Knowledge graph-based operation and maintenance method, system, equipment and medium

Abstract

The invention is suitable for the technical field of operation and maintenance, and provides an operation and maintenance overhaul operation method, system, equipment and medium based on a knowledge graph, and the method comprises the steps: dynamically constructing the knowledge graph based on element data which comprises a personnel entity, an equipment entity, a tool entity, a rule entity and an environment entity; collecting multi-source operation data of the equipment in real time, and outputting a fault feature vector through a multi-modal causal unwrapping model; and outputting an operation and maintenance scheduling scheme through the knowledge graph based on the fault feature vector. Through comprehensive integration and dynamic updating of human, machine, object, method and ring multi-source data, the practicability of operation and maintenance is improved, support is provided for operation, maintenance and management of a power plant, and personnel scheduling optimization and equipment fault prediction are realized in combination with a multi-modal causal unwrapping model.

Description

Technical Field

[0001] The present application belongs to the field of operation and maintenance technology, and in particular relates to an operation and maintenance method, system, equipment and medium based on knowledge graph. Background Art

[0002] With the continuous advancement of power system automation and intelligentization, the operation and maintenance model of power plant equipment is undergoing profound changes. Traditional operation and maintenance methods mainly rely on scheduled maintenance and passive fault handling. This approach is not only inefficient but also fails to meet the high reliability and safety requirements of modern power systems. In recent years, data-driven intelligent power plant operation and maintenance technology has made significant progress and gradually become the mainstream direction of industry development. This technology, centered on real-time data monitoring and predictive analysis, can achieve active perception and early warning of equipment operating status, thereby effectively reducing equipment failure rates and improving operation and maintenance efficiency.

[0003] However, the existing power plant intelligent operation and maintenance technology still faces many challenges in practical application. Power plant equipment is of various types and complex structures, and the data generated during its operation is diverse and complex. In addition, the operating environment of power plant equipment is complex and changeable, and is affected by many uncertain factors, such as equipment aging, load changes, ambient temperature, etc. Therefore, equipment failures often do not occur in isolation, but are closely related to the equipment's operating history, the status of other related equipment, and environmental factors. When performing fault diagnosis, the existing power plant intelligent operation and maintenance system often only focuses on the local data of the current equipment, ignoring the inherent connection between these data, resulting in the isolation of the diagnosis process and the inability to conduct a comprehensive and in-depth correlation analysis of abnormal events. Summary of the Invention

[0004] The embodiments of the present application provide a knowledge graph-based operation and maintenance method, system, equipment and medium, which can solve one of the above-mentioned existing technical problems.

[0005] In a first aspect, an embodiment of the present application provides an operation and maintenance method based on a knowledge graph, comprising: Dynamically construct a knowledge graph based on factor data, where the factor data includes personnel entities, equipment entities, tool entities, rule entities, and environment entities; Collect multi-source operating data of equipment in real time and output fault feature vectors through a multimodal causal disentanglement model; Based on the fault feature vector, an operation and maintenance scheduling plan is output through the knowledge graph.

[0006] Furthermore, the knowledge graph is dynamically constructed based on the element data, and the element data includes personnel entities, equipment entities, tool entities, rule entities and environment entities, including: Collecting element data, preprocessing the element data, and storing the preprocessed element data of different data types in different intermediate databases, wherein the intermediate databases include a time series database, a graph database, and a document database; An element ontology structure is constructed, and the attributes and entity relationships of each element are determined through the element ontology structure.

[0007] Furthermore, the dynamic construction of the knowledge graph includes: Based on timeliness, confidence, and manual feedback, a comprehensive scoring model is constructed for each entity, and the trust score of each entity is dynamically adjusted through the comprehensive scoring model; Based on the importance of the equipment and the sensitivity of the entity, the timeliness is quantified through a dynamic exponential decay model to obtain a timeliness score; For the confidence, a confidence score is obtained by quantifying the degree of trust of each element data to the corresponding entity; The operation feedback report of the staff is obtained in real time, the operation feedback report is processed, the manual feedback is quantified, and a feedback score is obtained.

[0008] Furthermore, the real-time acquisition of the operator's operational feedback report, processing the operational feedback report, quantifying the manual feedback, and obtaining a feedback score includes: Classifying the operation feedback report to determine a report type, wherein the report type includes a positive report and a negative report; Parsing the operation feedback report, and extracting entity content from the operation feedback report, wherein the entity content includes related entities and corresponding entity relationships; A feedback score is calculated based on the report type and the entity content, combined with a reinforcement coefficient of the report content of the operation feedback report, where the reinforcement coefficient specifically represents the confidence level of a positive report and the severity of a negative report.

[0009] Furthermore, the multi-source operating data of the real-time acquisition device is outputted as a fault feature vector through a multimodal causal disentanglement model, including: Real-time monitoring of multi-source operating data of each device, including vibration data, operation logs, and environmental parameters; The multi-source operation data is input into a pre-trained multimodal causal disentanglement model to output a fault feature vector of the device, where the fault feature vector includes mechanical features, human features, and environmental features.

[0010] Furthermore, outputting an operation and maintenance scheduling plan based on the fault feature vector through the knowledge graph includes: Based on the fault feature vector, searching in the knowledge graph to obtain candidate fault hypotheses; Calculate the failure probability value of each candidate fault hypothesis and sort the failure probability values ​​from large to small according to their numerical values; Calculating a probability difference between the first two failure probability values, and generating a verification instruction if the probability difference is less than a preset value; Perform fault verification through the verification instruction to determine the cause of the equipment failure; Based on the cause of the equipment failure and combined with the attributes and entity relationships of each element in the knowledge graph, a personnel scheduling plan and a safety management plan are output.

[0011] Furthermore, the method further comprises: Extract equipment knowledge rules from the power plant's knowledge graph and generate a meta-knowledge distillation framework; Through the meta-knowledge distillation framework, combined with the device adapter, the knowledge graph is migrated across plants, wherein the device adapter includes a local adapter and a rule calibrator, wherein the local adapter is used to receive device knowledge rules and load the local sample data of the target power plant, and the rule calibrator is used to adjust the device parameters according to the local sample data of the target power plant.

