Method and system for constructing geological digital twin model in petroleum exploration

Abstract

The present invention relates to the field of oil exploration technology, and discloses a method and system for constructing a geological digital twin model in oil exploration. The method comprises: receiving multi-source heterogeneous geological data streams in a target exploration area, performing multi-scale alignment and feature fusion through a geological feature decoupling model, and generating a three-dimensional geological attribute field distribution map; constructing an adaptive grid division model to generate a dynamic flow field simulation instruction set; simulating the dynamic response of the reservoir under production disturbance based on a multi-physics field coupling model and optimizing the simulation instruction set parameters; iteratively updating parameters through a quantum computing optimization framework, and outputting a model construction sequence to an exploration decision platform. The system comprises data reception, feature processing, grid instruction generation, simulation optimization, and output modules. The present invention integrates multi-source data, accurately simulates geological processes, optimizes exploration decisions, and improves oil exploration efficiency and production benefits.

Description

Technical Field

[0001] The present invention relates to the technical field of petroleum exploration, and in particular to a method and system for constructing a geological digital twin model in petroleum exploration. Background Art

[0002] In the field of oil exploration, accurately understanding the underground geological structure and reservoir characteristics is key to the efficient extraction of oil and gas resources. As global energy demand continues to grow, oil exploration is gradually expanding into areas with complex geology, posing numerous challenges to traditional exploration technologies.

[0003] Traditional oil exploration methods primarily rely on single data sources or simple data processing methods. For example, early methods relied solely on seismic reflection wave data to infer underground geological structures. However, this method provides limited information and cannot accurately describe complex geological structures. While core analysis can provide information on rock physical properties, the limited number of core samples cannot fully reflect the characteristics of the entire reservoir. Moreover, this data is often isolated and lacks effective integration and analysis, making it difficult for explorers to fully understand the geological situation.

[0004] With the development of technology, multi-source data collection methods are constantly emerging. Technologies such as geophysical exploration, geological drilling, and well logging are generating an ever-increasing amount of data. However, this data is heterogeneous, with significant differences in format, accuracy, and time scale between different types of data, making data integration and analysis extremely difficult. For example, seismic reflection wave data reflects the macroscopic characteristics of the underground geological structure, while the core porosity distribution matrix focuses on the microscopic rock properties. How to effectively integrate and analyze these data of different scales and properties has become a major issue restricting the accuracy of oil exploration.

[0005] Furthermore, when it comes to reservoir simulation, traditional models are often oversimplified and unable to accurately simulate the migration patterns of multiphase fluids under complex geological conditions. Fluid migration in reservoirs is influenced by numerous factors, such as formation pressure gradients, rock permeability anisotropy, and fluid viscosity. Traditional models struggle to comprehensively account for these factors, resulting in significant deviations between simulation results and actual conditions. Optimizing well placement lacks scientifically sound methods, often relying on empirical decision-making. This not only increases exploration costs but can also lead to missed opportunities for optimal production areas.

[0006] During oil production, reservoirs are disturbed by mining activities, causing changes in their physical properties and fluid distribution. However, existing technologies struggle to monitor and accurately predict these dynamic changes in real time, making it difficult to provide timely and effective support for production decisions. Faced with these challenges, there is an urgent need for a technology that can integrate multi-source heterogeneous geological data, accurately simulate geological processes, and optimize exploration decisions. Consequently, methods and systems for constructing geological digital twin models have emerged to meet the modern oil exploration industry's urgent need for high-precision, high-efficiency exploration technologies. Summary of the Invention

[0007] The purpose of the present invention is to provide a method and system for constructing a geological digital twin model in petroleum exploration to solve the problems raised in the above background technology.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a method for constructing a geological digital twin model in petroleum exploration, the method comprising:

[0009] Receiving a multi-source heterogeneous geological data stream of a target exploration area, the data stream including seismic reflection wave data, core porosity distribution matrix, and formation pressure gradient time series data;

[0010] Based on a preset geological feature decoupling model, multi-scale alignment and feature fusion are performed on the multi-source heterogeneous data to generate a three-dimensional geological attribute field distribution map and a reservoir heterogeneity tensor;

[0011] According to the attribute field distribution map, an adaptive grid partitioning model is constructed to generate a dynamic flow field simulation instruction set, wherein the instruction set includes a multiphase fluid migration path topology and a well location optimization deployment strategy;

[0012] Based on a preset multi-physics field coupling model, the dynamic response of the target reservoir under mining disturbance is simulated, and the multi-parameter collaborative constraint conditions in the simulation instruction set are optimized;

[0013] The collaborative constraints are iteratively updated through a quantum computing optimization framework, and the geological digital twin model construction sequence is output to the exploration decision-making platform.

