Karst area shale gas intelligent prediction method and system based on artificial intelligence

Abstract

The present application provides a kind of karst area shale gas intelligent prediction method and system based on artificial intelligence, it is related to artificial intelligence technical field, first, the multi-source exploration data set of target karst area is acquired, including three-dimensional seismic data body, well logging curve data and rock thin section microscopic image data;Then carry out geological entity and geological relationship double extraction, build karst geological knowledge network;Then input knowledge network with multi-source data to generate knowledge enhanced exploration feature representation;Call geological process coupling simulation model to process the feature representation, simulate karst and shale reservoir accumulation interaction, predict shale gas enrichment zone space distribution;Finally, based on the weight iteration optimization of knowledge network to prediction result.This application can comprehensively multi-source data, accurately predict shale gas enrichment zone.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and more specifically, to an intelligent prediction method and system for shale gas in karst areas based on artificial intelligence. Background Technology

[0002] In the field of energy exploration, shale gas, as a clean and efficient unconventional natural gas resource, plays a vital role in ensuring national energy security and optimizing the energy structure. Due to its unique geological structure and diagenetic environment, the occurrence and enrichment patterns of shale gas in karst areas differ significantly from those in conventional areas, making the prediction of shale gas in karst regions an extremely challenging task.

[0003] Traditional shale gas prediction methods primarily rely on single types of exploration data, such as analyzing only 3D seismic data or well logging data. However, the information contained in a single data source is limited, making it difficult to comprehensively and accurately reflect the complex geological characteristics of karst areas and the occurrence conditions of shale gas. For example, while 3D seismic data can provide a general outline of the subsurface structure, its ability to characterize microscopic geological features is insufficient; well logging data can obtain geological information around the wellbore, but lacks an understanding of the overall geological structure of the region. Furthermore, existing methods often lack effective fusion mechanisms when processing multi-source data, failing to fully leverage the synergistic effects between different data sources, thus affecting the accuracy and reliability of the prediction results. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide an intelligent prediction method and system for shale gas in karst areas based on artificial intelligence.

[0005] In conjunction with the first aspect of this application, an artificial intelligence-based intelligent prediction method for shale gas in karst areas is provided, applied to an artificial intelligence-based intelligent prediction system for shale gas in karst areas, the method comprising: Acquire a multi-source exploration data set of the target karst area, the multi-source exploration data set including a three-dimensional seismic data volume, a well logging curve data set, and a rock thin section microscopic image data set; The multi-source exploration data set is subjected to dual extraction operations of geological entities and geological relationships to obtain a set of karst geological entity units and a set of karst geological relationship units. Based on the set of karst geological entity units and the set of karst geological relationship units, a karst geological knowledge network is constructed. The network nodes in the karst geological knowledge network correspond to the geological entities in the set of karst geological entity units, and the network edges in the karst geological knowledge network correspond to the geological relationships in the set of karst geological relationship units. The multi-source exploration data set is input into the karst geological knowledge network for knowledge-guided collaborative feature representation to generate knowledge-enhanced exploration feature representations. The knowledge-enhanced exploration feature representation is processed by calling the geological process coupling simulation model to simulate the interaction between karst development and shale reservoir formation, generate karst reservoir coupling evolution simulation results, and predict the spatial distribution of shale gas enrichment zones in the target karst area based on the karst reservoir coupling evolution simulation results. Based on the predicted spatial distribution of shale gas enrichment zones, the weights of nodes and edges in the karst geological knowledge network are iteratively optimized to update the structure of the karst geological knowledge network.

[0006] In conjunction with the second aspect of this application, an artificial intelligence-based intelligent prediction system for karst shale gas is provided. The artificial intelligence-based intelligent prediction system for karst shale gas includes a machine-readable storage medium and a processor. The machine-readable storage medium stores machine-executable instructions. When the processor executes the machine-executable instructions, the artificial intelligence-based intelligent prediction system for karst shale gas implements the aforementioned artificial intelligence-based intelligent prediction method for karst shale gas.

[0007] In conjunction with a third aspect of this application, a computer-readable storage medium is provided, wherein computer-executable instructions are stored therein, and when the computer-executable instructions are executed, the aforementioned artificial intelligence-based intelligent prediction method for shale gas in karst areas is implemented.

[0008] Combining any of the above aspects, multi-source exploration data, including 3D seismic data, well logging curve data, and thin-section microscopic image data of the target karst area, are acquired. A karst geological knowledge network is constructed by performing dual extraction operations on geological entities and geological relationships. Multi-source exploration data is input into this knowledge network for knowledge-guided feature co-representation, generating knowledge-enhanced exploration feature representations. This effectively integrates information from different data sources. A geological process coupling simulation model is then used to process these knowledge-enhanced exploration feature representations, simulating the interaction between karst development and shale reservoir formation. This more realistically reflects the formation and enrichment mechanisms of shale gas, generating more reliable karst reservoir coupling evolution simulation results. Based on the prediction results, the weights of nodes and edges in the karst geological knowledge network are iteratively optimized, continuously updating the network structure to better adapt to actual geological conditions, further improving the accuracy and reliability of subsequent predictions. Attached Figure Description

[0009] Figure 1 This application provides a schematic flowchart of an artificial intelligence-based intelligent prediction method for shale gas in karst areas. Detailed Implementation

[0010] Figure 1This paper illustrates a flowchart of an artificial intelligence-based intelligent prediction method for shale gas in karst areas, provided in an embodiment of this application. The method includes, in detail: Step S110: Obtain a multi-source exploration data set for the target karst area, the multi-source exploration data set including a three-dimensional seismic data volume, a well logging curve data set, and a rock thin section microscopic image data set.

[0011] In this embodiment, the target karst area is a specific shale gas exploration region. The 3D seismic data volume was acquired through seismic exploration technology, and its data format is a 3D matrix, containing seismic wave reflection information at different depths and locations underground. The well logging curve data set comes from multiple existing wells in the region, and the well logging curve data for each well includes sequential data of various logging parameters such as natural gamma, resistivity, and sonic transit time as a function of depth. The rock thin section microscopic image data set is obtained by preparing thin sections from core samples collected in the region and then taking images using a microscope. The images have a specific resolution and contain information on the microstructure of the rocks. The acquisition process of the above data all follows relevant geological exploration specifications and data acquisition standards to ensure the accuracy and reliability of the data.

[0012] Step S120: Perform a dual extraction operation on the multi-source exploration data set to obtain a set of karst geological entity units and a set of karst geological relationship units. Based on the set of karst geological entity units and the set of karst geological relationship units, construct a karst geological knowledge network. The network nodes in the karst geological knowledge network correspond to the geological entities in the set of karst geological entity units, and the network edges in the karst geological knowledge network correspond to the geological relationships in the set of karst geological relationship units.

[0013] In this embodiment, the dual extraction of geological entities and geological relationships from the multi-source exploration data set is the foundation for the subsequent construction of the karst geological knowledge network. Through comprehensive analysis and processing of the 3D seismic data volume, well logging curve data set, and rock thin section microscopic image data set, the geological entity and geological relationship information contained therein is extracted.

[0014] Step S121: Perform seismic attribute extraction on the three-dimensional seismic data volume to obtain multiple seismic attribute data volumes.

[0015] In this embodiment, various seismic attribute extraction algorithms are used to process the acquired 3D seismic data volume. For example, the root mean square amplitude attribute is calculated by taking the square root of the average of the seismic amplitude values ​​within a certain time window around each sampling point in the seismic data volume; the coherence volume attribute is calculated by comparing the similarity between adjacent seismic traces to reflect the continuity of the underground geological body; and the instantaneous frequency attribute is calculated by processing the analytical signal of the seismic data. All of the above-mentioned seismic attribute data volumes are stored in the form of a 3D matrix, with its spatial dimension consistent with the original 3D seismic data volume. Each attribute data volume reflects different characteristics of the underground geological structure.

[0016] Step S122: Perform karst fracture and cavern spatial morphology identification on the various seismic attribute data volumes to identify the three-dimensional spatial boundaries of the karst fractures and caverns and generate karst fracture and cavern spatial morphology data; perform lithological stratification and physical property parameter interpretation on the well logging curve data set to obtain lithological stratification data and reservoir physical property parameter data.

[0017] In this embodiment, a model-based identification method is used to identify the spatial morphology of karst fissures and cavities for various seismic attribute data volumes. First, a geological model of the karst fissures and cavities is constructed to clarify their response characteristics in terms of seismic attributes. Then, pattern recognition algorithms, such as support vector machines or neural networks, are used to analyze the seismic attribute data volumes, identifying regions that conform to the characteristics of karst fissures and cavities, thereby determining their three-dimensional spatial boundaries. The generated spatial morphology data of karst fissures and cavities includes information such as the spatial coordinate range and volume of each fissure or cavity. For the well logging curve dataset, lithological stratification is performed by analyzing the characteristics of different well logging curves, combined with geological experience and interpretation models. The strata are divided into different lithological segments, resulting in lithological stratification data. This data is indexed by depth, recording the lithological type corresponding to each depth point. Simultaneously, based on the well logging curve data, reservoir porosity, permeability, and other physical property parameters are calculated using the interpretation model. These data also correspond to depth.

[0018] Step S123: Based on the set of rock thin section microscopic image data, perform microscopic pore structure and mineral component extraction operations to obtain microscopic pore structure data and mineral component data.

[0019] In this embodiment, image analysis techniques are used to extract the microscopic pore structure and mineral composition of the rock thin section microscopic image dataset. First, the microscopic images are preprocessed, including image enhancement and noise reduction, to improve image quality. Then, an image segmentation algorithm is used to separate pore and mineral grain regions, extracting pore morphological parameters such as pore size, pore distribution, and pore connectivity to form microscopic pore structure data. For mineral composition extraction, the color, morphology, and other characteristics of different minerals in the images are analyzed, and mineralogy is used to identify the types and contents of various minerals, obtaining mineral composition data. This data is expressed as the area or volume percentage of each mineral in the thin section.

[0020] Step S124: The spatial morphology data of the karst fissures and caves, the lithological stratification data, the reservoir physical property parameter data, the microscopic pore structure data, and the mineral composition data are associated and organized based on a unified spatial coordinate framework to form a fused exploration data volume.

