Oil and gas exploration heterogeneous data bloodline fusion management method based on knowledge graph

Abstract

The application discloses a knowledge graph-based oil and gas exploration heterogeneous data blood relationship fusion management method, which comprises the following steps: constructing a knowledge graph, defining cross-processing stage semantic identifiers in the knowledge graph, and associating the semantic identifiers to data object nodes and processing activity nodes; determining a semantic continuity state by comparing the semantic identifier sets of the upstream and downstream nodes of a propagation unit; determining a key blood relationship propagation boundary in combination with the semantic continuity state, change event attributes and activity sensitivity, and performing influence propagation along a semantic continuous and uninterrupted path to obtain an affected node set; and distinguishing semantic level mismatching from data content updating according to the key blood relationship propagation boundary and the affected node set, and respectively determining a minimum necessary repair range and a minimum necessary recalculation range. The application can improve the accuracy and traceability of oil and gas exploration heterogeneous data blood relationship management, and reduce invalid repair and excessive recalculation.

Description

Technical Field

[0001] This application relates to the field of knowledge graph application technology in oil and gas exploration, and more specifically, to a knowledge graph-based method for the lineage fusion and governance of heterogeneous oil and gas exploration data. Background Technology

[0002] Oil and gas exploration operations typically involve multiple stages, including seismic data acquisition and processing, velocity modeling, well-seismic calibration, stratigraphic interpretation, fault interpretation, reservoir prediction, trap evaluation, and reserve calculation. The data generated from these stages are diverse, including seismic bodies, well curves, interpreted stratigraphic levels, fault results, attribute volumes, and reservoir prediction rasters, as well as trap boundaries, reserve parameter tables, and related processing records. Complex derivation relationships, version inheritance relationships, and professional constraints exist between different data objects. Furthermore, with multiple rounds of processing, interpretation, and the parallel existence of multiple versions of results, professional bases such as interpretation domains, velocity models, well-seismic calibration results, stratigraphic standards, and coordinate references continuously change throughout the operation.

[0003] Existing data governance methods often rely on data catalogs, metadata registration, process logs, or task-level lineage records. These typically only reflect whether source relationships, generation relationships, or upstream / downstream call relationships exist between data objects, making it difficult to determine whether these lineage relationships remain valid at the professional semantic level. For example, after changes to upstream velocity models, well-seismic calibration versions, or stratigraphic standards, downstream results may still retain the original lineage connections in form, but the professional conditions upon which they are based are no longer consistent. Existing methods also struggle to distinguish the actual impact of different changes such as format conversion, parameter adjustment, interpretation domain conversion, and local revisions on downstream results, leading to unclear boundaries of impact propagation.

[0004] Therefore, when upstream data or professional basis changes, existing technologies often can only be used to address the issue through full-chain investigation, manual verification, or large-scale recalculation. This is not only inefficient, but also prone to problems such as invalid recalculation, missed repairs, or the continued spread of incorrect lineages. Summary of the Invention

[0005] This application provides a knowledge graph-based method for managing the lineage fusion of heterogeneous oil and gas exploration data, in order to at least solve some of the technical problems existing in the related technologies described above.

[0006] According to a first aspect of the embodiments of this application, a method for the lineage fusion and governance of heterogeneous oil and gas exploration data based on knowledge graphs is provided, comprising: constructing a knowledge graph, wherein the knowledge graph includes data object nodes, processing activity nodes, professional object nodes, and governance status nodes, and the data object nodes are associated with each other through processing activity nodes and connected by edges carrying semantic constraints; defining cross-processing stage semantic identifiers in the knowledge graph and associating them with data object nodes and processing activity nodes; splitting the lineage path in the knowledge graph into propagation units, each propagation unit consisting of an upstream data object node, a processing activity node, and a downstream data object node, comparing the semantic identifier sets of the upstream and downstream nodes of each propagation unit, and determining the semantic continuity state of each propagation unit; determining key lineage propagation boundaries in the lineage network based on the semantic continuity state, change event attributes, and activity sensitivity; performing influence propagation along semantically continuous paths that are not truncated by the key lineage propagation boundaries to obtain a set of affected nodes; and determining semantic mismatch and data content updates based on the key lineage propagation boundaries and the set of affected nodes, determining the minimum necessary repair range for semantic mismatch and the minimum necessary recalculation range for data content updates.

[0007] As an alternative approach, semantic identifiers are categorized into strongly constrained semantic identifiers, moderately constrained semantic identifiers, and weakly constrained semantic identifiers based on the degree to which conflicts affect lineage validity. Among these, conflicts in strongly constrained semantic identifiers are directly determined to be semantic mismatches when there is no explicit transformation relationship recorded by the processing activity node. Conflicts in moderately constrained semantic identifiers are determined to constitute semantic mismatches in conjunction with the activity type of the processing activity node. Weakly constrained semantic identifiers are used for candidate ranking and auxiliary verification.

[0008] As an optional approach, the semantic continuity status of each propagation unit is determined by: for each propagation unit, checking whether the values ​​of the upstream and downstream data object nodes in the strong constraint semantic identifier are consistent or whether there is an explicit transformation relationship recorded by the processing activity node; if the strong constraint semantic identifiers are consistent or there is an explicit transformation relationship, the propagation unit passes at the strong constraint level; if the strong constraint semantic identifiers conflict and there is no explicit transformation relationship, the propagation unit is marked as a semantic mismatch unit; after passing at the strong constraint level, checking whether there is a change in the intermediate constraint semantic identifier; if there is no change in the intermediate constraint semantic identifier, the propagation unit is marked as strongly continuous; if there is a change in the intermediate constraint semantic identifier, it is determined whether the change is an interpretable change based on the activity type of the processing activity node; if it is an interpretable change, it is marked as weakly continuous, and if it is not an interpretable change, it is marked as a semantic mismatch unit; if the semantic identifier in the propagation unit is missing and cannot be determined, it is marked as a continuity uncertainty unit and enters the supplementary verification or candidate repair process.

[0009] As an optional approach, for propagation units marked as semantic mismatch units, the mismatch is further subdivided into version reversal mismatch, interpretation domain mismatch, hierarchical standard conflict mismatch, coordinate reference mismatch, well-seismic calibration lag mismatch, velocity model replacement not propagated mismatch, local revision out-of-bounds propagation mismatch, or upstream basis missing mismatch, based on the semantic identifier conflict type that caused the mismatch. The upstream basis missing mismatch refers to the failure of the upstream data object identifier or processing activity identifier recorded in the lineage record to be associated with the corresponding node in the knowledge graph. The mismatch type serves as the input for subsequently determining the key lineage propagation boundary.

[0010] As an optional approach, key lineage propagation boundaries are determined, including forming propagation boundaries in the lineage network according to the following conditions: if the semantic continuity state of a propagation unit is semantic mismatch, a propagation boundary is formed at the downstream entry point of that propagation unit; if the change event involves an actual update of a strong constraint semantic identifier, a candidate propagation boundary is formed downstream of each processing activity node that takes the data object node carrying the actual update strong constraint semantic identifier as input; for the candidate propagation boundary, the sensitivity of the corresponding processing activity node to the current change event is determined based on activity sensitivity; if sensitive, the candidate propagation boundary is ineffective and affects continued propagation; if insensitive, the candidate propagation boundary becomes effective as a termination propagation boundary; if the impact range of the change event is limited to a local spatial region or local segment, a pruning propagation boundary is formed when the applicable scope of the downstream data object node does not intersect with the impact range; if a node already has a recalculation confirmation record or repair confirmation record for the current change source, a recovery propagation boundary is formed downstream of that node.