[0012] In a second aspect, an embodiment of the present application provides an operation and maintenance system based on a knowledge graph, including: The first processing module is used to dynamically construct a knowledge graph based on element data, where the element data includes personnel entities, equipment entities, tool entities, rule entities, and environment entities; The second processing module is used to collect multi-source operating data of the equipment in real time and output the fault feature vector through the multimodal causal disentanglement model; The third processing module is used to output an operation and maintenance scheduling plan through the knowledge graph based on the fault feature vector.

[0013] In a third aspect, an embodiment of the present application provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned knowledge graph-based operation and maintenance method when executing the computer program.

[0014] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, comprising a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned operation and maintenance method based on the knowledge graph.

[0015] Compared with the prior art, the embodiments of the present application have the following beneficial effects: The present application discloses an operation and maintenance inspection method based on a knowledge graph. By constructing a dynamic knowledge graph engine for five elements, namely, staff, power plant equipment, maintenance tools, work procedures, and environmental conditions, it realizes the comprehensive integration and dynamic update of multi-source data on people, machines, objects, methods, and environments, thereby improving the practicality of operation and maintenance inspection and providing support for the operation and management of power plants. Furthermore, the multimodal causal disentanglement model outputs fault feature vectors, and the equipment fault is accurately located and dynamically verified based on the fault feature vectors. Then, through the application of the knowledge graph, the corresponding operation and maintenance scheduling plan is generated, which facilitates the staff to carry the corresponding maintenance tools to the corresponding fault location to repair the fault and realize the operation and maintenance inspection of various equipment in the power plant. At the same time, through the migration of equipment knowledge across plants, the adaptability and generalization ability of the present application in different equipment and environments are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0017] Figure 1 This is a flowchart of a knowledge graph-based operation and maintenance method provided by one embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of an operation and maintenance system based on a knowledge graph provided by one embodiment of the present invention; Figure 3 It is a structural diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0019] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0020] It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0021] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0022] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0023] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0024] See also Figure 1 As shown, the present invention is an operation and maintenance method based on knowledge graph, which includes the following steps: S100, dynamically constructing a knowledge graph based on factor data, wherein the factor data includes personnel entities, equipment entities, tool entities, rule entities, and environment entities; In this application, a dynamic knowledge graph engine is constructed by combining five elements: staff, power plant equipment, maintenance tools, work procedures, and environmental conditions. This enables comprehensive integration and dynamic updating of multi-source data on people, machines, objects, methods, and the environment, thereby improving the practicality of operation and maintenance and providing support for the operation and management of power plants.

[0025] In some embodiments, step S100 includes: Collecting element data, preprocessing the element data, and storing the preprocessed element data of different data types in different intermediate databases, wherein the intermediate databases include a time series database, a graph database, and a document database; An element ontology structure is constructed, and the attributes and entity relationships of each element are determined through the element ontology structure.

[0026] In this application, a knowledge graph is constructed for the five elements of "people, machines, objects, laws, and environment", which are specifically personnel entities, equipment entities, tool entities, rule entities, and environment entities. By collecting data related to the entities of the above-mentioned elements in the power plant and pre-processing them, it is possible to effectively integrate and analyze data from different sources, realize the association of data between power plant equipment, operating environment, and staff, avoid information islands, thereby improving the practicality of operation and maintenance, and providing support for the operation and management of power plants. It can be understood that the personnel entity is specifically the staff of the power plant, the equipment entity is specifically the various types of power plant equipment in the power plant, the tool entity is specifically the maintenance tools used to repair equipment failures, the rule entity is specifically the procedural documents in the operation and maintenance process, such as power safety work procedures and maintenance process guidelines, etc., and the environmental entity is specifically the environmental data of the power plant, such as salt spray concentration and temperature and humidity.

[0027] For the above-mentioned element data, specifically, by installing equipment sensors on various equipment in the power plant, the operating status of the equipment can be monitored in real time, such as monitoring the vibration data and temperature data of the steam turbine, and obtaining the equipment ledger of each equipment to understand the historical status and current status of the equipment. The equipment ledger records the basic information, technical parameters, operating status, operation records, maintenance and repair records of the equipment, and obtains the GPS location, skill level, on-the-job status and qualification certificate of the staff to ensure the traceability of personnel qualifications and locations. In addition, the service life data of each maintenance tool is recorded through RFID technology, and the use and maintenance of the maintenance tools are managed to realize the life cycle management of the maintenance tools. In addition, by monitoring environmental parameters such as salt spray concentration, temperature and humidity in the environment, the impact of the environment on the equipment can be evaluated.

[0028] To be more specific, in this embodiment, the collected factor data are preprocessed to ensure the quality and consistency of the data. Specifically, missing values ​​and outliers in the numerical data are processed to ensure the integrity and accuracy of the data. The cleaned data is then converted into a standard format to facilitate subsequent processing and analysis. For the text data in the factor data, natural language processing technology is used to extract key entities therefrom, such as extracting key entity information such as equipment name, failure mode and safety clauses from maintenance records.

[0029] In some embodiments, the standard format in the data conversion process is specifically RDF triples. For example, sensor data is converted into RDF triples to facilitate the representation of devices and their attributes in the knowledge graph. Specifically, a sensor data is stored in JSON format after data cleaning, specifically: { "source": "TSI_VIB_001", "value": 7.2, "unit": "mm / s", "timestamp": "2025-06-04T08:23:15Z", "device": "#3HP_Turbine" } Furthermore, it is converted into an RDF triple: <#3HP_Turbine> <hasvibration> "7.2"^^ <xsd:decimal>, which is used to indicate that the No. 3 high-pressure steam turbine has the vibration attribute, and the attribute value is 7.2. Similarly, the timestamp data can be converted to: <#3HP_Turbine> <hasvibrationtimestamp>"2025-06-04T08:23:15Z"^^ <xsd:datetime>, and the data source can be converted to: <#3HP_Turbine> <hasvibrationsource>"TSI_VIB_001".