[0021] In this embodiment, the unified spatial coordinate framework adopts the geodetic coordinate system of the target karst area. First, the three-dimensional spatial boundaries of the karst fracture and cavern spatial morphology data are transformed to this coordinate system. For lithological stratification data and reservoir physical property parameter data, since they are depth-based, the depth coordinates need to be converted to geodetic coordinates using wellbore trajectory data. Microscopic pore structure data and mineral composition data correspond to specific core sample locations, and are associated with the geodetic coordinate system using the coordinate information of the core sampling location. During the association process, spatial interpolation and other methods are used to handle the spatial discontinuities of the data, ensuring the consistency and integrity of the data under the unified coordinate framework. The final fused exploration data volume is a three-dimensional data volume containing multiple geological information.

[0022] Step S125: Extract entities with clear geological significance from the fused exploration data volume. The entities include faults, fracture zones, caves, solution cavities, high-permeability strips, shale layers, interlayers, and unconformities, forming a set of karst geological entity units.

[0023] In this embodiment, a combination of rule-based and machine learning methods is used to extract geological entity objects from the fused exploration data volume. Faults are identified by analyzing features such as coherence changes and in-phase axis displacement in the seismic attribute data volume; fracture zones are extracted based on the response characteristics of fractures in seismic and well logging data; karst caves and pores are determined based on the spatial morphology and microstructure data of karst fractures and cavities; high-permeability strips are identified through permeability anomalies in reservoir physical property parameters; shale layers, interlayers, and unconformities are divided by combining lithological stratification data and seismic reflection characteristics. Each extracted geological entity object is assigned a unique identifier and corresponding attribute information, such as location, shape, and size. These entities collectively constitute a set of karst geological entity units.

[0024] Step S126: Perform geological relationship detection, type classification and integration processing based on the set of karst geological entity units to generate a set of karst geological relationship units.

[0025] In this embodiment, based on a set of karst geological entity units, the relationships between any two geological entity units are first detected. By analyzing their spatial relationships, formation time sequence, and physicochemical properties, potential geological relationships are identified. Then, based on different relationship characteristics, the detected geological relationships are categorized, such as spatial contact relationships, genetic connections, and property transfer relationships. Finally, the categorized relationships are integrated, removing duplicates and contradictions to form a complete set of karst geological relationship units. Each relationship unit in this set contains identifiers of the two related geological entity units and information such as the relationship type.

[0026] Step S1261: Detect the spatial contact relationship, genetic relationship and physical property transmission relationship between any two geological entities in the set of karst geological entities.

[0027] In this embodiment, the detection of spatial contact relationships is determined by calculating whether the coordinate ranges of two geological entity units overlap or are adjacent in three-dimensional space. For example, it is determined whether the spatial coordinate range of a karst cave entity overlaps with the coordinate range of a shale layer entity; if so, they are considered to have a spatial contact relationship. For genetic relationships, the formation process and geological background of the geological entity units are analyzed. For example, fault activity may lead to the formation of fracture zones, thus establishing a genetic relationship between the fault and the fracture zone. For property transmission relationships, the physical properties of the geological entity units, such as permeability and porosity, are used to determine whether fluid or stress transmission is possible between them. For example, a property transmission relationship may exist between a high-permeability strip and a karst cave.

[0028] Step S1262: Based on the spatial contact relationship, classify the adjacent, cutting, and enclosing relationship types between geological entity units to form a set of spatial contact relationships.

[0029] In this embodiment, the spatial contact relationships are further categorized based on their specific manifestations. When the spatial boundaries of two geological entity units are connected but not intersecting, they are considered adjacent; when the boundary of one geological entity unit passes through another, it is considered a cutting relationship; and when one geological entity unit is entirely within another, it is considered an enclosing relationship. These categorized relationships are recorded according to a specific format to form a spatial contact relationship set. Each element in this set contains the identifiers of the two geological entity units and their corresponding relationship type.

[0030] Step S1263: Based on the genetic relationships, classify the types of dissolution expansion, filling cementation, and tectonic modification relationships between geological entities to form a set of genetic relationships.

[0031] In this embodiment, the genetic relationships are categorized based on the type of geological process. When one geological unit expands in size due to dissolution, it is classified as a dissolution-expansion relationship; when one geological unit is filled with another substance and cemented, it is classified as an infilling-cementing relationship; when tectonic movements alter the morphology or properties of a geological unit, it is classified as a tectonic modification relationship. These relationships are recorded to form a set of genetic relationships.

[0032] Step S1264: Based on the described physical property transmission relationship, classify the fluid transport channels, pressure transmission paths, and stress transmission paths between geological entity units to form a set of physical property transmission relationships.

[0033] In this embodiment, the property transmission relationships are categorized based on the type of substance or energy being transmitted. When fluid can pass between geological entities, it is determined to be a fluid transport channel relationship; when formation pressure can be transmitted, it is determined to be a pressure transmission path relationship; and when tectonic stress can be transmitted, it is determined to be a stress transmission path relationship. These relationships are then organized to form a set of property transmission relationships.

[0034] Step S1265: Integrate the set of spatial contact relationships, the set of genetic relationships, and the set of physical property transmission relationships to form a complete set of karst geological relationship units.

[0035] In this embodiment, the sets of spatial contact relationships, genetic relationships, and property transmission relationships are integrated. First, the relationships in each set are checked for duplication or conflict. For duplicate relationships, only one record is retained, and for conflicting relationships, they are corrected through further geological analysis and verification. Then, different types of relationships are organized according to a unified format. Each relationship unit contains the identifiers of the two related geological entities, the relationship type, and other relevant attribute information, ultimately forming a complete set of karst geological relationship units.

[0036] Step S130: Based on the set of karst geological entity units and the set of karst geological relationship units, construct a karst geological knowledge network. The network nodes in the karst geological knowledge network correspond to the geological entities in the set of karst geological entity units, and the network edges in the karst geological knowledge network correspond to the geological relationships in the set of karst geological relationship units.

[0037] In this embodiment, a karst geological knowledge network is constructed based on a set of karst geological entity units and a set of karst geological relationship units. This is to organize and represent geological entities and geological relationships in a network form, facilitating subsequent knowledge-guided collaborative characterization and other operations. By using geological entities as network nodes and geological relationships as network edges, the complex connections between geological entities can be intuitively reflected.

[0038] Step S131: Create a network node for each geological entity unit in the set of karst geological entity units, and assign a unique node identifier and node type attribute to each network node.

[0039] In this embodiment, a corresponding network node is created for each geological entity unit in the karst geological entity unit set, such as faults, fracture zones, and caves. Each network node is assigned a unique node identifier, which can be a number or a string, to uniquely identify the node. Simultaneously, a node type attribute is set for each node, the value of which is the type of geological entity unit, such as "fault" or "cave," to facilitate node classification and management.

[0040] Step S132: Assign a relation type code to each geological relation type in the set of karst geological relation units.

[0041] In this embodiment, the karst geological relationship unit set contains various geological relationship types, such as adjacency, cutting, and dissolution expansion. A unique relationship type code is assigned to each relationship type; this code can be an integer or a string. For example, an adjacency relationship is encoded as "001," and a cutting relationship is encoded as "002," etc. Through relationship type encoding, the relationship types of network edges can be easily identified and distinguished.

[0042] Step S133: Based on the specific geological relationship instances recorded in the set of karst geological relationship units, establish a directed or undirected network edge between the network nodes corresponding to the two geological entity units.

[0043] In this embodiment, for each specific geological relationship instance in the karst geological relationship unit set, the two involved geological entity units are identified. Based on the identifiers of these two geological entity units in the karst geological entity unit set, the corresponding network nodes are found. Then, the nature of the relationship determines whether to establish directed or undirected network edges. For example, for a dissolution expansion relationship, since one geological entity unit acts on another, a directed network edge is established, with the direction pointing from the acting party to the acted party; for an adjacency relationship, since they are mutual, an undirected network edge is established.

[0044] Step S134: Assign a relation type attribute to each established network edge, the value of which is the relation type code of the corresponding geological relation type; calculate the initial relation strength weight for each established network edge, the initial relation strength weight being calculated based on the statistical frequency of the geological relation instance in the studied area and the expert prior knowledge of the influence of the relation type on shale gas enrichment.

[0045] In this embodiment, each network edge is assigned a relation type attribute, the value of which is the relation type code assigned to the corresponding geological relation type in step S132. For calculating the initial relation strength weight, the frequency of the geological relation instance appearing in the studied area is first statistically analyzed; the higher the frequency, the greater the initial weight. Simultaneously, the impact of the relation type on shale gas enrichment is considered in conjunction with expert assessments; the greater the impact, the greater the weight. The initial relation strength weight is obtained by comprehensively calculating the frequency and expert assessment results. For example, the frequency and expert assessment results are first normalized, and then weighted and summed according to a certain weight ratio to obtain the initial relation strength weight.

[0046] Step S135: Access the karst geological evolution constraint rule library and perform rule node construction, graph structure optimization, network embedding learning and network structure integration processing to generate a karst geological knowledge network structure with computable semantics.

[0047] In this embodiment, a karst geological evolution constraint rule base is accessed. This rule base contains a series of logical rules describing the state transformation of geological entities and the evolution of geological relationships. Through processing steps such as rule node construction, graph structure optimization, network embedding learning, and network structure integration, the above rules are integrated into the karst geological knowledge network, enabling the network to have computable semantics and better support subsequent knowledge-guided feature collaborative representation and other operations.

[0048] Step S1351: Access the karst geological evolution constraint rule library, which contains a series of logical rules describing the state transformation of geological entities and the evolution of geological relationships.

[0049] In this embodiment, the karst geological evolution constraint rule base is a pre-built database that stores a large number of logical rules based on geological theories and practical experience. These rules describe how geological entities undergo state transformations and how geological relationships evolve under different geological conditions. For example, the rule "When groundwater activity exists and the rock is soluble, the cave will undergo dissolution and expansion" is included. After accessing this rule base, the rules within can be used to optimize and improve the karst geological knowledge network.