[0011] As an optional approach, activity sensitivity is determined by an activity sensitivity matrix, which is arranged with change type as the row and activity type as the column, and the matrix element values ​​are sensitive or insensitive; the activity sensitivity matrix is ​​pre-configured by domain experts according to the oil and gas exploration and processing flow.

[0012] As an optional approach, influence propagation is performed along a semantically continuous path that is not truncated by the key lineage propagation boundary. This includes: classifying the lineage edges in the knowledge graph according to their semantic properties into content-equivalent edges, parameter-sensitive edges, interpretation-dependent edges, local mapping edges, and domain-transformation edges; starting from the upstream node that has changed, traversing each propagation unit layer by layer in the downstream direction; stopping propagation and recording the propagation boundary position when encountering a semantic mismatch unit; not amplifying the influence range when encountering a content-equivalent edge; determining whether to continue propagation based on the activity sensitivity when encountering a parameter-sensitive edge; trimming the influence range to the intersection with the influence range of the changed event when encountering a local mapping edge, stopping propagation if the intersection is empty; checking for the existence of a valid domain-transformation semantic identifier record when encountering a domain-transformation edge, stopping propagation if it does not exist; stopping propagation when encountering a node whose status has been recalculated and confirmed; after traversal, all reached downstream nodes are gathered into the affected node set, and each downstream node is labeled with the influence type of being affected in form or data content.

[0013] As an optional approach, determining the minimum necessary repair scope for semantic mismatch includes: in the graph subgraphs located upstream and downstream of the key lineage propagation boundary, retrieving candidate alternative links that satisfy strong constraint semantic identifier consistency or have an explicit transformation relationship, and medium constraint semantic identifier consistency or have an interpretable change. The candidate alternative links start from the effective data object node upstream of the propagation boundary, pass through other processing activity nodes and intermediate data object nodes to reach the data object node downstream of the propagation boundary, and each propagation unit along the way passes the semantic continuity judgment; the retrieved candidate alternative links are weighted and sorted according to the number of strong constraint semantic identifier matches, activity sequence completeness, temporal order rationality, and existing intermediate result coverage, with each weight being a non-negative number and the sum of the weights being 1. The weights are determined according to the pre-configured governance priority; when no governance priority is configured, all weights take the same value, and the candidate alternative link with the highest comprehensive score is selected as the repair link; the repair link is written into the knowledge graph, and the repair link is used as the alternative link or supplementary link of the original mismatch path, and the lineage path after governance is restored to a semantically continuous path.

[0014] As an optional approach, determining the minimum necessary recalculation range for data content updates includes: starting from the terminal result node in the affected node set, tracing back along the lineage path to the nearest stable propagation boundary, where the stable propagation boundary is either a recovery propagation boundary or a pruning propagation boundary; if no recovery propagation boundary or pruning propagation boundary exists during the backtracking process, the starting change node of this impact propagation is taken as the backtracking endpoint; the set of nodes covered by the path from the stable propagation boundary to the terminal result node is determined as the minimum necessary recalculation subgraph; the minimum necessary recalculation subgraph does not contain nodes with existing recalculation confirmation records, does not cross spatial regions or layer ranges that do not intersect with the impact range of the change event, and does not contain nodes marked as affected only because they are reached by content equivalence edges.

[0015] According to a second aspect of the embodiments of this application, a knowledge graph-based heterogeneous data lineage fusion and governance system for oil and gas exploration is also provided, comprising: a knowledge graph construction module for constructing a knowledge graph, wherein the knowledge graph includes data object nodes, processing activity nodes, professional object nodes, and governance status nodes, and the data object nodes are associated through processing activity nodes and connected by edges carrying semantic constraints; a semantic identifier definition module for defining cross-processing stage semantic identifiers in the knowledge graph and associating them with data object nodes and processing activity nodes; and a semantic continuity determination module for splitting the lineage path in the knowledge graph into propagation units, each propagation unit consisting of an upstream data object node, a processing activity node, and a governance status node. The downstream data object nodes are composed of semantic identifier sets that are compared with the upstream and downstream nodes of each propagation unit to determine the semantic continuity state of each propagation unit; the propagation boundary determination module is used to determine the key lineage propagation boundary in the lineage network based on the semantic continuity state, change event attributes, and activity sensitivity; the influence propagation execution module is used to execute influence propagation along a semantically continuous path that is not truncated by the key lineage propagation boundary to obtain the set of affected nodes; the governance scope determination module is used to determine semantic mismatch and data content update based on the key lineage propagation boundary and the set of affected nodes, determine the minimum necessary repair scope for semantic mismatch, and determine the minimum necessary recalculation scope for data content update.

[0016] This application unifies the organization of heterogeneous oil and gas exploration data objects, processing activities, professional constraints, and governance status into a knowledge graph, and introduces cross-processing stage semantic identifiers to determine the validity of formal lineage paths at the professional semantic level. This enables the identification of hidden problems such as version reversal, inconsistent interpretation domains, stratigraphic standard conflicts, inconsistent coordinate benchmarks, well-seismic calibration lags, and failure to propagate velocity model replacements. Simultaneously, based on semantic continuity status, change event attributes, and activity sensitivity, key lineage propagation boundaries are determined, ensuring that the impact of upstream changes propagates only along semantically continuous and dependent paths, avoiding indiscriminate diffusion along all formal lineage relationships. Furthermore, governance actions are categorized into semantic repair and data recalculation, with minimum necessary repair and recalculation ranges determined separately. This reduces manual checks and excessive recalculation, improving the accuracy, traceability, and processing efficiency of oil and gas exploration result governance, thereby ensuring the consistency and reliability of the professional conditions upon which downstream interpretation, evaluation, and reserve calculation results are based.

[0017] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Furthermore, no embodiment in this disclosure is required to achieve all the effects described above. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0019] Figure 1 A schematic diagram of a knowledge graph-based method for managing heterogeneous oil and gas exploration data lineage fusion, provided in this embodiment of the disclosure; Figure 2 This is a schematic diagram of the semantic identifier extraction and association process provided in the embodiments of this disclosure; Figure 3 This is a schematic diagram of the semantic continuity state determination process provided in the embodiments of this disclosure; Figure 4 This is a schematic diagram illustrating the process of obtaining the set of affected nodes provided in an embodiment of this disclosure; Figure 5 A flowchart illustrating the minimum necessary repair range and minimum necessary recalculation range provided for embodiments of this disclosure; Figure 6 A schematic diagram of a knowledge graph-based oil and gas exploration heterogeneous data lineage fusion and management system provided in this embodiment of the present disclosure; Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0020] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0021] According to embodiments of this disclosure, a knowledge graph-based method for the fusion and management of heterogeneous data lineage in oil and gas exploration is provided. This method addresses production scenarios in the oil and gas exploration field where multiple rounds of seismic processing, multiple rounds of geological interpretation, and multiple versions of supporting data coexist. In such scenarios, from the acquisition of raw seismic data to the final reserve evaluation, a large number of heterogeneous data objects are involved, including seismic bodies, well curves, stratigraphic interpretation surfaces, fault interpretation results, reservoir prediction grids, trap boundaries, and reserve parameter tables. These objects have complex derivation and version inheritance relationships. The interpretation domain, velocity model version, well-seismic calibration version, and stratigraphic standard coding used in different processing stages continuously change with each processing round, resulting in upstream and downstream data objects maintaining a formal lineage connection while their actual professional basis may no longer be consistent.

