Operation and maintenance decision driving method and system based on multi-modal data knowledge graph

Abstract

The invention discloses an operation and maintenance decision driving method and system based on a multi-modal data knowledge graph, and relates to the field of intelligent decision support, and the method comprises the steps: obtaining the multi-source and multi-modal data of a power distribution network, and carrying out the preprocessing, and obtaining a knowledge graph database; constructing a knowledge graph based on the knowledge graph database; constructing a fault decision mode based on the knowledge graph; performing decision analysis and output based on the fault decision mode; according to the method, the utilization efficiency of data is improved, the data quality and relevance are improved, the fault type and the influence range thereof can be quickly identified, fault early warning is realized, personalized decision suggestions are provided, operation and maintenance personnel are helped to deal with power grid faults in time, and the fault processing efficiency and accuracy are improved; risks can be prevented in advance, resource allocation is optimized, and stable operation of a power grid is guaranteed; the automatic execution of the decision is realized, and the overall operation and maintenance efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent decision support, and in particular to an operation and maintenance decision-making driving method and system based on a multimodal data knowledge graph. Background Art

[0002] During the operation of the distribution network, the emergency response capability and efficiency of faults are related to the reliable operation of the power grid and the power consumption of users. Traditional distribution network operation and maintenance decision-making methods mostly rely on a single data source and manual experience query, which cannot fully reflect the complexity and dynamics of the distribution network system, and it is difficult to fully explore the value contained in massive data. Because the distribution network data has the characteristics of large data volume, complex types, diverse semantic relationships, and unclear sharing mechanisms, conventional data analysis methods are difficult to discover many factors related to faults, which restricts the level of operation and maintenance decision-making.

[0003] Knowledge graph technology provides a new way to manage massive data. Knowledge graph is a structured semantic knowledge base that stores and organizes complex relationships between entities in the form of graphs. The construction of distribution network fault knowledge graph mainly refers to extracting knowledge composed of entities, relationships and attributes from multi-source and multi-modal fault-related data such as equipment, meteorology, and geographic location, and using graph database and other technologies to achieve structured representation of fault knowledge, and then mining data value on this basis to provide efficient fault handling decisions for distribution network operation and management. At present, the construction of distribution network fault knowledge graph is still in the construction stage. Most of the research combines deep learning and artificial intelligence analysis technologies to perform data analysis. However, due to the large number of distribution network faults, relatively scarce data, and many and scattered related equipment, conventional deep learning models are not effective, and the amount of knowledge constructed is difficult to drive the operation and maintenance decisions of power grid faults. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is that the traditional distribution network operation and maintenance decision-making method cannot fully cope with the problems of system complexity and dynamics.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides an operation and maintenance decision-making driving method based on a multimodal data knowledge graph, comprising:

[0008] Obtain multi-source and multi-modal data of the distribution network and pre-process it to obtain a knowledge graph database;

[0009] Build a knowledge graph based on the knowledge graph database;

[0010] Build a fault decision model based on the knowledge graph;

[0011] Perform decision analysis and output based on fault decision mode.

[0012] As a preferred solution for the operation and maintenance decision-making driven method based on multimodal data knowledge graph, it:

[0013] The construction of the knowledge graph based on the knowledge graph database includes:

[0014] Identify entities with specific meanings from unstructured text and classify them into entity types corresponding to the knowledge graph;

[0015] Identify the relationships between entities and extract the corresponding triples.

[0016] As a preferred solution for the operation and maintenance decision-making driven method based on multimodal data knowledge graph, it:

[0017] The construction of the knowledge graph based on the knowledge graph database further includes:

[0018] Use label attributes to build graph data models;

[0019] Establish an overall framework for distribution network fault auxiliary decision-making, which includes data layer, knowledge graph layer, knowledge reasoning layer and human-computer interaction layer.

