A multiscale geological model cross-scale nesting and fusion modeling method

Abstract

The application provides a multiscale geological model cross-scale nesting and fusion modeling method, belonging to the geological field, including four steps of multiscale data preprocessing, data fusion, cross-scale nesting and fusion and dynamic coupling. This method can integrate multi-source heterogeneous data, realize data fusion through joint probability space, variational assimilation framework, fuzzy logic conflict factor and alternating direction multiplier method, utilize deep learning auxiliary fusion technology such as multi-scale convolutional adversarial network to improve the stability and precision of the model. At the same time, the model parameters are determined through volume average upsampling algorithm and random field conditional simulation, and the consistency of the parameters between the models is realized through dynamic adjustment of ensemble Kalman filter and localized ensemble transform Kalman filter, bidirectional feedback and iterative optimization. Finally, through the double grid strategy and the restrictive interpolation, the multiscale convergence is realized, and the convergence and dynamic balance of the model on the scale of kilometers to microns are ensured.

Description

Technical Field

[0001] This invention belongs to the field of geology, and more specifically relates to a cross-scale nesting and fusion modeling method for multi-scale geological models. Background Technology

[0002] Geological modeling is a crucial step in oil exploration and development, CO2 geological storage, and geothermal development, and its quality and accuracy directly affect the final processing results and economic benefits. Traditional geological models are mainly based on information such as seismic reflections and drilling logs, and are based on the assumptions of homogeneity and continuity. Conventional modeling methods are difficult to achieve high accuracy and can only reflect large-scale geological features, with weak depiction of details.

[0003] With advancements in science and technology, seismic techniques, core analysis, and computer technology have made significant progress, providing a foundation for improving the accuracy of geological models. Furthermore, the types and scales of geological data are becoming increasingly diverse, such as scanning electron microscopy images at the microscopic pore scale, making the utilization of scale difference observation data increasingly important.

[0004] However, effectively utilizing these multi-scale, multi-source geological interpretation, geophysical, and geochemical data to construct a comprehensive geological model that simultaneously reflects large-scale geological features, depicts fine subsurface structures, and is conflict-free among the data remains a challenge. Especially when the data exhibits significant scale differences and complex data types, issues such as data preprocessing, spatial resolution unification, data fusion, and dynamic processing need to be addressed.

[0005] Furthermore, traditional models typically include only single geological elements, such as considering only sedimentary structures and neglecting geological processes, thus ignoring the complex interactions between geological elements and between geological factors and engineering activities. This means that traditional models have limited responsiveness to complex geological systems.

[0006] Therefore, proposing a geological model that can fully utilize multi-scale and multi-source data to establish cross-scale nesting and fusion, solve the above problems, improve the accuracy of oil exploration and extraction, reduce investment risks, and improve extraction efficiency has important practical significance and application value. Summary of the Invention

[0007] This invention aims to address the challenges in current geological modeling, including difficulties in integrating multi-source heterogeneous data, inconsistent resolution, data conflicts, inability to effectively reflect large-scale geological features and microscopic details, and neglect of complex interactions between geological elements. By proposing a novel cross-scale nested and fusion model, it achieves the construction of accurate geological models, improving the precision and efficiency of geological modeling and its applications.

[0008] To achieve the above objectives, the present invention employs the following technical solution, comprising the following steps:

[0009] Step 1: Multi-scale data preprocessing, unifying coordinate system, resolution and physical dimensions, and eliminating scale bias between data;

[0010] Step 2: Data fusion. Through optimization using joint probability space, variational assimilation framework, fuzzy logic conflict factor, and alternating direction multiplier method (ADMM), the integration of multi-source heterogeneous data is achieved.

[0011] Step 3: Cross-scale nesting and fusion, hierarchical modeling of geological models at different scales, and establishment of a two-way coupling mechanism through adaptive mesh technology and multi-physics coupling;

[0012] Step 4: Dynamic coupling, through cross-scale parameter transfer, data assimilation, ensemble Kalman filtering and localized ensemble transformation Kalman filtering, to achieve bidirectional feedback and iterative optimization between models.

[0013] In one approach, step 1 achieves coordinate system unification through Helmert transformation, which includes the combined application of scale factor, rotation matrix, and translation vector to eliminate spatial biases from different measurement systems. Simultaneously, an adaptive cubic convolution interpolation algorithm is used to achieve resolution unification. This algorithm utilizes cubic B-spline basis functions for interpolation to standardize the resolution of data at different scales while maintaining the geometric topology of the geological body.

