Braid river reservoir sand body configuration grading identification method and three-dimensional geologic model construction system

Abstract

The invention relates to the technical field of the crossing field of petroleum and natural gas geological exploration and development, geological modeling and computer science, and provides a braided river reservoir sand body configuration three-dimensional geological model construction system which comprises a system control center and four subsystems. The subsystems comprise a data acquisition and preprocessing subsystem, and the configuration identification subsystem comprises a three-dimensional modeling subsystem and a model optimization and verification subsystem; the data acquisition and preprocessing subsystem comprises a multi-source data acquisition module, a data cleaning and standardization module and a data storage and management module; and the configuration identification subsystem comprises a configuration unit period division module, a configuration unit superposition mode identification module and configuration prediction based on spatio-temporal evolution. In a configuration identification link, configuration prediction based on spatio-temporal evolution is introduced, multi-source data spatio-temporal feature fusion, a multi-scale spatio-temporal prediction model and a dynamic model considering a geological process are combined, and the depicting ability of the model to complex features of the sand body configuration is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of the intersection of oil and gas geological exploration and development, geological modeling, and computer science. Specifically, it relates to a method for hierarchical identification of braided river reservoir sandbody configurations and a three-dimensional geological model construction system. Background Art

[0002] Braided river reservoirs play an important role in global oil and gas resources. The accurate analysis of their sandbody configurations is crucial for the successful implementation of oil and gas exploration and development. The research on braided river reservoir configurations involves multiple disciplines such as geology, geophysics, and computer science. It aims to construct an accurate three-dimensional geological model through the collection and analysis of geological data of underground reservoirs, helping geologists and engineers understand the internal structure of the reservoir and the law of fluid migration, and then guiding the efficient development of oil and gas fields.

[0003] In the past, the research on braided river reservoir sandbody configurations mainly relied on single or limited data sources and used relatively simple analysis methods. In terms of data collection, usually only downhole data such as cores and well logs were used, ignoring the comprehensive use of multi-source information such as seismic data, outcrops, and modern river sedimentation characteristics. In the configuration analysis link, the configuration units were divided and identified based on simple geological theories, which was difficult to accurately reflect the complex spatio-temporal evolution characteristics of braided river reservoirs. In the three-dimensional modeling stage, static modeling methods were mostly used, and the dynamic changes of geological processes were not fully considered, resulting in the constructed models lacking an accurate description of the real situation of the reservoir.

[0004] Due to the single data source of traditional technologies, they cannot comprehensively reflect reservoir characteristics. The present invention introduces configuration prediction based on spatio-temporal evolution, which can better depict the dynamic changes of the reservoir. In contrast, the static analysis methods of traditional technologies are difficult to meet the need for understanding complex geological processes. In addition, the efficient cooperation and data feedback mechanism among the subsystems of the system of the present invention ensure the accuracy and reliability of the model, while traditional technologies lack an effective data interaction and model optimization mechanism, resulting in low-precision models that cannot provide a reliable basis for oil and gas development. Summary of the Invention

[0005] The present invention proposes a method for hierarchical identification of braided river reservoir sandbody configurations and a three-dimensional geological model construction system, which solves the technical problems of incomplete data integration, the difficulty of the model in reflecting the spatio-temporal evolution characteristics of sandbody configurations, and the inability to provide accurate and dynamic decision-making support for geological analysis and oil and gas development.

[0006] The technical solution of the present invention is as follows:

[0007] A method for hierarchical identification of braided river reservoir sandbody configurations includes the following steps:

[0008] Step 1. Multi-source information integration and analogy: Collect core, logging, and seismic data of the underground reservoir in the target area, as well as information on ancient fluvial outcrops and modern fluvial sedimentary characteristics to construct a multi-source database; compare the underground reservoir data with the sedimentary characteristics of ancient and modern rivers, and analyze the braided river sedimentation pattern and channel change law.

[0009] Step 2. Configuration unit stage division: According to the Miall configuration classification scheme, determine the single-stage channel as a fifth-level configuration unit; by analyzing core and logging data, identify the superimposed and cutting relationships of configuration elements between braided channels and between braided channels within a single stage, and construct a sand body general connectivity model.

[0010] Step 3. Identification of configuration unit superimposition patterns: Integrate well data and seismic data to identify isolated configuration sand bodies in braided channels; use logging curve analysis techniques to identify superimposed configuration sand bodies; combine core observations and logging curve analysis to identify cut-superimposed configuration sand bodies.