[0014] Preferably, the steps of constructing the geological feature decoupling model include:

[0015] Collect historical exploration data sets from multiple blocks and construct a three-dimensional feature training set including formation fracture patterns, sedimentary phase spatial distribution, and rock mechanical parameters;

[0016] Performing nonlinear dimensionality reduction on the three-dimensional feature training set using a spectral clustering algorithm to extract independent representations of dominant geological factors and secondary disturbances;

[0017] Combined with the geomechanical constitutive equation, the partial differential constraints of the dynamic correlation between factors are constructed;

[0018] The partial differential constraint condition is embedded in a Bayesian inference network to generate the geological feature decoupling model supporting online updating.

[0019] Preferably, the adaptive grid partitioning model includes:

[0020] Dynamically dividing the unstructured grid density levels according to the spatial gradient distribution of the three-dimensional geological attribute field distribution map;

[0021] Calculate the grid cell adaptive refinement weight based on the reservoir permeability anisotropy tensor and fluid viscosity coefficient;

[0022] Performing nonlinear mapping of the refinement weight and density level through a hyperbolic sine function to generate a multi-resolution nested grid partitioning scheme;

[0023] The initial boundary condition loading protocol for the fluid-structure interaction simulation is triggered according to the described scheme.

[0024] Preferably, the step of constructing the multi-physics field coupling model includes:

[0025] Collect the pressure-stress response correlation matrix from historical mining data and construct a multi-field coupling causal diagram dataset;

[0026] The interaction intensity parameters between the seepage field and the stress field are extracted using the random forest regression algorithm;

[0027] Combined with implicit solvers, the multi-physics dynamic equilibrium equations are constructed to quantify the relaxation time constants of energy transfer between fields;

[0028] The relaxation time constant and the real-time mining disturbance factor are input into a cellular automaton network to generate the multi-physics coupling model.

[0029] Preferably, the method further comprises:

[0030] Identifying reservoir dynamic response sensitive areas based on simulation results of the multi-physics field coupling model;

[0031] configuring a real-time data assimilation strategy for the region in the dynamic flow field simulation instruction set;

[0032] Based on the assimilation strategy, a multi-scale observation data fusion scheme is automatically generated, including seismic wave velocity field correction instructions and inter-well interpolation parameter update strategy.

[0033] Preferably, the calculation of the grid unit adaptive refinement weight includes:

[0034] Obtain formation fracture density distribution map and pore pressure diffusion coefficient matrix, and construct grid quality assessment hypercube;

[0035] Calculate the singular value contribution of each dimension feature through tensor decomposition algorithm;

[0036] Performing a Hadamard product operation on the contribution and the hypercube to obtain a comprehensive refinement weight;

[0037] The calculation formula of the comprehensive refinement weight is:

[0038]

[0039] Where, represents the comprehensive refinement weight value, Indicates the The singular values ​​of the dimensional features, represents the geological attribute field function, represents the spatial coordinate component, represents the nonlinear amplification exponent, represents the total number of feature dimensions, is the Hadamard product operator, which represents the product of corresponding elements.

[0040] Preferably, the embedding of the partial differential constraint condition includes:

[0041] Performing Lie group symmetry analysis on the nonlinear dimensionality reduction results to screen out correlation patterns that conform to the conservation laws of geomechanics;

[0042] Generate feature evolution paths that satisfy constitutive constraints through Markov chain Monte Carlo sampling;

[0043] The path data is used to regularize the hyperparameters of the Bayesian inference network to ensure that the model output conforms to the dynamic evolution law of the reservoir.

[0044] Preferably, the execution of the real-time data assimilation strategy includes:

[0045] Establish the spatiotemporal covariance matrix of multi-source observation data and define the data fusion confidence score;

[0046] Updating the posterior distribution of reservoir state estimates by integrating the Kalman filter algorithm;

[0047] The posterior distribution is input as the initial condition into the next simulation time step to complete the dynamic data closed-loop correction.

[0048] Preferably, the execution of the quantum computing optimization framework includes:

[0049] Define the Hamiltonian function for multi-objective optimization, including the dual constraints of increasing reservoir recovery and reducing energy consumption;

[0050] Solving the ground state configuration of the Hamiltonian by a quantum annealing algorithm to obtain an optimal cooperative constraint parameter set;

[0051] Wherein, the Hamiltonian function is:

[0052]

[0053] Where, represents the Hamiltonian, Representation parameters and The coupling strength, represents a binary decision variable, Represents the external field effect coefficient.