[0050] Step S1352: Convert each logical rule in the karst geological evolution constraint rule base into a rule node, add the rule node to the node set of the karst geological knowledge network; establish constraint edges between the rule node and the ordinary network node constrained by the logical rule, and assign rule trigger weights to the constraint edges.

[0051] In this embodiment, each logical rule in the karst geological evolution constraint rule base is converted into a rule node. The rule node contains information such as the condition and conclusion parts of the rule. These rule nodes are added to the node set of the karst geological knowledge network, so that the network includes not only geological entity nodes but also rule nodes. Then, the geological entities constrained by each rule are analyzed, and constraint edges are established between the rule nodes and the corresponding ordinary network nodes. The rule trigger weight of the constraint edge is set according to the importance and applicability of the rule; the higher the importance and the stronger the applicability of the rule, the greater its trigger weight.

[0052] Step S1353: Based on the initial relationship strength weight and the rule trigger weight, perform graph structure optimization operation on the full graph nodes and edges of the karst geological knowledge network, strengthen the connection between mutually supporting nodes and edges, and weaken or remove the connection between mutually contradictory nodes and edges.

[0053] In this embodiment, the graph structure optimization operation is based on initial relation strength weights and rule triggering weights. First, the support and conflict degree between nodes are calculated. Support is the sum of the products of the initial relation strength weight of the edge between two nodes and the rule triggering weight of the relevant rule node. The higher the support, the more reasonable the connection between the nodes. Conflict degree refers to the degree to which the relationship between two nodes conflicts with the rules in the rule base. For nodes and edges with high support, their weights are increased to strengthen the connection; for nodes and edges with high conflict degree, their weights are decreased. If the weight decreases below a certain threshold, the node or edge is removed. Through this method, the network structure is optimized, improving the accuracy and reliability of the network.

[0054] Step S1354: Perform network embedding representation learning on the optimized graph structure, mapping each network node and rule node to a low-dimensional continuous vector space to obtain the embedding vector representation of each node.

[0055] In this embodiment, a graph embedding algorithm is used to learn the network embedding representation for the optimized graph structure. Graph embedding algorithms map nodes in a network to a low-dimensional continuous vector space, allowing the distances between nodes in the vector space to reflect their relationships within the network. Commonly used graph embedding algorithms include DeepWalk and Node2Vec. During the learning process, the network's structural information and node attribute information are used as inputs to train the model and obtain the embedding vector representation for each node. These embedding vectors can then be used for subsequent feature calculations and analysis.

[0056] Step S1355: Integrate the embedding vector representation of the node, the connection relationship of the network edge, the relationship type attribute of the network edge, the initial relationship strength weight of the network edge, the rule node, the constraint edge and the rule trigger weight of the constraint edge to generate a karst geological knowledge network structure with computable semantics.

[0057] In this embodiment, the embedding vector representations of nodes, various attribute information of network edges, and rule nodes and constraint edges are integrated. This information is organized and stored according to a specific data structure to form a complete karst geological knowledge network structure with computable semantics. This network structure not only includes relationships between geological entities but also incorporates geological evolution rules, enabling knowledge-based reasoning and computation.

[0058] Step S140: Input the multi-source exploration data set into the karst geological knowledge network for knowledge-guided feature collaborative characterization operations to generate knowledge-enhanced exploration feature representations.

[0059] In this embodiment, a multi-source exploration data set is input into a karst geological knowledge network, and the knowledge within the network is used to perform collaborative feature representation of the data. Guided by this knowledge, the generated exploration feature representation not only includes the features of the original data but also incorporates geological knowledge, thereby improving the expressive power of the features and the accuracy of predicting shale gas enrichment zones.

[0060] Step S141: Divide the three-dimensional seismic data volume into multiple local seismic data blocks, and perform convolutional neural network feature extraction on each local seismic data block to obtain the original seismic feature vector of each local seismic data block.

[0061] In this embodiment, the 3D seismic data volume is divided into multiple local seismic data blocks according to a certain spatial size; for example, each data block has a specific 3D size. For each local seismic data block, a convolutional neural network (CNN) is used for feature extraction. The CNN includes multiple convolutional layers, pooling layers, and fully connected layers. First, the data block is convolved by the convolutional layers to extract local features; then, the features are reduced in dimensionality by the pooling layers; finally, the extracted features are mapped to a fixed-length original seismic feature vector by the fully connected layers. The original seismic feature vector of each local seismic data block contains the main seismic feature information of that data block.

[0062] Step S142: Sampling and aligning the well logging curve data set in the depth direction to form a well logging feature vector sequence corresponding to the spatial location of the local seismic data block.

[0063] In this embodiment, the well logging curve dataset is a sequence of data that varies with depth. To correspond to the spatial location of the local seismic data block, the well logging curve data needs to be sampled and aligned along the depth direction. Based on the depth range of the local seismic data block, data of the corresponding depth segment is extracted from the well logging curve and sampled at certain sampling intervals to obtain well logging data corresponding to the spatial location of the local seismic data block. Then, the above well logging data is organized into a vector sequence, namely a well logging feature vector sequence, where each vector contains multiple well logging parameter values ​​for that sampling point.

[0064] Step S143: Input the set of rock thin section microscopic image data into the image feature extraction network to obtain microscopic image feature vectors.

[0065] In this embodiment, each image in the rock thin section microscopic image dataset is input into an image feature extraction network. The image feature extraction network can employ a pre-trained convolutional neural network model, such as ResNet or VGG. Through the network's forward propagation process, high-level features of the image are extracted. The output of the last fully connected layer of the network is used as a microscopic image feature vector, which contains microscopic structural feature information of the rock thin section microscopic image.

[0066] Step S144: The original seismic feature vector, the well logging feature vector sequence, and the microscopic image feature vector are spliced ​​together to form a multimodal original feature vector corresponding to each spatial location.

[0067] In this embodiment, for each spatial location, the corresponding original seismic feature vector, well logging feature vector sequence, and microscopic image feature vector are concatenated. The concatenation operation connects the three vectors according to their feature dimensions, forming a longer vector. For example, if the length of the original seismic feature vector is A, the length of each vector in the well logging feature vector sequence is B, the sequence length is C, and the length of the microscopic image feature vector is D, then the length of the concatenated multimodal original feature vector is A + B * C + D. Through concatenation, data features from different modalities are fused together to form the multimodal original feature vector.

[0068] Step S145: Based on the original multimodal feature vectors, perform knowledge-guided vector generation, feature fusion, and graph attention network processing to output knowledge-enhanced exploration feature representations.

[0069] In this embodiment, based on the original multimodal feature vectors, a knowledge-guided vector is first generated, then fused with the original multimodal feature vectors, and further processed through a graph attention network to finally obtain a knowledge-enhanced exploration feature representation. This process fully utilizes the knowledge in the karst geological knowledge network to enhance and optimize the multimodal features.

[0070] For example, step S1451: In the karst geological knowledge network, find network nodes that are related to the geological background of the current spatial location, and extract the embedding vector representation of the relevant network nodes.

[0071] In this embodiment, the geological background information of the current spatial location can be obtained through the coordinates of that location and existing geological interpretation data. Based on this information, a search is performed in the karst geological knowledge network to find network nodes related to that geological background. For example, if the current spatial location is near a fault, network nodes related to that fault, as well as other geological entity nodes related to the fault, are searched. After finding the relevant nodes, the embedding vector representations of these nodes are extracted. These embedding vectors contain the semantic and structural information of the nodes.

[0072] Step S1452: Calculate the semantic relevance score between the original multimodal feature vector and the embedding vector representation of each relevant network node. Based on the semantic relevance score, perform a weighted summation operation on the embedding vector representations of the relevant network nodes to generate a knowledge guidance vector.

[0073] In this embodiment, the semantic relevance score can be calculated using methods such as cosine similarity. The cosine similarity is calculated between the original multimodal feature vector and the embedded vector representation of each relevant network node to obtain the semantic relevance score. A higher score indicates a stronger semantic association between the original multimodal feature vector and that node. Then, the embedded vector representations of the relevant network nodes are weighted and summed based on the semantic relevance score; the weights represent the semantic relevance score. The resulting vector after weighted summation is the knowledge guidance vector, which integrates knowledge information related to the geological background of the current spatial location.

[0074] Step S1453: Perform a feature fusion operation on the knowledge-guided vector and the multimodal original feature vector. The feature fusion operation includes vector concatenation and linear transformation to generate a preliminary knowledge-enhanced feature vector. Input the preliminary knowledge-enhanced feature vector into a graph attention network. The graph attention network uses the structure of the karst geological knowledge network as the prior graph structure.

[0075] In this embodiment, the knowledge-guided vector is first concatenated with the original multimodal feature vector to form a longer vector. Then, a linear transformation is performed on the concatenated vector, and it is mapped to a new feature space through a fully connected layer to generate a preliminary knowledge-enhanced feature vector. This preliminary knowledge-enhanced feature vector is then input into a graph attention network, which uses the structure of the karst geological knowledge network as its prior graph structure, meaning that the connection relationships between nodes and edges in the network are consistent with those in the karst geological knowledge network.

[0076] Step S1454: In the graph attention network, the preliminary knowledge-enhanced feature vectors are used for message passing and aggregation operations between nodes of the knowledge network, so that the features of each location can be fused with the neighboring node information of its associated geological entities.

[0077] In this embodiment, each node in the graph attention network performs message passing based on the features and attention weights of its neighboring nodes. For each node, its attention weight to its neighboring nodes is calculated, and the magnitude of the weight reflects the importance of the neighboring nodes to that node. Then, the features of the neighboring nodes are weighted and summed according to their attention weights to obtain the aggregated features. In this way, the features of each node incorporate information from its neighboring nodes, thereby achieving knowledge transfer and sharing.

[0078] Step S1455: After message passing and aggregation operations in the multi-layer graph attention network, the final knowledge-enhanced exploration feature representation is output. The knowledge-enhanced exploration feature representation includes both the original data features and the structured prior knowledge encoded in the karst geological knowledge network.

[0079] In this embodiment, the multi-layer graph attention network performs multiple message passing and aggregation operations. Each layer processes the output features of the previous layer, further fusing information from neighboring nodes. After multi-layer processing, the output feature vector not only contains the features of the original multimodal data but also fully encodes the structured prior knowledge in the karst geological knowledge network, such as the relationships between geological entities and geological evolution rules. This knowledge-enhanced exploration feature representation can more accurately reflect information related to underground geological structures and shale gas enrichment.