[0011] Furthermore, in Step 2, during the configuration unit stage division, further subdivide the single-stage channel, and clarify the boundaries and characteristics of each subdivision unit based on sedimentary characteristics and logging responses.

[0012] In Step 3, the identification of configuration unit superimposition patterns includes the identification of isolated sand bodies, superimposed sand bodies, and cut-superimposed sand bodies.

[0013] The three-dimensional geological model construction system for the braided river reservoir sand body configuration includes a system control center and four subsystems. The subsystems include: a data acquisition and preprocessing subsystem, a configuration identification subsystem, a three-dimensional modeling subsystem, and a model optimization and verification subsystem.

[0014] The data acquisition and preprocessing subsystem includes: a multi-source data acquisition module, a data cleaning and standardization module, and a data storage and management module.

[0015] The configuration identification subsystem includes: a configuration unit stage division module, a configuration unit superimposition pattern identification module, and a configuration prediction based on spatio-temporal evolution.

[0016] The three-dimensional modeling subsystem includes: a stratigraphic framework construction module and a sand body configuration modeling module.

[0017] The model optimization and verification subsystem includes: a model optimization module and a model verification module.

[0018] The data acquisition and preprocessing subsystem transmits the processed data to the configuration identification subsystem. The configuration identification subsystem transmits the identification results to the three-dimensional modeling subsystem. The three-dimensional modeling subsystem transmits the constructed model to the model optimization and verification subsystem. The model optimization and verification subsystem feeds back the optimized model to the three-dimensional modeling subsystem and feeds back the verification results to the data acquisition and preprocessing subsystem.

[0019] Furthermore, the configuration recognition subsystem introduces configuration prediction based on spatio-temporal evolution, and the specific implementation includes:

[0020] Stage 1: Spatio-temporal feature fusion of multi-source data;

[0021] Stage 2: Establish a multi-scale spatio-temporal prediction model;

[0022] Stage 3: Consider a dynamic model of geological processes.

[0023] Furthermore, in the above-mentioned Stage 1, an attention mechanism neural network is used to fuse spatio-temporal feature vectors. When constructing the spatio-temporal feature vectors of multi-source data, for the eigenvalues obtained after feature extraction of different types of geological data, when combining various eigenvalues into spatio-temporal feature vectors, weights are assigned according to the importance of the features. Let the eigenvalue of geological outcrop data be E, the eigenvalue of seismic data be S, and the eigenvalue of logging data be L, and the assigned weights be w E , w S , w L , and w E + w S + w L = 1, then the spatio-temporal feature vector V = w E E + w S S + w L L.

[0024] Furthermore, in the above-mentioned Stage 1, when fusing spatio-temporal feature vectors, the importance of each spatio-temporal feature vector is determined. After obtaining the attention score, it is converted into an attention weight, and the fused feature vector is calculated according to the attention weight. Through the method of weighted summation, the fused feature vector is obtained, and different feature vectors are combined according to their importance to generate a new feature vector.

[0025] Furthermore, in the above-mentioned Stage 2, a spatio-temporal prediction model including micro, meso, and macro scales is constructed.

[0026] Furthermore, in the above-mentioned Stage 2, the geological process parameters of braided rivers are incorporated into the model. Historical hydrological data and sediment source area information are collected, and their relationship with the spatio-temporal evolution of sandbody configuration is analyzed. By establishing a mathematical and physical model between geological processes and sandbody configuration, the formation, migration, and evolution processes of sandbodies under different geological conditions are simulated. Using a coupling algorithm, the geological process model is combined with the deep learning model.

[0027] Furthermore, in the above-mentioned Stage 2, the unit structure of the deep learning model involves the following specific formula:

[0028] Input gate output: i t = σ(W ii x t+b ii +W hi h t-1 +b hi )

[0029] Among them, W ii is the weight matrix from the input x t to the input gate, and b ii is the corresponding bias term; W hi is the weight matrix from the hidden state h t-1 at the previous moment to the input gate, and b hi is the corresponding bias term. The input gate determines how much information of the current input x t will be written into the memory cell;

[0030] Output of the forget gate: f t = σ(W if x t +b if +W hf h t-1 +b hf )

[0031] Among them, W if , b if , W hf , b hf are the corresponding weight matrices and bias terms respectively. The forget gate determines which information in the memory cell c t-1 will be retained in the memory cell c t at the current moment;

[0032] Output of the candidate value: g t = tanh(W ig x t +b ig +W hg h t-1 +b hg )

[0033] Among them, W ig , b ig , W hg , b hg are the corresponding parameters. The candidate value g t is a potential update value calculated based on the current input x t and the hidden state h t-1 at the previous moment;

[0034] Output of the output gate: o t = σ(W io x t +b io +W ho h t-1 +b ho)

[0035] Among them, W io , b io , W ho , b ho are the corresponding weights and biases.

