Construction method and system of knowledge graph in wind power field and electronic equipment

Abstract

The invention discloses a construction method and system for a knowledge graph in the wind power field and electronic equipment, and relates to the technical field of knowledge graph construction, and the method comprises the steps: carrying out the semantic fragmentation of received wind power data, and distributing sub-field labels; performing ontology modeling, entity and relationship extraction, alignment and disambiguation on the data fragments based on the ontology models of the sub-domains to generate sub-knowledge maps; performing global entity aggregation, cross-domain alignment, relationship reconstruction, consistency verification and conflict processing on each sub-knowledge graph to obtain a global knowledge graph; and continuously receiving new data, distributing and updating the corresponding sub-knowledge graph through the sub-domain labels, and then incrementally updating the global knowledge graph. According to the invention, data modularization and domain division construction are realized, and decoupling performance and maintenance efficiency are improved; through cross-domain integration and consistency verification, global semantic unification and reliability are guaranteed, incremental updating is supported, reconstruction cost is reduced, and expandability and availability are enhanced.

Description

Technical Field

[0001] This invention relates to the field of knowledge graph construction technology, and in particular to a method, system and electronic device for constructing a knowledge graph in the field of wind power. Background Technology

[0002] Throughout the wind power industry chain, there exists a wealth of heterogeneous data from diverse sources, with varying structures and frequent updates, including wind turbine equipment information, operational status records, maintenance logs, and real-time alarm data. Building knowledge graphs using this data can provide crucial support for intelligent analysis, fault prediction, and efficient operation and maintenance in the wind power industry. However, due to the wide range of data sources, complex structures, and continuous evolution, knowledge graphs need to possess the ability to be updated frequently and dynamically. Currently, most systems adopt a static construction approach, making it difficult to achieve continuous data access and incremental updates, and unable to respond promptly to changes in new equipment, standards, and technologies, severely limiting the applicability and real-time performance of knowledge graphs in actual business applications.

[0003] Currently, the construction of knowledge graphs in the wind power field mainly relies on two typical methods: The first is the static centralized graph construction method, which extracts entities and relationships from multi-source structured or semi-structured data manually or through rule engines and integrates them into a single centralized graph. Although this method is effective in the initial stages, its rigid structure and high coupling lead to a sharp increase in subsequent update and maintenance costs, and any change in local data can trigger global consistency issues. The second is the unified modeling method driven by hybrid data sources, which attempts to integrate multiple data sources to construct a unified ontology and model structure, achieving integrated knowledge modeling. However, this method lacks an effective domain partitioning mechanism, and the mixed processing of data with different business semantics easily leads to entity conflicts and semantic confusion. Furthermore, it does not support the independent evolution and incremental updates of substructures.

[0004] Both of the aforementioned existing methods have significant drawbacks: static centralized methods struggle to support high-frequency updates and dynamic expansion, requiring complete reconstruction upon error, resulting in low efficiency and high costs; while hybrid unified modeling methods, lacking clear domain partitioning and semantic fragmentation mechanisms, lead to cross-domain semantic conflicts and model chaos, failing to achieve true domain-specific construction and collaborative evolution. Therefore, existing methods generally face problems such as insufficient map update capabilities, high maintenance costs, and difficulty in ensuring consistency when dealing with multi-source, heterogeneous, and frequently changing data in the wind power field.

[0005] To address the aforementioned issues, there is an urgent need for a wind power knowledge graph construction method that can support domain-specific modeling, incremental updates, and global consistency maintenance, in order to improve the availability, scalability, and operational efficiency of the knowledge graph. Summary of the Invention

[0006] The technical problem to be solved by this invention is to address the shortcomings of existing technologies, specifically by providing a method, system, and electronic device for constructing a knowledge graph in the field of wind power, as detailed below: 1) In a first aspect, the present invention provides a method for constructing a knowledge graph in the field of wind power, the specific technical solution of which is as follows: The received wind power data is semantically segmented to obtain multiple data fragments, and each data fragment is assigned a subdomain label belonging to its subdomain. Based on the ontology model of each subdomain in the wind power field, ontology modeling, entity and relation extraction, entity alignment and disambiguation processing are performed sequentially on the data fragments with assigned subdomain labels to generate a sub-knowledge graph for each subdomain. For each sub-knowledge graph, global entity aggregation, cross-domain entity alignment, cross-domain relationship reconstruction, consistency rule verification and conflict handling are performed sequentially to obtain a global knowledge graph in the wind power field. It continuously receives new data from the wind power sector, performs semantic segmentation and sub-domain labeling on the new data, updates the corresponding sub-knowledge graphs based on the assigned sub-domain labels, and then updates the global knowledge graph based on all the updated sub-knowledge graphs.

[0007] The beneficial effects of the knowledge graph construction method in the field of wind power provided by this invention are as follows: By semantically segmenting the data and assigning sub-domain labels, modular organization of the data is achieved, laying the foundation for domain-specific construction. Sub-knowledge graphs are independently generated based on the ontology models of each sub-domain, significantly reducing the direct impact of local updates on the global structure and effectively improving construction and maintenance efficiency. Through global entity aggregation, cross-domain entity alignment, and cross-domain relationship reconstruction, global semantic integration and structural unification are achieved while ensuring the independent evolution of each sub-domain. Consistency rule verification and conflict handling mechanisms ensure the consistency and reliability of the global knowledge graph. This method supports continuous reception of new data and incremental updates without requiring overall graph reconstruction, significantly reducing update costs and computational overhead, and comprehensively improving the availability, scalability, and operational efficiency of the knowledge graph in the wind power field.

[0008] Based on the above scheme, the method for constructing a knowledge graph in the field of wind power according to the present invention can be further improved as follows.

[0009] Furthermore, before generating the sub-knowledge graph for each sub-domain, the following steps are also included: Each data fragment is distributed to the task queue of the corresponding subdomain according to the assigned subdomain label. The scheduling weight of the task queue of each subdomain is calculated, and the task queues are concurrently scheduled according to the order of scheduling weight from largest to smallest. Based on the ontology model of each subdomain in the wind power field, the data fragments with assigned subdomain labels are sequentially subjected to ontology modeling, entity and relation extraction, entity alignment, and disambiguation processing to generate a sub-knowledge graph for each subdomain, including: Based on the ontology model of the subdomain corresponding to the task queue scheduled by concurrent tasks, ontology modeling, entity and relation extraction, entity alignment and disambiguation processing are performed on all data fragments in the task queue in sequence to generate the sub-knowledge graph of the subdomain corresponding to the task queue, until the sub-knowledge graph of each subdomain is obtained.