[0054] Preferably, the present invention also includes a geological digital twin model construction system for petroleum exploration, the features of which include:

[0055] Data receiving module: used to receive multi-source heterogeneous geological data streams in the target exploration area, including seismic reflection wave data, core porosity distribution matrix and formation pressure gradient time series data;

[0056] Feature processing module: Based on a preset geological feature decoupling model, it performs multi-scale alignment and feature fusion on the multi-source heterogeneous data to generate a three-dimensional geological attribute field distribution map and a reservoir heterogeneity tensor;

[0057] Grid instruction generation module: constructs an adaptive grid partitioning model based on the attribute field distribution map and generates a dynamic flow field simulation instruction set, which includes a multiphase fluid migration path topology and a well location optimization deployment strategy;

[0058] Simulation Optimization Module: Based on a preset multi-physics field coupling model, it simulates the dynamic response of the target reservoir under mining disturbance and optimizes the multi-parameter collaborative constraints in the simulation instruction set;

[0059] Output module: Iteratively update the collaborative constraints through the quantum computing optimization framework, and output the geological digital twin model construction sequence to the exploration decision-making platform.

[0060] Compared with the prior art, the present invention has the following beneficial effects:

[0061] The method and system for constructing a geological digital twin model in petroleum exploration proposed in this invention significantly optimize the petroleum exploration process from multiple key aspects and bring many positive impacts to the development of the industry.

[0062] In terms of data processing and analysis, the invention is capable of receiving multi-source heterogeneous geological data streams from the target exploration area, including seismic reflection wave data, core porosity distribution matrices, and formation pressure gradient time series data. Using a pre-set geological feature decoupling model, these complex data are aligned and fused at multiple scales to generate a three-dimensional geological attribute field distribution map and reservoir heterogeneity tensor. This process overcomes the limitations of traditional single-data utilization by organically integrating data of different types and scales, providing explorers with more comprehensive and accurate geological information, enabling them to gain a deeper understanding of the underground geological structure and reservoir characteristics, thereby laying a solid foundation for subsequent exploration decisions.

[0063] At the model construction and simulation level, an adaptive meshing model is constructed based on the generated attribute field distribution map, and then a dynamic flow field simulation instruction set is generated that includes the multiphase fluid migration path topology and well location optimization deployment strategy. This adaptive meshing method fully considers the spatial variation of geological attributes and can more accurately simulate the migration of multiphase fluids under complex geological conditions. At the same time, combined with the preset multi-physics field coupling model, the dynamic response of the target reservoir under mining disturbance is simulated, and the multi-parameter collaborative constraints in the simulation instruction set are continuously optimized. This makes the simulation of reservoir dynamic changes closer to reality, effectively improves the accuracy of the prediction, and provides strong support for the rational planning of mining plans.

[0064] In terms of decision-making optimization, a quantum computing optimization framework iteratively updates collaborative constraints and outputs a geological digital twin model construction sequence to the exploration decision-making platform. This optimization framework can quickly find the optimal set of collaborative constraint parameters under multi-objective constraints, such as achieving a balance between increasing reservoir recovery and reducing energy consumption. Using these precise models and optimized decision sequences, explorers can more scientifically plan well locations, avoid blind drilling, and effectively reduce exploration costs. During the mining process, based on real-time monitoring data and model simulation results, mining strategies can be adjusted promptly to improve oil and gas resource extraction efficiency and increase production, thereby enhancing the economic benefits of the entire oil exploration and development project.

[0065] In real-time monitoring and feedback adjustment, sensitive areas of reservoir dynamic response are identified based on the simulation results of the multi-physics field coupling model, and a real-time data assimilation strategy is configured for them to automatically generate a multi-scale observation data fusion solution. This enables the system to track the dynamic changes of the reservoir in real time, integrate newly acquired data into the model in a timely manner, and continuously update the model parameters to ensure the accuracy and timeliness of the model. This closed-loop data processing and model update mechanism can promptly discover problems that arise during the mining process, and provide a scientific basis for adjusting the mining plan to ensure the safe and efficient conduct of mining activities. In short, this invention has comprehensively improved the technical level and efficiency of oil exploration from data integration, model construction, decision optimization to real-time monitoring and feedback, and is of great significance to promoting the intelligent and efficient development of the oil exploration industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 This is a working principle diagram of the method for constructing a geological digital twin model in petroleum exploration according to the present invention;

[0067] Figure 2 A diagram showing the working principle of the adaptive meshing model;

[0068] Figure 3 A diagram showing the steps involved in expanding the multi-physics coupling model.

[0069] Figure 4 Flowchart for the calculation of adaptive refinement weights for mesh cells. DETAILED DESCRIPTION

[0070] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0071] See also Figures 1-4 The present invention provides a method and system for constructing a geological digital twin model in petroleum exploration to achieve accurate simulation and analysis of the geological conditions in the target exploration area, providing strong support for exploration decision-making. The specific implementation steps are as follows:

[0072] Acquire multi-source heterogeneous geological data streams for the target exploration area, including seismic reflection wave data, core porosity distribution matrices, and formation pressure gradient time series data. Seismic reflection wave data, acquired through seismic exploration techniques, records seismic wave information reflected from different subsurface strata interfaces and can be used to infer the structure and architecture of the strata. The core porosity distribution matrix, obtained through analysis and testing of actual core samples, reflects the distribution of rock pores and is crucial for evaluating reservoir performance. Formation pressure gradient time series data records the changes in formation pressure over time, helping to understand the flow patterns of underground fluids.

