A method and device for electric power construction safety control based on knowledge graph
Through the safety control method of power construction based on knowledge graphs, the knowledge graph is dynamically updated with historical and real-time data to identify risk points at the construction site, the problems of blind spots and inaccurate risk prediction in traditional safety management methods are solved, and the construction site is improved Safety management efficiency.
Patent Information
- Application Number
- CN202411173704.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-26
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2044-08-26
AI Technical Summary
The traditional power construction safety management method relies on manual supervision, which has blind spots in supervision and is difficult to detect and deal with safety hazards in a timely manner, resulting in inaccurate risk prediction on construction sites and low system efficiency.
Adopt the safety control method of power construction based on knowledge graphs, pre-process by obtaining historical construction data, establishing initial knowledge graphs, real-time acquisition of construction site data, determining potential risk relationship sets, dynamically updating knowledge graphs, identifying risk event sets, and formulating safety control plan.
It realizes timely reflection of changes in the construction site, improves the accuracy and system efficiency of risk prediction, provides scientific decision-making support, and helps managers effectively deal with safety issues.
Smart Images

Figure CN118691096B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of knowledge graph construction, and specifically relates to a method and device for electric power construction safety management and control based on knowledge graph. Background Art
[0002] At present, the safety of construction sites is particularly important in power construction. As the scale and complexity of power engineering projects continue to increase, the safety risks in the construction process also increase. Traditional safety management methods rely on manual supervision and inspection, which not only consumes a lot of manpower and material resources, but also has blind spots in supervision, making it difficult to discover and deal with safety hazards in a timely manner, resulting in the inability to effectively ensure construction safety. Therefore, there is an urgent need for a new safety management method to improve the safety management level of construction sites. Knowledge graphs, as an emerging data analysis and management technology, can intuitively display complex data associations by constructing a network of entities and their relationships.
[0003] Patent application document CN116186282A discloses a method, device and non-volatile storage medium for safety detection of electric power construction behavior. A pre-trained target detection model is used to identify target electric power construction behavior in on-site construction behavior images to obtain behavior recognition results. The above behavior recognition results are matched with the above electric power construction behavior safety knowledge graph, and the safety detection results of the above target electric power construction behavior are determined according to the matching results. This solves the technical problems of high detection cost and poor real-time performance caused by the manual judgment of the safety detection results of electric power construction behavior in related technologies. However, it still cannot reflect changes in the construction site in a timely manner, resulting in inaccurate risk prediction of the construction site and low system efficiency. Summary of the invention
[0004] The purpose of the present invention is to solve the problem that changes in the construction site cannot be reflected in time, resulting in inaccurate risk prediction at the construction site and low system efficiency, and to propose a method and device for power construction safety management based on knowledge graph.
[0005] In the first aspect of the implementation of the present invention, a method for safety management and control of electric power construction based on knowledge graph is first proposed, and the method comprises:
[0006] A method for electric power construction safety management and control based on knowledge graph, characterized in that the method comprises:
[0007] Acquire historical construction data, preprocess the historical construction data to obtain an initial data set, and establish a knowledge graph based on the initial data set to obtain an initial knowledge graph;
[0008] Acquire the operation and construction data of the construction area in real time, and determine the potential risk relationship set according to the operation and construction data; obtain multiple head entities and query relationships of the head entities according to the operation and construction data and the initial knowledge graph;
[0009] For each head entity and the query relationship between the head entities, determine the target subgraph corresponding to the head entity according to the potential risk relationship set;
[0010] A risk event set in which risks may occur in the construction area is determined based on all target subgraphs and corresponding historical construction data, and a safety management and control plan is determined based on the risk event set.
[0011] Optionally, establishing a knowledge graph according to the initial data set to obtain an initial knowledge graph includes:
[0012] Identifying and extracting key entities from the initial data set by natural language processing technology, classifying the key entities, and obtaining relevant entities and irrelevant entities;
[0013] Extracting the relationships between the related entities from the initial data set through deep learning and a standard template, and defining different types of relationship sets; the relationship sets store the relationships between the related entities;
[0014] After normalizing entities, attributes, and attribute values with the same meaning but different character strings according to preset rules and the random forest algorithm, the knowledge in different data sources is integrated to obtain the target knowledge base;
[0015] For the target knowledge base, constructing a relation triple according to each relevant entity and the relation between the relevant entities;
[0016] Substitute all relationship triplets into the graph database to obtain the knowledge graph.
[0017] Optionally, obtaining a plurality of head entities and query relationships between head entities according to the operation construction data and the initial knowledge graph includes:
[0018] Cleaning the operation construction data to obtain a construction data set, and integrating data from different sources in the construction data set to obtain a target data view;
[0019] Performing natural language processing on the target data view to identify head entities to obtain a head entity set;
[0020] For the head entity set, the relationship triple corresponding to the head entity is obtained to obtain the query relationship corresponding to the head entity.
