A synchronous line loss intelligent diagnosis analysis system and method based on a power knowledge graph

Abstract

The present invention relates to an intelligent diagnosis and analysis system for synchronous line losses based on an electric power knowledge graph. The system comprises a data acquisition layer, a data access layer, a data storage layer, a data analysis layer, and a visualization layer. The data acquisition layer acquires grid node data; the data access layer dynamically synchronizes grid node data based on Apache Kafka; the data storage layer comprises a time series database and a graph database; the data analysis layer comprises a detection module and an inference engine; and the visualization layer dynamically displays the propagation paths and node characteristics of abnormal areas based on the topological structure of the knowledge graph. By clicking on a device node, the cause of the abnormality and its impact on the entire system can be queried. The present invention enables intelligent detection and diagnosis of abnormal line loss areas, rapidly identifying the root cause of the problem.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of line loss analysis, and particularly relates to a synchronous line loss intelligent diagnosis and analysis system and method based on a power knowledge graph. BACKGROUND

[0002] With the continuous expansion of the power system scale and the gradual improvement of power supply reliability requirements, line loss analysis and management in power grid operation has become one of the focuses of power enterprises. As an important indicator for measuring power transmission and distribution efficiency, the level of synchronous line loss directly reflects the degree of energy loss in the power transmission and distribution process, and also affects the economy of power grid operation and the fairness of user electricity consumption. However, due to the complexity of the power grid structure, the vastness of multi-source data, and the diversification of line loss causes (such as equipment aging, line loss, illegal electricity use, etc.), the traditional line loss detection and analysis method often has the following shortcomings: the data comes from multiple heterogeneous systems (such as SCADA system, GIS, metering system, etc.), which cannot be efficiently integrated, and lacks global semantic association and background information; multiple complex factors (such as equipment problems, unreasonable topology structure, or abnormal user behavior) may be involved behind high line loss, which need further intelligent analysis and evaluation; and with the dynamic changes of power grid load, the traditional static analysis method is difficult to capture abnormalities and locate problem sources in time. SUMMARY

[0003] In order to solve the above problems, the purpose of the present application is to provide a synchronous line loss intelligent diagnosis and analysis system based on a power knowledge graph, which realizes intelligent detection and diagnosis of line loss abnormal areas and quickly identifies problem root causes.

[0004] To achieve the above purpose, the present application adopts the following technical solutions:

[0005] The synchronous line loss intelligent diagnosis and analysis system based on the power knowledge graph comprises a data acquisition layer, a data access layer, a data storage layer, a data analysis layer, and a visualization layer.

[0006] The data acquisition layer acquires power grid node data.

[0007] The data access layer establishes a real-time streaming data analysis framework Apache Kafka to dynamically synchronize the power grid node data.

[0008] The data storage layer comprises a time series database and a graph database, the time series data is stored in the time series database, and the graph database stores the topology and semantic relationship of the power grid based on the power network knowledge graph.

[0009] The data analysis layer includes a detection module and an inference engine. The detection module extracts dynamic data from the time series database, identifies abnormal nodes and time periods, and extracts topological structures from the graph database to analyze the propagation path of the anomaly. The inference engine combines the power network knowledge graph to build a causal chain based on the detection results, infer the root cause of the anomaly, match the anomaly pattern, and identify the possible fault type.

[0010] The visualization layer, based on the topological structure of the knowledge graph, dynamically displays the propagation path and node characteristics of the abnormal area. By clicking on the device node, you can query the cause of the abnormality of the node and its impact on the entire system.

[0011] Furthermore, grid node data includes device sensor data, industrial protocol data, smart meter data, and grid operation and maintenance historical data; the industrial protocol data is accessed through a protocol adapter, and the device sensor data is accessed through a message middleware; smart meter data is accessed through the AMI system; and grid operation and maintenance historical data is imported in batches through a Rest API or ETL tool.

