Knowledge Graph Construction Method for Complex Geological Structures Based on Knowledge Reasoning

Abstract

The present invention discloses a method for constructing a knowledge graph of complex geological structures based on knowledge reasoning. Aiming at the problem that the traditional knowledge graph construction cannot be used for three-dimensional geological structure modeling, the present invention first establishes the schema layer of the knowledge graph of complex geological structures; then establishes the data layer of the knowledge graph of complex geological structures; secondly, under the constraints of the schema layer, the input original seismic interpretation data is reasoned to obtain the intersection point and intersection line entity information of the knowledge graph, as well as the topological position relationship between entities, to obtain the basic knowledge graph; finally, the complete topological structure knowledge graph is obtained through the improved VF2 subgraph isomorphism matching algorithm.

Description

Technical Field

[0001] The present invention belongs to the field of seismic data processing, and particularly relates to a technology for constructing a geological structure knowledge graph. Background Art

[0002] The three-dimensional model of the underground structure visually displays the geometric shapes and spatial relationships of underground geological interfaces and geological bodies such as strata and faults. The process of establishing the model is called structural modeling. Structural modeling is usually not the ultimate goal, but rather provides support for numerical and physical simulations of complex phenomena (such as seismic propagation and fluid migration), depth-domain imaging, lithology interpretation, reservoir modeling, etc. Based on structural modeling, work such as sequence modeling and attribute modeling can be carried out, directly supporting reserve calculation, well location deployment, and formulation of oil and gas development plans. It is one of the most important basic tasks in underground resource exploration and development. The development of computer three-dimensional visualization technology has promoted the progress and development of related technical methods for three-dimensional geological structure modeling. Due to the high complexity of geological phenomena, there are great uncertainties and deviations in three-dimensional structural modeling. In the actual process of structural modeling, mainly based on a small amount of existing original data, through geological surface reconstruction technology and visualization methods, a simulation result is obtained, but the actual geological situation cannot be fully reproduced. These uncertainties lead to uncertainties in the connection relationship between the geometric shape of the three-dimensional structural model itself and the geological interface. In actual structural modeling, due to the high cost of obtaining interpretation data, only limited data can be obtained within a certain research area, which requires more incorporation of expert experience and interpretations to construct a relatively accurate stratigraphic model.

[0003] The key to intelligent structure modeling lies in the construction of a knowledge graph, which provides modeling constraints and facilitates subsequent quality assessment by constructing a knowledge graph. The concept of a knowledge graph was proposed by Google in 2012 and widely applied. It is a structured semantic knowledge base. From the perspective of a graph, a knowledge graph is a concept network and also a symbolic representation in the physical world. Its nodes represent entities in the real world, and the edges connecting the entities represent the semantic relationships between the entities (Steiner et al., 2012). The construction process of a knowledge graph can be divided into two parts: schema layer construction and data layer construction. An ontology library is used to standardize the fact expression in the data layer (Alberti, 2016), (Poveda-Villalón et al., 2014). An ontology is a conceptual template for a structured knowledge base. The knowledge base constructed through the ontology library has advantages such as strong hierarchical structure and low redundancy (Tang et al., 2022). The data layer consists of a series of entities or concepts. Knowledge is stored in units of facts, and the data layer represents knowledge in the form of triples (entity 1 - relationship - entity 2) or (entity - attribute - attribute value). The topological structures in a geological structure model are all about the connections and relationships between objects. In a structure model, topological changes can be introduced in different ways: (1) adding or removing geological objects; (2) changing the truncation rules between two faults or bedding planes (Cherpeau et al., 2010). Zhan et al. (2022) constructed geometric constraints for a structure model through a knowledge graph, used the knowledge graph to represent the constraint relationships between knowledge, and when the structure model did not conform to expert cognition, modified the knowledge graph for quality assessment, avoiding the operation of repeated modeling (Zhan et al., 2022).