[0030] In this embodiment, the pre-processed element data are stored in different intermediate databases according to their data types to facilitate subsequent analysis and application. Specifically, the intermediate database includes a time series database, a graph database, and a document database. Among them, the time series database is used to store various sensor data, such as vibration data of the equipment or temperature data of the environment. The above data is usually time series data associated with time; the graph database is used to store graph data of entities and entity relationships, and the relationship between entities such as equipment, personnel, and environment is represented by a graph structure. The document database is used to store other unstructured data such as procedure documents, operation feedback reports, and maintenance records to facilitate the management and retrieval of documents or various maintenance reports.

[0031] In this embodiment, RDF / Turtle syntax is used to define the element ontology structure of each entity to semantically describe the knowledge graph of the power plant. In a possible embodiment, the definition of the element ontology structure is as follows: @prefix pp:<http: / / www.powerplantkg.org / ontology#> . pp:Human a owl:Class ; rdfs:subClassOf [ a owl:Restriction ; owl:onProperty pp:hasSkill; owl:allValuesFrom pp:SkillLevel ] . pp:Machine a owl:Class ; rdfs:subClassOf pp:CriticalEquipment; pp:hasFailureMode pp:VibrationFailure, pp:CorrosionFailure . pp:Material a owl:Class ; pp:lifetimeModel pp:HumidityStressModel . pp:Method a owl:Class ; pp:standardCode "DL / T 890.501-2018" . pp:Environment a owl:Class ; pp:hasParameter pp:Salinity, pp:Temperature . Specifically, by defining a prefix pp, each element points to the entity pp, which points to the namespace http: / / www.powerplantkg.org / ontology#. Therefore, all subsequent element definitions starting with pp: point to this namespace, thus improving the element ontology structure of the knowledge graph.

[0032] Furthermore, in the feature ontology structure in the above embodiment, pp:Human represents a personnel entity, which is defined as an OWL class. In this entity, each personnel must be associated with an instance of the pp:SkillLevel class through the pp:hasSkill attribute. It can be understood that the feature ontology structure also defines a SkillLevel class related to the skill level of each personnel, which covers the skill level of each maintenance personnel. Similarly, pp:Human can also be associated with an instance of the pp:Qualification class by setting the pp:hasQualification attribute, that is, pp:Human can add corresponding attributes and associate them with relevant instances according to actual needs to ensure data integrity.

[0033] Furthermore, pp:Machine represents the equipment entity, in which the pp:hasFailureMode attribute is associated with pp:VibrationFailure and pp:CorrosionFailure, indicating possible failure modes of the machine, such as vibration failure and corrosion failure. It is understood that the attribute value in pp:hasFailureMode can be increased based on actual needs. pp:Material represents the tool entity, in which the pp:lifetimeModel attribute is associated with pp:HumidityStressModel, indicating the lifespan of each maintenance tool. pp:Method represents the rule entity, and is associated with the relevant procedure document through the pp:standardCode attribute, such as the maintenance work procedure represented by the string "DL / T 890.501-2018" mentioned above. pp:Environment represents the environment entity, and is associated with pp:Salinity and pp:Temperature through the pp:hasParameter attribute, indicating that the environment has salinity and temperature parameters.

[0034] Specifically, the entities corresponding to the above-mentioned elements can add corresponding attribute classes according to actual production needs. By defining the element ontology structure for each element, it can be used to semantically describe the basic concepts such as personnel entities, equipment entities, tool entities, rule entities and environment entities in the power plant and the relationships between them, and convert them into corresponding entities and attributes, providing a basis for the construction and application of knowledge graphs.

[0035] In some embodiments, the dynamically constructing a knowledge graph includes: Based on timeliness, confidence, and manual feedback, a comprehensive scoring model is constructed for each entity, and the trust score of each entity is dynamically adjusted through the comprehensive scoring model; Based on the importance of the equipment and the sensitivity of the entity, the timeliness is quantified through a dynamic exponential decay model to obtain a timeliness score; For the confidence, a confidence score is obtained by quantifying the degree of trust of each element data to the corresponding entity; The operation feedback report of the staff is obtained in real time, the operation feedback report is processed, the manual feedback is quantified, and a feedback score is obtained.

[0036] In this embodiment, a comprehensive scoring model is set up to measure the trust score of each entity attribute value in the constructed knowledge graph, thereby realizing dynamic quantification of knowledge credibility and dynamically adjusting the weight of each path when making decisions in operation and maintenance scheduling, thereby improving the accuracy of obtaining corresponding information, such as the accuracy when retrieving candidate fault hypotheses in the knowledge graph.

[0037] Specifically, the comprehensive scoring model is measured by timeliness, confidence, and feedback from maintenance personnel during the maintenance process. The specific formula is: ,in, Represents the initial score of the corresponding entity attribute value and introduces a time decay factor , which is used to measure the impact of time on the initial score. That is, as time goes by, the correlation between the corresponding attribute values ​​of each attribute in the entity may decrease, that is, the trustworthiness decreases. Specifically, It is the time difference from a certain reference time point to the current time. Represents the dynamic attenuation coefficient, which is used to control the speed of aging attenuation. The larger the value of , the faster the score decays over time. In some embodiments, the dynamic decay coefficient It is determined by the information importance and entity sensitivity. The specific formula is: =a×information importance+b×entity sensitivity, where a represents the basic attenuation coefficient, which is set to 0.1 in one possible embodiment; b represents the environmental sensitivity factor, which ranges from 0.05 to 0.2. Furthermore, information importance can be evaluated and quantified based on the criticality, functional importance, and impact on business continuity of the entity's corresponding attribute value information during the operation and maintenance detection process, while entity sensitivity can be evaluated and quantified based on the likelihood of damage to the entity's corresponding attributes during environmental changes, such as the likelihood of loss of a steam turbine due to environmental changes. Specifically, for each evaluation quantitative indicator, a different quantitative level and corresponding score range are set. For example, each indicator is divided into three levels: low, medium, and high, corresponding to 1-3 points, 4-6 points, and 7-9 points, respectively. Then, a weighted average calculation is performed on each evaluation quantitative indicator to obtain the corresponding quantitative value of information importance or entity sensitivity.