[0022] Traditional data governance methods only track source relationships at the file or task level, lacking means to determine lineage validity at the professional semantic level, and also struggling to accurately define the scope of impact after upstream changes. This method can be implemented by a data governance service module deployed in an oil and gas exploration data management platform, or by an independently running governance engine. It organizes heterogeneous data objects and their professional constraints using a knowledge graph, determines the semantic continuity of lineage paths based on semantic identifiers across processing stages, thereby identifying key lineage propagation boundaries, controlling the scope of impact propagation, and determining the minimum necessary repair and minimum necessary recalculation scopes.

[0023] The implementation process of the method described in this application will be described in detail below with reference to specific embodiments. It should be noted that this embodiment is only used to explain this application and is not intended to limit the scope of protection of this application. Conventional adjustments or substitutions of each step by those skilled in the art without departing from the concept of this application should be included in the scope of protection of this application.

[0024] Please see Figure 1 , Figure 1 This is a schematic diagram of a knowledge graph-based method for managing heterogeneous oil and gas exploration data lineage fusion according to an embodiment of the present invention. Figure 1As shown, the method includes steps S1-S6: In step S1, a knowledge graph is constructed, which includes data object nodes, processing activity nodes, professional object nodes and governance status nodes. Data object nodes are associated with each other through processing activity nodes and connected by edges carrying semantic constraints.

[0025] In some embodiments, a knowledge graph for the governance of heterogeneous data in oil and gas exploration is first constructed. For example, this knowledge graph uses four types of nodes and various edges carrying semantic constraints as its basic structure, decomposing the derivation relationships between data objects into an interpretable structure with professional constraint information.

[0026] Specifically, the four types of nodes are data object nodes, processing activity nodes, professional object nodes, and governance status nodes. Data object nodes represent the input data or output results generated at each processing stage of oil and gas exploration. These include original seismic volumes, preprocessed volumes, pre-stack migration results, post-stack attribute volumes, well curves, well-seismic calibration results, stratigraphic interpretation surfaces, fault interpretation results, reservoir prediction grids, trap boundaries, and reserve parameter tables. Each data object node carries attributes such as version identifier, creation time, assigned work area, applicable stratigraphic range, and interpretation domain.

[0027] Processing activity nodes represent specific processing operations applied to data objects, including denoising, static correction, resampling, velocity modeling, migration imaging, attribute extraction, inversion, stratigraphic interpretation, fault interpretation, well-seismic calibration, depth conversion, trap delineation, and reserve recalculation. In some embodiments, after introducing processing activity nodes, the direct derivation relationship between data objects is decomposed into a ternary structure where upstream data object nodes generate downstream data object nodes through processing activity nodes. Processing activity nodes can carry information such as activity type, processing parameters, scope, and spatial range of action.

[0028] Specialized object nodes represent constrained entities with stable specialized meanings in the oil and gas exploration field, including work areas, blocks, wells, stratigraphic levels, sub-stratigraphic levels, fault blocks, structural units, interpretation domains, depth benchmarks, velocity models, time-depth relationship models, coordinate reference systems, etc. Specialized object nodes provide constraint context for data object nodes and processing activity nodes.

[0029] The governance status node records intermediate state information and judgment results generated during the lineage governance process, including semantic identifier integrity status, semantic continuity status, mismatch type marker, impact propagation status, propagation boundary status, recalculated and confirmed status, and repair candidate status. The governance status node persists the judgment results as part of the graph, which can be directly referenced in subsequent audits and incremental governance.

[0030] In some embodiments, for example, the following types of edge relationships are established between nodes: derived_from indicates that a data object is generated by an upstream object through a processing activity; generated_by indicates that a data object is generated by a processing activity; uses_input indicates that a processing activity uses a data object as input; constrained_by indicates that a data object or processing activity is constrained by a professional object; interprets indicates that an interpreting result points to a professional object; belongs_to_area indicates that an object belongs to a work area. Or spatial units; Horizontal applicability (applies_to_horizon) indicates that an object is applicable to a certain horizon or segment; Model usage (uses_model) indicates that a processing activity uses a velocity model or time-depth relation model; Version inheritance (version_succeeds) indicates that an object is a successor version of another object; Scope (affects_domain) indicates that a processing activity acts on the time domain or depth domain; Spatial applicability (valid_in_region) indicates that an object is only valid within a specific spatial region; Recomputed confirmation (recomputed_after) indicates that an object has been recomputed for a certain upstream change. Through the above relation design, each lineage edge in the knowledge graph carries the semantic constraints upon which its validity depends, including information such as the type of processing activity, constrained entities, spatial extent, and horizon extent.

[0031] In some embodiments, the data sources for constructing the knowledge graph include structured metadata in a data management system, task records and version records in the processing flow, and documents in exploration projects. Structured metadata can directly extract node attributes and relationships from database fields and version record tables; semi-structured data, such as file header information, processing parameter tables, and log files, have their corresponding fields extracted using preset parsing rules.

[0032] In step S2, semantic identifiers across processing stages are defined in the knowledge graph and associated with data object nodes and processing activity nodes.

[0033] In some embodiments, semantic identifiers refer to constraint fields or constraint entities that maintain relatively stable professional meaning across different processing stages, different data models, and different versions. Traditional lineage systems only determine whether a derivation relationship exists between data objects, while the role of semantic identifiers is to further determine whether that derivation relationship still holds true in the current professional context. When inconsistencies occur in the semantic identifiers between upstream and downstream data objects, even if the formal lineage path is complete, it is possible to identify situations where the professional basis has become invalid.

[0034] Specifically, please refer to Figure 2 , Figure 2 A schematic diagram of the semantic identifier extraction and association process provided in this embodiment is shown. In S201, according to the degree of impact of the conflict on the validity of bloodline, the semantic identifiers are divided into three levels: strongly constrained semantic identifiers, moderately constrained semantic identifiers, and weakly constrained semantic identifiers.

[0035] In cases where strongly constrained semantic identifiers conflict and there is no explicit transformation relationship recorded by the processing activity nodes, the semantic continuity state can usually be directly determined as semantic mismatch. Strongly constrained semantic identifiers include work area number, well number, stratigraphic standard code, interpretation domain identifier (i.e., time domain or depth domain), coordinate reference system identifier, velocity model version, time-depth relationship version, well-seismic calibration version, and key processing batch identifier. Changes in these identifier values ​​signify a fundamental change in the professional foundation upon which the data object is based. For example, different velocity model versions will directly affect the time-depth transformation relationship, thus causing the depth domain results to lose their reliable basis.

