A nuclear power secondary circuit equipment intelligent diagnosis method and system based on a knowledge graph

Abstract

This invention discloses an intelligent diagnostic method and system for nuclear power plant secondary loop equipment based on a knowledge graph. The diagnostic method comprises the following steps: First, constructing a nuclear power plant secondary loop knowledge graph; then, analyzing time-series data and identifying fault symptoms; next, mapping the fault symptom set to the equipment entities; finally, retrieving possible fault modes based on the fault symptoms using the knowledge graph, and then performing reasoning to determine the fault mode. This invention can capture potential equipment faults based on monitoring data, retrieve possible fault modes from the knowledge graph based on the captured fault symptoms, further perform reasoning to determine the fault mode, and provide corresponding fault causes and related fault handling suggestions based on the fault mode. This can assist operators in accurately locating and quickly troubleshooting faults, ensuring the safe operation of secondary loop equipment.

Description

Technical Field

[0001] This invention relates to the field of nuclear power plant secondary loop system monitoring, specifically to a knowledge graph-based intelligent diagnostic method and system for nuclear power plant secondary loop equipment. Background Technology

[0002] The secondary loop of a nuclear power plant contains numerous pieces of equipment, and the operating status of these pieces of equipment directly affects the operational safety of the nuclear power unit. Developing intelligent diagnostic systems for these equipment is an important task to ensure the safe operation of a nuclear power plant.

[0003] In existing nuclear power equipment condition monitoring systems, the monitoring of equipment anomalies, the methods for handling these anomalies, and the corresponding procedures are not interconnected, making it impossible to directly complete the continuous process from equipment fault detection to diagnosis and resolution. For abnormal events occurring in nuclear power equipment, the monitoring system often only captures the abnormal signal immediately; determining the cause of the fault and formulating subsequent action plans requires human intervention. In actual operation, a single fault symptom often corresponds to multiple fault modes, requiring operators to rely on experience or consult extensive data to analyze the abnormal event and determine the fault mode. The large number of devices in a nuclear power plant, the vast amount of data, and the coupling relationships between different devices severely impact the speed of fault diagnosis. Although numerous fault diagnosis methods have emerged to assist manual fault diagnosis, existing fault diagnosis models primarily focus on determining the cause of the fault, with limited connection to post-processing knowledge such as operating manuals and fault handling case studies.

[0004] Existing knowledge graph applications mostly focus on managing knowledge bases such as equipment data and case libraries, achieving interaction between nuclear power plant operators and the knowledge graph through semantic parsing of manual input. During fault diagnosis, manual intervention is required to identify fault symptoms and interact with the knowledge graph to locate and trace the fault, resulting in low automation. Alternatively, time-series data can be directly input into the graph and compared with standard values ​​to infer the cause of anomalies. However, nuclear power plant operation data has non-linear coupling relationships, and simply comparing time-series values ​​at a single measurement point cannot cover all faults or potential faults.

[0005] Current knowledge graph reasoning methods use predefined IF-THEN rule templates for reasoning. Based on the specific number of faults, corresponding reasoning templates are defined. However, the actual number of faults is large, often requiring the definition of a large number of templates, making the rule design process cumbersome. Furthermore, these methods need to iterate together with the graph iteration process. Alternatively, external inference engine programs can be used to complete the knowledge graph reasoning task. The use of graph inference engines involves new algorithmic languages, increasing the learning and maintenance costs for practical engineering applications. Summary of the Invention

[0006] The technical problem this invention aims to solve is to provide an intelligent diagnostic system based on knowledge graphs, addressing the shortcomings of existing fault monitoring systems for secondary loop equipment in nuclear power plants. This system can quickly locate possible fault causes based on abnormal data from single or multiple measurement points and provide corresponding fault handling measures, assisting operators in making equipment maintenance decisions more quickly. Simultaneously, a knowledge graph reasoning method is proposed based on the graph structure, simplifying the design process of graph reasoning rules.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] A knowledge graph-based intelligent diagnostic method for nuclear power plant secondary loop equipment includes the following steps:

[0009] S1. Construct a knowledge graph and store it in a graph database. The specific method is as follows:

[0010] S11. Model the key equipment of the nuclear power plant's secondary loop and construct a hierarchical relationship of the graph entity. The graph hierarchy is divided into various entities such as equipment, fault modes, fault symptoms, fault causes, fault handling measures, troubleshooting suggestions, and follow-up action suggestions.