[0020] As a preferred solution for the operation and maintenance decision-making driven method based on multimodal data knowledge graph, it:

[0021] The fault decision model constructed based on the knowledge graph includes:

[0022] Match the question and answer statements, use labels to normalize the dispatcher's questions, and process them uniformly;

[0023] When a dispatch failure occurs, the fault information will be transmitted back to the dispatcher's monitoring system in real time, and the fault plan for the area will be automatically recommended based on the location of the fault, and related fault cases will be matched.

[0024] As a preferred solution for the operation and maintenance decision-making driven method based on multimodal data knowledge graph, it:

[0025] In the above, based on the knowledge graph, constructing the fault decision model also includes:

[0026] The real-time fault information including faulty equipment, location, status, and phenomenon is matched with the fault phenomena and fault cases in the fault dispatch knowledge graph that fits the memory network. Initial constraints are formed based on the fault information, and entity links are established. Based on subgraph retrieval and path association, fault case clusters and fault processing knowledge are associated, and historical fault information is counted. Knowledge search, calculation, and recommendation are achieved through path recall.

[0027] As a preferred solution for the operation and maintenance decision-making driven method based on multimodal data knowledge graph, it:

[0028] The decision analysis and output based on the fault decision mode includes:

[0029] Use graph neural network algorithms to analyze patterns and relationships in knowledge graphs and predict potential operation and maintenance issues;

[0030] For the time series information in the operation and maintenance data, use the time series analysis method to predict the future trend of system performance indicators;

[0031] Automatically generate decision recommendations based on predefined business rules.

[0032] As a preferred solution for the operation and maintenance decision-making driven method based on multimodal data knowledge graph, it:

[0033] The decision analysis and output based on the fault decision mode also includes:

[0034] Generate reports including system status overview, abnormal events, predicted trends, and recommended measures;

[0035] Monitor performance indicators in real time and trigger alarms when indicators exceed preset thresholds;

[0036] Integrate the decision support system with the operation and maintenance automation platform through API to realize automatic execution of decisions;

[0037] Feed the execution results back to the knowledge graph to optimize the decision model.

[0038] In a second aspect, an embodiment of the present invention provides an operation and maintenance decision-making driving system based on a multimodal data knowledge graph, including:

[0039] The data acquisition and preprocessing module is used to acquire and preprocess the multi-source and multi-modal data of the distribution network to obtain a knowledge graph database;

[0040] A knowledge graph database construction module is used to construct a knowledge graph based on the knowledge graph database;

[0041] Fault decision model building module, used to build fault decision models based on knowledge graph;

[0042] The decision analysis and output module is used to perform decision analysis and output based on the fault decision mode.

[0043] In a third aspect, an embodiment of the present invention provides a computing device, including:

[0044] Memory and processor;

[0045] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the one or more programs are executed by the one or more processors, the one or more processors implement the operation and maintenance decision-driven method based on a multimodal data knowledge graph as described in any embodiment of the present invention.

[0046] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the operation and maintenance decision-driven method based on a multimodal data knowledge graph.

[0047] Beneficial effects of the present invention: The present invention solves the problems of scattered data sources and diversified formats by acquiring and integrating various data types of the distribution network, so that data can be uniformly processed and analyzed, thereby improving the efficiency of data utilization; the constructed knowledge graph can be dynamically adjusted, which can not only realize the structured storage of data, but also improve the data quality and relevance through technical means such as entity recognition, relationship extraction and normalization processing, providing a solid foundation for subsequent decision analysis; the fault decision model constructed based on the knowledge graph can quickly identify the fault type and its impact range, realize fault warning, and provide personalized decision suggestions to help operation and maintenance personnel respond to power grid faults in a timely manner and improve the efficiency and accuracy of fault handling; through Combined with advanced technologies such as graph neural networks, the system can perform complex relationship analysis and pattern recognition, support real-time operation and maintenance decisions, predict potential problems, and automatically recommend the best solutions, which helps to prevent risks in advance, optimize resource allocation, and ensure stable operation of the power grid; a dispatch-assisted question-and-answer processing flow is designed to significantly lower the threshold for non-professionals to operate the graph database. At the same time, the decision support system is integrated with the existing operation and maintenance automation platform to realize the automated execution of decisions and improve the overall operation and maintenance efficiency; the entire system is designed with a feedback mechanism to feed back the latest execution results to the knowledge graph, forming a closed loop, ensuring that the decision model can be continuously optimized, adapt to changes in the power grid operating environment, and maintain the advancement and effectiveness of decision support. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0049] Figure 1 It is an overall flow chart of the operation and maintenance decision-making driving method based on multimodal data knowledge graph described in the present invention;