[0036] When a conflict occurs in the mid-constraint semantic identifier, it is necessary to determine whether the conflict constitutes a semantic mismatch by considering the activity type of the processing node. The mid-constraint semantic identifier includes the sampling interval, frequency band range, raster resolution, time window definition, local spatial range, interpretation round identifier, and attribute calculation parameter set. Some processing activities may reasonably change these parameter values; for example, a resampling activity changing the sampling interval is a normal processing behavior, and in this case, the change in the mid-constraint semantic identifier does not constitute a mismatch.

[0037] Weak constraint semantic identifiers include file naming rules, project directory paths, historical aliases in documents, project abbreviations, etc., which are used for candidate sorting and auxiliary verification, but are not used as the sole basis for mismatch judgment.

[0038] In S202, semantic identifiers are extracted from multi-source data. In some embodiments, semantic identifier extraction involves three types of data sources. For structured metadata, semantic identifier values ​​are read directly from database fields, task log records, and version record tables. For semi-structured data, semantic identifiers are extracted from file headers, processing parameter tables, export templates, and log files according to preset parsing rules. For unstructured text, such as explanation reports, job descriptions, and result catalog descriptions, a sequence labeling model is used to assist in the extraction of semantic identifier entities.

[0039] In some embodiments, optionally, the sequence labeling model described above uses a pre-trained language model MacBERT as the encoder. The MacBERT encoder receives the text sequence as input and generates a context representation vector for each position in the sequence, forming a vector sequence of the same length as the input. A Conditional Random Field (CRF) layer is connected above the output layer of the MacBERT encoder. The CRF layer takes the vector sequence as input, models the dependency relationship between adjacent labels using the transition probability matrix between labels, and decodes and outputs the globally optimal label sequence using the Viterbi algorithm. The label set is defined according to the oil and gas exploration field and includes entity categories such as layer name, version number, well number, work area number, and processing batch identifier. The BIO labeling system is adopted, where B represents the starting position of the entity, I represents the internal position of the entity, and O represents the non-entity position. During training, manually labeled oil and gas exploration text samples are used as the training set, and a negative log-likelihood loss function is used for optimization, with the goal of maximizing the conditional probability of the correct label sequence.

[0040] In this embodiment, sequence labeling models can be used as an auxiliary means for semantic identifier extraction. The processing results of the above models are all used as the generation results of candidate semantic identifiers or candidate relations. The candidate semantic identifiers output by the models need to be verified by rules or confirmed by humans before they are associated with data object nodes and processing activity nodes in the knowledge graph. The rule verification includes verification of the category of candidate semantic identifiers, verification of value format, verification of applicable scope of work area or layer, and verification of consistency with existing professional object constraints of target nodes.

[0041] Optionally, for the terms extracted by the model, terminology standardization matching can be further used to map them to standard professional object nodes in the knowledge graph. In one implementation, a RoBERTa-wwm-ext pre-trained language model is used to encode candidate terms and entries in the standard terminology library into semantic vectors, respectively. The cosine similarity between the two is calculated, and the candidate terms are mapped to the standard terms with the highest similarity exceeding a preset threshold. This threshold is determined based on the validation results on the labeled dataset.

[0042] Optionally, for partially missing relations in the knowledge graph, a Relational Graph Convolutional Network (R-GCN) can be used to assist in inference. R-GCN maintains a set of convolutional parameters for each relation type. Nodes update their vector representations at each layer by aggregating the representations of neighboring nodes under each relation. Finally, a scoring function predicts whether a specific relation exists between two nodes. During training, existing relation triples in the graph are used as positive samples, and negative samples are constructed by randomly replacing the head or tail entity. The cross-entropy loss function is used for optimization. The prediction results of R-GCN also need to be validated according to rules before being written into the graph, and the written relations only serve as candidate completion relations in subsequent semantic continuity determination.

[0043] In step S203, the extracted semantic identifiers are associated with the corresponding data object nodes and processing activity nodes in the knowledge graph. In some embodiments, for each data object node and processing activity node, its corresponding strong constraint semantic identifier, medium constraint semantic identifier, and weak constraint semantic identifier are written into the knowledge graph as node attributes, or associated with the corresponding professional object node through constraint relationship edges. After the association is completed, the conditions for the establishment of each lineage edge in the knowledge graph can be verified by comparing the semantic identifier sets of upstream and downstream nodes. When a semantic identifier is associated with a professional object node through a constraint relationship edge, the data object node or processing activity node carrying the semantic identifier refers to the node associated with the corresponding professional object node through the constraint relationship edge.

[0044] In step S3, the lineage path in the knowledge graph is divided into propagation units. Each propagation unit consists of an upstream data object node, a processing activity node, and a downstream data object node. The semantic identifier sets of the upstream and downstream nodes of each propagation unit are compared to determine the semantic continuity status of each propagation unit.

[0045] In some embodiments, please refer to Figure 3 , Figure 3 A schematic diagram of the semantic continuity state determination process provided in an embodiment of this disclosure is shown. For example... Figure 3 As shown, in S301, the bloodline path is divided into propagation units.

[0046] In some embodiments, a complete lineage path in a knowledge graph from an upstream data object node to a downstream data object node is connected by several processing activity nodes. This path is divided into multiple propagation units, each consisting of an upstream data object node, a processing activity node, and a downstream data object node. The propagation unit is the basic unit for determining semantic continuity.

[0047] In S302, a semantic continuity determination is performed for each propagation unit. In some embodiments, for each propagation unit, its respective semantic identifier set is read from its upstream data object node, processing activity node, and downstream data object node, and the semantic continuity state of the propagation unit is determined according to the following determination logic.

[0048] Specifically, the determination process begins by checking the strong constraint semantic identifiers: The strong constraint semantic identifier values ​​on the upstream and downstream data object nodes are read and compared item by item. If all strong constraint semantic identifier values ​​are consistent, the propagation unit passes at the strong constraint level. If a particular strong constraint semantic identifier value differs, it is further checked whether an explicit conversion relationship for that identifier is recorded on the processing activity node. An explicit conversion relationship here refers to a record explicitly registered on the processing activity node showing a valid conversion of an identifier from one value to another. For example, a depth conversion activity converts the interpretation domain from the time domain to the depth domain; this conversion relationship is explicitly registered on the activity node. If an explicit conversion relationship exists, the propagation unit is considered to have passed at the strong constraint level; if not, the propagation unit is marked as a semantic mismatch unit, and the determination ends. The explicit conversion relationship should correspond to the conflicting strong constraint semantic identifier, and the identifier value before conversion, the identifier value after conversion, and the corresponding processing activity type should be recorded.

[0049] After passing the strong constraint level, the intermediate constraint semantic identifier is checked. If the intermediate constraint semantic identifier values ​​of the upstream and downstream nodes remain unchanged, the propagation unit is marked as strongly continuous. If the intermediate constraint semantic identifier changes, it is determined whether the change is interpretable based on the activity type of the processing activity node. An interpretable change refers to a change in the corresponding parameter value that would reasonably alter during normal execution of that type of processing activity.