[0011] S12. Based on the acquired historical cases of nuclear power equipment fault diagnosis, entity recognition and relation extraction are performed according to the hierarchical relationship of the graph ontology to obtain diagnostic knowledge triples, and the diagnostic knowledge triples are fused.

[0012] S13. Based on the obtained operation manuals and operation-related data of nuclear power plant operators, perform entity identification and relationship extraction to obtain operation and maintenance-related triples.

[0013] S14. Merge the triples obtained in steps S12 and S13 to link operation and maintenance knowledge with diagnostic knowledge.

[0014] S15. Based on step S14, construct a knowledge graph based on the obtained diagnostic knowledge triplet and operation and maintenance related triplet, and store the knowledge graph in a graph database.

[0015] S2. Read the monitoring data of the nuclear power plant's secondary loop equipment to obtain the corresponding time-series data. Analyze the time-series data of the nuclear power plant's secondary loop equipment, detecting the time-series status information of individual measurement points and making joint judgments based on multiple measurement points. If an abnormal state is detected, determine the fault symptoms. The method for determining fault symptoms is as follows: First, cluster historical fault data and fault simulation data to determine the cluster centers and boundaries, obtaining a fault data clustering space. Then, divide the clustering results into corresponding fault symptom types. Next, map the detected abnormal data into this fault data clustering space to obtain the corresponding fault symptoms. If no fault occurs, analyze the time-series data for the next time period according to a preset time interval.

[0016] S3. Perform entity mapping on the set of fault symptoms obtained in step S2 and the devices with relevant fault symptoms to be diagnosed, and match the fault symptoms and devices with entities in the knowledge graph.

[0017] S4. Based on the entities obtained in step S3 (including fault symptoms and devices exhibiting related fault symptoms), use Python to link to the graph database, retrieve possible fault modes based on the fault symptoms, and then perform reasoning to determine the fault mode; the reasoning specifically includes:

[0018] For each fault symptom, the set of all fault modes containing that fault symptom is obtained by retrieval; the intersection of all fault mode sets is taken, and the result includes three cases: an empty set, a set containing a single element, and a set containing multiple elements.

[0019] If the result is an empty set, then perform a reverse search on the fault mode set to obtain the fault symptom set corresponding to the relevant single fault mode; and perform a combination traversal on all fault symptom sets to find the fault mode combination that can cover the detected fault symptom set, sort them in ascending order of combination complexity and output the corresponding fault mode combination set.

[0020] If the result is a set containing a single element, then output the set of failure modes directly.

[0021] If the result is a set containing multiple elements, the failure mode retrieval results are sorted and output according to the combinatorial complexity from smallest to largest.

[0022] S5. Based on the fault mode determined in step S4, search the knowledge graph to find the corresponding fault cause, fault handling measures, troubleshooting suggestions, and follow-up action suggestions.

[0023] Furthermore, in the above technical solution, the fault mode, fault cause, fault handling measures, troubleshooting suggestions, and follow-up action suggestions are fed back to the operators, who can then carry out the follow-up processing.

[0024] Furthermore, in step S11, the key equipment of the nuclear power plant's secondary loop includes the turbine body, high-pressure heater, low-pressure heater, deaerator, condenser, high-pressure cylinder, and low-pressure cylinder.

[0025] Furthermore, the historical cases of nuclear power equipment fault diagnosis in step S12 specifically include equipment fault cases, equipment deterioration assessment reports, etc.

[0026] Furthermore, in step S14, during the fusion process, the entities in the diagnostic knowledge triplet obtained in step S2 and the operation and maintenance related triplet obtained in step S3 are subjected to referencing resolution and entity disambiguation.

[0027] Furthermore, in step S4, the fault symptom node and the faulty device node are connected through a fault mode.