[0021] Optionally, for each head entity and the query relationship between the head entities, determining the target subgraph corresponding to the head entity according to the potential risk relationship set includes:
[0022] Step 1: Obtain a first subgraph according to the head entity, the query relationship of the head entity and the initial knowledge graph;
[0023] Step 2: for the first subgraph, according to the potential risk relationship set, filter out neighbor entities related to the head entity to obtain a target neighbor entity set;
[0024] Step 3: For the target neighbor entity set, the attention coefficient and attention value of the connection relationship between each neighbor entity and the head entity are calculated through the dynamic subgraph attention mechanism;
[0025] Step 4: The neighbor entities whose attention coefficients and attention values are both greater than the preset thresholds are regarded as neighbor influencing entities, and all neighbor influencing entities are aggregated to obtain the target influencing entity;
[0026] Step 5: embed the target impact entity into the initial knowledge graph to obtain a target knowledge graph, and obtain the query relationship corresponding to the target impact entity in the target knowledge graph to obtain a target query relationship;
[0027] Step 6: Repeat the above steps 1-5 with the target impact entity as the head entity and the query relationship as the query relationship of the head entity until the preset conditions are met to obtain the target subgraph.
[0028] Optionally, the risk event set for determining the risk of the construction area according to all target subgraphs and corresponding historical construction data includes:
[0029] The target data and the historical construction data are similarly matched by similarity calculation to determine the target historical records; the similarity matching is to determine the historical records whose similarity with the target subgraph is greater than a preset value by a similarity calculation method; the target data is the query relationship corresponding to the entities and entities in all the target subgraphs;
[0030] Determine the potential risk factors in all target subgraphs, and count the number of occurrences of each risk event in the target historical records;
[0031] According to the number of occurrences of each risk event, the probability of each potential risk factor leading to a risk event under current conditions is calculated using the Bayesian theorem, and the risk event set in which risks occur in the construction area is determined based on the probability of each risk event.
[0032] In a second aspect of the present invention, a knowledge graph-based power construction safety management and control device is proposed, comprising:
[0033] An initial knowledge graph establishment module is used to obtain historical construction data, pre-process the historical construction data to obtain an initial data set, and establish a knowledge graph based on the initial data set to obtain an initial knowledge graph;
[0034] The operation and construction data analysis module is used to obtain the operation and construction data of the construction area in real time, determine the potential risk relationship set according to the operation and construction data; obtain multiple head entities and query relationships of the head entities according to the operation and construction data and the initial knowledge graph;
[0035] A target subgraph establishment module is used to determine the target subgraph corresponding to each head entity and the query relationship between the head entities according to the potential risk relationship set;
[0036] The safety management and control module is used to determine the risk event set of the construction area according to all target sub-graphs and corresponding historical construction data, and determine the safety management and control plan according to the risk event set.
[0037] Optionally, the initial knowledge graph building module includes:
[0038] A key entity classification module, used to identify and extract key entities from the initial data set through natural language processing technology, classify the key entities, and obtain relevant entities and irrelevant entities;
[0039] A relationship set determination module, used to extract the relationship between the related entities from the initial data set through deep learning and a standard template, and define different types of relationship sets; the relationship set stores the relationship between the related entities;
[0040] The knowledge fusion module is used to normalize entities, attributes, and attribute values with the same meaning but different character strings according to preset rules and the random forest algorithm, and then fuse the knowledge in different data sources to obtain the target knowledge base;
[0041] A relation triplet construction module, used for constructing a relation triplet according to each relevant entity and the relationship between the relevant entities for the target knowledge base;
[0042] The knowledge graph generation module is used to substitute all relationship triples into the graph database to obtain the knowledge graph.
[0043] Optionally, the operation construction data method analysis module includes:
[0044] A target data view generation module, used for performing data cleaning on the operation construction data to obtain a construction data set, and integrating data from different sources in the construction data set to obtain a target data view;
[0045] A header entity set generation module, configured to perform natural language processing on the target data view to identify the header entity and obtain a header entity set;
[0046] The query relationship generation module is used to obtain the relationship triplet corresponding to the head entity for the head entity set to obtain the query relationship corresponding to the head entity.
[0047] Optionally, the target subgraph establishing module includes:
[0048] A first subgraph generating module, configured to obtain a first subgraph according to the head entity, the query relationship of the head entity and the initial knowledge graph;
[0049] A target neighbor entity set generation module is used to filter out neighbor entities related to the head entity according to the potential risk relationship set for the first subgraph to obtain a target neighbor entity set;
[0050] An entity relationship generation module is used to calculate the attention coefficient and attention value of the connection relationship between each neighbor entity and the head entity through a dynamic subgraph attention mechanism for the target neighbor entity set;
[0051] A target influence entity generation module is used to take the neighbor entities whose attention coefficients and attention values are both greater than a preset threshold as neighbor influence entities, and aggregate all neighbor influence entities to obtain the target influence entity;
[0052] A target query relationship generation module, used to embed the target impact entity into the initial knowledge graph to obtain a target knowledge graph, and obtain the query relationship corresponding to the target impact entity in the target knowledge graph to obtain a target query relationship;
[0053] The target subgraph generation module is used to return the query relationship with the target influencing entity as the head entity and the query relationship as the head entity to the first subgraph generation module until a preset condition is met to obtain the target subgraph.