[0073] The pre-set geological feature decoupling model is used to process the collected multi-source heterogeneous data. Multi-scale alignment unifies data from different sources and scales in time and space, making them comparable. Feature fusion is then performed to integrate the characteristic information contained in various geological data sets, ultimately generating a 3D geological attribute field distribution map and a reservoir heterogeneity tensor. The 3D geological attribute field distribution map intuitively displays the distribution of various attributes of the subsurface geological body in three-dimensional space, such as lithology and porosity. The reservoir heterogeneity tensor quantitatively describes the reservoir heterogeneity, providing an important basis for subsequent reservoir evaluation and development.

[0074] An adaptive meshing model is constructed based on the generated three-dimensional geological attribute field distribution map. This model rationally divides the mesh based on the spatial variation characteristics of the geological attribute field, making the mesh finer in areas with drastic geological changes and relatively sparse in areas with gentle changes, thereby improving the accuracy and efficiency of the simulation. This model generates a dynamic flow field simulation instruction set, which includes a multiphase fluid migration path topology and a well location optimization deployment strategy. The multiphase fluid migration path topology describes the flow paths and interrelationships of various fluids such as oil, gas, and water in underground reservoirs; the well location optimization deployment strategy determines the optimal drilling location and production plan based on geological conditions and fluid migration patterns to improve recovery.

[0075] Using a pre-set multi-physics coupling model, the dynamic response of the target reservoir under mining disturbances is simulated. Considering the changes in physical fields such as formation pressure and stress caused by mining, the multi-physics coupling model comprehensively considers the interactions between these factors and accurately simulates the dynamic changes of the reservoir. During the simulation process, the multi-parameter collaborative constraints in the simulation instruction set are optimized to ensure that the simulation results are more consistent with actual conditions.

[0076] Leveraging a quantum computing optimization framework, collaborative constraints are iteratively updated. Quantum computing's powerful parallel computing capabilities enable the search for optimal solutions in a short period of time. Through continuous iteration, the optimal collaborative constraints are found, ultimately outputting a geological digital twin model construction sequence to the exploration decision-making platform. Based on this construction sequence, the exploration decision-making platform can formulate more scientific and rational exploration and development plans, improving exploration efficiency and economic benefits.

[0077] The implementation of the present invention will be further described below with reference to Examples 1 to 6. Example 1:

[0078] The construction of a geological feature decoupling model begins with extensive data collection. Historical exploration datasets are collected from multiple different blocks, containing a wealth of geological information. The fracture patterns of the formations record the distribution, orientation, and density of fractures within the formations, which significantly influence the migration of subsurface fluids and reservoir connectivity. Sedimentary facies spatial distribution describes the spatial distribution of rock phases formed under different depositional environments, with different sedimentary facies exhibiting distinct reservoir properties. Rock mechanical parameters, such as elastic modulus and Poisson's ratio, reflect the physical and mechanical properties of rocks, which are crucial for analyzing formation stability and deformation characteristics. This information is integrated to construct a three-dimensional feature training set encompassing fracture patterns, sedimentary facies spatial distribution, and rock mechanical parameters.

[0079] A spectral clustering algorithm is used to perform nonlinear dimensionality reduction on the three-dimensional feature training set. Spectral clustering, a graph-theoretic clustering algorithm, is capable of discovering complex clustering structures in high-dimensional data. This algorithm can map high-dimensional geological feature data into a low-dimensional space while preserving the data's key features. During the dimensionality reduction process, independent representations of dominant geological factors and secondary perturbations are extracted. Dominant geological factors represent the primary trends and patterns in geological features, while secondary perturbations reflect localized, less significant variations.

[0080] By combining geomechanical constitutive equations with the dynamic relationships between factors, partial differential constraints are constructed. Geomechanical constitutive equations describe the deformation and failure patterns of rocks under load. By combining these with the geological factors derived from dimensionality reduction, dynamic relationships between factors can be established. These partial differential constraints reflect the physical laws of geological processes and play a key role in the accuracy of subsequent models.