[0010] The beneficial effects of adopting the above-mentioned further scheme are as follows: By distributing data fragments with assigned subdomain labels to the corresponding subdomain task queues, modular organization of data processing tasks is achieved. By calculating the scheduling weight of each queue and scheduling concurrent tasks accordingly, subdomain tasks with large data volumes or high priority can be prioritized, significantly optimizing the utilization of computing resources and reducing overall processing waiting time. During the scheduling process, based on the ontology model of each subdomain, ontology modeling, entity and relation extraction, entity alignment, and disambiguation processing are performed on the dataset within the queue, ensuring the efficiency and accuracy of each sub-knowledge graph generation. This parallelized process effectively supports the rapid processing of massive heterogeneous data, providing a solid foundation for subsequent global knowledge graph integration and updates.

[0011] Furthermore, each data fragment is assigned a subdomain label belonging to its subdomain, including: Extract the keyword set, entity set, and semantic vector for each data segment. Based on the predefined keyword set, entity set, and semantic vector for each subdomain, calculate the matching score between each data segment and each subdomain, and assign the subdomain label corresponding to the highest score to each data segment.

[0012] The beneficial effects of adopting the above-mentioned further approach are as follows: by extracting the keyword set, entity set, and semantic vector of data fragments, and calculating matching scores with the features of each predefined subdomain, the domain to which the data fragment belongs can be accurately determined at the semantic level, thereby assigning it the most matching subdomain label. This quantitative allocation method based on multi-feature fusion effectively avoids the subjectivity of manual classification and the limitations of single-rule processing, ensuring the accuracy and consistency of subsequent sub-knowledge graph construction. This automated process provides data support for the rapid and accurate domain segmentation processing of massive, multi-source, and heterogeneous wind power data, laying a solid foundation for the incremental updates and efficient maintenance of the entire knowledge graph.

[0013] Furthermore, it also includes: connecting to various heterogeneous data sources in the wind power field through a streaming bus, receiving data from the wind power field, and standardizing and unifying the format of the received data to obtain processed data. Among them, the various heterogeneous data sources include: data sources for storing structured data, data sources for storing semi-structured data, and data sources for storing unstructured data. Semantic segmentation is performed on the received wind power data, including semantic segmentation of the processed data.

[0014] The beneficial effects of adopting the above-mentioned further solution are as follows: By connecting to various heterogeneous data sources in the wind power field, including structured, semi-structured, and unstructured data, a streaming bus is used to achieve real-time and unified access to multi-source data. The received data is standardized and formatted to generate consistent processed data, providing a high-quality data foundation for subsequent semantic segmentation. This mechanism significantly reduces the complexity and integration difficulty of the original data, ensures the diversity and real-time nature of data sources, and enables the entire knowledge graph construction process to efficiently and stably handle the continuous influx of massive amounts of data, providing reliable data support for the construction of sub-knowledge graphs and the incremental updates of the global graph.

[0015] 2) Secondly, the present invention also provides a knowledge graph construction system in the field of wind power, the specific technical solution of which is as follows: It includes a semantic segmentation module, a subdomain label allocation module, a sub-knowledge graph generation module, a global knowledge graph generation module, and a global knowledge graph update module; The semantic fragmentation module is used to perform semantic fragmentation on the received wind power data to obtain multiple data segments; The subdomain label assignment module is used to: assign a subdomain label to each data fragment to belong to its subdomain; The sub-knowledge graph generation module is used to: based on the ontology model of each sub-domain in the wind power field, perform ontology modeling, entity and relation extraction, entity alignment and disambiguation processing on the data fragments with assigned sub-domain labels in sequence to generate a sub-knowledge graph for each sub-domain. The global knowledge graph generation module is used to perform global entity aggregation, cross-domain entity alignment, cross-domain relationship reconstruction, consistency rule verification and conflict handling on each sub-knowledge graph in sequence to obtain a global knowledge graph in the wind power field. The global knowledge graph update module is used to: continuously receive new data in the wind power field, perform semantic segmentation and sub-domain label assignment on the new data, update the corresponding sub-knowledge graph according to the assigned sub-domain labels, and then update the global knowledge graph according to all the updated sub-knowledge graphs.

[0016] Based on the above scheme, the knowledge graph construction system in the field of wind power of the present invention can be further improved as follows.

[0017] Furthermore, it also includes a data scheduling module, which is used to: distribute each data fragment to the corresponding subdomain task queue according to the assigned subdomain label before generating the sub-knowledge graph of each subdomain, calculate the scheduling weight of the task queue of each subdomain, and perform concurrent task scheduling on the task queue according to the scheduling weight in descending order; The sub-knowledge graph generation module is specifically used to: based on the ontology model of the sub-domain corresponding to the task queue scheduled by concurrent tasks, perform ontology modeling, entity and relation extraction, entity alignment and disambiguation processing on all data fragments in the task queue in sequence, and generate the sub-knowledge graph of the sub-domain corresponding to the task queue, until the sub-knowledge graph of each sub-domain is obtained.

[0018] Furthermore, the subdomain label allocation module is specifically used to: extract the keyword set, entity set, and semantic vector of each data segment; calculate the matching score between each data segment and each subdomain based on the predefined keyword set, entity set, and semantic vector of each subdomain; and assign the subdomain label corresponding to the highest score to each data segment.

[0019] Furthermore, it also includes a data receiving and processing module, which is used to: connect to various heterogeneous data sources in the wind power field through a streaming bus, receive data from the wind power field, and perform standardization and format unification processing on the received data from the wind power field to obtain processed data. Among them, the various heterogeneous data sources include: data sources for storing structured data, data sources for storing semi-structured data, and data sources for storing unstructured data. The semantic sharding module is specifically used to perform semantic sharding on the processed data.

[0020] 3) In a third aspect, the present invention also provides an electronic device, the electronic device including a processor coupled to a memory, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor, so as to enable the electronic device to implement any of the above-mentioned methods for constructing a knowledge graph in the field of wind power.

[0021] 4) In a fourth aspect, the present invention also provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements any of the above-mentioned methods for constructing a knowledge graph in the field of wind power.

[0022] It should be noted that the beneficial effects of the technical solutions of the second to fourth aspects of the present invention and their corresponding possible implementations can be found in the above description of the technical effects of the first aspect and its corresponding possible implementations, and will not be repeated here. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments of the present invention will be briefly introduced below: Figure 1 This is a flowchart illustrating a method for constructing a knowledge graph in the field of wind power, according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating the task scheduling process. Figure 3 This is one of the structural schematic diagrams of a knowledge graph construction system in the field of wind power according to an embodiment of the present invention; Figure 4 This is a second schematic diagram of the structure of a knowledge graph construction system in the field of wind power, according to an embodiment of the present invention. Figure 5 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0024] The principles and features of the present invention are described below. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0025] The technical solution of the present invention and how the technical solution of the present invention solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of the present invention will now be described with reference to the accompanying drawings.