[0004] In recent years, knowledge graph technology has played an important role in the field of oil and gas exploration. Tang et al. (2022) proposed an engineering-based method to improve traditional natural language processing methods by leveraging the correlation of geological data, thereby constructing a domain knowledge graph based on the ontology of oil exploration and development (Tang et al., 2022). Wang (2018) et al. constructed a knowledge graph for knowledge management of geographical literature, indicating the importance of constructing a knowledge graph in the field of oil and gas exploration. In the face of geological structure modeling (Wang et al., 2018), traditional geological domain knowledge graphs only use text as the sole organizational form, which is monotonous and incomplete for presenting conceptual or entity information (Ma, 2022). Compared with text, the way of organizing and processing seismic data makes it easier for researchers to analyze and understand 3D geological models. Zhan (2022) et al. applied the knowledge graph to the field of structural modeling and demonstrated that the knowledge graph is not limited to the text domain. These studies prove that it is very important to mine knowledge from seismic data (Zhan et al., 2022).

[0005] In the process of constructing a structural knowledge graph, the knowledge graph can be regarded as a knowledge base containing a large amount of data, which is semantic for humans and computable for machines. Knowledge reasoning can be carried out relying on it, and the reasoning of the existing knowledge graph can be extended to new knowledge and conclusions, thus expanding the knowledge base. The construction of a complex knowledge reasoning method can provide services for oil and gas exploration users, oil and gas exploration experts, and oil and gas exploration software developers. Experts can more easily discover the connections between knowledge through the visualization of the knowledge graph. Constraint modeling through the knowledge graph can improve the efficiency of intelligent structural modeling. Geological experts can obtain a more complete knowledge base through knowledge graph reasoning to constrain structural modeling and discover new geological laws.

[0006] Three-dimensional structural modeling is a process of repeated updating. Due to the uncertainty of interpretation data, quality control is required throughout the process. Traditional structural modeling cannot modify incorrect models that do not conform to expert cognition at any time, which consumes a large amount of time during the construction process, is not conducive to the subsequent update of the model, and cannot meet the needs of geological comprehensive research in actual engineering. A knowledge graph can be constructed. Through human-computer interaction, a knowledge graph of the structural model is established, and then the knowledge graph is transformed into geometric constraints, which can be directly used for the reconstruction problem of geological surfaces in the modeling process. Compared with data-driven structural modeling methods, it is relatively easier to manually modify the knowledge graph. However, no general knowledge graph construction process is given, or it can only handle special geological structures. Traditional knowledge graph construction cannot be used in three-dimensional geological structure modeling, and the existing knowledge graph construction has the following problems:

[0007] 1. In the existing technology, there is no knowledge graph construction method for constraining 3D geological structure modeling to handle common geological structures;

[0008] 2. There is a lack of a rule base that can be used in actual engineering for the knowledge graph construction work of common structural patterns;

[0009] 3. There is a lack of an inference method for constructing a knowledge graph of complex geological structures, making it difficult to mine hidden entity and relationship data from the interpreted data. Summary of the Invention

[0010] To solve the above technical problems, the present invention proposes a method for constructing a knowledge graph of complex geological structures based on knowledge reasoning, establishing a large-scale structural modeling structure knowledge base for fusing the topological position relationships of geological surface space elements, and ultimately providing technical support for 3D geological structure modeling technology.

[0011] The technical solution adopted by the present invention is: a method for constructing a knowledge graph of complex geological structures based on knowledge reasoning, including:

[0012] S1. Establish a pattern layer of the knowledge graph of complex geological structures;

[0013] S2. Establish a data layer of the knowledge graph of complex geological structures; under the constraint of the pattern layer, infer the intersection point and intersection line entity information of the knowledge graph and the topological position relationships between entities from the input original seismic interpretation data to obtain a basic knowledge graph;

[0014] S3. Obtain a complete topological structure knowledge graph through an improved VF2 subgraph isomorphism matching algorithm.

[0015] The beneficial effects of the present invention: The geometric constraint knowledge graph constructed by the present invention can be used to constrain 3D geological structure modeling. Using knowledge graph technology to construct a knowledge graph in the geological field is convenient for geologists to query and analyze geological data, discover geological laws and evolution trends, constrain and optimize the model according to the geological knowledge in the knowledge graph, and use the constraint relationships in the knowledge graph to define the topological relationships and geometric properties of the intersection lines to improve the accuracy and controllability of the intersection lines, thereby providing the geometric constraints required for surface reconstruction to ensure that the surface is as smooth as possible while meeting the geological structure characteristics, and the optimized geological model is more in line with the actual geological situation Brief Description of the Drawings

[0016] Figure 1 is the method flow chart of the present invention;

[0017] Figure 2 is the pyramid hierarchical structure representing the knowledge graph;