[0028] This invention also provides a knowledge graph-based intelligent diagnostic system for nuclear power plant secondary loop equipment, the system comprising:

[0029] A knowledge graph construction module is used to construct a knowledge graph and store the knowledge graph in a graph database;

[0030] The fault symptom determination module is used to read monitoring data of the nuclear power plant's secondary loop equipment, obtain corresponding time-series data, and analyze the time-series data. During the analysis process, it not only detects the time-series status information of a single measuring point value, but also makes a joint judgment based on multiple measuring points to determine whether the equipment has an abnormal state. If an abnormal state exists, fault symptom determination is performed. If no fault occurs, the time-series data of the next time period is analyzed according to a preset time interval.

[0031] The entity mapping module is used to perform entity mapping on the fault symptom set output by the fault symptom determination module and the devices with relevant fault symptoms to be diagnosed, thereby mapping the fault symptoms and devices to entities in the knowledge graph.

[0032] The fault mode determination module is used to determine the fault mode by using Python language to link to the graph database based on the entity obtained by the corresponding module of the entity, searching for possible fault modes based on fault symptoms, and then performing reasoning.

[0033] The fault cause and handling suggestion determination module is used to search the knowledge graph based on the fault mode output by the fault mode retrieval and reasoning module to find the corresponding fault cause, fault handling measures, troubleshooting suggestions, and follow-up action suggestions.

[0034] The present invention also provides an electronic device, the electronic system comprising: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the above-described intelligent diagnostic method for nuclear power plant secondary loop equipment based on knowledge graphs.

[0035] The present invention also provides a computer-readable storage medium storing computer instructions thereon, the computer instructions being used to cause a computer to execute the steps of the above-described intelligent diagnostic method for nuclear power plant secondary loop equipment based on knowledge graphs.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] This invention links knowledge graphs and time-series data, enabling a comprehensive analysis of the status of nuclear power plant secondary loop equipment directly from the data source. It identifies equipment fault symptoms, uses the knowledge graph to pinpoint a list of possible causes for the current fault symptoms, and provides subsequent handling measures. The entire process eliminates the need for manual fault symptom identification and retrieval, and the reasoning process is clear, controllable, and highly interpretable. After locating a potential fault, it guides operators in further troubleshooting. In terms of knowledge graph reasoning, a programming language calls the graph database and retrieves possible fault patterns based on fault symptoms before performing reasoning to determine the fault mode. This eliminates the need for additional inference engine programs, resulting in good maintainability. Attached Figure Description

[0038] Figure 1 This is a flowchart of the intelligent diagnostic method for nuclear power plant secondary loop equipment based on knowledge graphs according to the present invention;

[0039] Figure 2 The knowledge graph constructed for this invention;

[0040] Figure 3 This is a flowchart of the intelligent diagnostic system of the present invention. Detailed Implementation

[0041] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0042] This invention uses the high-pressure heater fault diagnosis of the secondary loop in a nuclear power plant as an example to illustrate the method of this invention.

[0043] The intelligent diagnostic method for nuclear power plant secondary loop equipment based on knowledge graphs of this invention has the following process: Figure 1 As shown, the specific steps include:

[0044] Step S100: Construct a knowledge graph and store it in a graph database. The constructed graph is shown in the attached figure. Figure 2 As shown.

[0045] First, key equipment in the secondary loop of the nuclear power plant is modeled, and a hierarchical relationship of the graph is constructed. The graph hierarchy is divided into entities such as equipment, fault modes, fault symptoms, fault causes, fault handling measures, troubleshooting suggestions, and follow-up action suggestions. Key equipment in the secondary loop includes entities such as the turbine body, high-pressure heater, deaerator, high-pressure cylinder, low-pressure cylinder, low-pressure heater, and condenser.

[0046] Then, relevant technical data on nuclear power and case studies of nuclear power equipment fault diagnosis are acquired. Based on the acquired historical case studies of nuclear power equipment fault diagnosis, entity recognition and relation extraction are performed according to the hierarchical relationship of the graph ontology to obtain diagnostic knowledge triples, and the diagnostic knowledge triples are then fused.