[0014] In one approach, step 1 employs a feature scaling technique guided by dimensionality analysis to achieve uniformity of physical dimensions. Specifically, this includes dimensionless transformation of parameters such as permeability, porosity, and seismic wave velocity. By calculating the mean and standard deviation of these parameters, the comparability of data under different physical dimensions is ensured.

[0015] In one approach, step 2 uses joint likelihood function, cokriging interpolation, and Markov random field for data fusion, combined with fuzzy logic conflict factor and alternating direction multiplier method optimization, to resolve the conflict problem between multi-source heterogeneous data, thereby achieving effective data integration and consistency.

[0016] In one approach, step 2 involves constructing a multi-scale fusion model using a Gaussian mixture process to ensure statistical consistency across different scales and data types. This leverages the flexibility and adaptability of the Gaussian process to handle uncertainty, thereby improving the robustness and accuracy of the model.

[0017] In one approach, step 3 employs adaptive grid technology and multi-physics coupling, combined with deep learning-assisted fusion techniques, such as multi-scale convolutional adversarial networks, to improve the stability and accuracy of the model and ensure the effectiveness of bidirectional coupling and information transfer between geological models of different scales.

[0018] In one approach, step 4 uses a volume-average upscaling algorithm and random field condition simulation to ensure parameter consistency between models of different resolutions. It then uses ensemble Kalman filtering and localized ensemble transformation Kalman filtering for dynamic adjustment, enabling bidirectional feedback and iterative optimization between models.

[0019] In one scheme, step 4 achieves multi-scale convergence through a dual-grid strategy and restricted interpolation, forming a spatiotemporally continuous integrated geological-engineering model. The convergence and dynamic balance of the model at the kilometer to micrometer scale are ensured by controlling the joint criteria of relative parameter change rate and data residuals.

[0020] Beneficial effects of this invention:

[0021] It can effectively integrate geological data from multiple sources and at multiple scales, including seismic, well logging, core, outcrop, and dynamic production data. Data of different scales and types have good compatibility in the model, which greatly improves the realism and accuracy of geological modeling.

[0022] Dynamic coupling technology enables real-time adjustment of model parameters, ensuring the stability and accuracy of the model under complex geological conditions.

[0023] By utilizing geostatistics and rock physics models, physical and mathematical integration can be achieved between models of different scales, making the models more physically meaningful and improving their credibility.

[0024] The multi-scale geological modeling method of this invention, which involves cross-scale nesting and fusion modeling, can not only better understand the structure and characteristics of geological bodies, but also help scientists and engineers accurately predict the dynamic changes of geological bodies. Attached Figure Description

[0025] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0026] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Typical embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0027] Unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. To facilitate understanding, the invention will now be described more fully with reference to the accompanying drawings. Typical embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to make the disclosure of the invention more thorough and complete.

[0028] like Figure 1 As shown, a cross-scale nesting and fusion modeling method for multi-scale geological models includes:

[0029] The multi-scale geological model cross-scale nesting and fusion modeling method is a complex modeling method that combines different spatial resolutions (such as basin scale, oil and gas reservoir scale, and micropore scale) and data types (such as geological, geophysical, and geochemical data). It aims to construct a comprehensive geological model that can reflect both macroscopic geological features and microscopic details.

[0030] This technology, by integrating multidisciplinary data and methods, helps to significantly improve the ability of geological models to characterize complex geological systems, and has become one of the core technologies that urgently need to be solved in the fields of deep oil and gas exploration and development, unconventional resource extraction, CO2 geological storage, deep geological engineering, and geothermal development.

[0031] Step 1: Multi-scale data preprocessing: unify coordinate system, resolution and physical dimensions to eliminate scale bias between data.

[0032] The core of multi-scale data preprocessing lies in establishing a basis for comparability between data, and this process begins with unifying spatial references. For multi-source data such as satellite remote sensing, seismic exploration, and well logging, coordinate system normalization needs to be achieved through Helmert transformation, mathematically expressed as:

[0033]

[0034] Where s is the scale factor, R is the rotation matrix, and T is the translation vector, this transformation can eliminate spatial biases between different measurement systems (such as WGS84 and the local coordinate system). Resolution unification involves grid reconstruction techniques. For matching kilometer-level grids at the basin scale with micrometer-level data at the pore scale, an adaptive cubic convolution interpolation algorithm is adopted:

[0035]

[0036] In the formula, B3 is the cubic B-spline basis function, (Δx, Δy) is the target grid spacing, and this algorithm can achieve resolution standardization while preserving the geometric topology of the geological body. Physical dimensions are uniformly scaled using feature scaling guided by dimensionality analysis. For permeability (10... -15 -10 -12 m 2 Parameters such as porosity (0-1) and seismic wave velocity (2000-6000 m / s) are used to construct a dimensionless transformation:

[0037]

[0038] Where μ φ and σ φ These are the mean and standard deviation of porosity, respectively, K ref =1mD is the reference permeability. To eliminate scale bias, wavelet multiresolution analysis is used to construct a scale-space mapping function:

[0039]

[0040] The Daubechies wavelet basis function ψ is selected to decompose the multi-scale data at a specific scale level (e.g., a=2). j Feature alignment is achieved through energy spectrum matching. Finally, a scale-consistent data volume is reconstructed via inverse transformation, with its fidelity determined by the reconstruction error. The control requirement is that ω < 5% to meet the accuracy requirements of geological modeling. The entire process needs to be iteratively optimized until data at different scales are aligned with the gradient field. Statistical consistency is achieved at the level of geometric features such as curvature field (κ).

[0041] Step 2, Data Fusion: Integrate multi-source heterogeneous data (such as seismic, well logging, core, outcrop, and dynamic production data) to eliminate scale differences and data conflicts, and construct a self-consistent geological model.

[0042] The core of the data fusion phase lies in constructing a joint probability space for multi-source heterogeneous data, achieving self-consistent modeling through a variational assimilation framework under geological constraints. For data types such as seismic reflection amplitude, well logging curves, core thin section images, outcrop profile sketches, and production well flowing pressure, a joint likelihood function for multimodal data is first established:

[0043]

[0044] Where m represents the geological model parameters to be determined, and d i Let g represent the i-th type of observation data. i (·) represents the corresponding forward modeling operator (such as a seismic wave equation solver or well logging response function), C iThis represents the covariance matrix for each data type. To eliminate scale differences, a multi-resolution data matching strategy is employed: constrained cokriging interpolation is performed on kilometer-level seismic data.

[0045]

[0046] Where λ α and μ β For Kriging weights, Using the Laplace operator, this forced interpolation result satisfies the tectonic trends in seismic interpretation. For centimeter-level core CT data, a microstructure optimization model based on Markov random fields is constructed:

[0047]

[0048] In the formula V c (·) represents the cluster potential energy function that characterizes the topological constraints of the pores, φ p Here, γ is the local porosity calculation function, and γ is the matching weight of the core measurement data. Dynamic production data are fused through adjoint equation inversion to establish a historical matching objective function for the pressure-saturation field:

[0049]

[0050] Where Q is the observation error covariance, β is the regularization parameter, and R is the prior covariance of the model parameters. To address the data conflict problem, a fuzzy logic conflict factor is introduced:

[0051]

[0052] When ζ ij When the value is greater than 0.3, a conflict resolution mechanism is triggered, and optimization is performed using the Alternating Direction Multiplier Method (ADMM) under rock physics cross plot constraints.

[0053] min m,z f(m) + g(z)stAm + Bz = c

[0054] Where f(·) represents the fitting term between seismic and well logging data, g(·) is the core microstructure constraint term, and A and B are the scaling transformation matrices between different scales. Finally, a multi-scale fusion model is constructed using a Gaussian mixture process.

[0055]

[0056] in This represents the convolution coupling operator between scales, with weights w. k The signal-to-noise ratio is dynamically adjusted based on the data. This process needs to be executed iteratively until the residuals of each data satisfy the requirements. The statistical consistency criteria ensure the compatibility of geological rules in multiple dimensions, such as structural styles, sedimentary sequences, and fluid dynamics.

[0057] Step 3, Cross-scale nesting and fusion:

[0058] First, the models at different scales are modeled in layers: basin-scale structural framework models (such as faults and stratigraphic interfaces) are established based on seismic interpretation data, reservoir models (such as sedimentary facies and fracture networks) are embedded in the structural framework based on borehole data, and microstructural features such as pore structure and mineral distribution are characterized by digital core or CT scan data.

[0059] Then, cross-scale nesting and fusion of models at different scales are performed:

[0060] Geological models at different scales are correlated with each other through mathematical or physical constraints (physical constraint fusion), and seismic attributes and well logging data are combined through geostatistics (such as sequential Gaussian simulation and multi-point geostatistics), and microscopic pore structure and macroscopic reservoir parameters are linked by rock physics models (such as porosity-permeability relationship).

[0061] Adaptive meshing techniques (such as unstructured meshes or locally refined meshes) are employed to achieve high-resolution nesting in key areas (such as crack zones and fluid contact surfaces), forming a hierarchical cross-scale nested structure with different spatial resolutions.