[0036] The working principle and beneficial effects of the present invention are as follows:

[0037] 1. In the three-dimensional geological model construction system for braided river reservoir sandbody configuration constructed in the present invention, with the help of the multi-source data acquisition and preprocessing mechanism, geological data such as cores, logging, and seismic data, as well as geological process parameters such as hydrodynamic force and sediment supply, are integrated to provide comprehensive and accurate data support for model construction. In the configuration recognition link, the configuration prediction based on spatio-temporal evolution is introduced, combined with the spatio-temporal feature fusion of multi-source data, the multi-scale spatio-temporal prediction model, and the dynamic model considering geological processes, to comprehensively capture the spatio-temporal evolution laws of sandbody configuration at the micro, meso, and macro scales. Through the collaborative work of each subsystem, the system can construct a high-precision three-dimensional geological model, which effectively improves the ability of the model to depict the complex characteristics of sandbody configuration compared with traditional modeling methods, and provides a more reliable model basis for geological research.

[0038] 2. In the present invention, the high-precision three-dimensional geological model constructed by the system can intuitively display the braided river reservoir sandbody configuration, providing clear geological structure information for geologists to help them deeply analyze geological characteristics such as the sedimentation pattern and river channel change law of the braided river. In the field of oil and gas development, the system helps petroleum engineers to master the sandbody distribution and change trend in advance by dynamically predicting the spatio-temporal evolution of sandbody configuration, optimize the oil and gas exploitation plan, improve the exploitation efficiency, and reduce the development risk. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The present invention will be further described in detail below with reference to the drawings and specific embodiments.

[0040] Figure 1 It is a cross-relationship diagram of data sources for the recognition of the stacked pattern of braided river reservoir sandbody configuration units;

[0041] Figure 2 It is a system architecture and data flow diagram for the construction of a three-dimensional geological model of braided river reservoir sandbody configuration;

[0042] Figure 3 It is a flowchart for realizing the configuration prediction based on spatio-temporal evolution in the braided river reservoir sandbody configuration recognition subsystem. SPECIFIC EMBODIMENTS

[0043] Next, in combination with the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of the present invention.

[0044] Embodiment 1

[0045] A method for hierarchical identification of braided river reservoir sandbody configuration includes multi-source information integration and analogy, division of configuration unit periods, and identification of configuration unit stacking patterns;

[0046] Step 1: Multi-source information integration and analogy

[0047] Information collection: Collect geological data of the underground reservoir in the target area, including core, logging, seismic and other data. At the same time, widely collect paleo-river facies outcrop and modern river sedimentary characteristic information to construct a multi-source database. This is because the study of underground reservoirs is relatively abstract, and by analogy with intuitive outcrops and modern sedimentary characteristics, abstract concepts can be made concrete and provide a reference for subsequent analysis.

[0048] Feature analogy: Compare the underground reservoir data with the sedimentary characteristics of paleo-rivers and modern rivers, and deeply analyze the sedimentation patterns, river channel change laws, etc. of braided rivers to provide a basis for the analysis of configuration units.

[0049] Step 2: Division of configuration unit periods

[0050] Determination of the fifth-level configuration unit: According to the Miall configuration classification scheme, determine the single-stage river channel as the fifth-level configuration unit. Through the fine analysis of core and logging data, identify the stacking and cutting relationships between the two configuration elements of braided river channels and between braided river channels within a single stage, and construct a sandbody general connectivity model. For example, ZVx1-1-single sand layer can be longitudinally divided into ZVx-1-1a and ZVx-1-1b first-stage braided channel sandbodies and between braided river channels.

[0051] Subdivision of internal units: Further subdivide the single-stage river channel. For example, ZVx-1-1b is identified as a single braided river channel with a large scale; ZVx-1-a is mainly composed of sedimentation between braided river channels. According to sedimentary characteristics and logging responses, clarify the boundaries and characteristics of each subdivided unit.