[0081] Partial differential constraints are embedded in a Bayesian inference network to generate a geological feature decoupling model that supports online updates. A Bayesian inference network is a machine learning model based on probabilistic reasoning that continuously updates model parameters and predictions based on new observational data. By embedding partial differential constraints within this model, the model's feature decoupling not only considers the statistical characteristics of the data but also adheres to the physical laws of geomechanics. In practical applications, as new geological data is continuously collected, the model can be updated online, improving the accuracy of geological feature decoupling and better adapting to varying exploration scenarios. Example 2:

[0082] When building an adaptive meshing model, the unstructured mesh density is dynamically divided based on the spatial gradient distribution of the three-dimensional geological attribute field. The spatial gradient of the geological attribute field reflects the speed of change in the spatial distribution of geological attributes. In areas with large gradients, such as stratigraphic boundaries and near faults, geological attributes vary dramatically, requiring a denser mesh to accurately describe these changes. In areas with smaller gradients, geological attributes vary more gradually, and a sparser mesh can be used to reduce the computational effort.

[0083] Based on the reservoir permeability anisotropy tensor and fluid viscosity coefficient, adaptive grid cell refinement weights are calculated. A formation fracture density distribution map and a pore pressure diffusion coefficient matrix are obtained to construct a grid quality assessment hypercube. The formation fracture density distribution map reflects the distribution of fractures in the formation, which affect fluid flow. A higher fracture density leads to more complex fluid flow and requires a finer mesh. The pore pressure diffusion coefficient matrix describes the diffusion capacity of pore pressure in the medium. A higher value indicates a wider impact of pressure changes on the surrounding area, requiring a finer mesh to simulate.

[0084] The singular value contribution of each dimensional feature is calculated using a tensor decomposition algorithm. The tensor decomposition algorithm can decompose a high-order tensor into a combination of multiple low-order tensors. By calculating the singular value contribution, the importance of each dimensional feature to the overall mesh quality can be determined.

[0085] Perform Hadamard product operation on the contribution and the hypercube to obtain the comprehensive refinement weight, which is calculated as follows:

[0086]

[0087] In this formula, Represents the comprehensive refinement weight value, which comprehensively considers the influence of each dimensional feature on the grid refinement. The larger the value, the more refined the grid cell needs to be; Indicates the The singular value of the dimension feature. The larger the singular value, the greater the contribution of the dimension feature to the overall feature. It represents the geological attribute field function, which describes the distribution of geological attributes in space; Represents the spatial coordinate component, which is used to determine the feature changes in different directions; Represents the nonlinear amplification index, which is used to adjust the influence of different feature changes on the weight. The value can make the weight calculation more consistent with the actual geological conditions; Represents the total number of characteristic dimensions, covering all geological characteristic dimensions involved in the calculation; is the Hadamard product operator, which represents the product of corresponding elements. Through this operation, the characteristic contribution of each dimension is combined with the change rate of the geological attribute field function to obtain the comprehensive refinement weight of each grid unit.

[0088] A multi-resolution nested gridding scheme is generated by nonlinearly mapping refinement weights and density levels using a hyperbolic sine function. This function applies nonlinear transformations to the data, making the relationship between refinement weights and density levels more consistent with geological conditions. This multi-resolution nested gridding scheme effectively reduces computational effort while maintaining simulation accuracy.

[0089] This scheme triggers the initial boundary condition loading protocol for fluid-solid coupling simulation. Fluid-solid coupling simulation considers the interaction between fluids and solids. In oil exploration, complex interactions exist between reservoir rocks and the fluids within them. By loading appropriate initial boundary conditions, this interaction can be accurately simulated, providing a reliable foundation for subsequent dynamic flow field simulations. Example 3:

[0090] When building a multi-physics coupling model, we first collect the pressure-stress response correlation matrix from historical mining data and construct a multi-field coupling causal diagram dataset. Historical mining data contains information on changes in formation pressure and stress under different mining conditions. The pressure-stress response correlation matrix records the relationship between pressure and stress changes. By analyzing this data, we can discover the causal relationships between multiple physical fields. Constructing a multi-field coupling causal diagram dataset graphically represents these causal relationships, providing an intuitive data structure for the subsequent extraction of interaction strength parameters.

[0091] The random forest regression algorithm is used to extract the interaction strength parameters between the seepage and stress fields. This is a machine learning algorithm based on decision trees. By constructing multiple decision trees and performing comprehensive predictions, it can effectively handle high-dimensional data and complex nonlinear relationships. In multi-physics coupling problems, the seepage and stress fields interact in complex ways. The random forest regression algorithm can extract parameters reflecting the strength of these interactions from extensive historical data.

[0092] An implicit solver is used to construct the multi-physics dynamic equilibrium equations and quantify the relaxation time constant for energy transfer between fields. Implicit solvers offer excellent stability and high precision in numerical calculations, effectively solving multi-physics dynamic equilibrium equations. When constructing the equations, the energy transfer between the seepage and stress fields is taken into account, and a relaxation time constant is introduced to quantify the speed and efficiency of this energy transfer. The relaxation time constant reflects the response speed of one physical field to changes in another. By accurately determining this constant, the coupling process between multiple physical fields can be more accurately simulated.