[0047] Next, based on the obtained operation manuals and operation-related data of nuclear power plant operators, entity identification and relation extraction are performed to obtain operation and maintenance-related triples; the diagnostic knowledge triples and the operation and maintenance-related triples are fused to link operation and maintenance knowledge and diagnostic knowledge.

[0048] Finally, based on the previous step, a knowledge graph is constructed using the diagnostic knowledge triples and the operation and maintenance related triples, and the knowledge graph is stored in a graph database for subsequent retrieval.

[0049] This step integrates fault knowledge into the fault map. Specific devices in the map can be associated with one or more fault modes from technical documentation or historical fault cases. At the same time, the fault causes, fault symptoms, fault modes, fault solutions, troubleshooting suggestions, and follow-up action suggestions corresponding to the fault modes in the technical documentation or cases are associated with the fault modes in the map.

[0050] Step S200: Read the monitoring data of the nuclear power plant's secondary loop equipment, obtain the corresponding time-series data, and analyze the time-series data. During the analysis, not only is the time-series status information of a single measuring point value detected, but also a joint judgment is made based on multiple measuring points to determine whether the equipment has an abnormal state. If an abnormal state exists, fault symptoms are determined. If no fault occurs, the time-series data of the next time period is analyzed according to a preset time interval.

[0051] The method for determining the fault symptoms is as follows: First, cluster the fault data and fault simulation data in historical operation, determine the cluster centers and boundaries, obtain the fault data clustering space, and divide the clustering results into the corresponding fault symptom types. Then, map the monitored abnormal data into the fault data clustering space to obtain the corresponding fault symptoms.

[0052] S300: Perform entity mapping on the set of fault symptoms obtained in step S200 and the devices with relevant fault symptoms to be diagnosed, and match the fault symptoms and devices with entities in the knowledge graph.

[0053] S400. Based on the entities obtained in step S300, a graph database is linked using Python. Possible fault modes are retrieved based on fault symptoms, and then inference is performed to determine the fault mode. Specifically, the inference involves:

[0054] For each fault symptom, the set of all fault modes containing that fault symptom is obtained by retrieval; the intersection of all fault mode sets is taken, and the result includes three cases: an empty set, a set containing a single element, and a set containing multiple elements.

[0055] If the result is an empty set, then perform a reverse search on the fault mode set to obtain the fault symptom set corresponding to the relevant single fault mode; and perform a combination traversal on all fault symptom sets to find the fault mode combination that can cover the detected fault symptom set, sort them in ascending order of combination complexity and output the corresponding fault mode combination set.

[0056] If the result is a set containing a single element, then output the set of failure modes directly.

[0057] If the result is a set containing multiple elements, the failure mode retrieval results are sorted and output according to the combinatorial complexity from smallest to largest.

[0058] S500: Based on the fault mode determined in step S400, search the knowledge graph to find the corresponding fault cause, fault handling measures, troubleshooting suggestions, and follow-up action suggestions.

[0059] When operators receive the above diagnostic results, they should prioritize handling the fault modes listed first in the order of failure modes.

[0060] Another embodiment of the present invention provides a knowledge graph-based intelligent diagnostic system for nuclear power plant secondary loop equipment, the system comprising:

[0061] A knowledge graph construction module is used to construct a knowledge graph and store the knowledge graph in a graph database;

[0062] The fault symptom determination module is used to read monitoring data of the nuclear power plant's secondary loop equipment, obtain corresponding time-series data, and analyze the time-series data. During the analysis process, it not only detects the time-series status information of a single measuring point value, but also makes a joint judgment based on multiple measuring points to determine whether the equipment has an abnormal state. If an abnormal state exists, fault symptom determination is performed. If no fault occurs, the time-series data of the next time period is analyzed according to a preset time interval.

[0063] The entity mapping module is used to perform entity mapping on the fault symptom set output by the fault symptom determination module and the devices with relevant fault symptoms to be diagnosed, thereby mapping the fault symptoms and devices to entities in the knowledge graph.