[0038] More specifically, the confidence is obtained by quantifying the degree of trust of multiple factor data on the corresponding entity. Specifically, the attribute values ​​corresponding to the attributes in the corresponding entity are used to reflect the degree of trust of multiple information sources or data points on the entity or the entity attribute value. Specifically, a weight is assigned to each factor data according to the degree of influence of each factor data on the entity confidence. For example, for the equipment entity, the equipment sensor monitoring data may be the most critical for the evaluation of the equipment operation status, so it is given a higher weight, such as 0.4; the equipment ledger data can provide comprehensive information about the equipment, and the weight is set to 0.3; the environment-related factor data has a relatively small impact on the equipment, and the weight is set to 0.2; the personnel and tool-related factor data have a relatively limited role in the equipment entity confidence assessment, and the weights are each set to 0.05. Furthermore, due to The dimensions and value ranges of each factor data are different, and they need to be standardized to make them comparable. Then, based on the standardized data and the scoring criteria set by experts, the confidence score is calculated for each factor data. The scoring criteria can be divided according to the actual situation. For example, the standardized data is divided into three levels: low, medium, and high, corresponding to 1-3 points, 4-6 points, and 7-9 points respectively. Taking the equipment vibration data as an example, if the standardized vibration data is between 0-0.3, it is considered normal and the confidence score is 9 points; between 0.3-0.6, it is considered slightly abnormal and the confidence score is 6 points; above 0.6, it is considered seriously abnormal and the confidence score is 3 points. Finally, according to the weight and confidence score of each factor data, the comprehensive confidence of the entity is calculated. The specific calculation formula is: ,in represents the weight of the i-th element data, represents the confidence score of the i-th element data, and n is the total number of element data.

[0039] To be more specific, manual feedback refers to the operation feedback reports of equipment entities and tool entities provided by staff during the operation and maintenance or inspection process. By parsing and analyzing the operation feedback reports, a qualitative evaluation is performed on the corresponding information, and the information in the knowledge graph is dynamically updated to improve the credibility of the corresponding information in the knowledge graph.

[0040] Therefore, the comprehensive scoring model comprehensively considers the basic score, timeliness, confidence and manual feedback of each entity or entity attribute value to obtain a more comprehensive and accurate score. The introduction of dynamic time decay ensures that the contribution of older information to the final score gradually decreases, reflecting the dynamic changes in the attribute values ​​of each entity. The confidence provides a quantitative assessment of the reliability of information, and improves the accuracy of the score through the integration of multiple information sources. Manual feedback introduces the operational feedback reports of staff during the operation and maintenance or inspection process to correct or supplement the score. Through the above-mentioned comprehensive scoring model, the quality of relevant information such as each entity status or entity attribute value can be more effectively evaluated to provide support for decision-making.

[0041] In some embodiments, obtaining the operator's operational feedback report in real time, processing the operational feedback report, quantifying the manual feedback, and obtaining a feedback score include: Classifying the operation feedback report to determine a report type, wherein the report type includes a positive report and a negative report; Parsing the operation feedback report, and extracting entity content from the operation feedback report, wherein the entity content includes related entities and corresponding entity relationships; A feedback score is calculated based on the report type and the entity content, combined with a reinforcement coefficient of the report content of the operation feedback report, where the reinforcement coefficient specifically represents the confidence level of a positive report and the severity of a negative report.

[0042] In this embodiment, the operation feedback report is classified to determine its type. Specifically, the report type is divided into positive report and negative report. Among them, the positive report means that the information in the report is positive and is beneficial to the corresponding equipment entity or tool entity, such as discovering hidden dangers and handling them in advance, while the negative falsification report means that the information in the report is negative and is not conducive to the corresponding equipment entity or tool entity, such as unlicensed operation causing equipment damage or failure to perform inspections according to regulations. Therefore, it is necessary to perform semantic analysis on the operation feedback report to determine the report type to which the operation feedback report belongs.

[0043] More specifically, relevant entities and corresponding entity relationships are extracted from the report. It can be understood that entities refer to specific objects involved in the report, and entity relationships represent associations between specific object entities.

[0044] In this embodiment, the reinforcement coefficient of the report is determined by the report type and entity content of the operation feedback report, which is used to adjust the weights of entities and relationships in the knowledge graph. Specifically, for positive reports, the reinforcement coefficient is calculated based on the confidence level of the report, and the specific calculation formula is: adjustment = 0.3 * confidence_level, while for negative reports, the reinforcement coefficient is calculated based on the severity of the report, and the formula is: adjustment = -0.4 * severity_level.

[0045] In this embodiment, for each entity involved in the report, its corresponding feedback score is quantified according to the reinforcement coefficient, that is, the weight value with the reinforcement coefficient as a multiple is added or subtracted from the original weight of each entity, and it is used as the corresponding feedback score. In a preferred embodiment, for reports with positive feedback, the weight of each entity is increased by no more than 50%, that is, the new weight is no more than 1.5 times the current weight, and for reports with negative feedback, the weight of each entity is reduced by no more than 40%, that is, the new weight is no less than 60% of the current weight.