[0018] Figure 3Topological rule base for knowledge reasoning;

[0019] Among them, (a) is the intersection of the bedding plane and the fault plane, generating two intersection lines; (b) is the intersection of the fault plane and the fault plane, generating one intersection line; (c) is the intersection of the bedding plane and the bedding plane, generating one intersection line; (d) is the intersection of the fault plane and the boundary plane, generating one intersection line; (e) is the intersection of the bedding plane, the fault plane and the boundary plane, generating two intersection lines; (f) is the merging of intersection points; (g) is the geological block entity formed by the intersection of the bedding plane and the fault plane with the boundary plane; (h) is the geological block entity constructed by the intersection of the bedding plane and the bedding plane with the boundary plane; (i) is the geological surface entity constructed by the intersection of the bedding plane and the bedding plane with the boundary plane;

[0020] Figure 4 Common geological structure patterns;

[0021] Figure 5 Process for constructing the knowledge graph data layer;

[0022] Figure 6 Workflow of knowledge reasoning;

[0023] Figure 7 Process of the basic knowledge graph obtained after the prior knowledge is matched by the rule base;

[0024] Among them, (a) is the prior knowledge, (b) is the line entities, point entities and their topological position relationships obtained after the prior knowledge passes the rule matching, and (c) is the point entities obtained by the intersection of the line entities after passing the rule base matching;

[0025] Figure 8 Pure topological structure knowledge graph obtained after sub-graph reasoning;

[0026] Figure 9 Knowledge graph and its representation model;

[0027] Among them, (a) represents the hierarchical structure diagram of the knowledge graph, and (b) is the wireframe model of the knowledge graph corresponding to (a);

[0028] Figure 10 Pyramid hierarchical structure diagram and wireframe diagram constructed using actual work area data;

[0029] Among them, (a) is the pyramid hierarchical structure diagram of the structural model knowledge graph; (b) is the diagram representing the knowledge graph through the wireframe model; (c) is the wireframe model entity filling diagram, which is more in line with the actual geological conditions;

[0030] Figure 11 Actual work area results of 3D geological structure modeling based on knowledge reasoning. Specific implementation method

[0031] For the convenience of those skilled in the art to understand the technical content of the present invention, the following introduces the relevant prior art:

[0032] 1. GFO Ontology

[0033] GFO (General Formal Ontology) is a philosophical ontology framework that provides a general formal language and a set of basic entity types and relationships to support ontology modeling and knowledge representation. The architecture of the GFO ontology includes the following three parts:

[0034] Entity types: The GFO ontology provides a set of basic entity types, including material entities, abstract entities, and process entities, etc. Material entities are divided into basic materials, composite materials, spatial materials, and spatial regions. Abstract entities include mass, quantity, time, and space, etc. Process entities include events and actions, etc.; Entity relationships: The GFO ontology provides a set of basic entity relationships, including part-whole relationships, component-structure relationships, adjacent relationships, inclusion relationships, comparison relationships, classification relationships, participation relationships, occurrence relationships, result relationships, etc. These entity relationships can be used to describe the relationships between entities; Entity attributes: The GFO ontology also provides a set of basic entity attributes, including color, shape, size, location, state, and ability, etc. These entity attributes can be used to describe the characteristics and attributes of entities.

[0035] The GFO ontology provides an extensible modeling framework that can support the ontology modeling needs of multiple fields, and is particularly suitable for modeling in fields such as physics, space, and geography. In a topological structure knowledge graph, the relationships between entities are often defined based on geometric features such as their positions, distances, and directions, and the GFO ontology provides a mechanism to describe these geometric features, so it can be used as an effective ontology option.

[0036] 2. VF2 Subgraph Isomorphism Matching Algorithm

[0037] The VF2 algorithm is an algorithm for graph matching that can be used to solve the subgraph isomorphism problem, that is, to find a one-to-one mapping of nodes and edges between two graphs so that one graph can be transformed into another graph through this mapping. The VF2 algorithm is based on the idea of backtracking search and can efficiently handle the matching problem of large-scale graphs.