[0064] The fault mode determination module is used to determine the fault mode by using Python language to link to the graph database based on the entity obtained by the corresponding module of the entity, searching for possible fault modes based on fault symptoms, and then performing reasoning.

[0065] The fault cause and handling suggestion determination module is used to search the knowledge graph based on the fault mode output by the fault mode retrieval and reasoning module to find the corresponding fault cause, fault handling measures, troubleshooting suggestions, and follow-up action suggestions.

[0066] Specifically, the fault mode retrieval and inference module can be deployed to the API server, which can accept point-in-time data and entity data, and ultimately determine the fault mode.

[0067] The intelligent diagnostic system of this invention is used to diagnose nuclear power plant secondary loop equipment based on a knowledge graph. The process is as follows: Figure 3 :

[0068] The system is started to monitor the equipment's operating status. If abnormal data is detected, cluster analysis is performed to determine the type of fault symptom. For example, if abnormal terminal difference data is detected in the high-pressure heater, the measuring point data of the high-pressure heater is first read, the terminal difference is calculated, and the terminal difference data is compared with the fault data cluster space. The classification result is that the hydrophobic terminal difference is too large.

[0069] The detected fault symptoms, equipment, and fault time points are sent to the fault mode determination module. This module receives the data and maps the incoming fault symptom and equipment data to entities in the knowledge graph. Specifically, all anomaly detection results are organized into standard descriptions of the corresponding entities in the knowledge graph, and then output. For example, if the faulty equipment is a high-pressure heater, the fault symptom might be a large difference in condensate discharge. Based on the obtained entities, Python is used to connect to the graph database, retrieve possible fault modes based on the fault symptoms, and then perform reasoning to determine the fault mode. After determining the fault mode, the knowledge graph is searched based on the obtained fault mode to obtain the fault cause, fault handling measures, troubleshooting suggestions, and follow-up action suggestions.

[0070] The above fault diagnosis results (including fault causes, fault handling measures, troubleshooting suggestions, and follow-up action suggestions) are transmitted to the front end. The front end displays the fault diagnosis results at the time of the fault and stores the results in the database. Operators can then handle the equipment based on the diagnosis results.

[0071] Another embodiment of the present invention also provides an electronic device, the electronic system comprising: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the above-described intelligent diagnostic method for nuclear power plant secondary loop equipment based on knowledge graphs.

[0072] Another embodiment of the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to cause a computer to perform the steps of the above-described intelligent diagnostic method for nuclear power plant secondary loop equipment based on knowledge graphs.