[0054] Optionally, the security management and control module includes:
[0055] A target history record generation module is used to perform similarity matching between target data and historical construction data through similarity calculation to determine target history records; the similarity matching is to determine the history records with similarity greater than a preset value with the target subgraph through a similarity calculation method; the target data is the query relationships corresponding to entities and entities in all target subgraphs;
[0056] A risk event statistics module is used to determine the potential risk factors in all target subgraphs and count the number of occurrences of each risk event in the target historical records;
[0057] The risk event set generation module is used to calculate the probability of each potential risk factor causing a risk event under current conditions through Bayesian theorem according to the number of times each risk event occurs, and determine the risk event set of the construction area according to the probability of each risk event.
[0058] Beneficial effects of the present invention:
[0059] The present invention proposes a method for safety management and control of electric power construction based on a knowledge graph, which comprises the following steps: obtaining historical construction data, preprocessing the historical construction data to obtain an initial data set, and establishing a knowledge graph based on the initial data set to obtain an initial knowledge graph; obtaining on-site operation construction data in a construction area in real time, and determining a potential risk relationship set based on the operation construction data; obtaining a plurality of head entities and query relationships between the head entities based on the operation construction data and the initial knowledge graph; determining a target subgraph corresponding to each head entity and the query relationship between the head entities based on the potential risk relationship set; determining a risk event set in which risks occur in the construction area based on all target subgraphs and the corresponding historical construction data, and determining a safety management and control plan based on the risk event set. By acquiring the operation data of the construction site in real time and combining it with historical data, the knowledge graph is dynamically updated. This enables the system to promptly reflect changes on site and make risk predictions. With real-time data and the initial knowledge graph, the system can identify the query relationship between the potential risk relationship set and the head entity, and generate a target subgraph. This can more accurately identify the specific risk points on the construction site, and then conduct risk assessment by combining historical data and real-time data. The system can generate a set of risk events and a safety management and control plan, provide scientific decision-making support, and help managers more effectively deal with safety issues on the construction site. This allows the system to promptly reflect changes on the construction site, thereby making risk predictions and improving system efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0060] The present invention will be further described below in conjunction with the accompanying drawings.
[0061] Figure 1 A flow chart of a method for safety management and control of electric power construction based on a knowledge graph is provided for an embodiment of the present invention;
[0062] Figure 2 A structural schematic diagram of a power construction safety management and control device based on a knowledge graph is provided for an embodiment of the present invention. DETAILED DESCRIPTION
[0063] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative work shall fall within the scope of protection of the present invention.
[0064] The embodiment of the present invention provides a method for power construction safety management and control based on knowledge graph. Figure 1 , Figure 1 A flowchart of a method for power construction safety management and control based on a knowledge graph is provided in an embodiment of the present invention. The method comprises the following steps:
[0065] S101, obtaining historical construction data, preprocessing the historical construction data to obtain an initial data set, and establishing a knowledge graph based on the initial data set to obtain an initial knowledge graph.
[0066] S102, obtaining the on-site operation and construction data of the construction area in real time, determining a potential risk relationship set based on the operation and construction data; and obtaining multiple head entities and query relationships between the head entities based on the operation and construction data and the initial knowledge graph.
[0067] S103: For each head entity and the query relationship between the head entities, determine the target subgraph corresponding to the head entity according to the potential risk relationship set.
[0068] S104, determining a risk event set of risks occurring in the construction area based on all target subgraphs and corresponding historical construction data, and determining a safety management and control plan based on the risk event set.
[0069] A knowledge graph-based power construction safety management and control method is provided based on an embodiment of the present invention. By acquiring the operation data of the construction site in real time and combining it with historical data, the knowledge graph is dynamically updated. This enables the system to timely reflect the changes on the site and perform risk prediction. Based on the real-time data and the initial knowledge graph, the system can identify the query relationship between the potential risk relationship set and the head entity, and generate a target subgraph. This can more accurately identify the specific risk points of the construction site, and then conduct risk assessment by combining historical data and real-time data. The system can generate a risk event set and a safety management and control plan, provide scientific decision-making support, and help managers more effectively deal with safety issues on the construction site, so that the system can timely reflect the changes on the construction site, thereby performing risk prediction and improving system efficiency.
[0070] In one implementation, the historical construction data includes project progress data, construction process record data, equipment usage data, personnel configuration data, environmental data, construction site photos and video data, etc. during the historical power construction process.
[0071] In one implementation, preprocessing includes error correction, missing value processing, data filtering, and data aggregation.