[0052] Step 3: Identification of configuration unit stacking patterns

[0053] Isolated sand body identification: By comprehensively analyzing well data and seismic data, isolated configuration sand bodies in braided channels are identified. Given the characteristics of braided rivers, namely "wide bars and narrow channels", such sand bodies have a small lateral distribution scale, a short lateral extension distance, and are vertically wrapped by thick floodplain mudstones. In areas with dense well patterns, their maximum extension distance does not exceed two well spacings. This identification helps to accurately depict the distribution pattern of reservoir sand bodies.

[0054] Stacked sand body identification: Using logging curve analysis technology, stacked configuration sand bodies are identified. Such sand bodies represent the stacking of two or more relatively complete channel sedimentary cycles, and there are fine-grained interlayers or sedimentary discontinuity surfaces between sand bodies of different periods. The logging response characteristics show that the spontaneous potential and microelectrode curves of the upper and lower configuration sand bodies are box-shaped or bell-shaped, and the fine-grained sediments in the middle show significant return characteristics. By analyzing these characteristics, the stacking relationship and periods of the sand bodies are determined.

[0055] Cut-and-stack sand body identification: Combining core observation and logging curve analysis, cut-and-stack configuration sand bodies are identified. This mode is formed by the strong downcutting action of the river. The lower fine-grained sediments or sand bodies are scoured by later hydrodynamic forces, resulting in the direct contact of two-stage channel sand bodies. In the core, the interface can be determined by identifying the grain size difference and scouring surface of the two-stage configuration sand bodies; in the logging response, there is a weak return between the spontaneous potential and microelectrode curves of adjacent two-stage configuration sand bodies, and there is a physical property change section between the top of the lower configuration unit and the bottom of the upper configuration unit, which is used as the third or fourth level interface for configuration unit division. However, it is necessary to reasonably layer and interpret it in combination with the well data trend and sedimentation law of adjacent connected wells.

[0056] In the identification of isolated sand bodies, well data and seismic data are integrated; while for the logging curve analysis technology in the identification of stacked sand bodies, and the core observation and logging curve analysis combined in the identification of cut-and-stack sand bodies, there is an overlap in the above data sources. For example, seismic data can provide a macroscopic background of sand body distribution, indicating the direction for the identification of isolated sand bodies, and its results can also assist in judging the regional distribution of cut-and-stack sand bodies. At the same time, the core data and logging curves in well data can not only help in the identification of cut-and-stack sand bodies, but also provide the key curve analysis basis for the identification of stacked sand bodies.

[0057] Example 2

[0058] A three-dimensional geological model construction system for the configuration of braided river reservoir sand bodies integrates multi-source data to construct a high-precision three-dimensional geological model, intuitively displays the configuration of braided river reservoir sand bodies, and assists in geological analysis and oil and gas development decision-making. The system consists of four subsystems: data acquisition and preprocessing, configuration identification, three-dimensional modeling, and model optimization and verification. Each subsystem is both independent and collaborative to ensure the accuracy and efficiency of model construction;

[0059] 1. The data acquisition and preprocessing subsystem includes: a multi-source data acquisition module, a data cleaning and standardization module, and a data storage and management module;

[0060] 2. The configuration recognition subsystem includes: a configuration unit stage division module, a configuration unit superposition pattern recognition module, and a configuration prediction based on spatio-temporal evolution;

[0061] 3. The 3D modeling subsystem includes: a formation framework construction module, a sand body configuration modeling module;

[0062] 4. The model optimization and verification subsystem includes: a model optimization module, a model verification module;

[0063] The data acquisition and preprocessing subsystem transmits the collected, cleaned, and standardized data to the configuration recognition subsystem to provide data support for the configuration unit stage division and superposition pattern recognition; the configuration recognition subsystem transmits the recognition results to the 3D modeling subsystem for constructing a 3D geological model; the 3D modeling subsystem transmits the constructed model to the model optimization and verification subsystem for model optimization and verification; the model optimization and verification subsystem feeds back the optimized model to the 3D modeling subsystem and at the same time feeds back the verification results to the data acquisition and preprocessing subsystem to guide the further acquisition and processing of data. Each subsystem works collaboratively under the unified scheduling of the system control center; the system control center reasonably allocates tasks to each subsystem according to the user's operation instructions and monitors the running status of the subsystem in real time; when a certain subsystem fails or is abnormal, the system control center can make timely adjustments to ensure the stable operation of the system. This system integrates multi-source data, realizes high-precision 3D modeling of the braided river reservoir sand body configuration, comprehensively captures the evolution laws of the sand body configuration at the micro, meso, and macro scales by introducing a configuration prediction based on spatio-temporal evolution. At the same time, geological process parameters such as hydrodynamic conditions and sediment supply are incorporated into the model to realize the dynamic prediction of the sand body configuration, providing a reliable basis for 3D modeling;