[0080] Step S150: Call the geological process coupling simulation model to process the knowledge-enhanced exploration feature representation, simulate the interaction between karst development process and shale reservoir formation process, generate karst reservoir coupling evolution simulation results, and based on the karst reservoir coupling evolution simulation results, complete the spatial distribution prediction of shale gas enrichment zone in the target karst area.

[0081] In this embodiment, a geological process coupling simulation model is used to simulate the interaction between karst development and shale reservoir formation. Knowledge-enhanced exploration feature representations are input into the model, and simulation calculations yield the coupled evolution results of the karst reservoir. Based on the simulation results, the distribution of shale gas underground can be analyzed, thereby enabling the prediction of the spatial distribution of shale gas enrichment zones in the target karst area.

[0082] Step S151: The geological process coupled simulation model includes a karst development process simulation module and a shale reservoir formation process simulation module. The knowledge-enhanced exploration feature representation is input into the karst development process simulation module, which performs calculations based on the groundwater dynamics sub-model and the rock chemical dissolution sub-model.

[0083] In this embodiment, the coupled geological process simulation model consists of a karst development process simulation module and a shale reservoir formation process simulation module. Knowledge-enhanced exploration feature representations are first input into the karst development process simulation module. This module includes a groundwater dynamics sub-model and a rock chemical dissolution sub-model. Through the collaborative computation of these two sub-models, the karst development process is simulated. The groundwater dynamics sub-model primarily simulates groundwater flow, while the rock chemical dissolution sub-model primarily simulates rock dissolution processes.

[0084] Step S1511: The groundwater dynamics sub-model simulates the flow path and velocity changes of groundwater flow field in the fracture network based on the paleohydrogeological data in the knowledge-enhanced exploration feature representation.

[0085] In this embodiment, the knowledge-enhanced exploration feature representation includes paleohydrogeological data, such as paleogroundwater levels and paleoprecipitation. The groundwater dynamics sub-model utilizes this data, combined with the distribution characteristics of the fracture network, and employs groundwater flow equations for calculation. By solving the equations, the flow path and velocity variations of the groundwater flow field within the fracture network are obtained. The flow path reflects the direction of groundwater flow, while the velocity variations reflect the flow speed of groundwater at different locations.

[0086] Step S1512: The rock chemical dissolution sub-model calculates the dissolution rate and amount of rock along the water flow path based on the rock mineral composition data and groundwater chemical data in the knowledge-enhanced exploration feature representation.

[0087] In this embodiment, the rock mineral composition data in the knowledge-enhanced exploration feature representation includes the types and contents of various minerals, and the groundwater chemical data includes ion concentrations, pH values, etc., in the water. Based on this data and combined with the principles of chemical reaction kinetics, the rock chemical dissolution sub-model calculates the dissolution rate of the rock reacting with groundwater along the water flow path. The dissolution rate is related to time and reaction conditions, and the amount of dissolution over a certain time can be obtained through integral calculation.

[0088] Step S1513: The karst development process simulation module integrates the results of the groundwater dynamics sub-model and the rock chemical dissolution sub-model to dynamically simulate the evolution process of the karst fissure cave system from the initial state to the current state, and outputs the karst system evolution history data, which includes the volume distribution of dissolution space and the changes in the connectivity of dissolution channels in different geological periods.

[0089] In this embodiment, the karst development process simulation module integrates the groundwater flow field information obtained from the groundwater dynamics sub-model and the dissolution rate and amount information obtained from the rock chemical dissolution sub-model. Based on the water flow path and dissolution rate, it simulates the development of the karst fissure-cavity system in different geological periods. The initial state can be set to the original rock state, and over time, the spatial distribution and connectivity of the karst fissure-cavity system are updated based on the accumulation of dissolution. The final output karst system evolution history data records the distribution of dissolution space volume and changes in the connectivity of dissolution channels in different geological periods.

[0090] Step S1514: Simultaneously input the knowledge-enhanced exploration feature representation and the karst system evolution history data into the shale reservoir accumulation process simulation module; the shale reservoir accumulation process simulation module performs calculations based on the source rock hydrocarbon generation and expulsion sub-model and the fluid migration and aggregation sub-model; the source rock hydrocarbon generation and expulsion sub-model simulates the hydrocarbon generation intensity and expulsion history of the shale system based on the shale organic geochemical data and stratigraphic thermal history data in the knowledge-enhanced exploration feature representation; the fluid migration and aggregation sub-model simulates the hydrocarbon generation intensity and expulsion history of the shale system based on the karst system... The historical evolution data of the shale reservoir provides information on the evolution of reservoir space and migration channels, and the hydrocarbon fluid source provided by the hydrocarbon generation and expulsion sub-model of the source rock is used to simulate the migration path, accumulation location and accumulation amount of the generated hydrocarbon fluid in the evolving karst fracture-cavity system; the shale reservoir accumulation process simulation module integrates the results of the hydrocarbon generation and expulsion sub-model of the source rock and the fluid migration and accumulation sub-model to output historical data of shale gas accumulation, which includes hydrocarbon saturation distribution and formation pressure field evolution in different geological periods.

[0091] In this embodiment, the shale reservoir formation simulation module receives knowledge-enhanced exploration feature representations and karst system evolution history data. The source rock hydrocarbon generation and expulsion sub-model utilizes shale organic geochemical data (such as total organic carbon content and kerogen type) and formation thermal history data (such as paleothermal gradient and burial history) from the knowledge-enhanced exploration feature representations to calculate the hydrocarbon generation intensity and expulsion history of the shale system through a hydrocarbon generation kinetic model. The fluid migration and accumulation sub-model simulates the migration process of hydrocarbon fluids in the karst fracture-cavity system based on the reservoir space and migration channel information in the karst system evolution history data, as well as the hydrocarbon fluid sources provided by the source rock hydrocarbon generation and expulsion sub-model. By calculating factors such as flow resistance and buoyancy in different channels, the migration path, accumulation location, and accumulation amount are determined. The shale reservoir formation simulation module integrates the results of these two sub-models to obtain the hydrocarbon saturation distribution and formation pressure field evolution at different geological periods, forming shale gas accumulation evolution history data.

[0092] Step S1515: Perform bidirectional data feedback and iterative simulation between the two simulation modules until the output results reach a dynamic equilibrium state and are integrated to generate karst reservoir coupled evolution simulation results.

[0093] In this embodiment, the karst development process simulation module and the shale reservoir accumulation process simulation module influence each other. The evolution of the karst system affects the accumulation conditions of shale reservoirs, while the accumulation process of shale gas, in turn, affects the development of the karst system. Therefore, bidirectional data feedback and iterative simulation are required. In each iteration, the output of one module is used as the input of the other module for a new round of calculation. Through multiple iterations, the outputs of the two modules change very little, reaching a dynamic equilibrium. At this point, the outputs of the two modules are integrated to generate the coupled evolution simulation results of the karst reservoir.

[0094] For example, in step S15151: the historical data of the evolution of the karst system and the historical data of the shale gas accumulation at the current geological time are used as coupling feedback signals and input into the karst development process simulation module.

[0095] In this embodiment, during the iterative simulation process, when a certain geological moment is simulated, the historical data of karst system evolution and shale gas accumulation evolution at that moment are used as coupling feedback signals and fed back to the karst development process simulation module. The above data reflects the state of the karst system and shale gas accumulation at the current geological moment.

[0096] For example, step S15151-1: extract the spatial distribution data of hydrocarbon saturation at the current simulated geological moment from the shale gas accumulation and evolution history data; extract the spatial distribution data of formation pressure at the current simulated geological moment from the shale gas accumulation and evolution history data.

[0097] In this embodiment, the shale gas accumulation and evolution history data records the hydrocarbon saturation distribution and formation pressure field evolution at different geological moments. At the current simulated geological moment, the corresponding spatial distribution data of hydrocarbon saturation and spatial distribution data of formation pressure are extracted from this data. The above data are stored in the form of a three-dimensional grid, and each grid cell contains the corresponding hydrocarbon saturation value and formation pressure value.

[0098] Step S15151-2: Convert the spatial distribution data of hydrocarbon saturation into an influence coefficient matrix on the seepage capacity of groundwater phase. The influence coefficient value of groundwater phase seepage capacity in areas where hydrocarbon saturation values ​​exceed the set standard is reduced. Convert the spatial distribution data of formation pressure into a correction amount for the driving potential of groundwater flow. The correction amount of groundwater flow driving potential in areas where formation pressure values ​​exceed the set standard is increased.

[0099] In this embodiment, the hydrocarbon saturation level affects the permeability of the rock, and consequently, the phase flow capacity of groundwater. When the hydrocarbon saturation value exceeds a set standard, it indicates a high gas content in the area, reducing the space for groundwater flow and decreasing the phase flow capacity; therefore, the influence coefficient value for this area is reduced. The formation pressure affects the driving potential of groundwater flow. When the formation pressure value exceeds a set standard, the potential driving groundwater flow increases; therefore, the correction value for the driving potential of groundwater flow in this area is increased. The specific values ​​of the influence coefficient matrix and the correction value are determined through experiments and empirical formulas.

[0100] Step S15151-3: In the groundwater dynamics sub-model of the karst development process simulation module, the influence coefficient matrix of the groundwater phase seepage capacity is applied to the permeability parameter field to update the permeability parameters used to calculate the groundwater flow field; in the groundwater dynamics sub-model of the karst development process simulation module, the correction amount of the groundwater flow driving potential is superimposed on the original hydraulic head boundary conditions to update the potential function boundary conditions used to calculate the groundwater flow field.

[0101] In this embodiment, the permeability parameter field in the groundwater dynamics sub-model is a crucial parameter for calculating the groundwater flow field. The influence coefficient matrix of the groundwater phase seepage capacity is multiplied by the original permeability parameter field to obtain the updated permeability parameter field. Simultaneously, the correction amount of the groundwater flow driving potential is superimposed on the original hydraulic head boundary conditions to form new potential function boundary conditions. By updating the permeability parameters and boundary conditions, the groundwater dynamics sub-model can more accurately simulate groundwater flow.