[0046] Specifically, in one embodiment, there is an intelligent maintenance work order, which is specifically an operation feedback report of a staff member, as follows: ## Troubleshooting - **Equipment**: #3 Steam Turbine High Pressure Rotor Bearing - **Fault Type**: Bearing raceway spalling (92% confidence level) - **Level of Urgency**: (Processing required within 72 hours) ## Personnel Scheduling **Role** **STAFF** Qualifications **Arrive Time** Work leader Wang Gong Rotating Machinery Maintenance (Advanced) 08:30 Technical Guardian Li Gong Safety monitoring (special) 08:30 Surveyor Zhang Gong Precision Measurement (Intermediate) 08:45 ## Maintenance plan 1. Replace the high-pressure rotor bearing (spare part number: B-2037) 2. Clean the lubrication oil circuit and replace the filter element 3. Centering accuracy calibration (target value <0.03mm) ## Security Control **Compulsory measures**: - Execute LOTO procedure: disconnect 3A52 circuit breaker - Set up a protective fence (5m radius) **Personal protection**: - Class 4 arc protective clothing - Chemical resistant mask **Environmental Control**: - Start local dehumidification (target humidity < 60%) - Real-time salt spray monitoring (threshold 25mg / m³) ## Forecast Suggestions - Expected life after overhaul: >3 years (85% confidence level) - Oil analysis is recommended after 3 months Specifically, the above-mentioned operation feedback report was analyzed. The content of the report generally reflects the execution of the maintenance work. There is no negative information such as unlicensed operation and failure to perform inspections according to regulations. The fault handling was successfully completed and follow-up suggestions were given. It is a positive report. Further, entity content extraction is performed. Among them, the equipment entity is specifically the #3 steam turbine high-pressure rotor bearing, the personnel entity is specifically Mr. Wang (work leader), Mr. Li (technical guardian) and Mr. Zhang (measurement specialist), and the tool entity is specifically the high-pressure rotor bearing (spare part number: B-2037), lubricating oil line cleaning tool, filter element, LOTO procedure, protective fence, Class 4 arc protective clothing, chemical protection mask, local dehumidification equipment, and salt spray monitoring equipment. The corresponding entity relationship is that Mr. Wang, Mr. Li and Mr. Zhang jointly participated in the maintenance work of the #3 steam turbine high-pressure rotor bearing, and the maintenance work used the high-pressure rotor bearing (spare part number: B- 2037) was replaced, and the lubricating oil line was cleaned with a lubricating oil cleaning tool and the filter element was replaced. During the maintenance, LOTO procedures were implemented, and safety control measures such as protective fencing were set up. Workers used personal protective equipment such as Class 4 arc protection suits and chemical masks. Local dehumidification equipment was activated, and environmental control measures such as salt spray monitoring were implemented. Furthermore, the report was positive, and the reinforcement factor was calculated according to the formula adjustment = 0.3 * confidence_level. Since this report is a maintenance report, the overall confidence level of the report is related to the confidence level of the fault diagnosis. Given that the confidence level of bearing raceway spalling in the fault diagnosis is 92%, the confidence level of this positive report is 0.92, and the reinforcement factor adjustment = 0.3 * 0.92 = 0.276. Therefore, the original weights of the corresponding attributes of each of the above entities are obtained respectively, and then the corresponding feedback scores are calculated. For example, if Mr. Wang's original weight is assumed to be original_weight_wang = 80, the new weight new_weight_wang = 80 * (1 + 0.276) = 102.08, 102.08 < 80 * 1.5 = 120, then the final feedback score is 102.08. According to a similar method, the new weights of other entities are calculated according to their original weights to obtain feedback scores, which are then synchronized to the comprehensive scoring model.

[0047] Therefore, through the above analysis and processing of the operation feedback report, the knowledge graph can dynamically update its content according to the actual operation and feedback, so as to better reflect the actual status and relationship of each entity in the power plant.

[0048] S200, collect multi-source operating data of the equipment in real time, and output fault feature vectors through a multimodal causal disentanglement model; In some embodiments, step S200 includes: Real-time monitoring of multi-source operating data of each device, including vibration data, operation logs, and environmental parameters; The multi-source operation data is input into a pre-trained multimodal causal disentanglement model to output a fault feature vector of the device, where the fault feature vector includes mechanical features, human features, and environmental features.

[0049] In this embodiment, the operating status of each device object in the device entity is monitored through a pre-trained multimodal causal disentanglement model, thereby outputting a fault feature vector related to the fault for subsequent maintenance of the equipment fault.

[0050] Specifically, multi-source operational data, including vibration data, operation logs, and environmental parameters, is collected in real time by installing vibration sensors on key equipment in the power plant. This data reflects the operational status of the equipment's mechanical components, such as bearing wear and gear failures. Specifically, when bearings experience early wear, the frequency components of the vibration signal change, potentially leading to specific fault frequencies. Furthermore, operator operation records for each device, such as the intelligent maintenance work order in step S100 or the operation work order for each device during task execution, which includes information such as operation time, operation type, and operation parameters, can reflect the impact of human factors on equipment operation. For example, improper operation sequences or parameter settings can lead to equipment failure. Furthermore, environmental sensors collect parameters such as temperature, humidity, and salt spray concentration from the equipment's environment to measure the impact of environmental factors on equipment performance and reliability. For example, high temperatures can lead to poor heat dissipation, while excessive humidity can degrade insulation performance.

[0051] More specifically, the multi-source operation data collected above is preprocessed, such as filtering the collected vibration data to remove noise interference. Common filtering methods include low-pass filtering, high-pass filtering and band-pass filtering. The appropriate filtering method is selected according to the characteristics of the vibration signal. Then, feature extraction is performed on the filtered signal, such as extracting time domain features and frequency domain features, cleaning and normalizing the operation log to remove duplicate, erroneous and irrelevant information, and converting the operation log into a structured data format to facilitate subsequent analysis and processing. For example, the operation type and operation parameters are encoded to facilitate data mining and pattern recognition, and the environmental parameters are normalized to convert data of different dimensions into a unified range to facilitate subsequent fusion analysis. At the same time, outlier detection and processing are performed on the environmental parameters to remove data that obviously deviates from the normal range. The preprocessed multi-source operation data is then input into a pre-trained multimodal causal disentanglement model to output the corresponding fault feature vector, specifically mechanical features, human features and environmental features.

[0052] For the pre-trained multimodal causal disentanglement model, a deep learning architecture combining convolutional neural networks and recurrent neural networks is used. The convolutional neural network extracts the spatial features of vibration data, while the recurrent neural network processes the temporal features of operation logs and environmental parameters. Specifically, the model input layer receives the pre-processed vibration data, operation logs, and environmental parameters. The vibration data is converted into a two-dimensional image format and input into the convolutional neural network for feature extraction. The operation logs and environmental parameters are converted into sequence data and input into the recurrent neural network for feature extraction. In the middle layer of the model, the features extracted by the convolutional neural network and the recurrent neural network are fused using methods such as splicing and weighted summation to fully utilize the information of the multimodal data. Finally, the model output layer outputs a fault feature vector, which contains the fault-related feature information extracted from the multimodal data. During model training, a large amount of multimodal data from both normal and faulty equipment is collected to construct training and test datasets. The training dataset is annotated to clearly identify the fault type and location for each sample. The multimodal causal disentanglement model is trained using the training dataset. Using a backpropagation algorithm and gradient descent optimization, the model parameters are continuously adjusted to minimize the error between the model output and the true label. Simultaneously, the model's performance is evaluated using the test dataset, monitoring metrics such as accuracy, recall, and F1 score. Based on the evaluation results, the model's structure and parameters are adjusted and optimized to improve its generalization capabilities.