[0093] The relaxation time constant and the real-time mining perturbation factor are input into a cellular automaton network to generate a multiphysics coupling model. A cellular automaton network is a discrete dynamic system model composed of multiple cells, each of which updates its state based on the states of surrounding cells and certain rules. In this multiphysics coupling model, the relaxation time constant and the real-time mining perturbation factor are used as input. The cellular automaton network can simulate the dynamic changes of multiple physical fields in space and time, thereby generating a multiphysics coupling model that accurately reflects the actual situation. Example 4:

[0094] Based on the simulation results of the multiphysics coupling model, sensitive areas of reservoir dynamic response are identified. During the simulated production process, the multiphysics coupling model outputs the changes in physical quantities such as pressure and stress in various regions of the reservoir. By analyzing these simulation results, areas with large changes in physical quantities and high sensitivity to production disturbances are identified. These sensitive areas may be key pathways for fluid flow in the reservoir or areas prone to formation deformation or damage. Focusing on and studying these areas is crucial for optimizing production plans.

[0095] In the dynamic flow field simulation instruction set, a real-time data assimilation strategy is configured for sensitive areas. The purpose of the real-time data assimilation strategy is to combine real-time observation data with the simulation model to improve simulation accuracy. First, a spatiotemporal covariance matrix for multi-source observation data is established and a data fusion confidence score is defined. The spatiotemporal covariance matrix describes the correlation between observations at different times and spaces. By calculating the covariance matrix, the reliability of the observations and the degree of their mutual influence can be understood. Based on the covariance matrix, a data fusion confidence score is defined to measure the importance of each observation in the data fusion process.

[0096] The posterior distribution of the reservoir state estimate is updated using an ensemble Kalman filter algorithm. The ensemble Kalman filter algorithm is a commonly used data assimilation method that continuously updates the reservoir state estimate by combining observed data with model predictions. In this process, the observed data are weighted using an established spatiotemporal covariance matrix and data fusion confidence scores, and then incorporated into the model predictions to obtain the posterior distribution of the reservoir state estimate.

[0097] The posterior distribution is input as the initial condition into the next simulation time step, completing a closed-loop correction of the dynamic data. This allows the simulation results to be corrected using the latest observational data at each simulation time step, enabling the simulation model to better track the actual dynamic changes in the reservoir, improving the accuracy and reliability of the simulation. Furthermore, based on the results of the closed-loop correction, the extraction plan can be adjusted in a timely manner, optimizing the multiphase fluid migration path topology and well placement strategies, thereby improving the efficiency and safety of oil extraction. Example 5:

[0098] When embedding partial differential constraints into the Bayesian inference network, the results of the nonlinear dimensionality reduction are first subjected to Lie group symmetry analysis. Lie group symmetry analysis is a method used to study the symmetries of mathematical and physical systems. In the geological feature decoupling model, Lie group symmetry analysis of the nonlinear dimensionality reduction results can identify correlation patterns that conform to the conservation laws of geomechanics. These conservation laws, including conservation of mass, momentum, and energy, play an important role in constraining geological processes. By analyzing symmetry, it is possible to identify which correlations between geological factors conform to these conservation laws, thereby screening for effective correlation patterns.

[0099] Markov Chain Monte Carlo sampling is used to generate feature evolution paths that satisfy constitutive constraints. Markov Chain Monte Carlo sampling is a random sampling method based on Markov chains that enables efficient sampling in high-dimensional space. In this embodiment, Markov Chain Monte Carlo sampling is used to generate feature evolution paths that satisfy constitutive constraints based on geomechanical constitutive equations and selected correlation patterns. These feature evolution paths reflect the temporal and spatial evolution of geological features and are key to building an accurate geological feature decoupling model.

[0100] Path data is used to perform regularized training on the hyperparameters of the Bayesian inference network. The hyperparameters of the Bayesian inference network have a significant impact on the performance of the model. Regularized training can avoid model overfitting and improve the model's generalization ability. During the training process, the generated feature evolution path data is used as input to adjust and optimize the hyperparameters of the Bayesian inference network to ensure that the model output conforms to the dynamic evolution laws of the reservoir. For example, when adjusting hyperparameters, methods such as cross-validation can be used to select the optimal hyperparameter combination so that the model can perform well on different data sets. In this way, partial differential constraints are effectively embedded in the Bayesian inference network, enabling the geological feature decoupling model to better reflect the actual geological situation and provide reliable support for subsequent geological analysis and exploration decisions. Example 6:

[0101] When implementing the quantum computing optimization framework, the Hamiltonian function for multi-objective optimization is first defined. This function incorporates the dual constraints of increasing reservoir recovery and reducing energy consumption. The Hamiltonian function is a key concept in quantum computing, describing the energy state of a quantum system. In this paper, to achieve multi-objective optimization in oil exploration, increasing reservoir recovery and reducing energy consumption are selected as two key optimization objectives, and a Hamiltonian function is constructed.