[0026] like Figure 1 As shown in the figure, a method for constructing a knowledge graph in the field of wind power according to an embodiment of the present invention includes the following steps: S1. Semantically segment the received wind power data to obtain multiple data fragments, and assign a subdomain label to each data fragment. The system connects to various heterogeneous data sources in the wind power sector via a streaming bus, receives data from the wind power sector, and standardizes and unifies the format of the received data to obtain processed data. These heterogeneous data sources include: structured data sources for storing structured data (such as equipment databases and monitoring data tables), semi-structured data sources for storing semi-structured data (such as JSON and XML configuration files), and unstructured data sources for storing unstructured data (such as operation and maintenance logs, fault reports, and technical documents). The streaming bus collects the raw data from the wind power sector in real time and completes the data standardization and format unification processing to form a standardized and continuous data stream, i.e., the processed data, providing stable and consistent data input for subsequent semantic segmentation and sub-domain tag allocation.

[0027] The standardization and format unification of the received wind power data is implemented as follows: ① For structured data, data records are extracted through JDBC / ODBC interfaces and converted into a unified Avro or Parquet format; for semi-structured data, tools such as Jackson or Jsoup are used for parsing to extract valid data from key-value pairs or tree structures and map them into a standardized JSON-LD format; for unstructured text data, the original text content is extracted using the Apache Tika text extraction tool and converted into a unified UTF-8 encoding to remove irrelevant characters and noise data.

[0028] ② Handle invalid values ​​(such as null values, outlier values, or remove them) and convert data types (such as converting string-format numbers to Float or Integer types) for all extracted fields. Standardize timestamps (unify them to ISO 8601 format) and map and rename field names according to the predefined wind power metadata model. For example, unify "wind_speed" and "wind speed" to "windSpeed".

[0029] ③ Add source identifiers, timestamps, data quality scores and other metadata to the data, and standardize the coding of certain fields with clear semantics (such as equipment number and fault code).

[0030] The processed data is encapsulated into a unified data packet format (such as using Apache Avro as the serialization format, which includes schema description and data body), and pushed to the downstream semantic sharding module in the form of a continuous stream through a message middleware, forming a standardized, stable, and traceable data stream.

[0031] At this point, semantic segmentation is performed on the received wind power data to obtain multiple data fragments. Specifically, this includes: semantic segmentation of the processed data to obtain multiple data fragments, specifically: The processed data will be categorized according to its data type (structured data, unstructured data, etc.). Semantic fragmentation is performed. Specifically, for structured data, each record in the structured data is treated as an independent data fragment; for semi-structured data, it is divided according to its fields or data blocks to obtain multiple data fragments; for unstructured text data, it is divided into several data fragments according to the semantic boundaries of the unstructured text data, based on sentences or paragraphs. Through the above fragmentation methods, a set of data fragments is formed. , This lays the foundation for subsequent subdomain tag allocation and domain-specific processing, among which, This represents the first data segment. This represents the first data segment. This indicates the total number of data segments.

[0032] In S1, each data fragment is assigned a subdomain label belonging to its subdomain, including: S10. Extract the keyword set, entity set, and semantic vector for each data segment. Based on the predefined keyword set, entity set, and semantic vector for each subdomain, calculate the matching score between each data segment and each subdomain, and assign the subdomain label corresponding to the highest score to each data segment. This includes the following steps: Extract the first Data fragments Keyword set Entity set and semantic vectors Predefine the keyword set, entity set, and semantic vector for each subdomain, and then... Sub-fields The set of keywords is denoted as , will the Sub-fields The set of entities is denoted as , will the Sub-fields The semantic vector is denoted as The matching score is calculated using the following formula. Data fragments With the Sub-fields Match score : in, , and All are weights, and , , and Determined based on practical experience. express: and The number of identical keywords contained within. express: and The number of identical entities contained within.

[0033] Calculate the score using the matching score calculation formula. Data fragments The highest matching score is obtained by matching each subdomain separately. The subdomain label corresponding to the highest matching score is then determined as the [number]th subdomain. Data fragments The subdomain labels are used to iterate through each data segment and assign the subdomain label corresponding to the highest score to each data segment.

[0034] The specific implementation process for extracting the keyword set, entity set, and semantic vector for each data segment is as follows: ① Each data segment undergoes preprocessing and text cleaning. For unstructured or semi-structured text segments, word segmentation, stop word removal, lemmatization, or stemming are performed, and the text is uniformly converted to lowercase to reduce noise and improve the accuracy of feature extraction. Structured data segments are directly concatenated with their field names and values ​​to form a text sequence for subsequent processing.

[0035] ② TF-IDF (Term Frequency-Inverse Document Frequency) is used to calculate the weights of candidate words, and combined with the TextRank graph ranking algorithm, highly important words are identified from the preprocessed and cleaned data fragments. Finally, the top-K words with the highest weight rankings are selected to form the keyword set for that data fragment.

[0036] ③ Using the trained BERT-BiLSTM-CRF model, entities in the preprocessed and cleaned data segments are identified, including equipment components (such as "blade" and "gearbox"), fault types (such as "bearing overheating" and "voltage sag"), and operating parameters (such as "speed" and "power"). Simultaneously, a wind power entity dictionary is introduced for supplementary matching to improve the recall rate of entity recognition. Finally, all extracted entities constitute an entity set.

[0037] ④ Input the preprocessed and cleaned data fragments into the trained Sentence-BERT model or the trained SimCSE model to obtain the [CLS] position output or a fixed-dimensional semantic vector after pooling. This semantic vector can capture the deep semantic information of the fragments and is used for subsequent similarity calculation with subdomain semantic vectors.

[0038] The wind power sub-field includes sub-fields such as wind turbine components, operation monitoring, fault and alarm, operation and maintenance, and grid connection control. The wind turbine components sub-field covers the static information and attributes of various components of the wind turbine generator set, such as the specifications, models, materials, suppliers, and lifecycle data of components like blades, gearboxes, generators, pitch systems, yaw systems, towers, and foundation rings. The operation monitoring sub-field covers real-time and historical status data generated during wind turbine operation, such as SCADA monitoring parameters like power output, speed, wind speed, wind direction, temperature, and vibration, used to characterize the real-time operational health of the equipment. The fault and alarm sub-field includes various wind turbine fault events, alarm information, and anomaly diagnostic data, including fault codes, alarm levels, trigger times, durations, diagnostic conclusions, and historical fault records. The operation and maintenance sub-field covers data related to management and maintenance activities, including periodic maintenance plans, maintenance work orders, maintenance personnel records, spare parts replacement information, maintenance costs, and post-maintenance equipment status verification results. The grid connection control subdomain covers performance evaluation indicators for wind turbine units and wind farms, such as power generation, capacity factor, power curve consistency, energy availability, and loss analysis, used to assess power generation efficiency and economic benefits. Subdomain labels for each subdomain can be set according to actual conditions. For example, the subdomain label for the wind turbine component subdomain is "Wind Turbine Component," the subdomain label for the operation monitoring subdomain is "Operation Monitoring," the subdomain label for the fault and alarm subdomain is "Fault Alarm," the subdomain label for the operation and maintenance subdomain is "Operation and Maintenance," and the subdomain label for the grid connection control subdomain is "Grid Connection Control." Alternatively, the wind power domain can be divided into subdomains and subdomain labels can be set according to actual conditions. For example, the wind power domain subdomains also include the wind turbine fault subdomain, the operation and maintenance activity subdomain, and the component information subdomain, with corresponding subdomain labels of "Wind Turbine Fault," "Operation and Maintenance Activity," and "Component Information," respectively.