[0053] S300. Based on the fault feature vector, output an operation and maintenance scheduling plan through the knowledge graph.

[0054] This application outputs fault feature vectors through a multimodal causal disentanglement model, and achieves accurate positioning and dynamic verification of equipment faults based on the fault feature vectors. Then, through the application of knowledge graphs, a corresponding operation and maintenance scheduling plan is generated, which makes it convenient for staff to carry the corresponding maintenance tools to the corresponding fault location to repair the fault and realize the operation and maintenance of various equipment in the power plant.

[0055] In some embodiments, step S300 includes: Based on the fault feature vector, searching in the knowledge graph to obtain candidate fault hypotheses; Calculate the failure probability value of each candidate fault hypothesis and sort the failure probability values ​​from large to small according to their numerical values; Calculating a probability difference between the first two failure probability values, and generating a verification instruction if the probability difference is less than a preset value; Perform fault verification through the verification instruction to determine the cause of the equipment failure; Based on the cause of the equipment failure and combined with the attributes and entity relationships of each element in the knowledge graph, a personnel scheduling plan and a safety management plan are output.

[0056] In this embodiment, mechanical feature matching is performed in the knowledge graph, specifically, dynamic thresholds are set for mechanical features, such as a matching threshold of 90% for vibration data, and environmental adaptability filtering is performed based on environmental features, such as setting environmental thresholds for environmental features. Specifically, candidate fault hypotheses that meet the environmental thresholds are quickly screened through range queries. At the same time, spatial search is performed in the knowledge graph based on mechanical features to obtain candidate fault hypotheses for load action thresholds, and ultimately multiple candidate fault hypotheses based on feature similarity are generated. Furthermore, cosine similarity is used to compare the semantic similarity between the fault feature vector and the query vector preset in the knowledge graph, and ultimately multiple candidate fault hypotheses sorted by embedding similarity are generated. Thus, the multiple candidate fault hypotheses based on feature similarity and the multiple candidate fault hypotheses sorted by embedding similarity are the multiple candidate fault hypotheses finally generated.

[0057] In this embodiment, for each candidate fault hypothesis, the probability of its occurrence is calculated using the Bayesian theorem. The formula of the Bayesian theorem is: ,in, Indicates that under the condition that the fault characteristic vector E occurs, the fault hypothesis Probability of occurrence; Indicates that the fault hypothesis The probability of the fault feature vector E occurring under the condition of occurrence, which is obtained through historical fault data and model training; Indicates a fault hypothesis The prior probability can be estimated based on the equipment's fault history and operating conditions; P(E) represents the probability of the fault feature vector E occurring, which can be obtained by normalizing the probabilities of all fault hypotheses. Further, the calculated fault probability values ​​of each candidate fault hypothesis are sorted according to the numerical values ​​of the fault probability values ​​to obtain a list of fault hypothesis probabilities. The first two candidate fault hypotheses are selected and the probability difference between the two is calculated. , that is, the probability difference between the candidate fault hypothesis with the highest probability and the candidate fault hypothesis with the second highest probability. In a preferred embodiment, if When the value is less than 0.25, it indicates that the current fault location result has a large uncertainty and requires further verification. In this case, a verification instruction is generated to further verify the device to determine the cause of the device failure.

[0058] To be more specific, the verification instruction is generated based on the current device status information and must also meet the minimum cost principle and the maximum probability change principle, that is, the verification instruction must meet min∑k=1nC(ak), where C(ak) is the action cost, that is, the resources required to execute the verification instruction ak, such as time, manpower, equipment, etc. At the same time, the verification instruction must also meet , that is, after executing the verification instruction Ek, the probability of a certain fault hypothesis can be significantly changed, specifically, the probability value changes by more than 0.2, thereby improving the accuracy of fault location.

[0059] Furthermore, in some embodiments, the knowledge graph of the power plant is also provided with a knowledge base of achievable verification action instructions, as follows: Verify Instructions Cost coefficient Available evidence Applicable assumptions Infrared temperature measurement 0.5 Temperature distribution Bearing wear / inadequate lubrication Oil analysis 2.0 Metal particle concentration Bearing wear / fatigue spalling Laser centering 1.8 Axis deviation Poor alignment / loose foundation Vibration phase 1.2 Phase characteristics Unbalance / looseness Listening stick detection 0.3 Abnormal sound characteristics All mechanical failures Specifically, a budget cost threshold is set, and the verification instructions applicable to the candidate fault hypothesis are selected by traversing the above verification instruction knowledge base, and the verification instructions exceeding the budget cost threshold are skipped. Then, the utility value of each verification instruction is calculated. Specifically, the expected probability change of each verification instruction is calculated. When its expected probability change is greater than When the device is fault-checked, it is determined as the final verification instruction and used to verify the device fault.

[0060] More specifically, the generated verification instruction is sent to the corresponding execution unit, such as an infrared thermometer, a video surveillance system, and an environmental sensor calibration device. The execution unit executes the verification instruction and updates the probability of the corresponding candidate fault hypothesis based on the verification result. Specifically, if the verification result supports a candidate fault hypothesis, the probability of the fault hypothesis is increased; if the verification result denies a candidate fault hypothesis, the probability of the fault hypothesis is reduced, and then the fault hypothesis probability difference is recalculated to finally determine the cause of the equipment failure.