[0050] Figure 2It is a schematic diagram of the knowledge graph framework of the operation and maintenance decision-making driven method based on the multimodal data knowledge graph described in the present invention. DETAILED DESCRIPTION

[0051] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0052] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0053] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0054] Example 1

[0055] Reference Figure 1 , which is the first embodiment of the present invention, and provides an operation and maintenance decision-making driving method based on a multimodal data knowledge graph, comprising:

[0056] S1: Obtain multi-source and multi-modal data of the distribution network and pre-process it to obtain a knowledge graph database;

[0057] S2: Build a knowledge graph based on the knowledge graph database;

[0058] S3: Construct fault decision model based on knowledge graph;

[0059] S4: Perform decision analysis and output based on the fault decision model.

[0060] It should be noted that, through steps S1-S4, the emergency decision support technology driven by multimodal data and knowledge fusion integrates discrete and diversely formatted fault information into a unified knowledge representation, formalizes the fault-related entities and the relationships between them, thereby constructing a structured representation of fault knowledge, realizing structured storage, dynamic adjustment and data empowerment of data, and then carrying out research on distribution network fault warning technology, and realizing personalized decision-making recommendations, providing targeted decision support for regional grid distribution faults.

[0061] Example 2

[0062] Reference Figure 1-Figure 2 , is an embodiment of the present invention, and provides an operation and maintenance decision-making driving method based on a multimodal data knowledge graph based on the previous embodiment, including:

[0063] In the embodiment of the present application, in the above step S1, multi-source and multi-modal data of the distribution network is obtained and pre-processed to obtain a knowledge graph database including:

[0064] The multi-source and multi-modal data of the distribution network includes structured, semi-structured, and unstructured data, covering many factors related to distribution network failures;

[0065] Among them, structured data include distribution network operation data (voltage, current, power, etc.), equipment status data (operating status, fault records, maintenance logs, etc.), user power consumption data (power load, power consumption mode, power consumption history, etc.), geographic information system data (distribution network geographical location, line layout, substation location, etc.); semi-structured data include log files (system operation logs, fault logs, maintenance logs, etc.), XML / JSON data (equipment communication protocols, sensor data, etc.), statistical data or historical records, etc.; unstructured data include text data (technical documents, operating manuals, accident reports, maintenance guides, etc.), image data (equipment photos, on-site images, satellite images, etc.), audio data (voice call records, meeting recordings, etc.), video data (surveillance videos, on-site operation videos), etc.

[0066] The collected structured / semi-structured data is cleaned, deduplicated, formatted and other processing procedures are carried out to ensure the quality and consistency of the data. At the same time, unstructured data such as text is subjected to structural conversion, word segmentation, labeling, word vector construction and other operations to form structured data.

[0067] Specifically, data preprocessing mainly includes the following steps:

[0068] a) Missing value processing:

[0069] Identify missing values ​​in the data, delete non-fault-related records based on correlation analysis, and fill in missing values ​​using a spatiotemporal data interpolation method based on tensor decomposition. For example, for missing values ​​x ij , whose fill value is:

[0070]

[0071] in, is the core tensor, a i 、b j and c are spatial, temporal and characteristic pattern factors respectively.

[0072] This method can simultaneously consider the spatial and temporal correlation of the data and provide more accurate missing value estimates.

[0073] b) Outlier detection and removal

[0074] The improved local outlier factor (LOF) algorithm is used to identify and remove outlier data points. This method is particularly suitable for outlier detection in high-dimensional data. For feature j, the criterion for determining outliers is:

[0075]

[0076] in, and They represent the mean and standard deviation of the LOF value of feature j respectively.