[0038] The core idea of the VF2 algorithm is to mark the states of the nodes in two graphs, and then perform backtracking search to match the nodes until a set of matches is found or all possible matches are traversed. During the matching process, the VF2 algorithm uses state vectors to record the states of the matched nodes and the structural information of the subgraphs, and at the same time accelerates the search according to the predefined heuristic rules. Specifically, during the search process, the VF2 algorithm maintains two state vectors, one representing the state vector of the matched nodes and the other representing the state vector of whether the candidate nodes can be matched, and then determines whether the node can be matched with the current subgraph according to the state of the candidate node and the state of the previously matched nodes. The VF2 algorithm also uses techniques such as pruning and backtracking to accelerate the search process, thereby improving the matching efficiency. The VF2 algorithm itself cannot directly query the number of all isomorphic subgraphs, but this function can be achieved by making some modifications to the VF2 algorithm. Specifically, based on the VF2 algorithm, each pair of matched nodes can be marked, and then the search for the unmatched node pairs can be continued. Each time a new match is found, the marked node pairs are removed from the current search. When there are no unmatched nodes in the search, an isomorphic subgraph can be obtained.

[0039] As Figure 1 shown, the method of the present invention includes the following processes:

[0040] 1) Data preparation

[0041] According to the seismic data obtained from oilfields and geophysical service companies, interpreters perform horizon and fault interpretation on the seismic data to obtain the original seismic interpretation data. The interpretation data acts on the data layer. The interpretation data of a work area often contains the point cloud data of several horizons and faults. It is necessary to preprocess the interpretation data, including extracting the work area boundary and data normalization. After obtaining the interpretation data, the intersection relationship between the horizon plane and the fault plane is judged by means of a spatial bounding box, and it is used as prior knowledge data in the form of an adjacency matrix.

[0042] 2) Construction of the schema layer of the complex geological structure knowledge graph

[0043] 21) Select GFO as the top - level ontology for constructing the geological model. Design the domain ontology in the schema layer of the knowledge graph according to GFO. The entities in the schema layer of the constructed modeling knowledge graph specifically include point, line, surface, and block entities. Entities in the knowledge graph refer to geometric objects, and these constituent units are called n - cells, where n refers to the dimension of the element. They can be divided into the following four types: point (0 - cell), line (1 - cell), surface (2 - cell), and volume (3 - cell). Relations refer to the topological geometric relations, positional relations, and compositional relations (inclusion relations between target entities) between two entities. Attributes refer to the position and closure of geometric objects in construction modeling, with the GFO ontology as a broad concept. To facilitate the participation of expert knowledge in construction modeling, the knowledge graph is represented as a pyramid - like hierarchical structure, as Figure 2 shown.

[0044] 22) Construct a geological structure rule base Figure 3 , providing constraint conditions for subsequent knowledge reasoning. Specifically, the intersection relations between specific layer surfaces, fault surfaces, and boundary surfaces, the rules for point entities and line entities obtained through reasoning, and the topological positional relations between entities are Figure 3 (a) - (f) are the prior rules of the rule base (such as the intersection relations between layer surfaces and boundaries, layer surfaces and faults, faults and boundaries, etc.). The association relations between intersection lines, intersection points, layer surfaces, boundaries, and faults are established, and the simple diagrams of basic sub - surfaces and block entities are given as Figure 3 shown in (g) - (i) below, which are used to constrain subsequent sub - graph isomorphism reasoning research. The specific constraints are as follows: The intersection of two layer surfaces generates an intersection line, the intersection of a layer surface and a fault surface generates two intersection lines, the intersection of two fault surfaces generates an intersection line, the intersection of a layer surface and a boundary surface generates an intersection line, the intersection of two boundary surfaces generates an intersection line. The intersection of a layer surface, a fault surface, and a boundary surface requires intersection line segmentation, and the pairwise intersection of three boundary surfaces requires intersection point fusion. The surface entity refers to a closed sub - surface loop, and the block entity is the space composed of closed sub - surfaces, all of which can be composed of upper - layer entities. These rules form the basic rule base;

[0045] 23) Figure 4 List common geological structure patterns, such as faults, intrusions, and unconformity structures, etc. Most common geological structures exist in fault structures and unconformity structures. In the proposed rules, the above - mentioned models can be represented in the form of a knowledge graph through the knowledge graph.