[0061] In this embodiment, based on the above-determined cause of the equipment failure, the equipment failure location, required skills, and required repair tools are determined. Based on historical failure data, the required repair time and severity are determined. Then, candidate personnel are screened and the availability of repair tools is analyzed in the knowledge graph. At the same time, with minimizing travel time, maximizing skill matching, and minimizing safety risks as maintenance goals, a multi-objective genetic algorithm is used to determine the final operation and maintenance scheduling plan, that is, to generate a personnel scheduling plan and a safety management plan. For example, in one embodiment, based on the cause of the equipment failure, the information determined is as follows: { "fault_type": "Turbine bearing wear", # Fault location results "severity": 8.5, # Severity rating (1-10) "location": "#3 Unit Plant Area B", # Fault location "required_skills": ["Rotating Machinery Maintenance", "Precision Measurement"], # Required skills "required_tools": ["Laser Plummet", "Hydraulic Puller"], # Required tools "time_constraint": "4h" # Time limit } Based on the above information, the final operation and maintenance scheduling plan is as follows: { "team_leader": "Mr. Wang (Senior Technician)", "members": ["Engineer Zhang (precision measurement)", "Engineer Li (safety supervision)"], "tools": ["Laser Alignment Tool (S / N: GL-1023)", "Hydraulic Puller (S / N: HY-771)"], "path": "Maintenance Center → Unit #3: 8 minutes walk", "time_estimate": "2.5 hours", "safety_gear": ["High-temperature resistant gloves", "Protective mask"] } The above operation and maintenance scheduling plan will be fed back to the corresponding staff, so that the corresponding staff can carry the corresponding maintenance tools to the corresponding fault location, repair the fault, and realize the operation and maintenance of various equipment in the power plant.

[0062] In some other embodiments, the method further comprises: Extract equipment knowledge rules from the power plant's knowledge graph and generate a meta-knowledge distillation framework; Through the meta-knowledge distillation framework, combined with the device adapter, the knowledge graph is migrated across plants, wherein the device adapter includes a local adapter and a rule calibrator, wherein the local adapter is used to receive device knowledge rules and load the local sample data of the target power plant, and the rule calibrator is used to adjust the device parameters according to the local sample data of the target power plant.

[0063] Specifically, in the actual production application process, enterprises usually set up power plants in different regions. For different power plants, the knowledge graph-based operation and maintenance method of this application further proposes a method for migrating the corresponding equipment knowledge in the knowledge graphs of different power plants, thereby improving the adaptability and generalization ability of this application in different equipment and environments.

[0064] Specifically, a device feature vectorization encoding algorithm is used to numerically represent the equipment models, operating parameters, and environmental conditions of the source and target power plants. A standardized device feature matrix is ​​obtained through multidimensional feature space mapping technology. If the device feature similarity exceeds a preset threshold of 0.8, the device knowledge rule is considered to meet the basic conditions for migration. Based on the similarity calculation results of the device feature matrix, knowledge rule parsing technology is used to structurally decompose the parameter thresholds and judgment criteria in the source power plant. Through semantic parsing and rule pattern recognition, the core logical framework and parameter dependency relationships of the knowledge rules are obtained, and a meta-knowledge distillation framework is generated. Specifically, by traversing the device-related entity nodes in the knowledge graph, such as generators and transformers, associated attribute rules and relationship paths are extracted. Attribute rules are specifically thresholds for various attribute values, such as temperature thresholds and vibration amplitude limits. Relationship paths are relationships between attributes, such as the "bearing-lubrication-temperature association." Ultimately, a structured rule set is formed: {Equipment Type: [Rule 1, Rule 2...], ...}.

[0065] In this embodiment, based on the structured rule set generated above for expressing the meta-knowledge distillation framework, a cross-domain knowledge mapping algorithm is used to establish the equipment parameter correspondence between the source power plant and the target power plant, and the parameter mapping space corresponding to the target power plant is obtained by calculating the parameter domain conversion matrix. Based on the conversion result of the parameter mapping space, the adaptive parameter calibration technology is used to adjust the parameter threshold of the source power plant, and the calibration coefficient is determined according to the operating environment characteristics and equipment characteristic differences of the target power plant. Specifically, K=σtarget / σsource, where σtarget represents the standard deviation of the target power plant parameters, and σsource represents the standard deviation of the source power plant parameters.

[0066] Furthermore, based on the calculation result of the calibration coefficient K, the judgment criteria in the knowledge rules are dynamically adjusted. If the operating parameter P of the target power plant equipment satisfies the P×K≥threshold adjustment, it is judged that the knowledge rule is still valid in the target power plant environment. After the validity of the knowledge rule is verified, the rule adaptation engine is used to perform grammatical reconstruction and logical optimization on the adjusted knowledge rules. According to the specific operating environment of the target power plant, an adapted knowledge rule set is generated and the adapted knowledge rule set is obtained. The newly generated knowledge rules are checked for consistency with the existing rules of the target power plant through the rule conflict detection algorithm. If a rule conflict is found, the priority weight allocation mechanism is used to determine the final execution rule. According to the results of rule conflict detection and priority allocation, a complete rule base for cross-plant knowledge migration is established. Through the version management and update mechanism of the knowledge rules, a continuously optimized intelligent operation and maintenance knowledge system is obtained.

[0067] See also Figure 2 As shown, the present invention also provides an operation and maintenance system based on a knowledge graph, the system comprising: The first processing module 203 is used to dynamically construct a knowledge graph based on element data, where the element data includes personnel entities, equipment entities, tool entities, rule entities, and environment entities; The second processing module 202 is used to collect multi-source operation data of the equipment in real time and output the fault feature vector through the multimodal causal disentanglement model; The third processing module 201 is used to output an operation and maintenance scheduling plan through the knowledge graph based on the fault feature vector.

[0068] It is understandable that if Figure 1 The contents of the embodiment of the operation and maintenance method based on knowledge graph shown are applicable to the embodiment of the operation and maintenance system based on knowledge graph. The functions specifically implemented by the embodiment of the operation and maintenance system based on knowledge graph are the same as those in the embodiment of the operation and maintenance method based on knowledge graph shown. Figure 1 The embodiment of the operation and maintenance method based on the knowledge graph shown is the same as that of the embodiment of the operation and maintenance method based on the knowledge graph shown, and the beneficial effects achieved are the same as those of the embodiment of the operation and maintenance method based on the knowledge graph shown. Figure 1 The beneficial effects achieved by the embodiment of the operation and maintenance method based on knowledge graph shown are also the same.