[0062] The core of cross-scale nesting and fusion lies in establishing a bidirectional coupling mechanism between multi-resolution geological models. In the layered modeling stage, a basin-scale structural framework is first constructed based on the seismic interpretation surface, and then a radial basis function surface reconstruction technique with fault constraints is employed.

[0063]

[0064] Where φ(r)=r 3 For the cubic radial basis kernel function, p j (x,y) are polynomial terms used to eliminate the influence of rigid body displacement. The coefficients are determined by solving the linear equation system Φα+Pβ=z, ensuring that the formation interface maintains \(C^2\) continuity on both sides of the fault. Reservoir-scale modeling employs a multi-point geostatistical simulation algorithm, whose conditional probability is expressed as:

[0065]

[0066] Where T k To train the topological patterns extracted from the image, V c (·) represents the potential function based on distance transformation. Micropore modeling employs a fractal-Markov coupling algorithm, and the porosity field generation follows:

[0067]

[0068] In the formula D fa is the fractal dimension fitted to the scanned data. n The energy spectrum coefficient is used. Cross-scale fusion is achieved through multiphysics coupling, establishing a two-way constraint between the macroscopic Darcy flow and the microscopic Stokes flow.

[0069]

[0070] Establish the permeability tensor using homogenization theory. The bridging relationship. Adaptive meshing technology employs local refinement criteria for unstructured tetrahedral meshes:

[0071]

[0072] When the gradient index η of unit e e When the threshold is exceeded, h-refinement is triggered to achieve millimeter-level mesh nesting in key areas such as crack zones. Deep learning-assisted fusion employs a multi-scale convolutional adversarial network architecture. The generator G includes a macroscopic path (5 layers of dilated convolutions to extract basin features) and a microscopic path (3 layers of U-Net to process pore structures). The discriminator (D) improves stability through spectral normalization constraints. The loss function is designed as follows:

[0073]

[0074] Where F(·) is the rock physics forward modeling operator, and λ controls the strength of geological rule constraints. The final model is solved using the multi-scale finite element method within a unified framework, and the stiffness matrix assembly employs a scale-progressive strategy.

[0075]

[0076] In the formula C s,s+1 Let be the constraint matrix between scales s and s+1, ensuring that physical quantities such as pressure field and saturation field satisfy the flux continuity condition at the cross-scale interface. This fusion process must satisfy the multi-scale convergence criterion max(||R) s || / ||F s ||)<10 -4 Meanwhile, the consistency of Betti numbers of pore networks at different scales was verified through topological data analysis, ultimately forming a seamless nested geological model from kilometer-level basin structure to micron-level pore structure.

[0077] Step 4, Dynamic Coupling: Achieve bidirectional feedback and iterative optimization between models of different resolutions through cross-scale parameter transfer (e.g., upscaling and downscaling algorithms). Dynamically adjust model parameters through data assimilation (e.g., ensemble Kalman filtering) to ensure multi-scale consistency.

[0078] The core of dynamic coupling lies in establishing a bidirectional transfer mechanism and real-time correction system for multi-scale model parameters. Regarding cross-scale parameter transfer, a volume-averaged upscaling algorithm is used to map microscopic pore structure parameters to the reservoir scale.

[0079]

[0080] In the formula, w(x) is the microscopic perturbation potential function solved under periodic boundary conditions, obtained by discretizing the characteristic element problem using the finite element method. A pore-reservoir permeability correlation is achieved. The downscaling process is then simulated using random field conditions.

[0081] m fine =G down m coarse +Lz,z~N(0,I)

[0082] The scaling operator G down Constructed by wavelet multiresolution analysis, L satisfies LL T =C fine|coarse The Cholesky decomposition matrix ensures spatial consistency between fine-scale parameters and coarse-scale dominant trends. Dynamic feedback is achieved through a coupled adjoint system, constructing a joint macro-micro objective function:

[0083]

[0084] In the formula T s→s′ Q is the interscale transformation operator. ss′ Characterizing the covariance of parameter differences across scales. The dynamic assimilation process of ensemble Kalman filtering (EnKF) is achieved through multi-scale state vector expansion:

[0085]

[0086] The cross-scale covariance matrix It explicitly includes off-diagonal coupling terms for macroscopic and microscopic parameters. This is for the inter-scale observation operator H. s The nonlinearity of (·) is improved by using Localized Ensemble Transform Kalman Filter (LETKF):

[0087]

[0088] Weight matrix W k The observation perturbation term is determined by solving for the minimum value of the localized cost function. It includes multi-scale data covariance information. Iterative optimization employs a dual-grid strategy, solving for the global update on a coarse-scale grid.