[0064] The configuration prediction based on spatio-temporal evolution introduced in the configuration recognition subsystem includes the following stages:

[0065] Stage 1. Spatio-temporal feature fusion of multi-source data: For different types of geological data, such as geological outcrop data from a long time ago, seismic monitoring data from different periods, logging data over the years, etc., feature extraction is carried out in the time and space dimensions. For geological outcrop data, combined with its geographical location and collection time, the changing trend of the sand body configuration in spatial distribution over time is analyzed. For seismic data, time-frequency analysis technology is used to extract the changing characteristics of different frequency components in time and space, thereby constructing a spatio-temporal feature vector of multi-source data. The attention mechanism neural network is used to fuse the spatio-temporal feature vectors to highlight the key features and improve the model's learning ability of the spatio-temporal evolution law of the sand body configuration.

[0066] When constructing the spatio-temporal feature vector of multi-source data, the eigenvalues obtained from different types of geological data after feature extraction, as well as the relevant parameters for determining the dimension and weight of the feature vector, when combining various eigenvalues into a spatio-temporal feature vector, weights are assigned according to the importance of the features. Assume the eigenvalue of geological outcrop data is E, the eigenvalue of seismic data is S, and the eigenvalue of logging data is L, and the assigned weights are w E , w S , w L , and w E +w S +w L =1, then the spatio-temporal feature vector V = w E E + w S S + w L L.

[0067] When fusing spatio-temporal feature vectors, the following parameters are required, including spatio-temporal feature vectors, structural parameters of the attention mechanism neural network, and weight parameters (used to calculate attention scores and weights);

[0068] To determine the importance of each feature vector, first generate a query vector Q i and a key vector K j . For example, for the input feature vector V, the query vector and key vector are calculated through Q = WV and K = W'V, where W and W' are weight matrices.

[0069] After obtaining the attention score, it needs to be converted into an attention weight to determine the contribution of each feature vector in the fusion process. This conversion is achieved through the Softmax function:

[0070]

[0071] The Softmax function maps the attention scores to between 0 and 1, and the sum of the weights of all scores is 1;

[0072] According to the attention weights, calculate the fused feature vector. The value vector V j is the same as the input spatio-temporal feature vector, and the fused feature vector is obtained through weighted summation:

[0073]

[0074] Combine different feature vectors according to their importance to generate a new and more representative feature vector, providing better data for the subsequent model to learn the spatio-temporal evolution law of sand body configuration

[0075] Phase 2: Establish a multi-scale spatio-temporal prediction model: Construct a spatio-temporal prediction model that includes micro-scale, meso-scale, and macro-scale. At the micro-scale, based on core thin-section data, analyze the changes in the internal particle arrangement and pore structure of the sand body over time; at the meso-scale, combine well logging data to study the spatio-temporal evolution of the bedding structure and lithology combination within the sand body; at the macro-scale, based on seismic data, grasp the spatial distribution and large-scale migration law of the entire braided river reservoir sand body. Through a scale conversion module, realize the transfer and fusion of information between different scales, enabling the model to comprehensively reflect the spatio-temporal evolution characteristics of the sand body configuration at different scales.

[0076] Phase 3: Dynamic model considering geological processes: Incorporate geological process parameters such as hydrodynamic conditions and sediment supply of the braided river into the model. Collect historical hydrological data and sediment source area information, and analyze their relationship with the spatio-temporal evolution of the sand body configuration. By establishing a mathematical and physical model between geological processes and the sand body configuration, simulate the formation, migration, and evolution processes of the sand body under different geological conditions. Using a coupling algorithm, combine the geological process model with the deep learning model, enabling the model to dynamically predict the spatio-temporal evolution of the sand body configuration according to real-time changes in geological conditions.