[0072] In one implementation, the operation construction data includes real-time progress data of the current construction, equipment status data, personnel activity data, environmental monitoring data, safety monitoring data, material usage data, construction equipment sensor data, construction operation data, construction site communication data, etc.
[0073] In one implementation, a potential risk relationship set is determined based on operation construction data by acquiring operation construction data from the construction site in real time, integrating the real-time data with historical construction data to form a complete data set, and extracting key features after preprocessing the data set, such as construction progress deviation, equipment failure rate, personnel fatigue, environmental severity, etc.; using the Apriori algorithm in combination with the established knowledge graph to identify high-risk associations, such as increased equipment failures in high temperature environments, the association between personnel fatigue and accident rates, etc., and determining a potential risk relationship set based on association rules and relationship analysis results; for example, the relationship between high temperature environments and equipment failures, if the association rules show that the equipment failure rate in high temperature environments increases significantly, it will be marked as a potential risk relationship.
[0074] In one implementation, specific risk events are extracted from the risk event set, and each risk event is evaluated using risk assessment methods (FMEA, failure mode and effects analysis, etc.) to determine the likelihood of its occurrence and the extent of its impact. Based on the evaluation results, the risk events are classified according to their severity into high risk, medium risk, and low risk.
[0075] In one implementation method, for high-risk events, detailed response measures and emergency plans are formulated to ensure rapid response and handling. For medium-risk events, preventive measures and monitoring plans are formulated to reduce the possibility of their occurrence. For low-risk events, basic safety measures are formulated to ensure timely handling when they occur. According to the risk level, the priority of various security control measures is determined to ensure the rational allocation of resources. The handling plan for each risk event is determined by technical personnel.
[0076] In one implementation, by acquiring on-site construction data in the construction area in real time and combining it with historical data, the system can promptly reflect changes on site, thereby enabling dynamic updates so that the system can perform risk predictions in real time and effectively respond to changes on the construction site, thus solving the problem of being unable to perform dynamic updates based on the latest data.
[0077] In one implementation, based on real-time operation construction data and the initial knowledge graph, the system can accurately identify the set of potential risk relationships and determine the head entity and its query relationship, so as to more accurately identify the specific risk points on the construction site, provide targeted safety control measures, and improve the accuracy and effectiveness of risk management.
[0078] In one implementation, by constructing and analyzing target subgraphs, the system can determine the set of risk events that occur in the construction area based on all target subgraphs and historical construction data, and formulate a safety management and control plan based on the risk event set to effectively address safety issues at the construction site.
[0079] In one embodiment, establishing a knowledge graph according to an initial data set to obtain an initial knowledge graph includes:
[0080] Identify and extract key entities from the initial data set through natural language processing technology, classify the key entities, and obtain relevant entities and irrelevant entities;
[0081] Through deep learning and standard templates, the relationships between related entities are extracted from the initial data set, and different types of relationship sets are defined; the relationship sets store the relationships between related entities;
[0082] After normalizing entities, attributes, and attribute values with the same meaning but different character strings according to preset rules and the random forest algorithm, the knowledge in different data sources is integrated to obtain the target knowledge base;
[0083] For the target knowledge base, a relation triple is constructed according to each relevant entity and the relationship between the relevant entities;
[0084] Substitute all relationship triplets into the graph database to obtain the knowledge graph.
[0085] In one implementation, natural language processing technology is used to identify and extract key entities from an initial data set by first performing text cleaning on the data in the initial data set, including removing irrelevant characters, punctuation marks, and stop words. The continuous text is then segmented into meaningful units, usually words or phrases, and each word is assigned a part-of-speech tag, such as noun, verb, adjective, etc. A labeling model is used to identify named entities in the text. Common named entities include names of people, places, organizations, dates, etc. The identified entities are matched with corresponding entities in a knowledge base to determine their uniqueness and specific meanings; the knowledge base stores all entities.
[0086] In one implementation, relevant entities are specific equipment models, construction materials, construction personnel, construction methods, safety regulations, etc.; irrelevant entities are entities that are not directly related to the current task or field or do not contribute to analysis and decision-making; identified entities are classified according to their characteristics or attributes to determine whether they are relevant to the current task and thus distinguish relevant entities from irrelevant entities.
[0087] In one implementation, the specification template combines industry standards and specifications to define the relationship types between entities, which may include participation relationships, usage relationships, influence relationships, occurrence relationships, compliance relationships, cause relationships, monitoring relationships, etc.; the relationships between relevant entities include the operational relationship between construction steps and equipment, the usage relationship between personnel and equipment, the causal relationship between accidents and environmental factors, etc.; different types of relationship sets include causal relationships, operational relationships, usage relationships, etc., which describe the causal relationship between hazards, risks and accidents, and clarify the role and influence of each entity in the entire system.
[0088] In one implementation, the preset rules specify how entities with different expressions are mapped to a standard form; for example, "underground cable laying" and "underground cable construction" are normalized to "underground cable laying", and the graph database is Neo4j.