[0102] The expression of the Hamiltonian function is: In this formula, Represents the Hamiltonian, which represents the energy of the quantum system. In this application scenario, The smaller the value of is, the closer it is to the optimal combination of synergistic constraints, that is, it can better balance the two goals of increasing reservoir recovery and reducing energy consumption. Representation parameters and The coupling strength reflects the interaction between different parameters. Different geological parameters may be correlated with each other in affecting reservoir recovery and energy consumption. The larger the value of and The stronger the coupling between them, the more comprehensive their influence needs to be considered during the optimization process; Represents binary decision variables, which are used to represent the value states of different parameters. In actual optimization, the optimal parameter combination can be found by adjusting the values ​​of these binary decision variables; It represents the external field effect coefficient, which takes into account the impact of external factors on the system. In oil exploration, external factors may include mining technology, geological environment, etc. These factors will affect reservoir recovery and energy consumption. The value is determined according to specific external conditions.

[0103] The quantum annealing algorithm is used to solve the Hamiltonian's ground state configuration and obtain the optimal cooperative constraint parameter set. The quantum annealing algorithm is an optimization algorithm based on the principles of quantum mechanics, which can find the global optimal solution in a complex energy landscape. In this embodiment, the quantum annealing algorithm is used to solve the defined Hamiltonian function, and the state of the quantum system is continuously adjusted to gradually approach the ground state configuration. In the ground state configuration, the corresponding binary decision variable The value of is the optimal set of collaborative constraint parameters. These parameter sets can be used to optimize the multi-parameter collaborative constraints in the simulation instruction set, thereby improving the accuracy and practicality of geological digital twin models and providing a more scientific basis for oil exploration decision-making. For example, in practical applications, the obtained optimal collaborative constraint parameter sets can be applied to multi-physics coupling models and adaptive meshing models to further optimize the model simulation results, improve the prediction accuracy of reservoir dynamic response, and provide more reliable support for the formulation of oil production plans.

[0104] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0105] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for constructing a geological digital twin model in petroleum exploration, characterized in that: include: Receiving a multi-source heterogeneous geological data stream of a target exploration area, the data stream including seismic reflection wave data, core porosity distribution matrix, and formation pressure gradient time series data; Based on a preset geological feature decoupling model, multi-scale alignment and feature fusion are performed on the multi-source heterogeneous data to generate a three-dimensional geological attribute field distribution map and a reservoir heterogeneity tensor; According to the attribute field distribution map, an adaptive grid partitioning model is constructed to generate a dynamic flow field simulation instruction set, wherein the instruction set includes a multiphase fluid migration path topology and a well location optimization deployment strategy; Based on a preset multi-physics field coupling model, the dynamic response of the target reservoir under mining disturbance is simulated, and the multi-parameter collaborative constraint conditions in the simulation instruction set are optimized; Iteratively update the collaborative constraints through a quantum computing optimization framework, and output the geological digital twin model construction sequence to the exploration decision-making platform; The adaptive grid partitioning model includes: Dynamically dividing the unstructured grid density levels according to the spatial gradient distribution of the three-dimensional geological attribute field distribution map; Calculate the grid cell adaptive refinement weight based on the reservoir permeability anisotropy tensor and fluid viscosity coefficient; Performing nonlinear mapping of the refinement weight and density level through a hyperbolic sine function to generate a multi-resolution nested grid partitioning scheme; triggering the initial boundary condition loading protocol for the fluid-structure interaction simulation according to the scheme; The calculation of the grid unit adaptive refinement weight includes: Obtain formation fracture density distribution map and pore pressure diffusion coefficient matrix, and construct grid quality assessment hypercube; Calculate the singular value contribution of each dimension feature through tensor decomposition algorithm; Performing a Hadamard product operation on the contribution and the hypercube to obtain a comprehensive refinement weight; wherein the calculation formula of the comprehensive refinement weight is: ； Where, represents the comprehensive refinement weight value, Indicates the The singular values ​​of the dimensional features, represents the geological attribute field function, represents the spatial coordinate component, represents the nonlinear amplification exponent, represents the total number of feature dimensions, is the Hadamard product operator, which represents the product of corresponding elements.

2. The geological digital twin model construction method according to claim 1, characterized in that: The steps of constructing the geological feature decoupling model include: Collect historical exploration data sets from multiple blocks and construct a three-dimensional feature training set including formation fracture patterns, spatial distribution of sedimentary phases, and rock mechanical parameters; Performing nonlinear dimensionality reduction on the three-dimensional feature training set using a spectral clustering algorithm to extract independent representations of dominant geological factors and secondary disturbances; Combined with the geomechanical constitutive equation, the partial differential constraints of the dynamic correlation between factors are constructed; The partial differential constraint condition is embedded in a Bayesian inference network to generate the geological feature decoupling model supporting online updating.