[0077] This method can effectively identify local outliers and improve the reliability of data. At the same time, the kd tree algorithm is used to accelerate the calculation process of LOF to process large-scale data sets.

[0078] c) Identification and deletion of duplicate data

[0079] Checks for duplicate records in a dataset and removes or merges duplicates to avoid data redundancy.

[0080] d) Data type conversion and format unification

[0081] Convert data into a format suitable for analysis, such as converting strings into dates and converting categorical data into numeric data. Use label encoding to convert categorical features into numeric data.

[0082] e) Data normalization:

[0083] Using the improved Min-Max normalization method, nonlinear transformation is introduced to enhance the distinguishability of features:

[0084]

[0085] Among them, α is an adjustable nonlinear factor (usually set between 0.1 and 10), μ j and σ j are the mean and standard deviation of feature j respectively.

[0086] This nonlinear normalization method can better handle features of different scales and distributions, especially for features with long-tail distribution.

[0087] f) Data consistency check: ensure the logical consistency of data.

[0088] In another possible implementation, more types of sensors, such as temperature sensors, humidity sensors, vibration sensors, etc., may be added when acquiring multi-source and multi-modal data of the distribution network to comprehensively monitor the status of the equipment.

[0089] It is also possible to integrate external data sources such as meteorological data, grid policy changes, and market data to enrich the background information of the knowledge graph.

[0090] In the embodiment of the present application, in the above step S2, based on the knowledge graph database, constructing the knowledge graph includes:

[0091] Entity recognition and linking: Entity recognition and linking is to identify entities with specific meanings from unstructured text and classify them into entity types corresponding to the knowledge graph. The specific steps are as follows:

[0092] a) Determine the entity type: identify the fault type, fault time, fault response strategy, weather conditions, power supply station, geographical location, etc.

[0093] b) Data preparation: Select sample data and annotate them to form training data, including entity labels and entity boundaries

[0094] c) Feature extraction: The Transformer model is used to extract features from text data, including lexical features, syntactic features, semantic features, etc.; the principal component analysis method is used to extract features from non-text data, including fault type, occurrence time, location, maintenance plan, power outage time, weather, etc.

[0095] d) Model training: The model is trained using the conditional random field (CRF) machine learning algorithm, and the model parameters are adjusted using the optimization algorithm.

[0096] e) Entity linking: The similarity between entities is calculated through Euclidean distance, and the most similar entity is selected as the link target.

[0097] f) Post-processing: Post-processing steps such as standardization of identified entities and removal of stop words.

[0098] Relation extraction and normalization: Identify the relationship between entities and extract the corresponding triples using the End-to-End method;

[0099] Exemplarily, the entity relationships mainly include <fault-impact-weather>, <fault-impact-region>, <region-belongs to-power supply station>, <fault-handling-fault response strategy>, etc. The internship attributes take the fault entity as an example, including the fault occurrence time, recovery time, fault description, etc.; and the extracted relationships are normalized to eliminate the redundancy caused by different expressions.

[0100] Graph data model construction: The graph data model is constructed using Neo4j-based label attributes. From a macro perspective, it consists of two types of data: nodes and relationships. Nodes store entity information, relationships link entities, and the properties and labels of nodes and relationships are stored in the form of key-value pairs.

[0101] Entity=Node:Label{Property1,Property2···}

[0102] Relationship=start node-[rel:Property3]-end node

[0103] In the formula: Label, Property1, Property2 are the labels and multiple attribute values ​​of the entity. Labels are used to distinguish different types of nodes. Each entity can contain one or more labels and zero or more attributes. Startnode and endnode represent the head entity and the tail entity, rel represents the relationship name, Property3 represents the relationship attribute, and the relationship between entities is directional.

[0104] Framework design and implementation: The knowledge graph design should consider the semantic model and business model in combination with the scenario. The overall framework of distribution network fault auxiliary decision-making designed in this paper is as follows: Figure 2 As shown, it includes data layer, knowledge graph layer, knowledge reasoning layer and human-computer interaction layer.