[0039] S2. Based on the ontology model of each subdomain in the wind power field, perform ontology modeling, entity and relation extraction, entity alignment and disambiguation processing on the data fragments with assigned subdomain labels in sequence to generate a sub-knowledge graph for each subdomain.

[0040] With the first Sub-fields Taking all data fragments as an example, this illustrates how to call the first data fragment from the ontology model of each predefined and managed subdomain. Sub-fields The corresponding ontology model, where the ontology model of each subdomain defines the semantic structure of entity categories, attributes, and relation types for the corresponding subdomain, and then the first... Sub-fields After all data fragments are processed sequentially using the following procedure, the first data fragment is obtained. Sub-fields Sub-knowledge graph: S20, Ontology modeling, specifically, the first Sub-fields Information in the data fragment is mapped to the first Sub-fields Specifically, within the semantic framework defined by the corresponding ontology model: Based on data type, the first Sub-fields The data fragments are parsed. For structured and semi-structured data, field names and values ​​are extracted directly; for unstructured data, the extracted entity set is used. and keyword set By combining Named Entity Recognition (NER) and relation extraction models, the first... Sub-fields The data fragments are used to identify entities, attribute values, and potential relationships between them. Then, the entities identified in the data fragments are mapped to instances of a class in the ontology. For example, "gearbox" identified in the data fragment is mapped to creating a new instance of the `Gearbox` class and assigning it a unique URI. Attribute values ​​from the data are populated into the corresponding instance properties. For example, the value of the database field `rated_power` is appended to a `Generator` instance using the rule `hasRatedPower`. Relationships are established between instances based on implicit associations in the data, according to `ObjectProperty` defined in the ontology. For example, based on the text description "the gearbox is located in the generator compartment," an `isLocatedIn` relationship is established between the `Gearbox` instance and the `Nacelle` instance. After mapping, an intermediate representation conforming to the ontology's semantic specifications is output, typically a preliminary RDF graph containing instance data.

[0041] S21. Entity and Relationship Extraction: Specifically, from the semantic framework after mapping, identify and extract entities (such as devices, events, parameters) and the relationships between them to construct a preliminary set of triples. RDF graphs already contain nodes labeled as instances of ontology classes and some of their data attributes, but the object attribute relationships between instances may not be fully established, and deep relationships implied in unstructured text may not have been uncovered. For semantic frameworks mapped from structured and semi-structured data, many relationships are explicit. These relationships are identified using predefined relationship extraction rules (typically based on SPARQL queries or graph pattern matching). For example, a rule could be defined as: if the hasAssociatedComponent property of an Alert instance points to a component ID, and that ID matches the ID of a Gearbox instance, then a triple is automatically generated. Such rules directly infer the associations between entities from a known attribute value.

[0042] For information in the semantic framework that originates from text but is not yet fully structured (such as plain text descriptions in operation and maintenance logs), deep extraction using natural language processing techniques is required. The specific process is as follows: First, a trained named entity recognition model (such as one based on BERT-BiLSTM-CRF) is used for sequence labeling to identify entities in the data fragments. Then, entity linking technology is used to link the identified entities to existing entity instances with unique URIs within the semantic framework, thus resolving aliasing and referential resolution issues. For entity pairs identified in a sentence, a relation classification model is used to determine whether a predefined relation type exists between them. This can be implemented using a neural network-based classifier that takes the entity pair and its context as input and outputs the most likely relation type or "no relation".

[0043] Then, for each successfully identified entity relationship pair, a standard triple is constructed based on the object attributes defined in the ontology model. At the same time, each automatically extracted triple is assigned an extraction confidence score, which is derived from the probability output of the relationship classification model or the reliability of rule matching, providing a decision basis for subsequent disambiguation and fusion steps.

[0044] All the triples extracted in the above manner were summarized to form a preliminary set of triples for this subdomain.

[0045] S22. Entity alignment and disambiguation processing: Specifically, based on the extracted triples, and with the help of contextual semantics and attribute similarity calculation, the same entity referred to by different expressions is identified, and entity merging and disambiguation are completed to ensure that the entity references in the sub-knowledge graph are unique and the relationships are consistent. Traverse all entities in the sub-knowledge graph and generate a set of candidate entity pairs to be compared based on predefined rules (such as: entities of the same type, having the same or similar key attributes) to avoid the excessive computational complexity caused by fully connected comparisons.

[0046] The attribute values ​​of two entities are compared. Specifically, for string-type attributes, methods based on edit distance, Jaccard similarity, or the BERT model used to calculate semantic text similarity are employed. For numerical attributes, relative or absolute differences are calculated. The similarity of each attribute is obtained by weighted summation (weights can be pre-set based on attribute importance) to obtain a comprehensive attribute similarity. The adjacency structure of the two entities in the sub-knowledge graph is analyzed. Graph embedding models (such as TransE and Node2Vec) are used to map entities and their relational contexts to a low-dimensional vector space, and the similarity of their structural contexts is measured by calculating the cosine similarity between vectors. If an entity is accompanied by original text descriptions or processed semantic vectors, the cosine similarity of its semantic vectors is calculated. The above multi-dimensional similarities are weighted and fused to obtain the final comprehensive similarity score of the candidate entity pair, with a preset similarity threshold. If the final comprehensive similarity score is greater than the similarity threshold, the two entities are determined to point to the same object and should be merged.

[0047] Typically, the entity with the most complete attributes, the highest data source reliability, or the highest extraction confidence is selected as the retained entity. Attribute values ​​from redundant entities are merged into the retained entity. Conflicting attribute values ​​are handled based on the reliability of the data source, the age of the timestamp, or a predefined conflict resolution strategy (such as taking the maximum value, the latest value, or manual annotation). All incoming and outgoing edges of the merged entities are redirected to the retained entity to ensure the connectivity and consistency of the knowledge graph. After all merging operations are completed, a disambiguated set of triples with unique entity referencing is generated as the final output of the sub-domain knowledge graph.