3. The geological digital twin model construction method according to claim 1, characterized in that: The steps of constructing the multi-physics coupling model include: Collect the pressure-stress response correlation matrix from historical mining data and construct a multi-field coupling causal diagram dataset; The interaction intensity parameters between the seepage field and the stress field are extracted using the random forest regression algorithm; Combined with implicit solvers, the multi-physics dynamic equilibrium equations are constructed to quantify the relaxation time constants of energy transfer between fields; The relaxation time constant and the real-time mining disturbance factor are input into a cellular automaton network to generate the multi-physics coupling model.

4. The geological digital twin model construction method according to claim 3, characterized in that: Also includes: Identifying reservoir dynamic response sensitive areas based on simulation results of the multi-physics field coupling model; configuring a real-time data assimilation strategy for the region in the dynamic flow field simulation instruction set; Based on the assimilation strategy, a multi-scale observation data fusion scheme is automatically generated, including seismic wave velocity field correction instructions and inter-well interpolation parameter update strategy.

5. The geological digital twin model construction method according to claim 2, characterized in that: The embedding of the partial differential constraint condition includes: Performing Lie group symmetry analysis on the nonlinear dimensionality reduction results to screen out correlation patterns that conform to the conservation laws of geomechanics; Generate feature evolution paths that satisfy constitutive constraints through Markov chain Monte Carlo sampling; The path data is used to regularize the hyperparameters of the Bayesian inference network to ensure that the model output conforms to the dynamic evolution law of the reservoir.

6. The geological digital twin model construction method according to claim 4, characterized in that: The execution of the real-time data assimilation strategy includes: Establish the spatiotemporal covariance matrix of multi-source observation data and define the data fusion confidence score; Updating the posterior distribution of reservoir state estimates by integrating the Kalman filter algorithm; The posterior distribution is input as the initial condition into the next simulation time step to complete the dynamic data closed-loop correction.

7. The method for constructing a geological digital twin model according to claim 1, wherein: The execution of the quantum computing optimization framework includes: Define the Hamiltonian function for multi-objective optimization, including the dual constraints of increasing reservoir recovery and reducing energy consumption; Solving the ground state configuration of the Hamiltonian by a quantum annealing algorithm to obtain an optimal cooperative constraint parameter set; Wherein, the Hamiltonian function is: ； Where, represents the Hamiltonian, Representation parameters and The coupling strength, represents a binary decision variable, Represents the external field effect coefficient.

8. A geological digital twin model construction system for petroleum exploration, characterized by: include: Data receiving module: used to receive multi-source heterogeneous geological data streams in the target exploration area, including seismic reflection wave data, core porosity distribution matrix and formation pressure gradient time series data; Feature processing module: Based on a preset geological feature decoupling model, it performs multi-scale alignment and feature fusion on the multi-source heterogeneous data to generate a three-dimensional geological attribute field distribution map and a reservoir heterogeneity tensor; Grid instruction generation module: constructs an adaptive grid partitioning model based on the attribute field distribution map and generates a dynamic flow field simulation instruction set, which includes a multiphase fluid migration path topology and a well location optimization deployment strategy; Simulation Optimization Module: Based on a preset multi-physics field coupling model, it simulates the dynamic response of the target reservoir under mining disturbance and optimizes the multi-parameter collaborative constraints in the simulation instruction set; Output module: Iteratively updates the collaborative constraints through the quantum computing optimization framework and outputs the geological digital twin model construction sequence to the exploration decision-making platform; The adaptive grid partitioning model includes: Dynamically dividing the unstructured grid density levels according to the spatial gradient distribution of the three-dimensional geological attribute field distribution map; Calculate the grid cell adaptive refinement weight based on the reservoir permeability anisotropy tensor and fluid viscosity coefficient; Performing nonlinear mapping of the refinement weight and density level through a hyperbolic sine function to generate a multi-resolution nested grid partitioning scheme; triggering the initial boundary condition loading protocol for the fluid-structure interaction simulation according to the scheme; The calculation of the grid unit adaptive refinement weight includes: Obtain formation fracture density distribution map and pore pressure diffusion coefficient matrix, and construct grid quality assessment hypercube; Calculate the singular value contribution of each dimension feature through tensor decomposition algorithm; Performing a Hadamard product operation on the contribution and the hypercube to obtain a comprehensive refinement weight; wherein the calculation formula of the comprehensive refinement weight is: ； Where, represents the comprehensive refinement weight value, Indicates the The singular values ​​of the dimensional features, represents the geological attribute field function, represents the spatial coordinate component, represents the nonlinear amplification exponent, represents the total number of feature dimensions, is the Hadamard product operator, which represents the product of corresponding elements.

Images