[0105] The data layer unifies the physical model and knowledge model required in the fault dispatch decision of the distribution network, establishes the relationship link between multi-source data through the knowledge graph layer, and forms a knowledge graph for fault dispatch auxiliary decision. On the basis of constructing the knowledge graph, the knowledge reasoning layer conducts data logic association mining in combination with the business scenario, analyzes and calculates by methods such as key information matching, topological path search, and data mining statistics, and realizes the exploration of power knowledge application in refined scenarios. According to the design content of distribution network fault handling and the actual business processing needs of the control personnel, the present invention divides the knowledge graph of distribution network fault dispatch into an entity part and an event part. The entity part includes conceptual entities such as dispatching procedures and safety regulations, as well as physical entities such as power grid topology, equipment, and department personnel information; the event part includes fault handling, business logic, fault plans, and historical fault records. The entity part of fault dispatch knowledge mainly exists in documents such as dispatching procedures and safety regulations and the experience of dispatching experts, and the event part exists in the power grid structure and historical data texts. In view of the unstructured source data of the scheduling knowledge graph, the complex knowledge associations, and the need to meet the characteristics of high graph accuracy and good visualization effects, a combination of "bottom-up" and "top-down" methods is adopted to create it, and an artificial construction method based on the combination of key information extraction and expert experience is selected.

[0106] In another possible implementation, the entity recognition and linking stage can use pre-trained models such as BERT and RoBERTa to perform entity recognition to improve the recognition accuracy.

[0107] The relationship extraction stage can use a graph neural network (GNN)-based method: use graph neural network models, such as GCN, GAT, etc., to perform relationship extraction, considering the graph structure information between entities.

[0108] When building a knowledge graph, you can use distributed graph databases such as JanusGraph, TigerGraph, etc. to process large-scale knowledge graph data and improve query performance.

[0109] Represent the entities and relationships in the knowledge graph as low-dimensional vectors, using methods such as TransE and DistMult to facilitate subsequent graph analysis and recommendation tasks.

[0110] In the embodiment of the present application, in the above step S3, constructing a fault decision model based on the knowledge graph includes:

[0111] It should be noted that, similar to the MySQL language used in relational databases, the Neo4j database uses the Cypher graph database query language, which is a declarative graph data query language with rich expressiveness and high query efficiency. However, for scheduling staff to manage graph databases, learning Cypher requires a database language foundation and a lot of time and energy, and is inefficient for real-time decision-making operations. In order to meet development and professional operation needs, the present invention combines business needs and query question-and-answer habits, and designs a scheduling auxiliary question-and-answer processing flow based on AIML.

[0112] Specifically, by matching question and answer statements, the ambiguity of the dialogue and the ambiguity of the input text are considered in the process, and the dispatcher's questions are normalized by using labels, and then processed uniformly; for problems such as "220 substation" and "load adjustment" where the units are unclear and the terms are not standardized, a common ambiguous dictionary is established to ask and answer again to reduce ambiguity and standardize the terms. After matching, in the response process, the variables are stored by matching labels, and the custom function such as "shortest_path" is activated with the wildcard "#". Cypher operation statements are generated according to the label parameter transmission, and the results are output by using text and voice with the Neo4j interactive interface, so as to finally realize the intelligent retrieval and answering of the dispatcher and the dispatch knowledge graph, which greatly reduces the threshold of graph database operation and improves work efficiency. In addition, the question and answer module can be called at any time during the fault dispatch process to complete the retrieval of dispatch knowledge to assist in completing the fault handling task. When a dispatch fault occurs, the fault information is transmitted back to the dispatcher monitoring system in real time. According to the location of the fault, the fault plan for the area can be automatically recommended, and the relevant fault cases can be matched. At the same time, it fits human thinking and working mode, matches the real-time fault information including faulty equipment, location, status, phenomenon with the fault phenomenon and fault case in the fault dispatch knowledge graph that fits the memory network, from three aspects: the same equipment fault case cluster, the same equipment fault case cluster, and the same equipment fault handling principle. According to the fault information, the initial constraints are formed, the entity link is established, the fault case cluster and fault handling knowledge are associated based on subgraph retrieval and path, the historical fault information is counted, and the knowledge search, calculation and recommendation are further realized through path recall. Suppose there are m types of faults in the same type of equipment E (e1, e2, ..., en), and the set of all possible fault phenomena is set to A. Then, all m types of faults correspond to the experienced fault phenomenon subset Am{a1, a2, ..., am}, and each fault case cluster C corresponds to the fault phenomenon subset Ca. On the basis of constructing the fault case function module, the automatic generation, storage and display of case event clusters can be realized. The Neo4j front-end interface can provide a visual display of the fault case cluster, provide more fault information, and help to find the weak points of the power grid.