[0046] The topological structure knowledge graph can be used to represent and describe various geological structure patterns, including fractures, intrusions, unconformities, etc. By representing these structure patterns in the form of nodes and using edges to describe the relationships between them, a topological structure knowledge graph can be constructed. In the knowledge graph, the relationships between different nodes can be represented as topological relationships, which can more accurately represent the topological relationships in the geological structure and can automatically perform topological checks and constraints during the modeling process to ensure the correctness of the geological model.

[0047] Consider whether the geological processes and structure types covered in the rule base are extensive, covering common geological structure types such as faults, folds, intrusions, etc., and geological structures at different scales, such as large-scale topography and small-scale rock structures. The present invention describes the universality of the rule base from three perspectives: (1) Types of covered structure patterns: The types of structure patterns included in the rule base can be listed, such as fracture, intrusion, unconformity structure, etc. See specifically Figure 2 as shown. This rule base can cover most of the common structure patterns in the figure; (2) Spatial distribution of covered structure patterns: In addition to covering the types of structure patterns, the present invention also considers the distribution of these structure patterns in the geological space, and these structure patterns can describe various structure patterns in different geological periods and different geological regions; (3) Complexity of covered structure patterns: The complexity of geological structure patterns is different. Some simple patterns may be easier to describe and identify, while some complex patterns may require more rules to describe. In the subsequent simulation, the present invention has implemented relatively complex patterns (such as Figure 10 ). Therefore, the structure patterns that the rule base of the present invention can cover can handle structure patterns of different complexities. Constructing a rule base is also very important for the research and analysis of geological structure patterns in order to discover new structure patterns and laws, and thus continuously improve and optimize the rule base.

[0048] 3) Construction of the data layer of the complex geological structure knowledge graph

[0049] Figure 5Shows the workflow of the knowledge graph construction proposed by the present invention. In the first part of the workflow, the topological position relationships of the layer planes and fault planes (based on the spatial bounding box method) are automatically extracted from the input structural interpretations. The surface relationship data in the geological space is converted into topological geometric space relationships. In the second part, a general rule library for the topological relationship knowledge graph is constructed, and the rule library is used to constrain the actual data set through rule matching to construct the spatial topological knowledge graph; in the third part, we use the knowledge graph obtained in the second part for constraint, and represent the spatial topological knowledge graph in the form of a wireframe model. If the initially constructed knowledge graph is inconsistent with the model cognition of the earth scientist in charge of modeling, manual editing should be performed to update the knowledge graph, and the corresponding rules should be imported into the rule library.

[0050] Figure 6 Workflow of knowledge reasoning: In the first part of the workflow, we represent the rule library in the form of a graph and construct the entity and relationship information in the topological space knowledge graph through rule matching. In the second part, the hidden knowledge (face entities and block entities) in the knowledge graph is mined through subgraph isomorphism query.

[0051] First, the intersection relationship between the surfaces (layer planes and fault planes) is judged through the spatial bounding box technology for the interpretation data, and the data (layer planes and fault planes) is imported to construct the work area boundary (up and down, left and right, and front and back). Through rule matching with the constructed rule library, the topological position relationships of cross-layer entities such as the position relationship between the intersection point entity and the intersection line entity, and the position relationship between the intersection line entity and the sub-face entity are obtained, forming a basic knowledge graph. Then, through the improved VF2 subgraph isomorphism matching algorithm (specifically, on the basis of the VF2 algorithm, each matched node pair can be marked, and at the same time, the matching information of the edge labels between the nodes is included, and then the unmatched node pairs are continued to be searched. Each time a new match is found, the already marked node pairs are excluded from the current search to query the number of all isomorphic subgraphs.), the complete topological structure knowledge graph is obtained. As Figure 6 shown, the detailed steps are as follows:

[0052] 1. The fault plane and the layer plane are regarded as a thin slice composed of two layers of surfaces (i.e., adding twin sub-faces), and the intersection line between adjacent sub-volumes on the same section is split into two intersection line entities, with the number of nodes and the node order being exactly the same. When the intersection point is a node entity on the same fault intersection point, it is split into two node entities, and the preliminary topological relationship knowledge graph is constructed through the defined rules;

[0053] 2. Guided by the prior rule library, the intersection point and intersection line entity information of the knowledge graph, as well as the topological position relationships between the entities, are inferred. The simulation process is as Figure 7 shown, Figure 7(a) is the prior knowledge, where the intersection relationship between the layer plane and the fault plane is input by geological experts. (b) is the line entities, point entities and their topological position relationships obtained by rule matching of the prior knowledge. For example, Figure 7 (c) is the point entity obtained by matching the intersection of line entities through the rule library, and entities with the same attributes are merged through knowledge fusion to complete the knowledge graph reasoning in the first stage, and it is represented in a pyramid hierarchical structure.

