CN-122021830-A - Power equipment fault diagnosis and control strategy recommendation method based on knowledge graph

CN122021830ACN 122021830 ACN122021830 ACN 122021830ACN-122021830-A

Abstract

The invention discloses a power equipment fault diagnosis and control strategy recommendation method based on a knowledge graph, which relates to the technical field of power equipment diagnosis and comprises the steps of collecting power equipment operation parameters, defining node types and relation types of the knowledge graph, and constructing the power equipment fault diagnosis knowledge graph; the method comprises the steps of extracting node characteristics of the power equipment, taking the space distance of adjacent node vectors as a target, storing the obtained node vectors into a vector database to realize similarity query and semantic matching among the nodes, taking an alarm node as a starting point to carry out multi-hop reasoning in a knowledge graph, screening a high-confidence path as a diagnosis conclusion to generate a power equipment strategy recommendation list, evaluating the execution effect of the power equipment strategy recommendation list, adjusting the corresponding relation edge weight of the knowledge graph, dynamically updating the structure of the power equipment fault diagnosis knowledge graph and the node vectors, and realizing the self-adaptive optimization of the power equipment fault diagnosis knowledge graph. The invention improves the fault diagnosis capability of the power equipment.

Inventors

CHENG CHENG
YU DALIN
WANG JIECHANG
SHEN BO
Chen Zhannan
ZHANG CHE
ZHAO TIANYE
LIN ZIYANG
HAN WENYU
SU RUIZHI

Assignees

北京华能新锐控制技术有限公司
西安热工研究院有限公司
华能吉林发电有限公司新能源分公司

Dates

Publication Date: 20260512
Application Date: 20260203

Claims (10)