[0050] For example, resampling activities can explain changes in the sampling interval; depth transformation activities can explain changes in the interpretable domain from the time domain to the depth domain; attribute extraction activities can explain changes in the attribute calculation parameter set, but cannot explain changes in the hierarchical standard encoding; when a depth transformation activity causes a change in the interpretable domain from the time domain to the depth domain, it is judged according to the explicit transformation relationship of the aforementioned strongly constrained semantic identifiers, and is not treated as a change in the moderately constrained semantic identifiers. Propagation units that are judged to be interpretable changes are marked as weakly continuous, and those that are not interpretable changes are marked as semantically mismatched units.

[0051] If a key semantic identifier is missing in a propagation unit, preventing the aforementioned comparison from being performed, the propagation unit is marked as a continuity uncertainty unit. The handling method for continuity uncertainty units is as follows: Search the knowledge graph for other reachable paths between the upstream and downstream data object nodes of the propagation unit. If such paths exist and the propagation units on the alternative paths have been determined to be strongly or weakly continuous, then the missing semantic identifier value is obtained from the corresponding node of the alternative path and a new determination is made. If no available alternative path exists, it is marked as awaiting manual confirmation, and when the upstream data object identifier or processing activity identifier cannot be associated with a graph node, it is recorded as an upstream basis missing mismatch. The category and location of the missing identifier are output in the governance results.

[0052] In some embodiments, the method further includes, in S303, subdividing the propagation units marked as semantic mismatch units according to the semantic identifier conflict type that caused the mismatch.

[0053] Specifically, semantic identifier conflict types include: version reversal mismatch, where the downstream data object node uses an upstream version that is earlier than the currently valid version; interpretation domain mismatch, where the interpretation domains of upstream and downstream data object nodes are inconsistent and lack valid conversion records; hierarchical standard conflict mismatch, where upstream and downstream data object nodes reference different hierarchical standard codes; coordinate reference mismatch, where coordinate reference system identifiers are inconsistent; well-seismic calibration lag mismatch, where the well-seismic calibration version relied upon by the downstream data object node is outdated; velocity model replacement not propagated mismatch, where the upstream velocity model has been updated but the downstream data object node is still associated with the old version; local revision out-of-bounds propagation mismatch, where revisions within a local scope are incorrectly propagated to objects outside the scope; and upstream basis missing mismatch, where the upstream data object identifier or processing activity identifier recorded in the lineage record fails to be associated with the corresponding node in the knowledge graph. These mismatch types serve as inputs for subsequently determining key lineage propagation boundaries.

[0054] For example, suppose that in the lineage path of a certain trap boundary result, one of the propagation units is a stratigraphic surface result generated from the original stratigraphic interpretation surface through stratigraphic interpretation activities. When judging this propagation unit, if it is found that the well-seismic calibration version used in the stratigraphic interpretation activities is still an earlier version, and the well-seismic calibration version on which the stratigraphic surface result depends is still associated with that earlier version, while the currently valid well-seismic calibration version has been updated, and no version conversion relationship is registered on the stratigraphic interpretation activity node, then this propagation unit is marked as a semantically mismatched unit, and the mismatch type is well-seismic calibration lag mismatch.

[0055] In step S4, based on the semantic continuity state, change event attributes, and activity sensitivity, key kinship propagation boundaries are determined in the kinship network.

[0056] In some embodiments, the critical lineage propagation boundary refers to the set of nodes and edges in a lineage network that, after a semantic mismatch, upstream update, or local revision event occurs, can truncate, allow, or prune the propagation of influence. Simply identifying the location of the semantic mismatch is insufficient to control the scope of influence; the mismatch information must also be combined with the nature of the change event and the sensitivity of processing activities to that change to determine where the propagation of influence should be blocked, where it should be allowed to pass, and where it should be pruned to a local extent. The critical lineage propagation boundary is jointly determined by the semantic continuity state, the attributes of the change event, and the sensitivity of the activity.

[0057] Specifically, a propagation boundary is formed in the kinship network under the following conditions. The first condition is: if the semantic continuity state of a propagation unit is semantically mismatched, a propagation boundary is formed at the downstream entry point of that propagation unit to prevent the influence from continuing to propagate along that path.

[0058] The second condition: If the change event involves an actual update of the strong constraint semantic identifier, such as a velocity model version replacement, a well-seismic calibration version update, or a stratigraphic standard encoding revision, a candidate propagation boundary is formed downstream of each processing activity node that takes the data object node carrying the actual updated strong constraint semantic identifier as input. Whether the candidate propagation boundary ultimately takes effect depends on the assessment of the sensitivity of subsequent activities.

[0059] The third condition works in conjunction with the second: For the processing activity node where the candidate propagation boundary is located, its sensitivity to the current change event is determined based on the activity's sensitivity. If sensitive, the candidate propagation boundary is ineffective, and the impact continues to propagate downstream; if insensitive, the candidate propagation boundary becomes the termination boundary, and the impact ends there.

[0060] Activity sensitivity is determined using an activity sensitivity matrix. For example, the activity sensitivity matrix is ​​a two-dimensional rule table with change type as rows and processing activity type as columns, where each element is either sensitive or insensitive. This matrix is ​​pre-configured by domain experts based on their expertise in oil and gas exploration processing workflows. For instance, velocity model version changes are marked as sensitive to depth conversion activities and depth domain reserve calculation activities, but insensitive to activities involving browsing pure time-domain attributes; well-seismic calibration version updates are marked as sensitive to stratigraphic interpretation activities and well-seismic alignment results, but insensitive to file format conversion activities; stratigraphic alias corrections are marked as sensitive to terminology alignment operations, but insensitive to geometric calculation activities. In actual implementation, entries in the matrix can be added, deleted, or adjusted according to project requirements.

[0061] The fourth condition: If the impact of a change event is limited to a local spatial region or a local layer, and there is no spatial or layeral intersection between the applicable range of the downstream data object node and the impact range of the change event, a clipping propagation boundary is formed at the entrance of the downstream data object node. The intersection is determined by comparing the spatial coordinate range or layeral interval carried by the change event with the spatial or layeral applicable range recorded in the attributes of the downstream data object node. If there is no overlap, it is considered no intersection.

[0062] The fifth condition: If a node already has a recalculation confirmation record or repair confirmation record for the current source of change in the knowledge graph, and a recovery propagation boundary is formed downstream of that node, it means that the impact has been absorbed by that node.

[0063] In actual implementation, the above five conditions are checked sequentially from upstream to downstream for nodes on the same propagation path. Except for candidate propagation boundaries, which need to be determined to be effective after activity sensitivity judgment, the corresponding type of effective propagation boundary is formed when the other conditions are met. The candidate propagation boundary becomes the termination propagation boundary when the processing activity node is not sensitive to the current change event, and does not serve as the boundary to cut off propagation when it is sensitive. Once any condition is met, the corresponding type of propagation boundary is formed.

[0064] In step S5, influence propagation is performed along a semantically continuous path that is not truncated by the key lineage propagation boundary to obtain the set of affected nodes.