[0102] Step S15151-4: Based on hydrocarbon component data, perform the selection of dissolution reaction kinetic parameters, update of chemical reaction rate constants, and simulation module driven processing.

[0103] In this embodiment, hydrocarbon composition data affects the kinetics of rock dissolution. Based on different hydrocarbon compositions, corresponding dissolution kinetic parameters are selected, and the chemical reaction rate constant is updated to accurately simulate the rock dissolution process, thereby driving the karst development process simulation module to perform further calculations.

[0104] For example, step S15151-4-1: Extract hydrocarbon component data for the current simulated geological time from the shale gas accumulation and evolution history data, and determine whether hydrocarbons are mainly in the gas phase.

[0105] In this embodiment, hydrocarbon composition data for the current simulated geological moment is extracted from shale gas accumulation and evolution history data, including the content of various hydrocarbon gases and liquids. By comparing the content ratio of gaseous hydrocarbons and liquid hydrocarbons, it is determined whether hydrocarbons are predominantly gaseous. If the content ratio of gaseous hydrocarbons exceeds a certain threshold, it is determined that the hydrocarbons are predominantly gaseous.

[0106] Step S15151-4-2: If hydrocarbons are mainly in the gas phase, then select the kinetic parameters of rock dissolution reaction in a gaseous environment from the chemical kinetic parameter library of the rock chemical dissolution sub-model; if hydrocarbons are not mainly in the gas phase, then select the kinetic parameters of rock dissolution reaction in a water-rock environment from the chemical kinetic parameter library of the rock chemical dissolution sub-model.

[0107] In this embodiment, the chemical kinetic parameter library of the rock chemical dissolution sub-model stores kinetic parameters under different environments. When hydrocarbons are mainly in the gas phase, the rock dissolution reaction mainly takes place in a gaseous environment, so kinetic parameters under a gaseous environment are selected; when hydrocarbons are not mainly in the gaseous phase, the dissolution reaction mainly takes place in a water-rock environment, so kinetic parameters under a water-rock environment are selected. These parameters include reaction activation energy, reaction order, etc.

[0108] Step S15151-4-3: Update the chemical reaction rate constant controlling the dissolution rate in the rock chemical dissolution sub-model using the selected kinetic parameters.

[0109] In this embodiment, the chemical reaction rate constant is an important parameter for calculating the dissolution rate in the rock chemical dissolution sub-model. Based on the selected kinetic parameters, a new chemical reaction rate constant is calculated using the corresponding formula, and the corresponding parameters in the model are updated to reflect the influence of different hydrocarbon environments on the dissolution rate.

[0110] Step S15151-4-4: Use the updated permeability parameters, updated potential function boundary conditions, and updated chemical reaction rate constant as new input parameters to drive the karst development process simulation module to calculate the next simulation time step.

[0111] In this embodiment, the updated permeability parameters, potential function boundary conditions, and chemical reaction rate constants are input into the karst development process simulation module as calculation parameters for the next simulation time step. This method realizes the feedback influence of the shale reservoir formation process on the karst development process, making the simulation more consistent with actual geological conditions.

[0112] Step S15152: In the karst development process simulation module, the hydrocarbon saturation distribution and formation pressure field evolution in the shale gas accumulation and evolution history data affect the boundary conditions of the groundwater flow field and the physicochemical properties of the rocks.

[0113] In this embodiment, the hydrocarbon saturation distribution in the shale gas accumulation history data alters the pore structure and permeability of rocks, thereby affecting groundwater flow. The evolution of the formation pressure field changes the driving potential of groundwater flow, thus influencing the boundary conditions of the groundwater flow field. Simultaneously, the presence of hydrocarbons may also alter the physicochemical properties of rocks, such as hardness and solubility, all of which affect karst development processes.

[0114] Step S15153: Based on the updated groundwater flow field boundary conditions and rock physicochemical properties, the karst development process simulation module recalculates the karst evolution path for subsequent geological periods, and feeds back the recalculated subsequent karst evolution path data to the shale reservoir accumulation process simulation module to update the reservoir space and channel conditions for fluid migration.

[0115] In this embodiment, the karst development process simulation module recalculates groundwater flow and rock dissolution based on updated groundwater flow field boundary conditions and rock physicochemical properties, obtaining karst evolution path data for subsequent geological periods. This data includes the expansion direction of dissolution space and the connectivity of dissolution channels. This data is then fed back to the shale reservoir formation process simulation module to update the reservoir space and channel conditions for fluid migration, making the simulation of the shale reservoir formation process more accurate.

[0116] Step S15154: After multiple bidirectional data exchanges and iterative simulations, until the output results of the two simulation modules reach a dynamic equilibrium state, the final karst spatial structure data and shale gas distribution data obtained at this time are integrated into the karst reservoir coupled evolution simulation results.

[0117] In this embodiment, through multiple bidirectional data exchanges and iterative simulations, the output results of the karst development process simulation module and the shale reservoir accumulation process simulation module will gradually stabilize. When the changes in the output results of the two modules are less than the set threshold in several consecutive iterations, a dynamic equilibrium state is determined to be reached. At this point, the final karst spatial structure data and shale gas distribution data obtained from the simulation are integrated to form the karst reservoir coupled evolution simulation results.

[0118] For example, step S151541: Set the iteration simulation termination condition, which includes the rate of change of the volume distribution of dissolution space in the karst system evolution history data being less than a preset threshold, and the rate of change of the hydrocarbon saturation distribution in the shale gas accumulation evolution history data being less than another preset threshold.

[0119] In this embodiment, the termination condition of the iterative simulation is the basis for determining whether the simulation has reached a dynamic equilibrium state. The rate of change of the dissolution space volume distribution is the difference between the dissolution space volume obtained in the current iteration and the dissolution space volume obtained in the previous iteration, divided by the dissolution space volume in the previous iteration. The rate of change of the hydrocarbon saturation distribution is similar. The preset threshold is set according to the accuracy requirements of the simulation and the geological conditions. When both rates of change are less than their respective preset thresholds, the iteration stops.

[0120] Step S151542: After each complete bidirectional data exchange and iterative simulation cycle, calculate the difference measure between the karst system evolution history data output in the current cycle and the karst system evolution history data output in the previous cycle; after each complete bidirectional data exchange and iterative simulation cycle, calculate the difference measure between the shale gas accumulation evolution history data output in the current cycle and the shale gas accumulation evolution history data output in the previous cycle.

[0121] In this embodiment, the difference measurement can employ methods such as root mean square error (RMSE) and mean absolute error (MAE). For karst system evolution history data, the difference in the volume distribution of dissolution space between two cycles is calculated; for shale gas accumulation evolution history data, the difference in hydrocarbon saturation distribution between two cycles is calculated. The difference measurement can intuitively reflect the changes in the simulation results.

[0122] Step S151543: Determine whether the difference measure of the karst system evolution history data is less than the preset threshold, and determine whether the difference measure of the shale gas accumulation evolution history data is less than the other preset threshold.

[0123] In this embodiment, the calculated difference metric is compared with a preset threshold. If the difference metric of the karst system evolution history data is less than the preset threshold, and the difference metric of the shale gas accumulation evolution history data is less than another preset threshold, it indicates that the simulation results have stabilized and reached a dynamic equilibrium state; otherwise, iterative simulation needs to continue.

[0124] Step S151544: If the difference measure of the karst system evolution history data and the difference measure of the shale gas accumulation evolution history data are both less than their respective preset thresholds, it is determined that the output results of the two simulation modules have reached a dynamic equilibrium state; if the difference measure of the karst system evolution history data or the difference measure of the shale gas accumulation evolution history data are not less than their respective preset thresholds, the next round of bidirectional data exchange and iterative simulation cycle is started.

[0125] In this embodiment, the decision to continue iteration is made based on the judgment result. If a dynamic equilibrium state is reached, the iteration stops; otherwise, the next round of bidirectional data exchange and iterative simulation cycle is started, repeating the above feedback and calculation process.

[0126] Step S151545: In the next cycle, the historical data of karst system evolution generated in the current cycle is used as the new input of the shale reservoir accumulation process simulation module to update the reservoir space and channel condition parameters of the shale reservoir accumulation process simulation module.

[0127] In this embodiment, during the next cycle, the historical evolution data of the karst system obtained in the current cycle is input into the shale reservoir formation process simulation module. This data contains the latest karst spatial structure information. Based on this information, the shale reservoir formation process simulation module updates the size, shape, and connectivity of the reservoir space, as well as the parameters of the fluid migration channels, such as permeability and porosity.

[0128] Step S151546: Execute the next round of loop parameter update, coupled simulation calculation, iterative loop and result data integration processing to generate complete karst reservoir coupled evolution simulation results.

[0129] In this embodiment, in the next iteration, in addition to updating the parameters of the shale reservoir formation simulation module, the parameters of the karst development simulation module also need to be updated. Then, coupled simulation calculations are performed to obtain new historical data on the evolution of the karst system and the shale gas accumulation. The bidirectional data exchange, difference measurement calculation, and judgment process are repeated until the iterative simulation termination condition is met. When the termination condition is met, the final historical data on the evolution of the karst system and the shale gas accumulation are extracted and integrated into a complete coupled karst reservoir evolution simulation result.

[0130] Step S1515461: In the next cycle, the historical data of shale gas accumulation and evolution generated in the current cycle is used as the new input of the karst development process simulation module to update the groundwater flow boundary and rock physicochemical property parameters of the karst development process simulation module.

[0131] In this embodiment, the historical data on shale gas accumulation and evolution generated in the current cycle includes information such as hydrocarbon saturation distribution and formation pressure field evolution. This data is input into the karst development process simulation module to update groundwater flow boundary conditions, such as hydraulic head and flow rate boundaries, as well as the physicochemical properties of rocks, such as permeability and solubility, making the simulation of the karst development process more accurate.

[0132] Step S1515462: Based on the updated parameters, drive the karst development process simulation module and the shale reservoir formation process simulation module to perform a new round of coupled simulation calculations.

[0133] In this embodiment, updated parameters are used to drive the karst development process simulation module and the shale reservoir accumulation process simulation module to perform calculations. The karst development process simulation module calculates new karst system evolution history data, and the shale reservoir accumulation process simulation module calculates new shale gas accumulation evolution history data.