1. The power equipment fault diagnosis and control strategy recommendation method based on the knowledge graph is characterized by comprising the following steps of: S1, based on a power equipment multidimensional system, collecting power equipment operation parameters, alarm logs, maintenance records and target area meteorological data, defining node types and relation types of a knowledge graph, and constructing a power equipment fault diagnosis knowledge graph; S2, extracting node characteristics of the power equipment fault diagnosis knowledge graph based on the power equipment fault diagnosis knowledge graph, mapping the nodes into low-dimensional vectors, taking the optimization of the space distance of adjacent node vectors as a target, and storing the obtained node vectors into a vector database to realize similarity query and semantic matching between the nodes; S3, immediately triggering an alarm signal based on the monitoring system to detect the abnormality of the power equipment, performing multi-hop reasoning in a power equipment fault diagnosis knowledge graph by taking an alarm node as a starting point, screening a high-confidence path as a diagnosis conclusion, and generating a prioritized power equipment strategy recommendation list; and S4, based on the generation of the prioritized power equipment strategy recommendation list, evaluating the execution effect of the power equipment strategy recommendation list, adjusting the edge weight of the corresponding relation of the power equipment fault diagnosis knowledge graph, dynamically updating the structure of the power equipment fault diagnosis knowledge graph and the node vector, and realizing the self-adaptive optimization of the power equipment fault diagnosis knowledge graph.
2. The knowledge-graph-based power equipment fault diagnosis and control strategy recommendation method according to claim 1, wherein step S1 specifically comprises: Based on the power equipment monitoring system, acquiring power equipment operation parameters including current, voltage, temperature, oil chromatographic data and partial discharge signals in real time; Acquiring power equipment standing accounts, models, technical parameters and connection topology based on a power equipment management system; Acquiring a historical fault record, an overhaul report, test data and a replacement part record of the power equipment based on the power equipment maintenance recording system; acquiring meteorological data of a target area based on a weather forecast APP of the target area, wherein the meteorological data comprise temperature, humidity, wind speed, rainfall and lightning data; Integrating the data acquired by the power equipment monitoring system, the equipment management system, the maintenance recording system and the weather forecast APP, and preprocessing the data; Constructing a Nebula Graph database architecture by taking power equipment, components, symptoms, faults, strategies and meteorological data as entities and taking the types of directed relations among entity nodes as relations; The directed relation type among the entity nodes comprises the following steps: Component-belongs to-device, device-has-symptom, symptom-causes-faults, meteorological data-induces-faults, faults-repairs-strategies, symptom-concurrence-symptoms.
3. The knowledge-graph-based power equipment fault diagnosis and control strategy recommendation method according to claim 2, wherein step S1 further comprises: Acquiring structural data of the power equipment in real time based on a multidimensional system of the power equipment; Based on the power equipment management system and the maintenance recording system, acquiring the power equipment standing book, the model, the technical parameters, the connection topology, the history fault record, the maintenance report, the test data and the replacement part record text data in real time, extracting the power equipment text data by utilizing a APACHE TIKA tool, performing word segmentation and part-of-speech labeling on the power equipment text data, and identifying and extracting the entity and relation of the power equipment text data; Based on a weather forecast APP of a target area, a weather data API interface is called regularly, and an XML parsing technology is utilized to extract weather data fields of the target area and convert the weather data fields into a structured format; Based on the structural data, text data and meteorological data of the power equipment, extracting data entities and corresponding relations, designing entity-relation-entity triplets, importing a Nebula Graph database architecture, and constructing a power equipment fault diagnosis knowledge Graph.
4. The knowledge-graph-based power equipment fault diagnosis and control strategy recommendation method according to claim 3, wherein step S2 specifically comprises: Extracting all node information including node types, attributes, identifiers and states based on the power equipment fault diagnosis knowledge graph, and extracting all side information including side types, connected source nodes, side weights and relationship attributes; Unifying node types; the nodes are entities, and the edges are relations; and calculating the degree of each node in the power equipment fault diagnosis knowledge graph, establishing a node adjacency matrix, generating the shortest path information among the nodes, and obtaining the structural characteristics of the power equipment fault diagnosis knowledge graph.
5. The knowledge-graph-based power equipment fault diagnosis and control strategy recommendation method according to claim 4, wherein step S2 further comprises: dividing a power equipment fault diagnosis knowledge graph into a training set, a testing set and a verification set; Based on the power equipment fault diagnosis knowledge graph in the training set, training a GNN graph neural network model, learning each node characteristic representation, aggregating each node and neighbor node characteristics, mapping each node to a low-dimensional dense vector representation, and capturing structural information and semantic information of each layer of nodes in the power equipment fault diagnosis knowledge graph; continuously optimizing the learning rate and the training round number of the GNN graph neural network model by using a random gradient descent algorithm and taking the optimization of the space distance of the adjacent node vectors as a target; Monitoring a training process by using the power equipment fault diagnosis knowledge graph in the verification set, and adjusting the learning rate and the network layer number of the GNN graph neural network model to prevent overfitting; evaluating the trained GNN graph neural network model by using the power equipment fault diagnosis knowledge graph in the test set, and calculating similarity and recall index between nodes for quantitative evaluation; based on the trained GNN graph neural network model, correlating the paired power equipment fault diagnosis knowledge graphs, and generating a stability vector of all nodes reflecting semantic correlation; Based on generating stable vectors of all nodes reflecting semantic association, vector standardization processing is carried out, and a node vector database is established; Storing the node vectors to a Milvus vector database based on all the node vectors in the node vector database; and taking any node vector as input, carrying out k-nearest neighbor query on Milvus vector databases to obtain k most similar nodes and similarity scores in a vector space, and realizing similarity query and semantic matching among the nodes.
6. The knowledge-graph-based power equipment fault diagnosis and control strategy recommendation method according to claim 1, wherein step S3 specifically comprises: Based on a monitoring system, monitoring the operation parameters of the power equipment in real time, and selecting a section of stable operation parameters of the power equipment as an equipment operation health reference range by combining a known historical database of the power equipment; If the real-time operation parameters of the power equipment deviate from the equipment operation health reference range, judging that the power equipment is in an abnormal state at the moment, and immediately triggering an alarm signal; Analyzing the alarm signal through an API interface, extracting an entity and abnormal symptoms of the power equipment at the moment, combining a power equipment fault diagnosis knowledge graph, and positioning corresponding equipment nodes and symptom nodes of the power equipment at the moment to obtain a symptom node of the power equipment starting node at the moment, wherein the symptom node is the power equipment fault diagnosis knowledge graph; And taking a symptom node of the power equipment fault diagnosis knowledge graph as an initial node of the power equipment at the moment, traversing all dependent edges of the node at the moment in the power equipment fault diagnosis knowledge graph to carry out multi-hop reasoning, and gradually searching all fault nodes associated with the alarm abnormality to obtain an initial reasoning path set of the power equipment.
7. The knowledge-graph-based power equipment fault diagnosis and control strategy recommendation method according to claim 6, wherein step S3 further comprises: Calculating all edge weight products of each reasoning path based on an initial reasoning path set of the power equipment, introducing path occurrence frequency as a priori constraint to carry out score correction, and obtaining confidence scores of each reasoning path; screening the reasoning path with the highest confidence score as a main diagnosis result according to descending order to obtain a path list with the confidence score ordering; And aiming at each path with the confidence score, acquiring a control strategy node connected with the fault node, integrating the historical success rate and the current equipment state of the control strategy, calculating the execution feasibility score of each control strategy, and generating a prioritized power equipment strategy recommendation list.
8. The knowledge-graph-based power equipment fault diagnosis and control strategy recommendation method according to claim 7, wherein step S4 specifically comprises: Based on the generation of a prioritized power equipment strategy recommendation list, collecting result data of on-site execution of a power equipment control strategy, and evaluating the execution effect of the result data; Based on the evaluation of the execution effect of the control strategy of the power equipment, the side weight of the corresponding relation of the fault diagnosis knowledge graph of the power equipment is automatically adjusted.
9. The knowledge-graph-based power equipment fault diagnosis and control strategy recommendation method according to claim 8, wherein step S4 further comprises: and establishing an effect evaluation mechanism, quantitatively judging the success of each executed control strategy, marking positive feedback if the failure is successfully solved, improving the corresponding relation side weight of the power equipment failure diagnosis knowledge graph, marking negative feedback if the new problem is not solved or caused, and reducing the corresponding relation side weight of the power equipment failure diagnosis knowledge graph.
10. The knowledge-graph-based power equipment fault diagnosis and control strategy recommendation method according to claim 8, wherein step S4 further comprises: Aiming at the new power equipment fault type and the control strategy execution effect, adding the new power equipment fault type and the new control strategy execution effect as new nodes and relations into a power equipment fault diagnosis knowledge graph, and dynamically updating a power equipment fault diagnosis knowledge graph structure and a node vector representation; Based on the newly generated node vector, the Milvus vector database is synchronously updated, so that the self-adaptive optimization of the power equipment fault diagnosis knowledge graph is realized.