[0048] By traversing all data fragments in each subdomain, a sub-knowledge graph for each subdomain is obtained.

[0049] S3. For each sub-knowledge graph, perform global entity aggregation, cross-domain entity alignment, cross-domain relationship reconstruction, consistency rule verification, and conflict resolution sequentially to obtain the global knowledge graph of the wind power field, specifically including: S30, Global Entity Aggregation, which means aggregating entities from each sub-knowledge graph. Specifically: A global entity index table is constructed, collecting feature information such as the name, subdomain label, and semantic vector of all entities in all sub-knowledge graphs. Multiple methods, including string similarity, vector similarity, and attribute matching, are used for fusion judgment. Entities with fusion scores higher than a set threshold are determined to point to the same object and are uniformly assigned a globally unique entity ID, achieving cross-subdomain entity alignment. Specifically: The system traverses all sub-knowledge graphs, collecting features such as the name, core attributes (e.g., device ID, model), sub-domain tags, and semantic vectors of each entity. A centralized global entity index table is then constructed, recording the original local ID and the aforementioned feature information for each entity. Subsequently, pairwise entity comparisons are performed based on this index table, employing multiple matching methods for fusion judgment: string similarity calculation based on edit distance or Jaccard coefficient to compare entity names and text attributes; semantic vector similarity calculation based on cosine similarity to measure deep semantic associations; and attribute matching based on precise matching of key attribute values ​​(e.g., serial number, specifications). Next, a weighted fusion strategy is used to combine these similarity scores into a comprehensive matching score, which is compared with a preset threshold. If the score is higher than the threshold, entities from different sub-domains are determined to point to the same real-world object. Finally, the system assigns a globally unique entity ID (e.g., UUID) to these successfully matched entity clusters and updates this ID in the index table and global graph, thereby achieving cross-sub-domain entity alignment and ensuring the uniqueness and consistency of entity referencing within the global knowledge graph.

[0050] S31. Cross-domain relationship reconstruction, that is, based on entity alignment, analyzing the semantic context and event logic of aligned entities in different subdomains to identify potential cross-domain semantic associations. The specific implementation method is as follows: By employing pattern matching, rule-based reasoning, or embedding-based relation prediction models (such as TransE) to automatically complete cross-domain connections in sub-knowledge graphs and construct bridging relationships between entities, the connectivity and semantic reasoning capabilities of the global knowledge graph are improved. Specifically: Based on the global entity index table with completed entity alignment, entity pairs belonging to different subdomains but potentially related are identified. Subsequently, a strategy combining multiple techniques is employed to predict and complete the bridging relationships between them. Specifically, ontology-based pattern matching is used, i.e., predefined cross-subdomain association rules are used, and rule reasoning engines (such as Jena Inference Engine or Drools) are used to execute these rules to generate new relationships. An embedding-based relationship prediction model is used to map entities and relationships in the global knowledge graph to a low-dimensional vector space through representation learning models (such as TransE, DistMult, or RotatE). Vector operations are used to predict the potential relationship type between any two entities, and relationship prediction results with scores higher than a threshold are adopted. Finally, all new relationship triples generated by the above automatic completion mechanism must undergo consistency verification before being incorporated into the global knowledge graph, thereby significantly enhancing the connectivity density and cross-domain semantic reasoning capabilities of the graph.

[0051] S32. Consistency rule verification, specifically: Based on predefined ontology constraint rules or graph validation languages ​​(such as OWL or SHACL), a systematic check is performed on the type attribution, attribute structure, and relation constraints of entities in the global knowledge graph. For example, it verifies whether the same wind turbine equipment number is incorrectly used by multiple entities, whether the event sequence is reasonable, and whether the entity type conforms to the ontology model definition. If conflicts or anomalies are found during the detection process, error logs are automatically generated for repair and tracking. Specifically: The system loads predefined ontology constraint rules written in the OWL ontology language or the SHACL shape constraint language. These rules formally define the class hierarchy that entity types must conform to, the data types and value ranges that attributes must satisfy, and the constraints that relations must adhere to (such as relation type, domain, value range, and cardinality). Subsequently, the system calls a graph verification engine (such as an inference engine based on Jena or the OWL API, or a dedicated SHACL verifier) ​​to perform a systematic logical check on all triples in the global knowledge graph. The checks include, but are not limited to: verifying whether the entity type correctly belongs to its OWL class, checking whether the same key attribute (such as wind turbine number) is illegally shared by multiple entities, determining the logical order of events based on temporal attributes, and auditing whether the types of entities at both ends of the relation satisfy the domain and range constraints defined in the ontology model. During the detection process, any triple or entity that violates the above rules is marked as a conflict or anomaly. The verification engine automatically generates a structured error log, which records in detail the ID of the violating element, the specific rule clause violated, and its context information. This log is used to support subsequent repair and tracking processes, thereby ensuring the logical integrity and consistency of the global knowledge graph.

[0052] S33. Conflict resolution, specifically: To address entity or relationship conflicts that arise during the fusion process, a multi-level processing strategy is implemented. Specifically, data items are prioritized based on indicators such as data source credibility, entity frequency, and extraction confidence, retaining the most representative data items. For ambiguous entities that the automatic mechanism cannot determine, multiple candidate versions are retained and labeled for subsequent manual review, ensuring that the graph fusion results are both interpretable and controllable while maintaining automation. When multiple conflicting attribute values ​​or relationships are detected for the same real-world object, the system calculates a priority score for each conflicting item based on a predefined conflict resolution strategy. This score comprehensively considers multiple indicators, including the credibility weight of its data source, the frequency of the entity or relationship in the original data, and the confidence level assigned to it in the previous extraction and alignment stages. It is then quantified and ranked using a weighted scoring model. Subsequently, the system automatically selects the data item with the highest priority score as the most representative version and retains it, while covering or discarding other conflicting items. For some conflicting entities that cannot be determined by the above automatic scoring mechanism (such as those with extremely close scores or involving high ambiguity), the system does not force a merger but retains multiple candidate versions and adds detailed conflict annotation information (including source, confidence level, and conflict points) to each version. These are then stored in a pending review queue for subsequent manual review and decision-making. This ensures both the automation efficiency of the fusion process and the interpretability and controllability of the final graph results.

[0053] Ultimately, the above process achieves the integration and unification of multi-source sub-knowledge graphs, forming a global knowledge graph in the wind power field with consistent structure and semantic fusion.