[0113] In another possible implementation, a model such as a convolutional neural network (CNN) or a recurrent neural network (RNN) may be used to learn failure modes from historical cases and recommend similar cases.

[0114] The graph neural network (GNN) model is used to extract relevant entities and relationships from the knowledge graph and recommend similar fault cases.

[0115] In the embodiment of the present application, the decision analysis and output based on the fault decision mode in the above step S4 includes:

[0116] Use the structured knowledge and reasoning capabilities stored in the graph to support complex operation and maintenance decisions, and use graph neural network algorithms to analyze patterns and relationships in the knowledge graph to predict potential operation and maintenance problems;

[0117] For the time series information in the operation and maintenance data, the time series analysis method is used to predict the future trend of the system performance indicators.

[0118] Identify behaviors that differ significantly from normal patterns and quickly respond to potential system anomalies or security threats.

[0119] Based on the structure and node attributes of the distribution network knowledge graph, the similarity between fault entities is calculated to find other faults similar to the target fault entity.

[0120] Automatically generate decision recommendations based on predefined business rules.

[0121] Find the best balance between multiple conflicting objectives, such as cost, efficiency and service quality.

[0122] Evaluate the impact of different decision options by simulating possible outcomes before implementing them.

[0123] Decision output is the process of converting analytical results into specific actions, including generating reports, triggering alarms, and integrating into automation systems.

[0124] The generated report includes system status overview, abnormal events, predicted trends, recommended measures, etc. The decision analysis results are intuitively displayed in the form of charts, dashboards, etc., which is easy to understand and operate.

[0125] Key performance indicators are monitored in real time and alerts are triggered immediately when anomalies are detected.

[0126] For common problems, automatically execute preset repair scripts or adjust configurations.

[0127] Integrate the decision support system with the existing operation and maintenance automation platform through API to realize automatic execution of decisions.

[0128] Feed the execution results back to the knowledge graph to continuously optimize the decision-making model.

[0129] Example 3

[0130] The above is a schematic scheme of the operation and maintenance decision-driven method based on the multimodal data knowledge graph of this embodiment. It should be noted that the technical scheme of the operation and maintenance decision-driven system based on the multimodal data knowledge graph and the technical scheme of the operation and maintenance decision-driven method based on the multimodal data knowledge graph above belong to the same concept. The details of the technical scheme of the operation and maintenance decision-driven system based on the multimodal data knowledge graph in this embodiment that are not described in detail can all be referred to the description of the technical scheme of the operation and maintenance decision-driven method based on the multimodal data knowledge graph above.

[0131] This embodiment also provides a system for an operation and maintenance decision-making driving method based on a multimodal data knowledge graph, including:

[0132] The data acquisition and preprocessing module is used to acquire and preprocess the multi-source and multi-modal data of the distribution network to obtain a knowledge graph database;

[0133] A knowledge graph database construction module is used to construct a knowledge graph based on the knowledge graph database;

[0134] Fault decision model building module, used to build fault decision models based on knowledge graph;

[0135] The decision analysis and output module is used to perform decision analysis and output based on the fault decision mode.

[0136] This embodiment further provides a computing device, which is applicable to the case of an operation and maintenance decision-driving method based on a multimodal data knowledge graph, including:

[0137] Memory and processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the operation and maintenance decision-driven method based on multimodal data knowledge graph as proposed in the above embodiment.