[0089] In one implementation, by extracting the relationships between entities through deep learning and standardized templates, multiple types of relationship sets can be defined. These relationship sets not only preserve the direct connections between the relevant entities, but may also reveal the potential connections between them, thereby accurately providing a basis for subsequent risk analysis and prediction.
[0090] In one implementation method, the knowledge graph constructed through natural language processing technology, deep learning, and data fusion can significantly improve the safety management level of power construction, which not only improves the efficiency and accuracy of risk identification, but also reduces safety management costs.
[0091] In one embodiment, multiple head entities and query relationships of head entities are obtained according to the operation construction data and the initial knowledge graph, including:
[0092] Clean the operation construction data to obtain the construction data set, and integrate the data from different sources in the construction data set to obtain the target data view;
[0093] Perform natural language processing on the target data view to identify the head entity and obtain the head entity set;
[0094] For the head entity set, the relationship triple corresponding to the head entity is obtained to obtain the query relationship corresponding to the head entity.
[0095] In one implementation, data cleaning is performed to remove noise data and outliers, thereby improving data accuracy and reliability. Data from different sources are integrated to ensure data consistency and integrity, forming a unified view for subsequent analysis.
[0096] In one implementation, natural language processing technology can be used to accurately identify key entities related to construction safety, ensuring the accuracy of subsequent analysis.
[0097] In one embodiment, for each head entity and the query relationship between the head entities, determining the target subgraph corresponding to the head entity according to the potential risk relationship set includes:
[0098] Step 1: Obtain a first subgraph according to the head entity, the query relationship of the head entity and the initial knowledge graph;
[0099] Step 2: For the first subgraph, according to the potential risk relationship set, filter out the neighbor entities related to the head entity to obtain the target neighbor entity set;
[0100] Step 3: For the target neighbor entity set, the attention coefficient and attention value of the connection relationship between each neighbor entity and the head entity are calculated through the dynamic subgraph attention mechanism;
[0101] Step 4: The neighbor entities whose attention coefficients and attention values are both greater than the preset thresholds are regarded as neighbor influencing entities, and all neighbor influencing entities are aggregated to obtain the target influencing entity;
[0102] Step 5: Embed the target impact entity into the initial knowledge graph to obtain the target knowledge graph, and obtain the query relationship corresponding to the target impact entity in the target knowledge graph to obtain the target query relationship;
[0103] Step 6: Repeat the above steps 1-5 with the target impact entity as the head entity and the query relationship as the query relationship of the head entity until the preset conditions are met to obtain the target subgraph.
[0104] In one implementation, the first subgraph is a subset extracted from the initial knowledge graph based on the head entity and the query relationship, including the head entity and its directly related entities and relationships; the target neighbor entity set is the neighbor entities directly related to the head entity, which are filtered out from the first subgraph.
[0105] In one implementation, the preset threshold is determined by a technical staff, and the preset conditions may be a maximum number of iterations (setting a maximum number of iterations, and stopping iteration when the number is reached), a subgraph size limit (setting a maximum number of nodes or edges of a subgraph, and stopping iteration when the subgraph reaches the size), a stable risk assessment result (setting a stability condition for the risk assessment result, and stopping iteration when the risk assessment result of multiple consecutive iterations changes little), a fixed target entity and relationship set (setting a stability condition for the target entity and relationship set, and stopping iteration when the target entity and relationship set no longer change), etc., which are specifically determined by the technical staff; the maximum number of iterations.
[0106] In one implementation, by extracting subgraphs, the scale of data processing is reduced and the computational efficiency is improved.
[0107] In one implementation, neighbor entities are screened based on the potential risk relationship set to ensure that only risk-related entities are focused on to avoid interference from irrelevant information. The dynamic subgraph attention mechanism can dynamically adjust the connection weights between entities based on real-time data to reflect changes in actual conditions. By calculating the attention coefficient and attention value, neighbor entities that have a significant impact on the head entity can be accurately identified to improve the accuracy of risk assessment.
[0108] In one implementation, by setting a threshold, only neighbor entities that have a significant impact on risk assessment are retained, noise and irrelevant information are removed, and the aggregated target impact entity can better represent the actual risk factors, thereby improving the reliability of the assessment results.
[0109] In one implementation, the target influencing entity is embedded in the initial knowledge graph and the knowledge graph is updated in real time to ensure the latest data and risk information. The updated query relationship can reflect the latest risk association under current conditions, providing a basis for the next step of analysis.
[0110] In one implementation, the target subgraph is continuously refined and optimized through an iterative process to ensure that all relevant risk factors are fully analyzed and evaluated before the preset conditions are met, thereby improving the comprehensiveness and depth of risk management.