Description

Power equipment fault diagnosis and control strategy recommendation method based on knowledge graph Technical Field The invention relates to the technical field of power equipment diagnosis, in particular to a power equipment fault diagnosis and control strategy recommendation method based on a knowledge graph. Background In the prior art, the fault diagnosis and strategy recommendation of the power equipment are mostly dependent on expert experience libraries or static systems based on fixed rules, alarm information and historical data are processed in isolation generally, and effective fusion and utilization of deep semantic association among multi-source and heterogeneous data such as operation parameters, meteorological conditions and maintenance records are lacked, so that a diagnosis view angle is on one side, a system knowledge base is difficult to update once established, and cannot be independently learned from follow-up execution feedback, so that a recommendation strategy can be delayed or invalid, adaptability is poor, the reasoning capability of the traditional method is stiff, flexible and deep association reasoning is difficult to carry out when complex new faults or multiple faults are concurrent, and accuracy of diagnosis and accuracy of strategy recommendation are affected. Disclosure of Invention In order to solve the technical problems, a power equipment fault diagnosis and control strategy recommendation method based on a knowledge graph is provided. In order to achieve the above purpose, the invention adopts the following technical scheme: The power equipment fault diagnosis and control strategy recommendation method based on the knowledge graph comprises the following steps: S1, based on a power equipment multidimensional system, collecting power equipment operation parameters, alarm logs, maintenance records and target area meteorological data, defining node types and relation types of a knowledge graph, and constructing a power equipment fault diagnosis knowledge graph; S2, extracting node characteristics of the power equipment fault diagnosis knowledge graph based on the power equipment fault diagnosis knowledge graph, mapping the nodes into low-dimensional vectors, taking the optimization of the space distance of adjacent node vectors as a target, and storing the obtained node vectors into a vector database to realize similarity query and semantic matching between the nodes; S3, immediately triggering an alarm signal based on the monitoring system to detect the abnormality of the power equipment, performing multi-hop reasoning in a power equipment fault diagnosis knowledge graph by taking an alarm node as a starting point, screening a high-confidence path as a diagnosis conclusion, and generating a prioritized power equipment strategy recommendation list; and S4, based on the generation of the prioritized power equipment strategy recommendation list, evaluating the execution effect of the power equipment strategy recommendation list, adjusting the edge weight of the corresponding relation of the power equipment fault diagnosis knowledge graph, dynamically updating the structure of the power equipment fault diagnosis knowledge graph and the node vector, and realizing the self-adaptive optimization of the power equipment fault diagnosis knowledge graph. Preferably, step S1 specifically includes: Based on the power equipment monitoring system, acquiring power equipment operation parameters including current, voltage, temperature, oil chromatographic data and partial discharge signals in real time; Acquiring power equipment standing accounts, models, technical parameters and connection topology based on a power equipment management system; Acquiring a historical fault record, an overhaul report, test data and a replacement part record of the power equipment based on the power equipment maintenance recording system; acquiring meteorological data of a target area based on a weather forecast APP of the target area, wherein the meteorological data comprise temperature, humidity, wind speed, rainfall and lightning data; Integrating the data acquired by the power equipment monitoring system, the equipment management system, the maintenance recording system and the weather forecast APP, and preprocessing the data; Constructing a Nebula Graph database architecture by taking power equipment, components, symptoms, faults, strategies and meteorological data as entities and taking the types of directed relations among entity nodes as relations; The directed relation type among the entity nodes comprises the following steps: Component-belongs to-device, device-has-symptom, symptom-causes-faults, meteorological data-induces-faults, faults-repairs-strategies, symptom-concurrence-symptoms. Preferably, step S1 further includes: Acquiring structural data of the power equipment in real time based on a multidimensional system of the power equipment; Based on the power equipment management system and the