[0134] Step S1515463: Repeat the bidirectional data exchange, difference measurement calculation and judgment process until the iterative simulation termination condition is met.

[0135] In this embodiment, the newly calculated karst system evolution history data and shale gas accumulation evolution history data are exchanged bidirectionally, the difference metric is recalculated, and compared with a preset threshold. If the termination condition is not met, the next round of parameter updates and calculations continues until the termination condition is met.

[0136] Step S1515464: When the iterative simulation termination condition is met, extract the final karst system evolution history data output by the karst development process simulation module in the last loop; when the iterative simulation termination condition is met, extract the final shale gas accumulation evolution history data output by the shale reservoir accumulation process simulation module in the last loop.

[0137] In this embodiment, when the iterative simulation termination condition is met, it indicates that the simulation results have reached a dynamic equilibrium state. At this point, the final karst system evolution history data output by the karst development process simulation module and the final shale gas accumulation evolution history data output by the shale reservoir accumulation process simulation module are extracted from the last iteration. These data represent the final simulation results.

[0138] Step S1515465: Extract the volume distribution data of the dissolution space and the interconnection data of the fissure-cavity system at the final geological moment from the final karst system evolution history data, and integrate them into the final karst spatial structure data.

[0139] In this embodiment, the final karst system evolution history data includes karst evolution information from different geological periods. Data on the volume distribution of dissolution spaces and the interconnectedness of the fissure-cavity system at the final geological moment are extracted; these data reflect the current spatial structure of the karst system. These are then integrated to form the final karst spatial structure data.

[0140] Step S1515466: Extract hydrocarbon saturation distribution data and formation pressure field data at the final geological moment from the final shale gas accumulation and evolution history data, integrate them into the final shale gas distribution data, merge the final karst spatial structure data and the final shale gas distribution data to generate a complete karst reservoir coupled evolution simulation result.

[0141] In this embodiment, hydrocarbon saturation distribution data and formation pressure field data at the final geological moment are extracted from the final shale gas accumulation and evolution history data. These data reflect the current distribution of shale gas. The final karst spatial structure data and the final shale gas distribution data are merged to form a complete data volume containing information on karst structure and shale gas distribution, namely, the karst reservoir coupled evolution simulation result.

[0142] Step S152: Based on the karst reservoir coupled evolution simulation results, complete the spatial distribution prediction of shale gas enrichment zones in the target karst area.

[0143] In this embodiment, the enrichment of shale gas underground is analyzed based on the results of coupled evolution simulation of karst reservoirs. By comprehensively evaluating various parameters in the simulation results, the spatial location and range of shale gas enrichment zones are determined, and the spatial distribution prediction of shale gas enrichment zones in the target karst area is completed.

[0144] Step S1521: Extract shale gas distribution data at the final geological time from the karst reservoir coupled evolution simulation results. The shale gas distribution data includes spatial distribution data of gas saturation and spatial distribution data of formation pressure coefficient. Extract karst spatial structure data at the final geological time from the karst reservoir coupled evolution simulation results. The karst spatial structure data includes spatial distribution data of dissolution porosity and spatial distribution data of fracture-cavity system connectivity.

[0145] In this embodiment, the coupled evolution simulation results of karst reservoirs contain a wealth of information. Shale gas distribution data and karst spatial structure data at the final geological time are extracted. The spatial distribution data of gas saturation reflects the gas content at different locations, and the spatial distribution data of formation pressure coefficient reflects the magnitude of formation pressure. The spatial distribution data of dissolution porosity reflects the degree of porosity development in the rock, and the spatial distribution data of fracture-cavity system connectivity reflects the connectivity of the fracture-cavity system.

[0146] Step S1522: Compare the spatial distribution data of gas saturation with a preset gas saturation threshold value to obtain spatial regions where the gas saturation value exceeds the gas saturation threshold value, and mark them as gas saturation compliance regions; compare the spatial distribution data of formation pressure coefficient with a preset pressure coefficient threshold value to obtain spatial regions where the formation pressure coefficient value exceeds the pressure coefficient threshold value, and mark them as pressure coefficient compliance regions; compare the spatial distribution data of dissolution porosity with a preset porosity threshold value to obtain spatial regions where the dissolution porosity value exceeds the porosity threshold value, and mark them as porosity compliance regions.

[0147] In this embodiment, the preset threshold values ​​for gas saturation, pressure coefficient, and porosity are set based on experience and standards in shale gas exploration and development. The corresponding spatial distribution data are compared with these threshold values; areas exceeding these thresholds are considered to have shale gas enrichment potential and are marked as areas meeting the gas saturation, pressure coefficient, and porosity standards, respectively.

[0148] Step S1523: Perform three-dimensional connectivity analysis on the spatial distribution data of the slit-hole system to identify the spatially interconnected slit-hole system and calculate the total volume and geometric complexity of each connectivity.

[0149] In this embodiment, a three-dimensional connectivity analysis is performed on the spatial distribution data of the slot-hole system. Using a connectivity analysis algorithm, spatially interconnected slot-hole systems are identified, with each connected slot-hole system constituting a connectivity volume. The total volume of each connectivity volume is calculated, representing the spatial volume contained within that volume. Geometric complexity is also calculated, such as by measuring parameters like the ratio of surface area to volume and fractal dimension.

[0150] Step S1524: Perform multi-regional spatial overlay analysis, enrichment favorability index calculation and enrichment zone delineation to generate spatial distribution prediction results for shale gas enrichment zones.

[0151] In this embodiment, a multi-regional spatial overlay analysis is first performed, spatially superimposing regions meeting gas saturation, pressure coefficient, and porosity standards, along with the interconnected fracture-vuggy system, to identify their overlapping areas. Then, the enrichment favorability index for each spatial grid cell is calculated, taking into account the influence of various parameters. Finally, shale gas enrichment zones are delineated based on the enrichment favorability index, forming the prediction results.

[0152] For example, step S15241: Perform spatial superposition analysis on the gas saturation compliance area, the pressure coefficient compliance area, the porosity compliance area and the interconnected body of the crevice system.

[0153] In this embodiment, spatial overlay analysis involves superimposing the spatial extents of multiple regions to identify their intersections, unions, and other relationships. Through spatial overlay analysis, regions that simultaneously meet multiple conditions can be identified; these regions are often favorable areas for shale gas enrichment. For example, regions that simultaneously meet the criteria for gas saturation, pressure coefficient, and porosity, and are located within a fracture-vuggy system, can be identified.

[0154] Step S15242: Calculate the enrichment favorability index on each spatial grid cell; the calculation process of the enrichment favorability index includes: first, normalizing the gas saturation, formation pressure coefficient, dissolution porosity, and development scale index of the fracture-cavity system to eliminate differences in dimensions and numerical ranges; then, based on preset geological rules, performing nonlinear comprehensive calculation on the normalized parameter values ​​to generate the enrichment favorability index.

[0155] In this embodiment, the development scale index can be calculated based on parameters such as the total volume and geometric complexity of the fracture-cavity system. Normalization transforms the parameter values ​​to the range [0,1], for example, using the min-max normalization method. The preset geological rules are formulated based on geological theory and practical experience. For example, the normalized gas saturation, formation pressure coefficient, dissolution porosity, and development scale index are weighted and summed according to certain weights, or more complex nonlinear combinations are performed, such as product or exponential operations, to obtain the enrichment favorability index.

[0156] Step S15243: Based on the enrichment favorability index of all spatial grid cells, generate a three-dimensional enrichment favorability index data volume, perform three-dimensional spatial clustering analysis on the three-dimensional enrichment favorability index data volume, and identify regions where the enrichment favorability index value exceeds the set index standard and is continuously distributed in space.

[0157] In this embodiment, the three-dimensional enrichment favorability index data volume is a three-dimensional data matrix containing the enrichment favorability index of each spatial grid cell. Three-dimensional spatial clustering analysis employs clustering algorithms such as K-means clustering and density clustering to group spatial grid cells with similar enrichment favorability indices together. An index standard is set, and regions with enrichment favorability index values ​​exceeding this standard and spatially continuous are identified; these regions are potential shale gas enrichment zones.

[0158] Step S15244: Delineate the areas where the enrichment favorability index values ​​continuously distributed in space exceed the set index standard as shale gas enrichment zones, extract the spatial boundary coordinates, average enrichment favorability index and estimated gas-bearing volume of each enrichment zone, and form the spatial distribution prediction results of shale gas enrichment zones.

[0159] In this embodiment, spatially continuous areas with enrichment favorability indices exceeding a set standard are delineated as shale gas enrichment zones. For each enrichment zone, its spatial boundary coordinates are extracted to determine the zone's extent; the average enrichment favorability index of all grid cells within the zone is calculated to reflect the zone's enrichment level; and the gas-bearing volume is estimated based on data such as the zone's volume and gas saturation.

[0160] Step S160: Based on the predicted spatial distribution of shale gas enrichment zones, perform iterative optimization of node and edge weights on the karst geological knowledge network to update the structure of the karst geological knowledge network.

[0161] In this embodiment, the predicted spatial distribution of shale gas enrichment zones may differ from the actual situation. By comparing the predicted results with the actual exploration results, analyzing the reasons for the differences, and then iteratively optimizing the weights of nodes and edges in the karst geological knowledge network, the network structure is updated to improve the accuracy and reliability of the knowledge network, so as to obtain more accurate results in subsequent predictions.

[0162] Step S161: Compare the predicted spatial distribution of shale gas enrichment zones with the actual exploration results, which include the production capacity data of drilled wells and the test results data.

[0163] In this embodiment, the production capacity data of drilled wells includes daily gas production and cumulative gas production of a single well; the test results data includes formation pressure test data, core analysis data, etc. The predicted spatial distribution of shale gas enrichment zones is compared with these actual exploration results to compare whether the predicted enrichment zones are consistent with the locations of actual high-yield wells, and whether the predicted gas content matches the actual test results, etc.

[0164] Step S162: Based on the comparison results, calculate the consistency index between the predicted results and the actual situation. The consistency index includes spatial location consistency and enrichment degree consistency.