[0054] S4. Continuously receive new data from the wind power sector, perform semantic segmentation and sub-domain label assignment on the new data, update the corresponding sub-knowledge graph based on the assigned sub-domain labels, and then update the global knowledge graph based on all the updated sub-knowledge graphs. Specifically: New data from various heterogeneous data sources is continuously input via a streaming bus and standardized and formatted. Semantic sharding is performed on the processed new data, dividing it into multiple data segments based on data type. For each data segment, its keyword set, entity set, and semantic vector are extracted, and its matching score with each predefined subdomain is calculated. The subdomain label with the highest score is assigned. For each subdomain to be updated, only the newly added data segment is processed by calling its ontology model and performing ontology modeling, entity and relation extraction, entity alignment, and disambiguation processes again to generate new knowledge content for that subdomain. This new content is then fused with the atomic knowledge graph to update the subknowledge graph. After the affected subknowledge graph is updated, there is no need to rebuild the global graph. Instead, an incremental global fusion process is triggered. Specifically, focusing on the subknowledge graph that has changed, for newly added or modified entities, global entity aggregation, cross-domain entity alignment, and cross-domain relation reconstruction operations are performed again to associate and integrate them with existing entities in the global graph. Subsequently, consistency rule verification was performed on the affected parts, and conflict handling strategies were applied to ensure that the global knowledge graph maintained semantic consistency and structural integrity after the update. An efficient incremental update mechanism was implemented that processes only changed data, significantly reducing computational and time costs. This ensures that the global knowledge graph in the wind power field can reflect changes in real-world wind power data in a timely and accurate manner. The completed global knowledge graph is stored in a graph database, supporting efficient querying, graph computation, and subsequent knowledge reasoning.

[0055] Optionally, before generating the sub-knowledge graph for each subdomain, the following steps are also included: S020. Distribute each data fragment to the corresponding subdomain task queue according to the assigned subdomain label, calculate the scheduling weight of each subdomain task queue, and perform concurrent task scheduling on the task queues in descending order of scheduling weight. Specifically: Initialize a task queue for each subdomain. Distribute each data fragment to the corresponding subdomain's task queue based on the assigned subdomain label. Calculate the scheduling weight of each subdomain's task queue using the following scheduling weight formula: in, Indicates: the Sub-fields The scheduling weight of the task queue. Indicates: the Sub-fields The number of data segments to be processed in the task queue. Indicates: the Sub-fields The scheduling priority value corresponds to the task queue. The higher the scheduling priority, the larger the corresponding value. The specific value can be set according to the actual situation.

[0056] Let the maximum number of concurrent builds be denoted as The current number of active threads is recorded as ,like From all non-empty task queues, select the task queue with the highest current scheduling weight, and take a task from the head of that task queue (the selected data fragment with assigned subdomain labels and a series of standardized processing instructions to be executed; the series of standardized processing instructions refers to the instructions S20 to S22 used to execute the data fragment with assigned subdomain labels) and assign it to an idle thread for execution. At the same time, update the number of active threads. Once the sub-knowledge graph construction task corresponding to the sub-domain of the task queue is completed, release the thread resources and update the number of active threads. This updates the current length of the task queue to recalculate its scheduling weight, ensuring that subsequent scheduling decisions are based on the latest state. This indicates that the concurrency limit has been reached, and task distribution will be temporarily suspended until a thread completes its work and releases resources. Figure 2 As shown.

[0057] In S2, based on the ontology model of each subdomain in the wind power field, the data fragments with assigned subdomain labels are sequentially subjected to ontology modeling, entity and relation extraction, entity alignment, and disambiguation processing to generate a sub-knowledge graph for each subdomain, including: Based on the ontology model of the subdomain corresponding to the task queue scheduled by concurrent tasks, all data fragments in the task queue are sequentially subjected to ontology modeling, entity and relation extraction, entity alignment and disambiguation processing to generate a sub-knowledge graph of the subdomain corresponding to the task queue, until the sub-knowledge graph of each subdomain is obtained. Specifically: First, a sub-knowledge graph construction environment is initialized for each task queue (corresponding to a specific subdomain) allocated by the scheduler. The processing thread sequentially retrieves data fragments from the scheduled task queue and loads the predefined ontology model (OWL / RDFS file) for the corresponding subdomain. Then, for each data fragment, three core processes are strictly executed sequentially: ontology modeling (implemented via S20) for semantic mapping based on the ontology model; entity and relation extraction (implemented via S21) for extracting entities and relations from the mapped framework; and entity alignment and disambiguation based on attribute and context similarity (implemented via S22). This process is executed cyclically, data fragment by data fragment, until all data fragments in the current task queue are processed. During this process, extracted and disambiguated triples are continuously merged into the temporary graph storage for that subdomain. Finally, when all data fragments in a subdomain's task queue have been processed, the system generates a preliminary, internally consistent sub-knowledge graph. This process is repeated for all subdomain task queues until all subdomain sub-knowledge graphs are constructed.

[0058] Although the steps have been numbered in the above embodiments, they are only specific embodiments given by the present invention. Those skilled in the art can adjust the execution order of the steps according to the actual situation, which is also within the protection scope of the present invention. It can be understood that some embodiments may include some or all of the above embodiments.

[0059] like Figure 3 As shown, an embodiment of the present invention provides a knowledge graph construction system 200 in the field of wind power, which includes a semantic segmentation module 201, a sub-domain tag allocation module 202, a sub-knowledge graph generation module 203, a global knowledge graph generation module 204, and a global knowledge graph update module 205. The semantic fragmentation module 201 is used to: perform semantic fragmentation on the received wind power data to obtain multiple data segments; Subdomain label assignment module 202 is used to: assign a subdomain label to each data fragment to belong to its subdomain; The sub-knowledge graph generation module 203 is used to: based on the ontology model of each sub-domain in the wind power field, perform ontology modeling, entity and relation extraction, entity alignment and disambiguation processing on the data fragments with assigned sub-domain labels in sequence to generate a sub-knowledge graph for each sub-domain. The global knowledge graph generation module 204 is used to: sequentially perform global entity aggregation, cross-domain entity alignment, cross-domain relationship reconstruction, consistency rule verification and conflict handling on each sub-knowledge graph to obtain a global knowledge graph in the wind power field; The global knowledge graph update module 205 is used to: continuously receive new data in the wind power field, perform semantic segmentation and sub-domain label allocation on the new data, update the corresponding sub-knowledge graph according to the allocated sub-domain labels, and then update the global knowledge graph according to all the updated sub-knowledge graphs.