[0065] In some embodiments, to achieve differentiated propagation control, the lineage edges in the knowledge graph are first classified according to their semantic properties. Content equivalence edges correspond to operations that do not change the substantive content of the data, such as file format conversion, naming revision, and storage migration; parameter-sensitive edges correspond to processing operations that depend on specific parameters, such as attribute extraction, velocity modeling, and inversion; interpretation-dependent edges correspond to the dependence of the closed boundary on the interpretation results of the stratigraphic plane and fault; local mapping edges correspond to operations that only act on local spatial regions or local stratigraphic segments; and domain transformation edges correspond to transformation operations from the time domain to the depth domain.

[0066] Specifically, please refer to Figure 4 , Figure 4 A schematic diagram illustrating the process for obtaining the set of affected nodes provided in an embodiment of this disclosure is shown. Figure 4 As shown, in S401, starting from the upstream node where the change occurred, each propagation unit is traversed layer by layer along the downstream direction.

[0067] In one implementation, the traversal process employs a breadth-first strategy. Starting with a changed node, it is added to a processing queue. Each time, a node is retrieved from the queue, and all downstream propagation units that node participates in are obtained. Each propagation unit is processed according to the following rules: If the propagation unit is marked as a semantic mismatch unit, propagation stops, and that position is recorded as the propagation boundary position; downstream nodes are not added to the queue. If the corresponding lineage edge is a content-equivalent edge, the downstream node is marked as having its form affected but not its impact level amplified; the downstream node is added to the queue, and traversal continues. If it is a parameter-sensitive edge, the activity sensitivity matrix is ​​queried to determine whether the corresponding processing activity node is sensitive to the current change event. If sensitive, the downstream node is added to the queue; otherwise, propagation in that direction stops. If it is a locally mapped edge, the spatial impact range of the change event and the applicable range of the downstream data object node are calculated for intersection. If the intersection is empty, propagation stops; if the intersection is not empty, the impact range is trimmed to the intersection portion, and traversal continues. If it is a domain transformation edge, check if there is an explicit transformation relationship or a valid domain transformation semantic identifier record corresponding to the interpreted domain identifier on the processing active node. If it exists, continue propagation; otherwise, stop. If the reached node has a recalculated confirmation status, stop further propagation.

[0068] In S402, after the traversal is complete, all reached downstream nodes are aggregated into an affected node set. Each node in this set is labeled with the path to which it was affected and the type of impact. Nodes that are only reached via content-equivalent edges and are not further amplified by parameter-sensitive edges, interpretation-dependent edges, local mapping edges, or domain transformation edges are labeled as formally affected nodes and do not directly enter the minimum necessary recalculation range.

[0069] In step S6, based on the key lineage propagation boundary and the set of affected nodes, semantic mismatch and data content update are determined. The minimum necessary repair range for semantic mismatch and the minimum necessary recalculation range for data content update are determined.

[0070] In some embodiments, after identifying the set of affected nodes, a uniform recalculation is not performed on all affected nodes. Instead, governance actions are divided into two categories based on the cause of the impact: repair and recalculation, and the minimum necessary scope for each is determined.

[0071] Specifically, please refer to Figure 5 , Figure 5 A schematic diagram illustrating the minimum necessary repair range and minimum necessary recalculation range provided in the embodiments of this disclosure is shown. Figure 5 As shown in S501, the minimum necessary repair range is determined for semantic mismatch.

[0072] In some embodiments, when the mismatch in the lineage path is caused by incomplete or missing semantic identifiers, but the actual content of the upstream data object has not changed, a repair operation is performed. For example, applicable scenarios for the repair operation include: missing layer alias mapping relationships in the atlas, unregistered well-seismic calibration version relationships, unwritten interpretation domain conversion relationships in the atlas, and incomplete link semantics due to missing upstream processing activity nodes.

[0073] In some embodiments, the repair operation is performed as follows: In the graph subgraphs located upstream and downstream of the critical lineage propagation boundary, candidate alternative links are retrieved that satisfy strong constraint semantic identity consistency or have an explicit transformation relationship, and medium constraint semantic identity consistency or have an interpretable change. The candidate alternative links start from the valid data object node upstream of the propagation boundary, pass through other existing processing activity nodes and intermediate data object nodes, and reach the data object node downstream of the propagation boundary, with each propagation unit along the way passing semantic continuity determination.

[0074] In some embodiments, when multiple candidate alternative links are retrieved, the optimal link is selected by weighting and sorting according to the following four indicators: the number of strongly constrained semantic identifier matches and the completeness of explicit transformation relationships, the completeness of the activity sequence, the rationality of the temporal order, and the coverage of existing intermediate results. The completeness of the activity sequence is the ratio of the actual number of processing activity nodes in the link to the number of processing activity nodes required for that type of link; the rationality of the temporal order is assessed based on whether the timestamps of each node in the link satisfy a monotonically increasing relationship; and the coverage of existing intermediate results is the ratio of the number of existing intermediate data object nodes in the link to the total number of intermediate nodes required for the link.

[0075] In some embodiments, the four indicators are assigned weights and then summed to obtain a comprehensive score, with each weight being a non-negative number and the sum of the weights being 1. Domain experts pre-determine the relative importance of each indicator based on governance priorities and normalize them to their corresponding weights. When governance priorities are not configured, the weights of all four indicators are set to 0.25. Weight configuration and adjustment records are written to the audit log. The candidate alternative link with the highest comprehensive score is selected as the repair link, and the repair link is used as the alternative link or supplementary link of the original mismatched path, so that the lineage path after governance is restored to a semantically continuous path.

[0076] In S502, the minimum necessary recalculation range is determined for data content updates. In some embodiments, when the cause of the impact is an actual update of the data content, such as a change in velocity model parameters, a revision of fault geometry, or a recalibration of well-seismic calibration results, the corresponding downstream data object node needs to be recalculated.

[0077] For example, the process of determining the minimum necessary recalculation range is as follows: starting from the terminal result node in the affected node set, backtrack along the lineage path to the nearest stable propagation boundary. The stable propagation boundary includes the recovery propagation boundary or the trimming propagation boundary. If no recovery propagation boundary or trimming propagation boundary is encountered during the backtracking process, the starting change node of this impact propagation is taken as the backtracking endpoint. The set of nodes covered by the path from the stable propagation boundary or the starting change node to the terminal result node is the minimum necessary recalculation subgraph. This recalculation subgraph satisfies the following constraints: it covers all truly affected terminal result nodes; it does not contain nodes with existing recalculation confirmation records; it does not cross spatial regions or hierarchical ranges that do not intersect with the scope of the change event's impact; and it does not contain nodes marked as affected only because of content equivalence edges.

[0078] In S503, when semantic mismatch and data content update exist simultaneously in the same lineage network, repair and recalculation are performed collaboratively.

[0079] In some embodiments, exemplarily, the collaborative strategy specifically involves first performing a repair operation on semantic mismatches to restore effective propagation paths; then, redetermining the minimum necessary recalculation range on the repaired effective propagation paths. After the repair is completed, if some nodes in the affected node set have met the semantic identity consistency requirements through alternative paths, and these nodes have not been affected by data content updates, or were only included in the affected node set due to semantic mismatches, then these nodes are excluded from the minimum necessary recalculation range.