[0138] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the operation and maintenance decision-making driven method based on a multimodal data knowledge graph as proposed in the above embodiment is implemented.

[0139] The storage medium proposed in this embodiment and the operation and maintenance decision-driven method based on multimodal data knowledge graph proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0140] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An operation and maintenance decision-making method based on a multimodal data knowledge graph, characterized in that: include: Obtain multi-source and multi-modal data of the distribution network and pre-process it to obtain a knowledge graph database; Build a knowledge graph based on the knowledge graph database; Build a fault decision model based on the knowledge graph; Perform decision analysis and output based on fault decision mode.

2. The operation and maintenance decision-making driving method based on multimodal data knowledge graph according to claim 1, characterized in that: The construction of the knowledge graph based on the knowledge graph database includes: Identify entities with specific meanings from unstructured text and classify them into entity types corresponding to the knowledge graph; Identify the relationships between entities and extract the corresponding triples.

3. The operation and maintenance decision-making driving method based on multimodal data knowledge graph according to claim 2 is characterized in that: The construction of the knowledge graph based on the knowledge graph database further includes: Use label attributes to build graph data models; Establish an overall framework for distribution network fault auxiliary decision-making, which includes data layer, knowledge graph layer, knowledge reasoning layer and human-computer interaction layer.

4. The operation and maintenance decision-making driving method based on multimodal data knowledge graph according to claim 3 is characterized in that: The fault decision model constructed based on the knowledge graph includes: Match the question and answer statements, use labels to normalize the dispatcher's questions, and process them uniformly; When a dispatch failure occurs, the fault information will be transmitted back to the dispatcher's monitoring system in real time, and the fault plan for the area will be automatically recommended based on the location of the fault, and related fault cases will be matched.

5. The operation and maintenance decision-making driving method based on multimodal data knowledge graph according to claim 4, characterized in that: In the above, based on the knowledge graph, constructing the fault decision model also includes: The real-time fault information including faulty equipment, location, status, and phenomenon is matched with the fault phenomena and fault cases in the fault dispatch knowledge graph that fits the memory network. Initial constraints are formed based on the fault information, and entity links are established. Based on subgraph retrieval and path association, fault case clusters and fault processing knowledge are associated, and historical fault information is counted. Knowledge search, calculation, and recommendation are achieved through path recall.

6. The operation and maintenance decision-making driving method based on multimodal data knowledge graph according to claim 5, characterized in that: The decision analysis and output based on the fault decision mode includes: Use graph neural network algorithms to analyze patterns and relationships in knowledge graphs and predict potential operation and maintenance issues; For the time series information in the operation and maintenance data, use the time series analysis method to predict the future trend of system performance indicators; Automatically generate decision recommendations based on predefined business rules.

7. The operation and maintenance decision-making driving method based on multimodal data knowledge graph according to claim 6, characterized in that: The decision analysis and output based on the fault decision mode also includes: Generate reports including system status overview, abnormal events, predicted trends, and recommended measures; Monitor performance indicators in real time and trigger alarms when indicators exceed preset thresholds; Integrate the decision support system with the operation and maintenance automation platform through API to realize automatic execution of decisions; Feed the execution results back to the knowledge graph to optimize the decision model.

8. A system using the operation and maintenance decision-making driving method based on multimodal data knowledge graph as described in any one of claims 1 to 7, characterized in that: include: The data acquisition and preprocessing module is used to acquire and preprocess the multi-source and multi-modal data of the distribution network to obtain a knowledge graph database; A knowledge graph database construction module is used to construct a knowledge graph based on the knowledge graph database; Fault decision model building module, used to build fault decision models based on knowledge graph; The decision analysis and output module is used to perform decision analysis and output based on the fault decision mode.

9. A computing device comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the operation and maintenance decision-driven method based on a multimodal data knowledge graph as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the operation and maintenance decision-driven method based on a multimodal data knowledge graph as described in any one of claims 1 to 7.

Images