[0111] In one embodiment, the risk event set for determining the occurrence of risks in the construction area according to all target subgraphs and corresponding historical construction data includes:
[0112] By similarity calculation, the target data and historical construction data are matched similarly to determine the target historical records;
[0113] Identify potential risk factors in all target subgraphs and count the number of occurrences of each risk event in the target history;
[0114] According to the number of occurrences of each risk event, the probability of each potential risk factor leading to a risk event under current conditions is calculated using the Bayesian theorem, and the risk event set that will cause risks in the construction area is determined based on the probability of each risk event.
[0115] In one implementation, similarity matching is to determine historical records whose similarity to the target subgraph is greater than a preset value through a similarity calculation method; the target data is entities in all target subgraphs and corresponding query relationships between entities.
[0116] In one implementation, the cosine similarity is used to calculate the similarity between the target subgraph and the historical record, and the historical record with a similarity greater than 85% is selected as the target historical record.
[0117] In one implementation, the potential risk factors in all target subgraphs are determined by extracting all nodes and edges from the target subgraph, using NLP technology to analyze risk events in historical construction data, extracting relevant entities and relationships, applying the PageRank algorithm to calculate the importance score of each entity and relationship, preliminarily identifying high-risk entities and relationships based on the importance score and historical risk records, calculating the attention coefficient and attention value, dynamically screening key entities and relationships, and combining real-time environmental parameters and operating conditions to further confirm potential risk factors and form a final list of potential risk factors.
[0118] In one implementation, the Bayesian theorem is used to calculate the probability of each risk factor leading to a specific risk event. For example, the probability formula for high temperature leading to equipment failure is P(R|E) = ; R is equipment failure, E is high temperature, P (R|E) is the probability of equipment failure caused by high temperature, , is the probability that the equipment failure occurs at high temperature, The probability of high temperature.
[0119] In one implementation, the Bayesian theorem is used to calculate the probability of each potential risk factor leading to a risk event, which can comprehensively consider the frequency of risk events in historical data and changes in current construction conditions, thereby obtaining a more scientific and reasonable risk prediction result.
[0120] Based on the same inventive concept, the embodiment of the present invention also provides a power construction safety management and control device based on knowledge graph. Figure 2 , Figure 2 An electric power construction safety management and control device based on a knowledge graph provided in an embodiment of the present invention includes:
[0121] An initial knowledge graph establishment module is used to obtain historical construction data, pre-process the historical construction data to obtain an initial data set, and establish a knowledge graph based on the initial data set to obtain an initial knowledge graph;
[0122] The operation and construction data analysis module is used to obtain the operation and construction data of the construction area in real time, determine the potential risk relationship set based on the operation and construction data; obtain multiple head entities and query relationships of the head entities based on the operation and construction data and the initial knowledge graph;
[0123] A target subgraph establishment module is used to determine the target subgraph corresponding to each head entity and the query relationship between the head entities according to the potential risk relationship set;
[0124] The safety management and control module is used to determine the risk event set of risks in the construction area based on all target sub-graphs and corresponding historical construction data, and determine the safety management and control plan based on the risk event set.
[0125] Based on an electric power construction safety management and control device based on a knowledge graph provided by an embodiment of the present invention, the operation data of the construction site is obtained in real time, and combined with historical data to dynamically update the knowledge graph, so that the system can timely reflect the changes on the site and perform risk prediction. Based on real-time data and the initial knowledge graph, the system can identify the query relationship between the potential risk relationship set and the head entity, and generate a target subgraph, so that the specific risk points of the construction site can be identified more accurately, and then by combining historical data and real-time data for risk assessment, the system can generate a risk event set and a safety management and control plan, provide scientific decision-making support, and help managers more effectively deal with safety issues on the construction site, so that the system can timely reflect changes on the construction site, thereby performing risk prediction and improving system efficiency.
[0126] In one embodiment, the initial knowledge graph building module includes:
[0127] The key entity classification module is used to identify and extract key entities from the initial data set through natural language processing technology, classify the key entities, and obtain relevant entities and irrelevant entities;
[0128] A relationship set determination module is used to extract the relationships between related entities from the initial data set through deep learning and a standard template, and define different types of relationship sets; the relationship set stores the relationships between related entities;
[0129] The knowledge fusion module is used to normalize entities, attributes, and attribute values with the same meaning but different character strings according to preset rules and the random forest algorithm, and then fuse the knowledge in different data sources to obtain the target knowledge base;
[0130] A relation triple building module is used to build a relation triple according to each related entity and the relationship between the related entities for the target knowledge base;
[0131] The knowledge graph generation module is used to substitute all relationship triples into the graph database to obtain the knowledge graph.
[0132] In one embodiment, the operation construction data analysis module includes:
[0133] A target data view generation module is used to clean the operation construction data to obtain a construction data set, and integrate data from different sources in the construction data set to obtain a target data view;
[0134] A head entity set generation module is used to perform natural language processing on the target data view to identify the head entity and obtain the head entity set;
[0135] The query relationship generation module is used to obtain the relationship triplet corresponding to the head entity set to obtain the query relationship corresponding to the head entity.