[0080] For example, suppose the velocity model of a certain work area is updated, and the alias mapping relationship of a certain layer within that work area is missing in the graph. During remediation, the semantic mismatch caused by the missing alias mapping is first repaired. After the correct alias mapping relationship is completed in the graph, the semantic continuity of the relevant lineage paths is restored. Subsequently, an impact propagation analysis of the velocity model version update is performed on the restored effective paths. Since the velocity model version change is marked as sensitive to depth transformation activities and insensitive to pure time-domain attribute browsing activities in the activity sensitivity matrix, the depth transformation results and depth-domain reserve calculation results are entered into the minimum necessary recalculation subgraph. Because the corresponding processing activities are insensitive to velocity model version changes, downstream result nodes generated by pure time-domain attribute browsing activities are excluded. If a certain depth transformation result has been recalculated by other processing flows after the velocity model update and has a recalculation confirmation record, that node is also excluded.

[0081] In some embodiments, after the repair scope and recalculation scope are determined, a complete governance result is output, including a semantic continuity state graph, a set of key lineage propagation boundaries and the type and triggering conditions of each propagation boundary, a set of affected nodes after propagation pruning, the minimum necessary repair scope and the corresponding repair link order, the indicator weights used for repair link ordering, the minimum necessary recalculation scope and a suggested recalculation execution order, and audit records and recovery confirmation status. For mismatch paths that are repaired through alternative or supplementary links, the governance result records the original mismatch path, the repair link, and the alternative or supplementary relationship between them. When the semantic identifier or relationship originates from the model candidate result, the audit record also includes the candidate result source, rule validation result, or manual confirmation result.

[0082] The recalculation execution order is determined based on the dependencies in the lineage path: among nodes with dependencies, the upstream node that is depended upon performs recalculation before the downstream node that depends on it; nodes without mutual dependencies can be recalculated in parallel. In some embodiments, the audit log includes the semantic identifier value used for each semantic continuity determination, the determination result, the applicable rule identifier, and the triggering conditions and propagation boundary type for each propagation boundary formation, facilitating manual review and post-event traceability.

[0083] This invention addresses the challenges of multiple rounds of processing and parallel existence of multiple versions of evidence in oil and gas exploration. By combining semantic identification constraints across processing stages with lineage propagation boundary recognition based on knowledge graphs, it identifies parts of the formal lineage path where professional semantics have become invalid. On the other hand, it controls the actual propagation range of upstream changes to prevent the impact from spreading unconditionally along all formal lineage edges. Simultaneously, it categorizes governance actions into semantic repair and content recalculation. Through a collaborative strategy of repairing first and then recalculating, it compresses the governance scope, reducing unnecessary over-recalculation while preventing the continued propagation of erroneous evidence.

[0084] Please see Figure 6 , Figure 6 This is a schematic diagram of a knowledge graph-based heterogeneous data lineage fusion and management system for oil and gas exploration, provided in an embodiment of this application. For example... Figure 6 As shown, the system includes: The knowledge graph construction module 601 is used to construct a knowledge graph, which includes data object nodes, processing activity nodes, professional object nodes and governance status nodes. The data object nodes are associated with each other through the processing activity nodes and connected by edges carrying semantic constraints. The semantic identifier definition module 602 is used to define cross-processing stage semantic identifiers in the knowledge graph and associate them with data object nodes and processing activity nodes. The semantic continuity determination module 603 is used to split the lineage path in the knowledge graph into propagation units. Each propagation unit consists of an upstream data object node, a processing activity node, and a downstream data object node. The semantic continuity status of each propagation unit is determined by comparing the semantic identifier sets of the upstream and downstream nodes of each propagation unit. The propagation boundary determination module 604 is used to determine the key kinship propagation boundary in the kinship network based on the semantic continuity state, change event attributes, and activity sensitivity. The influence propagation execution module 605 is used to perform influence propagation along a semantically continuous path that is not truncated by the key lineage propagation boundary, and obtain a set of affected nodes; The governance scope determination module 606 is used to determine semantic mismatch and data content update based on the key bloodline propagation boundary and the set of affected nodes, determine the minimum necessary repair scope for semantic mismatch, and determine the minimum necessary recalculation scope for data content update.

[0085] Each processing unit and / or module in the embodiments of this application can be implemented by an analog circuit that implements the functions described in the embodiments of this application, or by software that executes the functions described in the embodiments of this application.

[0086] Based on the same inventive concept, this application also provides an electronic device, the method corresponding to which can be the method in the foregoing embodiments, and its problem-solving principle is similar to that method. For example... Figure 7 As shown, Figure 7 This is a schematic diagram of an electronic device structure provided in an embodiment of the present disclosure. The device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the methods and / or technical solutions of the foregoing embodiments of the present application.

[0087] In particular, the methods and / or embodiments in this application can be implemented as computer software programs. For example, the embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowchart. When the computer program is executed by a processor, it performs the functions defined in the methods of this application.

[0088] Another embodiment of this application provides a storage medium storing computer program instructions thereon, which can be executed by a processor to implement the methods and / or technical solutions of any one or more embodiments of this application.

[0089] In the above embodiments, the descriptions of each embodiment have different focuses. Parts not described in detail in a particular embodiment can be found in the relevant descriptions of other embodiments. The above descriptions are merely preferred embodiments of this application and explanations of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to the technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalent features without departing from the inventive concept.

Claims

1. A knowledge graph-based method for managing the lineage fusion of heterogeneous oil and gas exploration data, characterized in that, include: A knowledge graph is constructed, which includes data object nodes, processing activity nodes, professional object nodes, and governance status nodes. Data object nodes are associated with each other through processing activity nodes and connected by edges carrying semantic constraints. In the knowledge graph, semantic identifiers across processing stages are defined and associated with data object nodes and processing activity nodes. The lineage paths in the knowledge graph are divided into propagation units, each consisting of an upstream data object node, a processing activity node, and a downstream data object node. The semantic continuity state of each propagation unit is determined by comparing the semantic identifier sets of its upstream and downstream nodes. Based on the semantic continuity state, change event attributes, and activity sensitivity, key lineage propagation boundaries are determined in the lineage network. Influence propagation is performed along semantically continuous paths that are not truncated by the key lineage propagation boundaries to obtain a set of affected nodes. Based on the key lineage propagation boundaries and the set of affected nodes, semantic mismatches and data content updates are determined. The minimum necessary repair range for semantic mismatches and the minimum necessary recalculation range for data content updates are determined.

2. The method according to claim 1, characterized in that, Semantic identifiers are classified into strong-constraint semantic identifiers, medium-constraint semantic identifiers, and weak-constraint semantic identifiers according to the degree of impact of conflicts on lineage validity. Among them, conflicts of strong-constraint semantic identifiers directly determine the semantic continuity state as semantic mismatch when there is no explicit transformation relationship recorded by the processing activity node. Conflicts of medium-constraint semantic identifiers are combined with the activity type of the processing activity node to determine whether they constitute semantic mismatch. Weak-constraint semantic identifiers are used for candidate ranking and auxiliary verification.