[0054] 3. Use the VF2 algorithm for isomorphic graph matching to obtain the knowledge graph (surface entity and block entity) that matches the prior rules of the Figure 3 (a)-(f) in the rule library. State labels are assigned to the nodes of the two graphs, and then backtracking search is used to match the nodes until a set of matches is found or all possible matches are traversed. During the matching process, the VF2 algorithm uses state vectors to record the states of the matched nodes and the structural information of the subgraph, and at the same time accelerates the search according to the pre-defined heuristic rules. Specifically, the VF2 algorithm maintains two state vectors during the search process, one representing the state vector of the matched nodes and the other representing the state vector of whether the candidate nodes can be matched, and then determines whether the node can be matched with the current subgraph according to the state of the candidate node and the state of the previously matched nodes. The VF2 algorithm itself cannot directly query the number of all isomorphic subgraphs, but this function can be achieved by making some modifications to the VF2 algorithm. Specifically, on the basis of the VF2 algorithm, each pair of matched nodes can be marked, including the matching information of the edge labels between the nodes, and then the unmatched node pairs are continued to be searched. Each time a new match is found, the marked node pairs are removed from the current search to query the number of all isomorphic subgraphs, so as to construct a complete knowledge graph and represent it in a pyramid hierarchical structure, as Figure 8 shown;

[0055] 4. The quality of the knowledge obtained through knowledge reasoning is usually not guaranteed. As Figure 9 (a) shows, so there needs to be a quality assessment process before adding it to the knowledge base. The wireframe model is used to achieve the quality control of the knowledge graph in the construction modeling, as Figure 9 (a) and (b) show. When the constructed knowledge graph does not conform to the expert's cognition, knowledge update is carried out;

[0056] Figure 9 In (a), the nodes represent geological entities, and the edges connecting the nodes are the relationships between the entities (from bottom to top are the intersection point entity, intersection line entity, sub-surface entity and geological block entity in turn. The edges are the topological semantic relationships between the entities). Figure 9 (b) is the same Figure 9(a) The wireframe model of the corresponding knowledge graph, which improves the readability of the knowledge graph through the wireframe model and can judge whether it conforms to human cognition (theoretical data) through expert knowledge;

[0057] 5. Knowledge graph update. In an incremental update manner, using the currently newly added data (interpretation data from other work areas) as input, different geological structures will appear. By adding rules to the existing rule base to constrain the update of the knowledge graph in the data layer, new entity and relationship information can be added. This method consumes less resources and realizes the update of the knowledge graph by continuously supplementing geological structure rules. After constructing the knowledge graph and representing it with a wireframe model, it is easy for experts to see the models in the knowledge graph that do not conform to cognition, and an accurate knowledge graph can be obtained by querying and modifying some data. This is a process that needs to be carried out during the initial construction of the knowledge graph. When the rule base is complete, the rule base can guide the automatic update of the knowledge graph without the need for expert judgment every time.

[0058] Taking an actual work area (a certain work area in Sichuan) as an example, the effects of using the method of the present invention are described:

[0059] The knowledge graph processes complex seismic data into structured knowledge through an inference algorithm, and the represented knowledge can be displayed graphically to generate valuable research references. Figure 10 (a) is the pyramid hierarchical structure diagram of the tectonic model knowledge graph, where the nodes represent geological entities, and the edges connecting the nodes are the relationships between the entities. Figure 10 (b) is the diagram of the knowledge graph represented by the wireframe model. Figure 10 The nodes in the first layer in (a) represent Figure 10 the line entities in (b), the second layer corresponds to Figure 10 the line entities in (b), the third layer corresponds to Figure 10 the surface entities in (b). The fourth layer represents Figure 10 the geological block entities in (b). The edges connecting the nodes represent the relationships between the entity knowledge nodes, such as: adjacency, intersection, inclusion, etc. Figure 10 The geological dip angle in (b) is used as attribute information. The dotted line represents the entity information that actually exists after the mapping from the ontology concept node to the entity node but cannot be seen in the front view. Each knowledge entity node has a mapping in the wireframe model.