[0069] It should be noted that the information interaction, execution process and other contents between the above-mentioned systems are based on the same concept as the embodiment of the method of the present invention. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0070] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0071] See also Figure 3 As shown, an embodiment of the present invention further provides a computer device 3, comprising: a memory 302 and a processor 301 and a computer program 303 stored on the memory 302. When the computer program 303 is executed on the processor 301, an operation and maintenance method based on a knowledge graph as described in any one of the above methods is implemented.

[0072] The computer device 3 may be a desktop computer, a notebook computer, a PDA, a cloud server or other computing devices. The computer device 3 may include, but is not limited to, a processor 301 and a memory 302. Those skilled in the art will understand that Figure 3 This is merely an example of the computer device 3 and does not constitute a limitation on the computer device 3 . The computer device 3 may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the computer device 3 may also include input and output devices, network access devices, etc.

[0073] The processor 301 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor may be a microprocessor or any conventional processor.

[0074] In some embodiments, the memory 302 may be an internal storage unit of the computer device 3, such as a hard drive or memory of the computer device 3. In other embodiments, the memory 302 may also be an external storage device of the computer device 3, such as a plug-in hard drive, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash memory card, etc. equipped on the computer device 3. Furthermore, the memory 302 may include both an internal storage unit of the computer device 3 and an external storage device. The memory 302 is used to store an operating system, application programs, a boot loader, data, and other programs, such as the program code of the computer program. The memory 302 may also be used to temporarily store data that has been output or is about to be output.

[0075] An embodiment of the present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, it implements the operation and maintenance method based on the knowledge graph as described in any one of the above methods.

[0076] In this embodiment, if the integrated unit is implemented as a software functional unit and sold or used as a standalone product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application can implement all or part of the process steps in the above-mentioned method embodiments by using a computer program to instruct the relevant hardware. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a camera / computer device, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signals, telecommunication signals, and software distribution media. Examples include USB flash drives, removable hard drives, magnetic disks, or optical disks. In some jurisdictions, based on legislation and patent practice, computer-readable media cannot be electric carrier signals or telecommunication signals.

[0077] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.< / hasvibrationsource> < / xsd:datetime> < / hasvibrationtimestamp> < / xsd:decimal> < / hasvibration>

Claims

1. A knowledge graph-based operation and maintenance method, characterized in that: include: Dynamically construct a knowledge graph based on factor data, where the factor data includes personnel entities, equipment entities, tool entities, rule entities, and environment entities; Collect multi-source operating data of equipment in real time and output fault feature vectors through a multimodal causal disentanglement model; Based on the fault feature vector, an operation and maintenance scheduling plan is output through the knowledge graph.

2. The method according to claim 1, wherein The knowledge graph is dynamically constructed based on the element data, and the element data includes personnel entities, equipment entities, tool entities, rule entities and environment entities, including: Collecting element data, preprocessing the element data, and storing the preprocessed element data of different data types in different intermediate databases, wherein the intermediate databases include a time series database, a graph database, and a document database; An element ontology structure is constructed, and the attributes and entity relationships of each element are determined through the element ontology structure.

3. The method according to claim 1, wherein The dynamic construction of the knowledge graph includes: Based on timeliness, confidence, and manual feedback, a comprehensive scoring model is constructed for each entity, and the trust score of each entity is dynamically adjusted through the comprehensive scoring model; Based on the importance of the equipment and the sensitivity of the entity, the timeliness is quantified through a dynamic exponential decay model to obtain a timeliness score; For the confidence, a confidence score is obtained by quantifying the degree of trust of each element data to the corresponding entity; The operation feedback report of the staff is obtained in real time, the operation feedback report is processed, the manual feedback is quantified, and a feedback score is obtained.

4. The method according to claim 3, wherein The real-time acquisition of the operator's operation feedback report, processing the operation feedback report, quantifying the manual feedback, and obtaining a feedback score includes: Classifying the operation feedback report to determine a report type, wherein the report type includes a positive report and a negative report; Parsing the operation feedback report, and extracting entity content from the operation feedback report, wherein the entity content includes related entities and corresponding entity relationships; A feedback score is calculated based on the report type and the entity content, combined with a reinforcement coefficient of the report content of the operation feedback report, where the reinforcement coefficient specifically represents the confidence level of a positive report and the severity of a negative report.

5. The method according to claim 1, wherein The real-time acquisition of multi-source operating data of the equipment and output of the fault feature vector through the multi-modal causal disentanglement model include: Real-time monitoring of multi-source operating data of each device, including vibration data, operation logs, and environmental parameters; The multi-source operation data is input into a pre-trained multimodal causal disentanglement model to output a fault feature vector of the device, where the fault feature vector includes mechanical features, human features, and environmental features.

6. The method according to claim 1, wherein Outputting an operation and maintenance scheduling plan based on the fault feature vector through the knowledge graph includes: Based on the fault feature vector, searching in the knowledge graph to obtain candidate fault hypotheses; Calculate the failure probability value of each candidate fault hypothesis and sort the failure probability values ​​from large to small according to their numerical values; Calculating a probability difference between the first two failure probability values, and generating a verification instruction if the probability difference is less than a preset value; Perform fault verification through the verification instruction to determine the cause of the equipment failure; Based on the cause of the equipment failure and combined with the attributes and entity relationships of each element in the knowledge graph, a personnel scheduling plan and a safety management plan are output.

7. The method according to claim 1, wherein The method further comprises: Extract equipment knowledge rules from the power plant's knowledge graph and generate a meta-knowledge distillation framework; Through the meta-knowledge distillation framework, combined with the device adapter, the knowledge graph is migrated across plants, wherein the device adapter includes a local adapter and a rule calibrator, wherein the local adapter is used to receive device knowledge rules and load the local sample data of the target power plant, and the rule calibrator is used to adjust the device parameters according to the local sample data of the target power plant.

8. An operation and maintenance system based on knowledge graph, characterized in that: include: The first processing module is used to dynamically construct a knowledge graph based on element data, where the element data includes personnel entities, equipment entities, tool entities, rule entities, and environment entities; The second processing module is used to collect multi-source operating data of the equipment in real time and output the fault feature vector through the multimodal causal disentanglement model; The third processing module is used to output an operation and maintenance scheduling plan through the knowledge graph based on the fault feature vector.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images