[0060] Optionally, the above scheme also includes a data scheduling module, which is used to: distribute each data fragment to the task queue of the corresponding subdomain according to the assigned subdomain label before generating the sub-knowledge graph of each subdomain, calculate the scheduling weight of the task queue of each subdomain, and perform concurrent task scheduling on the task queue according to the scheduling weight in descending order. The sub-knowledge graph generation module 203 is specifically used to: based on the ontology model of the sub-domain corresponding to the task queue scheduled by concurrent tasks, perform ontology modeling, entity and relation extraction, entity alignment and disambiguation processing on all data fragments in the task queue in sequence, and generate the sub-knowledge graph of the sub-domain corresponding to the task queue, until the sub-knowledge graph of each sub-domain is obtained.

[0061] Optionally, in the above scheme, the subdomain label allocation module 202 is specifically used to: extract the keyword set, entity set and semantic vector of each data segment, calculate the matching score between each data segment and each subdomain based on the predefined keyword set, entity set and semantic vector of each subdomain, and assign the subdomain label corresponding to the highest score to each data segment.

[0062] Optionally, the above scheme also includes a data receiving and processing module, which is used to: connect to various heterogeneous data sources in the wind power field through a streaming bus, receive data from the wind power field, and perform standardization and format unification processing on the received data from the wind power field to obtain processed data. The various heterogeneous data sources include: data sources for storing structured data, data sources for storing semi-structured data, and data sources for storing unstructured data. The semantic sharding module 201 is specifically used to perform semantic sharding on the processed data.

[0063] In another embodiment, a knowledge graph construction system for the wind power field according to the present invention includes a domain controller, a construction engine, a data labeling machine, a streaming bus, and a domain integrator, such as... Figure 4 As shown.

[0064] Among them, the streaming bus, as the data access layer, is responsible for connecting to various heterogeneous data sources in the wind power field, including structured data sources, semi-structured data sources and unstructured data sources. It performs real-time acquisition, standardization and format unification processing of the received raw data to form a standardized and continuous data stream, providing stable input to the downstream.

[0065] The data segmentation machine performs semantic segmentation on the data processed by the streaming bus, dividing the data into multiple data segments according to the data type: structured data is segmented by record, semi-structured data is segmented by field or data block, and unstructured data is segmented by sentence or paragraph semantic boundaries. It also extracts a set of keywords, a set of entities, and a semantic vector for each data segment, and assigns subdomain labels based on the matching score with the predefined subdomain, thus completing the initial division of the data into subdomains.

[0066] The domain controller is responsible for subdomain task scheduling and resource coordination. Specifically, it receives labeled data fragments from the data labeling machine and distributes them to the task queues of the corresponding subdomains (e.g., subdomain-A, subdomain-B) according to the subdomain labels. Simultaneously, it dynamically calculates scheduling weights based on queue length and predefined priorities, managing the execution order of subknowledge graph construction tasks through concurrent task scheduling strategies to ensure that high-load subdomains are processed first. Under the domain controller's scheduling, the "construction engine," as the core execution module for subknowledge graph generation, loads its predefined ontology model for each subdomain and strictly performs three processes sequentially on the data fragments in the task queue: "ontology modeling" (mapping data fragment information to the semantic framework of the ontology model), "entity and relation abstraction" (extracting entities and relations from the mapping framework to construct preliminary triples), and "entity alignment and reduction processing" (merging entities and eliminating ambiguity based on attribute and context similarity), ultimately generating a consistent subknowledge graph for each subdomain.

[0067] The domain integrator receives sub-knowledge graphs generated by the engine from all sub-domains and sequentially performs global entity aggregation (achieving cross-sub-domain entity alignment and unique ID allocation based on a global entity index table), cross-domain relationship reconstruction (completing bridging relationships through pattern matching, rule reasoning, or embedded model prediction), consistency rule verification (performing logical checks based on OWL or SHACL constraints), and conflict resolution (prioritizing and resolving conflicts based on data source credibility, frequency, and confidence). This ultimately forms a structurally consistent and semantically fused global knowledge graph for the wind power domain. The overall architecture embodies a closed-loop process from multi-source data access and domain-specific processing to global knowledge fusion, ensuring semantic consistency, structural integrity, and efficient updates in the graph construction through the collaboration of various modules.

[0068] It should be noted that the beneficial effects of the knowledge graph construction system 200 in the wind power field provided in the above embodiments are the same as those of the knowledge graph construction method in the wind power field described above, and will not be repeated here. Furthermore, the system provided in the above embodiments is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the system can be divided into different functional modules according to the actual situation to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process is detailed in the method embodiments, and will not be repeated here.

[0069] The knowledge graph construction system in the field of wind power of the present invention can be a computer program (including program code) running on a computer device. For example, the knowledge graph construction system in the field of wind power of the present invention is an application software that can be used to execute the corresponding steps in the knowledge graph construction method in the field of wind power of the present invention.

[0070] In some embodiments, the knowledge graph construction system for the wind power field of the present invention can be implemented in a combination of hardware and software. As an example, the knowledge graph construction system for the wind power field of the present invention can be a processor in the form of a hardware decoding processor, which is programmed to execute the knowledge graph construction method for the wind power field of the present invention. For example, the processor in the form of a hardware decoding processor can be one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0071] The modules described in the embodiments of this invention can be implemented in software or hardware. The names of the modules are not, in some cases, limiting the scope of the module itself.

[0072] An electronic device according to an embodiment of the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any of the above-mentioned methods for constructing a knowledge graph in the wind power field. That is, an electronic device according to an embodiment of the present invention may include, but is not limited to: a processor and a memory; the memory is used to store the computer program; the processor is used to execute the method for constructing a knowledge graph in the wind power field shown in any embodiment of the present invention by calling the computer program.

[0073] In one alternative embodiment, an electronic device is provided, such as Figure 5 As shown, Figure 5 The illustrated electronic device 4000 includes a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, for example, via a bus 4002. Optionally, the electronic device 4000 may further include a transceiver 4004, which can be used for data interaction between the electronic device and other electronic devices, such as sending and / or receiving data. It should be noted that in practical applications, the transceiver 4004 is not limited to one type, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present invention.

[0074] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this invention. Processor 4001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0075] Bus 4002 may include a path for transmitting information between the aforementioned components. Bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 4002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5The bus 4002 is represented by only one thick line, but this does not mean that there is only one bus or one type of bus.

[0076] The memory 4003 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0077] The memory 4003 stores application code (computer program) for executing the present invention, and its execution is controlled by the processor 4001. The processor 4001 executes the application code stored in the memory 4003 to implement the content shown in the foregoing method embodiments.

[0078] Among them, electronic devices can also be terminal devices, which can be any device that can install applications, including at least one of smartphones, tablets, laptops, desktop computers, smart speakers, smartwatches, smart TVs, and smart in-vehicle devices.

[0079] It should be noted that, Figure 5 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0080] An embodiment of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements any of the above-described methods for constructing a knowledge graph in the field of wind power.

[0081] Alternatively, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, a floppy disk, and an optical data storage device, etc.