[0077] When combining the geological process model with the deep learning model, determine the coupling method and related parameters. In data fusion, determine the weights of the output of the geological process model and the input of the deep learning model to balance the contributions of both in prediction;

[0078] Input gate output: i t = σ(W ii x t + b ii + W hi h t-1 + b hi )

[0079] Where, W ii is the weight matrix from the input x t to the input gate, b ii is the corresponding bias term; W hi is the weight matrix from the previous hidden state h t-1 to the input gate, b hi is the corresponding bias term. The input gate determines how much information of the current input x t will be written into the memory cell. In the prediction of the sand body configuration, the input gate controls the influence degree of new geological data on the update of the memory cell.

[0080] Forget gate output: f t = σ(W if x t + b if + W hf h t-1 + bhf )

[0081] Among them, W if , b if , W hf , b hf are the corresponding weight matrices and bias terms respectively. The forget gate determines which information in the memory cell c t-1 will be retained in the memory cell c t at the current moment. During the long-term evolution of sand body configuration, the forget gate can judge whether some sand body deposition pattern information saved before is still applicable to the current geological conditions, so as to decide whether to retain this information in the memory cell.

[0082] Candidate value output: g t =tanh(W ig x t +b ig +W hg h t-1 +b hg )

[0083] Among them, W ig , b ig , W hg , b hg are the corresponding parameters. The candidate value g t is a potential update value calculated based on the current input x t and the hidden state h t-1 at the previous moment, and is used to fuse with the information in the memory cell c t-1 . When predicting the sand body configuration, the candidate value g t calculates a possible sand body configuration change trend value according to the current geological data for subsequent integration with the memory cell information.

[0084] Output gate output: o t =σ(W io x t +b io +W ho h t-1 +b ho )

[0085] Among them, W io , b io , W ho , b ho are the corresponding weights and biases. The output gate controls how much information in the memory cell c t will be output to the hidden state h t at the current moment. When outputting the prediction result of the sand body configuration, the output gate determines the content and proportion of the final output prediction information about the sand body configuration.

[0086] By integrating multi-source data, considering geological processes and spatio-temporal evolution, accurately identify the spatio-temporal characteristics of sandbody architecture, provide high-quality input data for the subsequent 3D modeling subsystem, and ultimately achieve the function of high-precision 3D modeling of the system;

[0087] Through the spatio-temporal feature fusion of multi-source data, mine the spatio-temporal information in different types of geological data; establish a multi-scale spatio-temporal prediction model to comprehensively reflect the spatio-temporal evolution characteristics of sandbody architecture at different scales; consider a dynamic model of geological processes, combine geological process parameters with a deep learning model, so that the model can dynamically predict the spatio-temporal evolution of sandbody architecture according to real-time geological condition changes. The above measures aim to improve the model's learning and prediction ability of the spatio-temporal evolution law of sandbody architecture, and provide a more reliable basis for geological analysis and oil and gas development decision-making.

[0088] The above are only the preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for hierarchical identification of braided river reservoir sandbody configuration, characterized in that It includes the following steps: Step 1, Multi-source information integration and analogy: Collect core, logging, and seismic data of the underground reservoir in the target area, as well as paleo-river facies outcrop and modern river sedimentation characteristic information to construct a multi-source database; Compare the underground reservoir data with the sedimentation characteristics of paleo-rivers and modern rivers, and analyze the braided river sedimentation pattern and river channel change law; Step 2, Configuration unit stage division: According to the Miall configuration classification scheme, determine that a single-stage river channel is a fifth-level configuration unit; By analyzing core and logging data, identify the superimposed cutting relationship between braided river channels and configuration elements between braided river channels within a single-stage river channel, and construct a sand body general connectivity model; Step 3, Configuration unit superimposed pattern recognition: Integrate well data and seismic data to identify isolated configuration sand bodies in braided river channels; Use logging curve analysis technology to identify superimposed configuration sand bodies; Combine core observation and logging curve analysis to identify cut-overlapping configuration sand bodies.

2. The hierarchical identification method of braided river reservoir sand body configuration according to claim 1, wherein, In Step 2, during the configuration unit stage division, further subdivide the single-stage river channel, and clarify the boundaries and characteristics of each subdivision unit based on sedimentation characteristics and logging responses; In Step 3, the configuration unit superimposed pattern recognition includes isolated sand body recognition, superimposed sand body recognition, and cut-overlapping sand body recognition.