[0136] In one embodiment, the target subgraph creation module includes:
[0137] A first subgraph generating module, configured to obtain a first subgraph according to the head entity, the query relationship of the head entity and the initial knowledge graph;
[0138] A target neighbor entity set generation module is used to filter out neighbor entities related to the head entity according to the potential risk relationship set for the first subgraph to obtain a target neighbor entity set;
[0139] The entity relationship generation module is used to calculate the attention coefficient and attention value of the connection relationship between each neighbor entity and the head entity through the dynamic subgraph attention mechanism for the target neighbor entity set;
[0140] A target influence entity generation module is used to take the neighbor entities whose attention coefficients and attention values are both greater than a preset threshold as neighbor influence entities, and aggregate all neighbor influence entities to obtain the target influence entity;
[0141] A target query relationship generation module is used to embed the target impact entity into the initial knowledge graph to obtain the target knowledge graph, and obtain the query relationship corresponding to the target impact entity in the target knowledge graph to obtain the target query relationship;
[0142] The target subgraph generation module is used to return the query relationship with the target influencing entity as the head entity and the query relationship as the head entity to the first subgraph generation module until the preset conditions are met to obtain the target subgraph.
[0143] In one embodiment, the security management module includes:
[0144] The target history record generation module is used to perform similarity matching between the target data and the historical construction data through similarity calculation to determine the target history records; similarity matching is to determine the history records whose similarity with the target subgraph is greater than a preset value through a similarity calculation method; the target data is the entities in all target subgraphs and the corresponding query relationships between the entities;
[0145] The risk event statistics module is used to determine the potential risk factors in all target subgraphs and count the number of occurrences of each risk event in the target history records;
[0146] The risk event set generation module is used to calculate the probability of each potential risk factor leading to a risk event under current conditions based on the number of times each risk event occurs, using the Bayesian theorem, and determine the risk event set in which risks occur in the construction area based on the probability of each risk event.
[0147] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.
Claims
1. A method for power construction safety management and control based on knowledge graph, It is characterized in that The method comprises: Acquire historical construction data, preprocess the historical construction data to obtain an initial data set, and establish a knowledge graph based on the initial data set to obtain an initial knowledge graph; Acquire the operation and construction data of the construction area in real time, and determine the potential risk relationship set according to the operation and construction data; obtain multiple head entities and query relationships of the head entities according to the operation and construction data and the initial knowledge graph; For each head entity and the query relationship between the head entities, determine the target subgraph corresponding to the head entity according to the potential risk relationship set; Determine a risk event set where risks may occur in the construction area based on all target subgraphs and corresponding historical construction data, and determine a safety control plan based on the risk event set; Establishing a knowledge graph according to the initial data set to obtain an initial knowledge graph includes: Identifying and extracting key entities from the initial data set by natural language processing technology, classifying the key entities, and obtaining relevant entities and irrelevant entities; Extracting the relationships between the related entities from the initial data set through deep learning and a standard template, and defining different types of relationship sets; the relationship sets store the relationships between the related entities; After normalizing entities, attributes, and attribute values with the same meaning but different character strings according to preset rules and the random forest algorithm, the knowledge in different data sources is integrated to obtain the target knowledge base; For the target knowledge base, constructing a relation triple according to each relevant entity and the relation between the relevant entities; Substitute all relationship triplets into the graph database to obtain the knowledge graph; For each head entity and the query relationship between the head entities, determining the target subgraph corresponding to the head entity according to the potential risk relationship set includes: Step 1: Obtain a first subgraph according to the head entity, the query relationship of the head entity and the initial knowledge graph; Step 2: for the first subgraph, according to the potential risk relationship set, filter out neighbor entities related to the head entity to obtain a target neighbor entity set; Step 3: For the target neighbor entity set, the attention coefficient and attention value of the connection relationship between each neighbor entity and the head entity are calculated through the dynamic subgraph attention mechanism; Step 4: The neighbor entities whose attention coefficients and attention values are both greater than the preset thresholds are regarded as neighbor influencing entities, and all neighbor influencing entities are aggregated to obtain the target influencing entity; Step 5: embed the target impact entity into the initial knowledge graph to obtain a target knowledge graph, and obtain the query relationship corresponding to the target impact entity in the target knowledge graph to obtain a target query relationship; Step 6: Repeat the above steps 1-5 with the target impact entity as the head entity and the query relationship as the query relationship of the head entity until the preset conditions are met to obtain the target subgraph.
2. According to the knowledge graph-based power construction safety management method of claim 1, It is characterized in that The multiple head entities and the query relationships of the head entities obtained according to the operation construction data and the initial knowledge graph include: Cleaning the operation construction data to obtain a construction data set, and integrating data from different sources in the construction data set to obtain a target data view; Performing natural language processing on the target data view to identify head entities to obtain a head entity set; For the head entity set, the relationship triple corresponding to the head entity is obtained to obtain the query relationship corresponding to the head entity.