[0089]

[0090] Subsequently, a restricted interpolation operator was used. Transfer to a finer-scale mesh for local correction:

[0091]

[0092] Where S f For fine-scale smoothing operators, r f =d f -H f (m f ( ) represents local residuals. Multi-scale convergence is controlled by a joint criterion of relative parameter change rate and data residuals:

[0093]

[0094] Simultaneously, a cross-scale compatibility index was introduced. As the iteration termination condition, when χ cross <δ tol The system is determined to reach dynamic equilibrium at multiple scales. This coupling process ultimately forms a geo-engineering integrated model with spatiotemporal continuity, whose parameter field simultaneously satisfies rock physics laws and dynamic production response at the kilometer to micrometer scale.

[0095] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0096] It should be understood that the above detailed description of the technical solutions of the present invention with reference to preferred embodiments is illustrative and not restrictive. Those skilled in the art can modify the technical solutions described in the embodiments or make equivalent substitutions for some of the technical features based on reading this specification; however, these modifications or substitutions do not cause the essence of the corresponding technical solutions to depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for cross-scale nesting and fusion modeling of multi-scale geological models, characterized in that, The method comprises the following steps: Step 1: multi-scale data preprocessing, unified coordinate system, resolution and physical dimension, eliminating the scale deviation between data; Step 2: data fusion, through joint probability space, variational assimilation framework, fuzzy logic conflict factor and alternating direction multiplier method optimization, realizing the integration of multi-source heterogeneous data; Step 3: cross-scale nesting and fusion, hierarchical modeling of different scale geological models, and establishing a two-way coupling mechanism through adaptive grid technology and multi-physical field coupling; Different scale geological models are related to each other through mathematical or physical constraint conditions, and through geostatistics, sequential Gaussian simulation, multi-point geostatistics joint seismic attribute and logging data, and rock physics model linking micro pore structure and macro reservoir parameters; Adaptive grid technology is used to realize high-resolution nesting in key areas, forming a hierarchical cross-scale nested structure with different spatial resolutions; The rock physics model is a porosity-permeability relationship model; The adaptive grid technology is an unstructured grid or local encryption grid; The key area is a fracture zone or a fluid contact surface; Step 4: dynamic coupling, through cross-scale parameter transmission, data assimilation, ensemble Kalman filter and localized ensemble transform Kalman filter, realizing the two-way feedback and iterative optimization between models.

2. The method of claim 1, wherein, In step 1, the unified coordinate system is realized by Helmholtz transformation, which includes the comprehensive application of scale factor, rotation matrix and translation vector to eliminate the spatial deviation between different measurement systems; at the same time, adaptive cubic convolution interpolation algorithm is used to realize the unification of resolution, which uses cubic B-spline basis function for interpolation to realize the resolution standardization of different scale data under the premise of maintaining the geometric topology of geological body.

3. The method of claim 1, wherein, In step 1, the feature scaling technology guided by dimension analysis is used to realize the unification of physical dimension, which includes the dimensionless conversion of permeability, porosity and seismic wave velocity parameters, and the calculation of the mean and standard deviation of these parameters to ensure the comparability of data under different physical dimensions.

4. The method of claim 1, wherein, In step 2, joint likelihood function, co-Kriging interpolation and Markov random field are used for data fusion, combined with fuzzy logic conflict factor and alternating direction multiplier method optimization to solve the conflict problem between multi-source heterogeneous data, so as to realize the effective integration and consistency of data.

5. The method of claim 1, wherein, In step 2, a multi-scale fusion model is constructed by mixing Gaussian process to ensure the statistical consistency under different scales and data types, and the flexibility and adaptability of Gaussian process are used to deal with uncertainty, thereby improving the robustness and precision of the model.

6. The method of claim 1, wherein, In step 3, adaptive grid technology and multi-physical field coupling are used, combined with deep learning assisted fusion technology, multi-scale convolutional adversarial network, to improve the stability and precision of the model, and ensure the effectiveness of two-way coupling and information transmission between different scale geological models.

7. The method of claim 1, wherein, In step 4, volume average upscaling algorithm and random field conditional simulation are used to ensure the parameter consistency between models with different resolutions, and ensemble Kalman filter and localized ensemble transform Kalman filter are used for dynamic adjustment to realize the two-way feedback and iterative optimization between models.

8. The method of claim 1, wherein, The step 4 realizes multi-scale convergence through a double-mesh strategy and limited interpolation, forms a geology-engineering integrated model with continuous space and time, and controls the model convergence and dynamic balance on the scale of kilometers to microns through a joint criterion of relative parameter change rate and data residual error.

Images