3. A three-dimensional geological model construction system for braided river reservoir sandbody configuration, characterized in that It includes a system control center and four subsystems. The subsystems include: a data acquisition and preprocessing subsystem, a configuration recognition subsystem, a 3D modeling subsystem, and a model optimization and verification subsystem; The data acquisition and preprocessing subsystem includes: a multi-source data acquisition module, a data cleaning and standardization module, and a data storage and management module; The configuration recognition subsystem includes: a configuration unit stage division module, a configuration unit superimposed pattern recognition module, and a configuration prediction based on spatio-temporal evolution; The 3D modeling subsystem includes: a stratigraphic framework construction module and a sand body configuration modeling module; The model optimization and verification subsystem includes: a model optimization module and a model verification module; The data acquisition and preprocessing subsystem transmits the processed data to the configuration recognition subsystem, the configuration recognition subsystem transmits the recognition results to the 3D modeling subsystem, the 3D modeling subsystem transmits the constructed model to the model optimization and verification subsystem, and the model optimization and verification subsystem feeds back the optimized model to the 3D modeling subsystem and feeds back the verification results to the data acquisition and preprocessing subsystem.

4. The three-dimensional geological model construction system for braided river reservoir sandbody architecture according to claim 3, characterized in that In the configuration recognition subsystem, a configuration prediction based on spatio-temporal evolution is introduced. The specific implementation includes: Stage 1, Spatio-temporal feature fusion of multi-source data; Stage 2, Establish a multi-scale spatio-temporal prediction model; Stage 3, A dynamic model considering geological processes.

5. The three-dimensional geological model construction system for braided river reservoir sandbody configuration according to claim 4, wherein In the first stage, an attention mechanism neural network is used to fuse spatio-temporal feature vectors. When constructing spatio-temporal feature vectors of multi-source data, for the eigenvalues obtained after feature extraction of different types of geological data, weights are assigned according to the importance of the features. Let the eigenvalue of geological outcrop data be E, the eigenvalue of seismic data be S, and the eigenvalue of logging data be L. The assigned weights are w E , w S , w L , and w E + w S + w L = 1. Then the spatio-temporal feature vector V = w E E + w S S + w L L.

6. The three-dimensional geological model construction system for braided river reservoir sand body configuration according to claim 5, characterized in that, In Stage 1, when fusing spatio-temporal feature vectors, determine the importance of each spatio-temporal feature vector. After obtaining the attention score, convert it into an attention weight, and calculate the fused feature vector according to the attention weight. By means of weighted summation, obtain the fused feature vector, and combine different feature vectors according to their importance to generate a new feature vector.

7. The three-dimensional geological model construction system for braided river reservoir sand body configurations according to claim 4, characterized in that In Stage 2, construct a spatio-temporal prediction model including micro, meso, and macro scales.

8. The three-dimensional geological model construction system for braided river reservoir sandbody configuration according to claim 4, characterized in that In the second stage, the geological process parameters of braided rivers are incorporated into the model, historical hydrological data and sediment source area information are collected, the relationship between them and the spatio-temporal evolution of sandbody configuration is analyzed, and by establishing a mathematical and physical model between geological processes and sandbody configuration, the formation, migration and evolution processes of sandbodies under different geological conditions are simulated. Using a coupling algorithm, the geological process model is combined with the deep learning model.

9. The three-dimensional geological model construction system for braided river reservoir sand body configuration according to claim 8, characterized in that, In the second stage, the unit structure of the deep learning model involves the following specific formulas: Input gate output: i t = σ(W ii x t + b ii + W hi h t-1 + b hi ) Among them, W ii is the weight matrix from the input x t to the input gate, and b ii is the corresponding bias term; W hi is the weight matrix from the hidden state h t-1 at the previous moment to the input gate, and b hi is the corresponding bias term. The input gate determines how much information of the current input x t will be written into the memory cell; Forgotten gate output: f t = σ(W if x t + b if + W hf h t-1 + b hf ) Among them, W if , b if , W hf , b hf are the corresponding weight matrix and bias term respectively. The forget gate determines which information in the memory cell c t-1 will be retained in the memory cell c t at the current moment; Candidate value output: g t = tanh(W ig x t + b ig + W hg h t-1 + b hg ) Among which W ig , b ig , W hg , b hg are corresponding parameters, and the candidate value g t is a potential update value calculated based on the current input x t and the previous hidden state h t-1 ; Output gate output: o t = σ(W io x t + b io + W ho h t-1 + b ho ) where W io , b io , W ho , b ho are the corresponding weights and biases.

Images