Claims

1. A knowledge graph-based intelligent diagnosis method for nuclear power secondary circuit equipment, characterized in that, Includes the following steps: S1. Construct a knowledge graph and store the knowledge graph in a graph database; S2. Read the monitoring data of the nuclear power plant's secondary loop equipment, obtain the corresponding time-series data, and analyze the time-series data. During the analysis, not only is the time-series status information of a single measuring point value detected, but also a joint judgment is made based on multiple measuring points to determine whether the equipment is in an abnormal state. If an abnormal state is found, fault symptoms are identified; if no fault is found, the time series data for the next time period is analyzed according to the preset time interval. S3. Perform entity mapping on the set of fault symptoms obtained in step S2 and the devices with relevant fault symptoms to be diagnosed, and match the fault symptoms and devices with the entities in the knowledge graph. S4. Based on the entities obtained in step S3, use Python to link to the graph database, retrieve possible fault modes based on fault symptoms, and then perform inference to determine the fault mode; the inference specifically involves: For each fault symptom, the set of all fault modes containing that fault symptom is obtained by retrieval; the intersection of all fault mode sets is taken, and the result includes three cases: an empty set, a set containing a single element, and a set containing multiple elements. If the result is an empty set, then perform a reverse search on the fault mode set to obtain the fault symptom set corresponding to the relevant single fault mode; and perform a combination traversal on all fault symptom sets to find the fault mode combination that can cover the detected fault symptom set, sort them in ascending order of combination complexity and output the corresponding fault mode combination set. If the result is a set containing a single element, then directly output the set of failure modes; If the result is a set containing multiple elements, the failure mode retrieval results are sorted and output according to the combinatorial complexity from smallest to largest. S5. Based on the fault mode determined in step S4, search the knowledge graph to find the corresponding fault cause, fault handling measures, troubleshooting suggestions, and follow-up action suggestions. Step S1 specifically includes the following steps: S11. Model the key equipment of the nuclear power plant's secondary loop and construct a hierarchical relationship of the graph entity; the graph hierarchy is divided into entities such as equipment, fault mode, fault symptom, fault cause, fault handling measures, troubleshooting suggestions, and follow-up action suggestions. S12. Based on the acquired historical cases of nuclear power equipment fault diagnosis, entity recognition and relation extraction are performed according to the hierarchical relationship of the graph ontology to obtain diagnostic knowledge triples, and the diagnostic knowledge triples are fused. S13. Based on the obtained operation manuals and operation-related data of nuclear power plant operators, perform entity identification and relationship extraction to obtain operation and maintenance-related triples; S14. The diagnostic knowledge triplet and the operation and maintenance related triplet are fused together to link operation and maintenance knowledge and diagnostic knowledge. S15. Based on step S14, construct a knowledge graph based on the diagnostic knowledge triplet and the operation and maintenance related triplet, and store the knowledge graph in a graph database. The determination of the fault symptoms specifically includes the following steps: First, cluster the historical fault data and fault simulation data to determine the cluster centers and boundaries, thus obtaining the fault data clustering space. Then, divide the clustering results into the corresponding fault symptom types. Finally, map the monitored abnormal data into the fault data clustering space to obtain the corresponding fault symptoms. 2.The knowledge graph-based nuclear power secondary circuit equipment intelligent diagnosis method according to claim 1, characterized in that, In step S11, the key equipment of the nuclear power plant's secondary loop includes the turbine body, high-pressure heater, low-pressure heater, deaerator, high-pressure cylinder, low-pressure cylinder, and condenser.

3. The intelligent diagnostic method for nuclear power plant secondary loop equipment based on knowledge graphs according to claim 1, characterized in that, In step S12, the historical cases of nuclear power equipment fault diagnosis include equipment fault cases and equipment deterioration assessment reports.

4. The intelligent diagnostic method for nuclear power plant secondary loop equipment based on knowledge graphs according to claim 1, characterized in that, In step S14, during the fusion process, the entities in the diagnostic knowledge triplet and the operation and maintenance related triplet are subjected to referencing resolution and entity disambiguation.

5. A knowledge graph-based nuclear power secondary circuit equipment intelligent diagnosis system for performing the method of claim 1, characterized in that, include A knowledge graph construction module is used to construct a knowledge graph and store the knowledge graph in a graph database; The fault symptom determination module is used to read monitoring data of the nuclear power plant's secondary loop equipment, obtain corresponding time-series data, and analyze the time-series data. During the analysis process, it not only detects the time-series status information of a single measuring point value, but also makes a joint judgment based on multiple measuring points to determine whether the equipment has an abnormal state. If an abnormal state is found, fault symptoms are identified; if no fault is found, the time series data for the next time period is analyzed according to the preset time interval. The entity mapping module is used to perform entity mapping on the fault symptom set output by the fault symptom determination module and the devices with relevant fault symptoms to be diagnosed, thereby mapping the fault symptoms and devices to entities in the knowledge graph. The fault mode retrieval and reasoning module is used to retrieve possible fault modes based on the entities obtained from the corresponding modules of the entities, using Python language to link to the graph database, retrieve possible fault modes based on fault symptoms, and then perform reasoning to determine the fault mode. The fault cause and handling suggestion determination module is used to search the knowledge graph based on the fault mode output by the fault mode retrieval and reasoning module to find the corresponding fault cause, fault handling measures, troubleshooting suggestions, and follow-up action suggestions.

6. An electronic device, comprising: include: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-4.

7. A computer-readable storage medium storing computer instructions thereon, characterized in that, The computer instructions are used to cause the computer to perform the steps of the method as described in any one of claims 1-4.

Images