scale weight function data, the geostress inversion constraint strategy and the optimized parameter set are adjusted to generate an updated inversion parameter configuration, and the updated inversion parameter configuration is used to iteratively execute steps S2 to S5. If the fitting result determines that the label is satisfied, then the current optimized three-dimensional dynamic distribution result of the geostress field will be output as the final three-dimensional dynamic inversion data of the geostress field.

Images