[0060] The workflow of knowledge graph participating in 3D structural modeling. In the first part of the 3D structural modeling workflow, the knowledge graph we established is converted into constraint conditions that can be directly used for modeling. Here, the constraint conditions refer to the wireframe of the model. First, preprocess the original structural interpretation, including initial smoothing and extraction of extreme points of the horizon. Then, use the knowledge graph of spatial topology to estimate the intersection points of the geological surface using the preprocessed data. Then, automatically modify the original data according to the reliability of the intersection points to avoid conflicts between the intersection points and the original data. The second part uses the intersection points obtained in the first part as constraint conditions and the modified original data to reconstruct the surface by interpolation between sections, establish a closed geological body, and finally obtain a structural model that conforms to the knowledge graph, as Figure 11 shown. The intersection point estimation and automatic modification are to make the finally reconstructed surface conform to the actual geological situation, ensuring the authenticity and reliability of the structural model. At the same time, the knowledge graph as a constraint condition also makes the modeling process more accurate and efficient.

[0061] Those of ordinary skill in the art will realize that the embodiments described herein are for helping readers understand the principles of the present invention and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. For those skilled in the art, various changes and modifications can be made to the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the scope of the claims of the present invention.

Claims

1. A method for constructing a knowledge graph of complex geological structures based on knowledge reasoning, characterized in that, Including: S1. Establish the schema layer of the complex geological structure knowledge graph. Specifically, step S1 is as follows: S11. Select GFO as the top-level ontology of the tectonic geological model. According to GFO, design the domain ontology in the schema layer of the knowledge graph, and construct the semantic entities in the schema layer of the modeling knowledge graph. The entities specifically include point, line, surface, and block entities, and the relationships are topological semantic relationships of covering, intersecting, equal, and containing. S12. Construct a geological structure rule base, specifically including: When a layer plane intersects with a layer plane, a intersection line is generated; when a layer plane intersects with a fault plane, two intersection lines are generated; when a fault plane intersects with a fault plane, an intersection line is generated; when a layer plane intersects with a boundary surface, an intersection line is generated; when a boundary surface intersects with a boundary surface, an intersection line is generated. When a layer plane, a fault plane, and a boundary surface intersect, the intersection line needs to be segmented, and when three boundary surfaces intersect pairwise, the intersection points need to be fused. S13. Represent the common geological structure models in the form of a knowledge graph in the geological structure rule base constructed in step S12. The common geological structure models include: faults, intrusions, and unconformity structures. S2. Establish the data layer of the complex geological structure knowledge graph. Under the constraint of the schema layer, infer the intersection point and intersection line entity information of the knowledge graph and the topological position relationship between entities from the input original seismic interpretation data to obtain a basic knowledge graph. Specifically, step S2 includes: S21. Use the spatial bounding box technology to judge the intersection relationship between surfaces, input the original seismic interpretation data, and construct the work area boundary. S22. Perform rule matching through the constructed rule base to determine the topological position relationship of cross-layer entities and form a basic knowledge graph. S3. Obtain a complete topological structure knowledge graph through the improved VF2 subgraph isomorphism matching algorithm. Specifically, step S3 is as follows: Use the VF2 algorithm for isomorphic graph matching. During the search process, the VF2 algorithm maintains two state vectors. One represents the state vector of the already matched nodes, and the other represents the state vector of whether the candidate nodes can be matched. Then, according to the state of the candidate nodes and the state of the previously matched nodes, it is judged whether to match the node with the current subgraph. It also includes: Mark each pair of already matched nodes, including the matching information of the edge labels between the nodes, and then continue to search for pairs of unmatched nodes. Each time a new match is found, the already marked pairs of nodes are excluded from the current search to query the number of all isomorphic subgraphs, thereby constructing a complete topological structure knowledge graph.

2. A method for constructing a knowledge graph of complex geological structures based on knowledge reasoning according to claim 1, characterized in that The topological position relationship of the cross-layer entities described in step S22 includes: the position relationship between the intersection point entity and the intersection line entity, and the position relationship between the intersection line entity and the sub-surface entity.

3. A method for constructing a knowledge graph of complex geological structures based on knowledge reasoning according to claim 2, characterized in that The topological structure knowledge graph in step S3 is represented in a pyramid hierarchical structure.

Images