3. According to the knowledge graph-based power construction safety management and control method of claim 1, It is characterized in that The risk event set for determining the risk of the construction area based on all target subgraphs and corresponding historical construction data includes: The target data and the historical construction data are similarly matched by similarity calculation to determine the target historical records; the similarity matching is to determine the historical records whose similarity with the target subgraph is greater than a preset value by a similarity calculation method; the target data is the query relationship corresponding to the entities and entities in all the target subgraphs; Determine the potential risk factors in all target subgraphs, and count the number of occurrences of each risk event in the target historical records; According to the number of occurrences of each risk event, the probability of each potential risk factor leading to a risk event under current conditions is calculated using the Bayesian theorem, and the risk event set in which risks occur in the construction area is determined based on the probability of each risk event.
4. A power construction safety control device based on knowledge graph, It is characterized in that The device comprises: An initial knowledge graph establishment module is used to obtain historical construction data, pre-process the historical construction data to obtain an initial data set, and establish a knowledge graph based on the initial data set to obtain an initial knowledge graph; The operation and construction data analysis module is used to obtain the operation and construction data of the construction area in real time, determine the potential risk relationship set according to the operation and construction data; obtain multiple head entities and query relationships of the head entities according to the operation and construction data and the initial knowledge graph; A target subgraph establishment module is used to determine the target subgraph corresponding to each head entity and the query relationship between the head entities according to the potential risk relationship set; A safety control module is used to determine a risk event set where risks may occur in the construction area according to all target subgraphs and corresponding historical construction data, and determine a safety control plan according to the risk event set; The initial knowledge graph establishment module includes: A key entity classification module, used to identify and extract key entities from the initial data set through natural language processing technology, classify the key entities, and obtain relevant entities and irrelevant entities; A relationship set determination module, used to extract the relationship between the related entities from the initial data set through deep learning and a standard template, and define different types of relationship sets; the relationship set stores the relationship between the related entities; The knowledge fusion module is used to normalize entities, attributes, and attribute values with the same meaning but different character strings according to preset rules and the random forest algorithm, and then fuse the knowledge in different data sources to obtain the target knowledge base; A relation triplet construction module, used for constructing a relation triplet according to each relevant entity and the relationship between the relevant entities for the target knowledge base; The knowledge graph generation module is used to substitute all relationship triples into the graph database to obtain the knowledge graph; The target subgraph establishment module comprises: A first subgraph generating module, configured to obtain a first subgraph according to the head entity, the query relationship of the head entity and the initial knowledge graph; A target neighbor entity set generation module is used to filter out neighbor entities related to the head entity according to the potential risk relationship set for the first subgraph to obtain a target neighbor entity set; An entity relationship generation module is used to calculate the attention coefficient and attention value of the connection relationship between each neighbor entity and the head entity through a dynamic subgraph attention mechanism for the target neighbor entity set; A target influence entity generation module is used to take the neighbor entities whose attention coefficients and attention values are both greater than a preset threshold as neighbor influence entities, and aggregate all neighbor influence entities to obtain the target influence entity; A target query relationship generation module, used to embed the target impact entity into the initial knowledge graph to obtain a target knowledge graph, and obtain the query relationship corresponding to the target impact entity in the target knowledge graph to obtain a target query relationship; The target subgraph generation module is used to return the query relationship with the target influencing entity as the head entity and the query relationship as the head entity to the first subgraph generation module until a preset condition is met to obtain the target subgraph.
5. According to claim 4, a knowledge graph-based power construction safety management and control device, It is characterized in that The operation construction data method analysis module includes: A target data view generation module, used for performing data cleaning on the operation construction data to obtain a construction data set, and integrating data from different sources in the construction data set to obtain a target data view; A header entity set generation module, configured to perform natural language processing on the target data view to identify the header entity and obtain a header entity set; The query relationship generation module is used to obtain the relationship triplet corresponding to the head entity for the head entity set to obtain the query relationship corresponding to the head entity.
6. According to claim 4, a knowledge graph-based power construction safety management and control device, It is characterized in that The security control module includes: A target history record generation module is used to perform similarity matching between target data and historical construction data through similarity calculation to determine target history records; the similarity matching is to determine the history records with similarity greater than a preset value with the target subgraph through a similarity calculation method; the target data is the query relationships corresponding to entities and entities in all target subgraphs; A risk event statistics module is used to determine the potential risk factors in all target subgraphs and count the number of occurrences of each risk event in the target historical records; The risk event set generation module is used to calculate the probability of each potential risk factor causing a risk event under current conditions through Bayesian theorem according to the number of times each risk event occurs, and determine the risk event set of the construction area according to the probability of each risk event.
Citation Information
Cited By
Building construction safety monitoring method based on artificial intelligence
CN120011992A
Hydraulic engineering tunnel construction safety assessment method
CN120013312A