3. The method according to claim 2, characterized in that, Determining the semantic continuity status of each propagation unit includes: for each propagation unit, checking whether the values ​​of the strong constraint semantic identifiers of the upstream data object node and the downstream data object node are consistent or whether there is an explicit transformation relationship recorded by the processing activity node; if the strong constraint semantic identifiers are consistent or there is an explicit transformation relationship, the propagation unit passes at the strong constraint level; if the strong constraint semantic identifiers conflict and there is no explicit transformation relationship, the propagation unit is marked as a semantic mismatch unit; after passing at the strong constraint level, checking whether there is a change in the intermediate constraint semantic identifier; if there is no change in the intermediate constraint semantic identifier, the propagation unit is marked as strongly continuous; if there is a change in the intermediate constraint semantic identifier, judging whether the change belongs to an interpretable change according to the activity type of the processing activity node, if it belongs to an interpretable change, it is marked as weakly continuous, and if it does not belong to an interpretable change, it is marked as a semantic mismatch unit; if the semantic identifier in the propagation unit is missing and cannot be judged, it is marked as a continuity uncertainty unit and enters the supplementary verification or candidate repair process.

4. The method according to claim 3, characterized in that, For propagation units marked as semantic mismatch units, the mismatch is further subdivided into version reversal mismatch, interpretation domain mismatch, hierarchical standard conflict mismatch, coordinate reference mismatch, well-seismic calibration lag mismatch, velocity model replacement not propagated mismatch, local revision out-of-bounds propagation mismatch, or upstream basis missing mismatch, based on the semantic identifier conflict type that caused the mismatch. The upstream basis missing mismatch refers to the failure of the upstream data object identifier or processing activity identifier recorded in the lineage record to be associated with the corresponding node in the knowledge graph. The mismatch type serves as the input for subsequently determining the key lineage propagation boundary.

5. The method according to claim 1, characterized in that, The critical lineage propagation boundary is formed by one or more propagation boundaries in the following ways: If the semantic continuity state of a propagation unit is semantic mismatch, a propagation boundary is formed at the downstream entry point of that propagation unit; if the change event involves the actual update of a strong constraint semantic identifier, a candidate propagation boundary is formed downstream of each processing activity node that takes the data object node carrying the actual update strong constraint semantic identifier as input; for the candidate propagation boundary, the corresponding processing activity node is judged to be sensitive to the current change event based on the activity sensitivity. If it is sensitive, the candidate propagation boundary is ineffective and affects the continued propagation; if it is not sensitive, the candidate propagation boundary becomes effective as a termination propagation boundary; if the impact range of the change event is limited to a local spatial region or a local segment, a pruning propagation boundary is formed when the applicable scope of the downstream data object node does not intersect with the impact range; if a node already has a recalculation confirmation record or repair confirmation record for the current change source, a recovery propagation boundary is formed downstream of that node.

6. The method according to claim 5, characterized in that, Activity sensitivity is determined by an activity sensitivity matrix, which is arranged with change type as the row and activity type as the column, and the matrix element value is either sensitive or insensitive. The activity sensitivity matrix is ​​pre-configured by domain experts according to the oil and gas exploration and processing procedures.

7. The method according to claim 1, characterized in that, The influence propagation is performed along a semantically continuous path that is not truncated by the key lineage propagation boundary, including: classifying the lineage edges in the knowledge graph according to their semantic properties into content equivalence edges, parameter-sensitive edges, interpretation-dependent edges, local mapping edges, and domain transformation edges; starting from the upstream node that has changed, traversing each propagation unit layer by layer in the downstream direction; stopping propagation and recording the propagation boundary position when encountering a semantic mismatch unit; not amplifying the influence range when encountering a content equivalence edge; determining whether to continue propagation based on the activity sensitivity when encountering a parameter-sensitive edge; trimming the influence range to the intersection with the influence range of the changed event when encountering a local mapping edge, stopping propagation if the intersection is empty; checking for the existence of a valid domain transformation semantic identifier record when encountering a domain transformation edge, stopping propagation if it does not exist; stopping propagation when encountering a node whose status has been recalculated and confirmed; after traversal, all reached downstream nodes are gathered into the affected node set, and each downstream node is labeled with the influence type of being affected in form or data content.

8. The method according to claim 1, characterized in that, Determining the minimum necessary repair scope for semantic mismatches includes: in the graph subgraphs located upstream and downstream of the key lineage propagation boundary, retrieving candidate alternative links that satisfy strong constraint semantic identifier consistency or have an explicit transformation relationship, and medium constraint semantic identifier consistency or have an interpretable change. The candidate alternative links start from the effective data object node upstream of the propagation boundary, pass through other processing activity nodes and intermediate data object nodes to reach the data object node downstream of the propagation boundary, and each propagation unit along the way passes the semantic continuity judgment; the retrieved candidate alternative links are weighted and sorted according to the number of strong constraint semantic identifier matches, activity sequence completeness, temporal order rationality, and existing intermediate result coverage, with each weight being a non-negative number and the sum of the weights being 1. The weights are determined according to the pre-configured governance priority; when no governance priority is configured, all weights take the same value, and the candidate alternative link with the highest comprehensive score is selected as the repair link; the repair link is written into the knowledge graph, and the repair link is used as the alternative link or supplementary link of the original mismatch path, and the lineage path after treatment is restored to a semantically continuous path.

9. The method according to claim 1, characterized in that, Determining the minimum necessary recalculation range for data content updates includes: starting from the terminal result node in the affected node set, tracing back along the lineage path to the nearest stable propagation boundary, where the stable propagation boundary is either a recovery propagation boundary or a pruning propagation boundary; if no recovery propagation boundary or pruning propagation boundary exists during the backtracking process, the starting change node of this impact propagation is taken as the backtracking endpoint; the set of nodes covered by the path from the stable propagation boundary to the terminal result node is determined as the minimum necessary recalculation subgraph; the minimum necessary recalculation subgraph does not contain nodes with existing recalculation confirmation records, does not cross spatial regions or layer ranges that do not intersect with the impact range of the change event, and does not contain nodes marked as affected only because they are reached by content equivalence edges.

10. A knowledge graph-based system for the fusion and management of heterogeneous oil and gas exploration data lineage, characterized in that, include: The knowledge graph construction module is used to construct a knowledge graph, which includes data object nodes, processing activity nodes, professional object nodes, and governance status nodes. Data object nodes are associated with each other through processing activity nodes and connected by edges carrying semantic constraints. The semantic identifier definition module is used to define cross-processing stage semantic identifiers in the knowledge graph and associate them with data object nodes and processing activity nodes. The semantic continuity determination module is used to divide the lineage path in the knowledge graph into propagation units. Each propagation unit consists of an upstream data object node, a processing activity node, and a downstream data object node. The semantic continuity status of each propagation unit is determined by comparing the semantic identifier sets of the upstream and downstream nodes. The propagation boundary determination module is used to determine the key lineage propagation boundaries in the lineage network based on the semantic continuity status, change event attributes, and activity sensitivity. The influence propagation execution module is used to perform influence propagation along a semantically continuous path that is not truncated by the key lineage propagation boundary, and obtain a set of affected nodes; the governance scope determination module is used to determine semantic mismatch and data content update based on the key lineage propagation boundary and the set of affected nodes, determine the minimum necessary repair scope for semantic mismatch, and determine the minimum necessary recalculation scope for data content update.

Images