[0165] In this embodiment, spatial location matching refers to the degree of overlap between the predicted enrichment zone and the actual high-yield well location, which can be measured by calculating the ratio of the overlapping area to the predicted zone area. Enrichment degree matching refers to the degree of closeness between the predicted gas content and the actual tested gas content, which can be measured by calculating the relative error or absolute error between the two.

[0166] Step S163: Identify the regions in the prediction results where the consistency index meets the set consistency criteria and the regions where the consistency index does not meet the set consistency criteria.

[0167] In this embodiment, the matching criteria are set based on exploration accuracy requirements and actual conditions. When the matching index is greater than or equal to the set criteria, the area is considered to meet the criteria; otherwise, it is considered not to meet the criteria. By identifying these areas, areas with accurate and inaccurate predictions can be determined.

[0168] Step S164: In the karst geological knowledge network, locate the network nodes and network edges related to the regions whose consistency index meets the set consistency standard, and the network nodes and network edges constitute a knowledge subgraph that supports correct prediction; In the karst geological knowledge network, locate the network nodes and network edges related to the regions whose consistency index does not meet the set consistency standard, and the network nodes and network edges constitute a knowledge subgraph that causes prediction deviation.

[0169] In this embodiment, based on the spatial location of areas that meet or do not meet the set criteria according to the consistency index, the relevant geological entity nodes and geological relationship edges are searched in the karst geological knowledge network. Nodes and edges in the knowledge subgraph that support correct predictions play a positive role in the prediction results, while nodes and edges in the knowledge subgraph that cause prediction deviations may contain errors or deficiencies.

[0170] Step S165: Perform incremental weight calculation, network edge weight update, and iterative optimization of network structure based on the knowledge subgraph.

[0171] In this embodiment, based on the knowledge subgraphs that support correct predictions and the knowledge subgraphs that cause prediction bias, the weight increment of the network edges is calculated, the weights of the network edges are updated, and the network structure is iteratively optimized to improve the performance of the karst geological knowledge network.

[0172] For example, in step S1651: based on the specific value of the consistency index, calculate the positive reinforcement weight increment for the network edges in the knowledge subgraph that supports correct prediction. The closer the consistency index value is to the set optimal standard, the larger the value of the positive reinforcement weight increment.

[0173] In this embodiment, the optimal standard is set as the ideal value of the consistency index. The calculation of the positive reinforcement weight increment is related to the specific value of the consistency index. The closer the consistency index is to the set optimal standard, the greater the contribution of the network edge to the correct prediction, and therefore the larger the value of the positive reinforcement weight increment. The weight increment can be calculated based on the consistency index by setting a functional relationship, such as a linear function or an exponential function.

[0174] Step S1652: Based on the specific value of the consistency index, calculate the negative weakening weight increment for the network edge in the knowledge subgraph that causes the prediction deviation. The greater the deviation of the consistency index value from the set optimal standard, the larger the absolute value of the negative weakening weight increment.

[0175] In this embodiment, the calculation of the negative weakening weight increment is related to the degree to which the consistency index deviates from the set optimal standard. The greater the deviation of the consistency index from the set optimal standard, the greater the impact of the network edge on the prediction bias, and therefore the larger the absolute value of the negative weakening weight increment. The weight increment can also be calculated using a functional relationship, except that the weight increment is negative.

[0176] Step S1653: The positive reinforcement weight increment is superimposed on the initial relation strength weight of the corresponding network edge in the knowledge subgraph that supports correct prediction, thereby enhancing the influence of the network edge in the knowledge network; the negative weakening weight increment is superimposed on the initial relation strength weight of the corresponding network edge in the knowledge subgraph that causes prediction bias, thereby weakening the influence of the network edge in the knowledge network.

[0177] In this embodiment, positive reinforcement weight increments are added to the initial weights of network edges that support correct predictions, increasing the weights of these edges and making their role in the knowledge network more prominent. Negative weakening weight increments are added to the initial weights of network edges that cause prediction bias, decreasing the weights of these edges and reducing their influence.

[0178] Step S1654: Relearn the network embedding representation of the karst geological knowledge network after weight update to obtain the updated node embedding vector representation. Use the updated node embedding vector representation and the updated network edge weights to replace the corresponding parts in the original karst geological knowledge network to complete the iterative optimization and update of the karst geological knowledge network structure. The updated karst geological knowledge network will be used to make the next round of predictions for the new target area.

[0179] In this embodiment, the network structure and relationships between nodes change after the weights are updated. The network embedding representation is relearned to obtain updated node embedding vector representations, enabling the node vectors to reflect the new network structure and weight information. Then, the corresponding parts in the original network are replaced with the updated node embedding vector representations and network edge weights, completing the iterative optimization and update of the network structure. The updated karst geological knowledge network has higher accuracy and can be used to predict the spatial distribution of shale gas enrichment zones in new target areas.

[0180] In some embodiments, the AI-based intelligent prediction system for karst shale gas in karst areas used to perform the above methods can be any electronic device with data computing, processing, and storage capabilities. This AI-based intelligent prediction system for karst shale gas in karst areas can be used to implement the text processing methods or text processing model processing methods provided in the above embodiments.

[0181] Typically, an AI-based intelligent prediction system for shale gas in karst areas includes a processor and memory. The processor may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor can be implemented using at least one hardware form of DSP (Digital Signal Processing), FPGA (Field Programmable Gate Array), or PLA (Programmable Logic Array). The processor may also include a main processor and coprocessors. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the active state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor may also include an AI (Artificial Intelligence) processor, which handles computational operations related to machine learning.

[0182] The memory may include one or more computer-readable storage media, which may be non-transitory. The memory may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In some embodiments, the non-transitory computer-readable storage media in the memory is used to store a computer program configured to be executed by one or more processors to implement the above-described text processing method, or the processing method of the text processing model.

[0183] In an illustrative embodiment, a computer-readable storage medium is also provided, wherein a computer program is stored in the storage medium, and the computer program, when executed by a processor of a computer device, implements the above-described text processing method or text processing model. Optionally, the above-described computer-readable storage medium may be ROM (Read-Only Memory), RAM (Random Access Memory), CD-ROM (CompactDisc Read-Only Memory), magnetic tape, floppy disk, and optical data storage device, etc.

[0184] This application provides a computer program product, which includes computer-executable instructions or a computer program. When the computer-executable instructions or the computer program are executed by a processor, the processor will execute the content recommendation method provided in this application.

[0185] This application provides a computer-readable storage medium storing computer-executable instructions or computer programs. When the computer-executable instructions or computer programs are executed by a processor, the processor will execute the content recommendation method provided in this application.

[0186] In some embodiments, the computer-readable storage medium may be a read-only memory (ROM), random access memory (RAM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory, magnetic surface memory, optical disk, or CD-ROM, etc.; or it may be a device that includes one or any combination of the above-mentioned memories.

[0187] In some embodiments, computer-executable instructions may take the form of programs, software, software modules, scripts, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as stand-alone programs or as modules, components, subroutines, or other units suitable for use in a computing environment.

[0188] As an example, computer-executable instructions may, but do not necessarily, correspond to files in a file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a HyperText Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple co-located files (e.g., files that store one or more modules, subroutines, or code sections).

[0189] As an example, computer-executable instructions can be deployed to execute on a single electronic device, or on multiple electronic devices located at one location, or on multiple electronic devices distributed across multiple locations and interconnected via a communication network.

[0190] Finally, it should be noted that the above-disclosed embodiments are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent prediction method for shale gas in karst areas based on artificial intelligence, characterized in that, The method includes: Acquire a multi-source exploration data set of the target karst area, the multi-source exploration data set including a three-dimensional seismic data volume, a well logging curve data set, and a rock thin section microscopic image data set; The multi-source exploration data set is subjected to a dual extraction operation of geological entities and geological relationships to obtain a set of karst geological entity units and a set of karst geological relationship units; Based on the set of karst geological entity units and the set of karst geological relationship units, a karst geological knowledge network is constructed. The network nodes in the karst geological knowledge network correspond to the geological entities in the set of karst geological entity units, and the network edges in the karst geological knowledge network correspond to the geological relationships in the set of karst geological relationship units. The multi-source exploration data set is input into the karst geological knowledge network for knowledge-guided collaborative feature representation to generate knowledge-enhanced exploration feature representations. The knowledge-enhanced exploration feature representation is processed by calling the geological process coupling simulation model to simulate the interaction between karst development and shale reservoir formation, generate karst reservoir coupling evolution simulation results, and predict the spatial distribution of shale gas enrichment zones in the target karst area based on the karst reservoir coupling evolution simulation results. Based on the predicted spatial distribution of shale gas enrichment zones, the weights of nodes and edges in the karst geological knowledge network are iteratively optimized to update the structure of the karst geological knowledge network.

2. The intelligent prediction method for shale gas in karst areas based on artificial intelligence according to claim 1, characterized in that, The process of performing dual extraction operations on the multi-source exploration data set, extracting both geological entities and geological relationships, yields a set of karst geological entity units and a set of karst geological relationship units, including: Seismic attribute extraction is performed on the three-dimensional seismic data volume to obtain multiple seismic attribute data volumes; The spatial morphology of karst fissures and cavities is identified by performing a karst fissure and cavity spatial morphology identification operation on the various seismic attribute data volumes, thereby identifying the three-dimensional spatial boundaries of the karst fissures and cavities and generating karst fissure and cavity spatial morphology data; lithological stratification and physical property parameter interpretation operations are performed on the well logging curve data set to obtain lithological stratification data and reservoir physical property parameter data; Based on the aforementioned rock thin section microscopic image data set, microscopic pore structure and mineral composition extraction operations are performed to obtain microscopic pore structure data and mineral composition data; The spatial morphology data of the karst fissures and caves, the lithological stratification data, the reservoir physical property parameter data, the micropore structure data, and the mineral composition data are associated and organized based on a unified spatial coordinate framework to form a fused exploration data volume. Extract entities with clear geological significance from the fused exploration data volume. These entities include faults, fracture zones, caves, solution cavities, high-permeability strips, shale layers, interlayers, and unconformities, forming a set of karst geological entity units. Based on the set of karst geological entity units, perform geological relationship detection, type classification and integration processing to generate a set of karst geological relationship units.