[0082] In an exemplary embodiment, a computer program product or computer program is also provided, which includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform any of the aforementioned methods for constructing a knowledge graph in the wind power field.

[0083] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0084] It should be understood that the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0085] The computer-readable storage medium provided in this invention can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EEPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0086] The aforementioned computer-readable storage medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the method shown in the above embodiments.

[0087] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this invention is not limited to 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 equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this invention.

[0088] It should be noted that the terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and represent a limitation on a specific order or sequence. Where appropriate, the order of use for similar objects can be interchanged so that the embodiments of this application described herein can be implemented in an order other than that shown or described.

[0089] Those skilled in the art will recognize that this invention can be implemented as a system, method, or computer program product. Therefore, this invention can be specifically implemented in the following forms: it can be entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, this invention can also be implemented as a computer program product contained in one or more computer-readable media, which includes computer-readable program code.

[0090] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for constructing a knowledge graph in the field of wind power, characterized in that, The method comprises the following steps: performing semantic segmentation on the received data in the wind power field to obtain a plurality of data segments, and assigning each data segment with a sub-field label of a sub-field to which the data segment belongs; performing ontology modeling, entity and relationship extraction, entity alignment and disambiguation processing on the data segments with the assigned sub-field labels according to the ontology model of each sub-field in the wind power field to generate a sub-knowledge graph of each sub-field; performing global entity aggregation, cross-domain entity alignment, cross-domain relationship reconstruction, consistency rule verification and conflict processing on each sub-knowledge graph to obtain a global knowledge graph of the wind power field; continuously receiving new data in the wind power field, performing semantic segmentation and sub-field label assignment on the new data, updating the corresponding sub-knowledge graph according to the assigned sub-field label, and updating the global knowledge graph according to all updated sub-knowledge graphs. 2.The method of claim 1, wherein, Before generating the sub-knowledge graph of each sub-field, the method further comprises the following steps: distributing each data segment to a task queue of the corresponding sub-field according to the assigned sub-field label, calculating the scheduling weight of each task queue of the sub-field, and performing concurrent task scheduling on the task queue according to the scheduling weight from large to small; performing ontology modeling, entity and relationship extraction, entity alignment and disambiguation processing on the data segments with the assigned sub-field labels according to the ontology model of each sub-field in the wind power field to generate a sub-knowledge graph of each sub-field, comprising: performing ontology modeling, entity and relationship extraction, entity alignment and disambiguation processing on all data segments in the task queue according to the ontology model of the sub-field corresponding to the task queue to generate a sub-knowledge graph of the sub-field corresponding to the task queue, until a sub-knowledge graph of each sub-field is obtained. 3.The method of claim 1 or 2, wherein, Assigning each data segment with a sub-field label of a sub-field to which the data segment belongs comprises: extracting a keyword set, an entity set and a semantic vector of each data segment, calculating a matching score of each data segment and each sub-field based on a pre-defined keyword set, entity set and semantic vector of each sub-field, and assigning each data segment with a sub-field label corresponding to the highest score.

4. The method of claim 1 or 2, wherein, Further comprising: connecting a plurality of heterogeneous data sources in the wind power field through a streaming bus, receiving data in the wind power field, and performing standardization and format unification processing on the received data in the wind power field to obtain processed data, wherein the plurality of heterogeneous data sources comprise: a data source for storing structured data, a data source for semi-structured data and a data source for unstructured data; performing semantic segmentation on the received data in the wind power field, comprising: performing semantic segmentation on the processed data. 5.A system for constructing a knowledge graph in the field of wind power, characterized in that, The method comprises a semantic segmentation module, a sub-field label assignment module, a sub-knowledge graph generation module, a global knowledge graph generation module and a global knowledge graph updating module; The semantic segmentation module is configured to perform semantic segmentation on the received data in the wind power field to obtain a plurality of data segments; The sub-field label assignment module is configured to assign each data segment with a sub-field label of a sub-field to which the data segment belongs; The sub-knowledge graph generation module is configured to perform ontology modeling, entity and relationship extraction, entity alignment and disambiguation processing on the data segments with the assigned sub-field labels according to the ontology model of each sub-field in the wind power field to generate a sub-knowledge graph of each sub-field. The sub-knowledge graph generation module is configured to: according to the ontology model of each sub-field in the wind power field, sequentially perform ontology modeling, entity and relation extraction, entity alignment and disambiguation processing on the data segments assigned with the sub-field labels, and generate a sub-knowledge graph of each sub-field. The global knowledge graph generation module is configured to: sequentially perform global entity aggregation, cross-domain entity alignment, cross-domain relation reconstruction, consistency rule verification and conflict processing on each sub-knowledge graph, and obtain a global knowledge graph of the wind power field. The global knowledge graph updating module is configured to: continuously receive new data of the wind power field, perform semantic segmentation and sub-field label assignment on the new data, update the corresponding sub-knowledge graph according to the assigned sub-field label, and update the global knowledge graph according to all updated sub-knowledge graphs. 6.The knowledge graph construction system for wind power field of claim 5, wherein, The data scheduling module is further configured to: before generating the sub-knowledge graph of each sub-field, distribute each data segment to a task queue of the corresponding sub-field according to the assigned sub-field label, calculate a scheduling weight of each task queue of the sub-field, and perform concurrent task scheduling on the task queues according to the scheduling weight from large to small. The sub-knowledge graph generation module is specifically configured to: according to the ontology model of the sub-field corresponding to the task queue through the concurrent task scheduling, sequentially perform ontology modeling, entity and relation extraction, entity alignment and disambiguation processing on all data segments in the task queue, generate a sub-knowledge graph of the sub-field corresponding to the task queue, and obtain the sub-knowledge graph of each sub-field.

7. The system according to claim 5 or 6, wherein, The sub-field label assignment module is specifically configured to: extract a keyword set, an entity set and a semantic vector of each data segment, calculate a matching score of each data segment and each sub-field based on the predefined keyword set, entity set and semantic vector of each sub-field, and assign a sub-field label corresponding to the highest score to each data segment. 8.The system of claim 5 or 6, wherein, The data receiving and processing module is further configured to: connect multiple heterogeneous data sources of the wind power field through a streaming bus, receive data of the wind power field, and perform standardization and format unification processing on the received data of the wind power field to obtain processed data, wherein the multiple heterogeneous data sources include: a data source for storing structured data, a data source for semi-structured data, and a data source for unstructured data. The semantic segmentation module is specifically configured to: perform semantic segmentation on the processed data.

9. An electronic device, comprising: The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the method for constructing a knowledge graph of the wind power field according to any one of claims 1 to 4.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the method for constructing a knowledge graph of the wind power field according to any one of claims 1 to 4.

Images