3. The intelligent prediction method for shale gas in karst areas based on artificial intelligence according to claim 2, characterized in that, The process of performing geological relationship detection, type classification, and integration based on the set of karst geological entity units generates a set of karst geological relationship units, including: Detect the spatial contact relationship, genetic relationship, and physical property transmission relationship between any two geological entities in the karst geological entity unit set; Based on the spatial contact relationships, the types of adjacent, cutting, and enclosing relationships between geological entity units are classified to form a set of spatial contact relationships; Based on the aforementioned genetic relationships, the types of relationships between geological entities, such as dissolution expansion, filling cementation, and tectonic modification, are classified to form a set of genetic relationships. Based on the aforementioned physical property transmission relationships, the types of fluid transport channels, pressure transmission paths, and stress transfer paths between geological entity units are classified to form a set of physical property transmission relationships; Integrating the set of spatial contact relationships, the set of genetic relationships, and the set of physical property transmission relationships constitutes a complete set of karst geological relationship units.

4. The intelligent prediction method for shale gas in karst areas based on artificial intelligence according to claim 1, characterized in that, The karst geological knowledge network is constructed based on the set of karst geological entity units and the set of karst geological relationship units. Network nodes in the karst geological knowledge network correspond to geological entities in the set of karst geological entity units, and network edges in the karst geological knowledge network correspond to geological relationships in the set of karst geological relationship units. This includes: Create a network node for each geological entity unit in the set of karst geological entity units, and assign a unique node identifier and node type attribute to each network node; Assign a relationship type code to each geological relationship type in the set of karst geological relationship units; Based on the specific geological relationship instances recorded in the set of karst geological relationship units, a directed or undirected network edge is established between the network nodes corresponding to the two geological entity units. Each established network edge is assigned a relation type attribute, the value of which is the relation type code of the corresponding geological relation type; an initial relation strength weight is calculated for each established network edge, the initial relation strength weight is calculated based on the statistical frequency of the geological relation instance in the studied area and the expert prior knowledge of the influence of the relation type on shale gas enrichment. By accessing the karst geological evolution constraint rule base and performing rule node construction, graph structure optimization, network embedding learning, and network structure integration processing, a karst geological knowledge network structure with computable semantics is generated.

5. The intelligent prediction method for shale gas in karst areas based on artificial intelligence according to claim 4, characterized in that, The process of accessing the karst geological evolution constraint rule base and performing rule node construction, graph structure optimization, network embedding learning, and network structure integration processing to generate a karst geological knowledge network structure with computable semantics includes: Access the karst geological evolution constraint rule library, which contains a series of logical rules describing the state transformation of geological entities and the evolution of geological relationships; Each logical rule in the karst geological evolution constraint rule base is converted into a rule node, and the rule node is added to the node set of the karst geological knowledge network; constraint edges are established between the rule node and the ordinary network node constrained by the logical rule, and rule trigger weights are assigned to the constraint edges. Based on the initial relationship strength weight and the rule trigger weight, a graph structure optimization operation is performed on all nodes and edges of the karst geological knowledge network to strengthen the connection between mutually supporting nodes and edges, and weaken or remove the connection between mutually contradictory nodes and edges. The optimized graph structure is subjected to network embedding representation learning, which maps each network node and rule node to a low-dimensional continuous vector space to obtain the embedding vector representation of each node. The embedding vector representation of the nodes, the connection relationship of the network edges, the relationship type attribute of the network edges, the initial relationship strength weight of the network edges, the rule nodes, the constraint edges, and the rule trigger weight of the constraint edges are integrated to generate a karst geological knowledge network structure with computable semantics.

6. The intelligent prediction method for shale gas in karst areas based on artificial intelligence according to claim 1, characterized in that, The step of inputting the multi-source exploration data set into the karst geological knowledge network for knowledge-guided collaborative feature representation to generate knowledge-enhanced exploration feature representations includes: The three-dimensional seismic data volume is divided into multiple local seismic data blocks, and a convolutional neural network feature extraction operation is performed on each local seismic data block to obtain the original seismic feature vector of each local seismic data block. The well logging curve data set is sampled and aligned along the depth direction to form a well logging feature vector sequence corresponding to the spatial location of the local seismic data block; The set of rock thin section microscopic image data is input into an image feature extraction network to obtain microscopic image feature vectors; The original seismic feature vector, the well logging feature vector sequence, and the microscopic image feature vector are concatenated to form a multimodal original feature vector corresponding to each spatial location; Based on the multimodal original feature vectors, knowledge-guided vector generation, feature fusion, and graph attention network processing are performed to output knowledge-enhanced exploration feature representations.

7. The intelligent prediction method for shale gas in karst areas based on artificial intelligence according to claim 1, characterized in that, The process of calling the geological process coupled simulation model to process the knowledge-enhanced exploration feature representation simulates the interaction between karst development and shale reservoir formation, generating coupled evolution simulation results of karst reservoirs, including: The geological process coupled simulation model includes a karst development process simulation module and a shale reservoir formation process simulation module. The knowledge-enhanced exploration feature representation is input into the karst development process simulation module, which performs calculations based on a groundwater dynamics sub-model and a rock chemical dissolution sub-model. The groundwater dynamics sub-model simulates the flow path and velocity changes of groundwater in the fracture network based on the paleohydrogeological data in the knowledge-enhanced exploration feature representation; the rock chemical dissolution sub-model calculates the dissolution rate and amount of rock along the flow path based on the rock mineral composition data and groundwater chemical data in the knowledge-enhanced exploration feature representation. The karst development process simulation module integrates the results of the groundwater dynamics sub-model and the rock chemical dissolution sub-model to dynamically simulate the evolution of the karst fissure system from its initial state to its current state, and outputs historical data on the evolution of the karst system, which includes the volume distribution of dissolution space and changes in the connectivity of dissolution channels in different geological periods. The knowledge-enhanced exploration feature representation and the karst system evolution history data are simultaneously input into the shale reservoir accumulation process simulation module; the shale reservoir accumulation process simulation module performs calculations based on the source rock hydrocarbon generation and expulsion sub-model and the fluid migration and accumulation sub-model; the source rock hydrocarbon generation and expulsion sub-model simulates the hydrocarbon generation intensity and expulsion history of the shale system based on the shale organic geochemical data and stratigraphic thermal history data in the knowledge-enhanced exploration feature representation; the fluid migration and accumulation sub-model simulates the hydrocarbon generation intensity and expulsion history of the shale system based on the karst system evolution history... The historical data provides information on the evolution of reservoir space and migration channels, as well as the hydrocarbon fluid source provided by the source rock hydrocarbon generation and expulsion sub-model. It simulates the migration path, accumulation location, and accumulation amount of the generated hydrocarbon fluid in the evolving karst fracture-cavity system. The shale reservoir formation simulation module integrates the results of the source rock hydrocarbon generation and expulsion sub-model and the fluid migration and accumulation sub-model to output historical data on shale gas formation evolution. This historical data includes hydrocarbon saturation distribution and formation pressure field evolution at different geological periods. The two simulation modules perform bidirectional data feedback and iterative simulation until the output results reach a dynamic equilibrium state and are integrated to generate karst reservoir coupled evolution simulation results.

8. The intelligent prediction method for shale gas in karst areas based on artificial intelligence according to claim 1, characterized in that, The step of predicting the spatial distribution of shale gas enrichment zones in the target karst area based on the coupled evolution simulation results of the karst reservoir includes: The shale gas distribution data at the final geological time is extracted from the coupled evolution simulation results of the karst reservoir. The shale gas distribution data includes spatial distribution data of gas saturation and spatial distribution data of formation pressure coefficient. The karst spatial structure data at the final geological time is also extracted from the coupled evolution simulation results of the karst reservoir. The karst spatial structure data includes spatial distribution data of dissolution porosity and spatial distribution data of fracture-cavity system connectivity. The spatial distribution data of gas saturation is compared with a preset gas saturation threshold value to obtain spatial regions where the gas saturation value exceeds the threshold value, and these regions are marked as areas where the gas saturation meets the standard. The spatial distribution data of formation pressure coefficient is compared with a preset pressure coefficient threshold value to obtain spatial regions where the formation pressure coefficient value exceeds the threshold value, and these regions are marked as areas where the pressure coefficient meets the standard. The spatial distribution data of dissolution porosity is compared with a preset porosity threshold value to obtain spatial regions where the dissolution porosity value exceeds the threshold value, and these regions are marked as areas where the porosity meets the standard. The spatial distribution data of the slot system is analyzed in three dimensions to identify the spatially interconnected slot systems and calculate the total volume and geometric complexity of each connected body. Multi-regional spatial overlay analysis, enrichment favorability index calculation, and enrichment zone delineation were performed to generate spatial distribution prediction results for shale gas enrichment zones.

9. The intelligent prediction method for shale gas in karst areas based on artificial intelligence according to claim 1, characterized in that, The method of iteratively optimizing the weights of nodes and edges in the karst geological knowledge network based on the predicted spatial distribution of shale gas enrichment zones, and updating the structure of the karst geological knowledge network, includes: The predicted spatial distribution of shale gas enrichment zones is compared with the actual exploration results, which include the production capacity data of drilled wells and the test results data. Based on the comparison results, the consistency index between the predicted results and the actual situation is calculated. The consistency index includes spatial location consistency and enrichment degree consistency. Identify the regions in the prediction results where the consistency index meets the set consistency criteria and the regions where the consistency index does not meet the set consistency criteria; In the karst geological knowledge network, network nodes and edges related to areas whose location and conformity index meet the set conformity standards constitute a knowledge subgraph that supports correct prediction; in the karst geological knowledge network, network nodes and edges related to areas whose location and conformity index do not meet the set conformity standards constitute a knowledge subgraph that leads to prediction deviation. The process involves calculating incremental weights, updating network edge weights, and iteratively optimizing the network structure based on knowledge subgraphs.

10. An intelligent prediction system for shale gas in karst areas based on artificial intelligence, characterized in that, The invention includes a processor and a computer-readable storage medium storing machine-executable instructions, which, when executed by a computer, implement the artificial intelligence-based intelligent prediction method for shale gas in karst areas as described in any one of claims 1-9.

Images