[0012] Furthermore, we established a real-time streaming data analysis framework, Apache Kafka, to dynamically synchronize grid node data. The details are as follows:

[0013] A distributed MQTT message broker is used to decode real-time data into a standardized format. KafkaProducer is used to convert data into message units in a Topic message queue for subsequent consumption. Each type of data is mapped to a different Kafka Topic to separate and manage real-time data streams. Multiple Topics are divided according to data sources and task requirements. Topic-node-data is responsible for real-time node data of the power grid, including voltage, current, and load fluctuations; topic-user-behavior is responsible for data related to user load and abnormal electricity consumption behavior; topic-ops-data is responsible for the operation and maintenance history of power grid equipment. Kafka partitions are designed based on data scale and load.

[0014] Furthermore, the graph database stores the topology and semantic relationships of the power grid based on the power network knowledge graph, as follows:

[0015] The power network knowledge graph G = (V, E, A) includes entity V, relationship E and semantic attribute A, as follows:

[0016] Entity V includes power grid equipment, regional divisions, and user nodes;

[0017] Relationship E includes the physical connection between devices, regional affiliation, and power usage behavior;

[0018] Semantic attribute A, which describes the characteristics of each entity or relationship;

[0019] Entities and relationships:

[0020] Device Collection V e ={v1,v2,…,v i ,…,v n}; Region set V r ={r1,r2,…,r m}; Connection relationship set , indicating the physical connection between devices:

[0021] E c ={(v i ,v j ) | Device v i With device v j There is a connection}

[0022] Attribute Modeling:

[0023] For any v i ∈V e , define its attribute as A(v i )={a1,a2,…,a k};

[0024] Computing device i and device v j The shortest connection path L(i,j):

[0025] ;

[0026] Where d(e) is the weight of edge e; P is the set of connection relations E in the graph. c A subset of represents all possible connection paths between node i and node j; path represents the actual specific path;

[0027] Simulation of upstream propagation impacts:

[0028] For a certain device failure, determine its impact on the upstream node V upstream Impact:

[0029] ;

[0030] in, Represents the slave node v i There is a path that can reach node f;

[0031] Divide load subnets based on regions and connection relationships , for partition analysis:

[0032] ;

[0033] wherein, denotes that node v is directly connected to nodes in the set V e through edges e in the graph;

[0034] and associate the measurement data di(t) in the time series database and the node attributes A(vi) in the graph database.

[0035] Further, the detection module extracts dynamic data from the time series database, identifies abnormal nodes and time periods, and extracts the topology structure from the graph database, analyzes the propagation path of the anomaly, as follows:

[0036] Extract dynamic data from the time series database, identify abnormal nodes and time periods;

[0037] Extract time series data of each power grid node from the time series database ;

[0038] Apply threshold detection, sliding window statistics and prediction error method to determine whether each node is abnormal at each time;

[0039] Define the anomaly indicator function :

[0040]

[0041] Collect all nodes that exist at any time t : :

[0042] ;

[0043] wherein, is the value interval of time t;

[0044] For each node collected , according to the continuous time segment , determine the abnormal time period set :

[0045] ;

[0046] Get the marked abnormal node V abnormal and the abnormal time period T abnormal ;

[0047] According to the abnormal node, according to the power network knowledge graph, calculate the shortest path from the abnormal node to other nodes; analyze to get the upstream or downstream propagation influence range.

[0048] ​Further, according to the abnormal node, the shortest path from the abnormal node to other nodes is calculated according to the power network knowledge graph; the propagation influence range of the upstream or downstream is analyzed, and the specific implementation is as follows:

[0049] From the abnormal node v k ∈V abnormal , the shortest path from the abnormal node to other nodes v j is calculated:

[0050] ;

[0051] If a certain abnormal node v f is set as a fault point, all nodes v f connected to v i through the path are considered to be affected by the upstream of the fault point, and the upstream propagation range is obtained:

[0052] ;

[0053] Downstream propagation range:

[0054] ;

[0055] For the propagation path total cost C ( P f,v ) from the abnormal node v f to any affected node v∈V affected :

[0056]

[0057] Where, P f,v is the shortest path from v f to v.

[0058] Further, combined with the power network knowledge graph, the cause-effect chain is constructed based on the detection result, the root cause of the anomaly is inferred, and the abnormal mode is matched to identify the possible fault type, and the specific implementation is as follows:

[0059] For each detected abnormal node v f , the upstream related entities and relationships are queried by using the power network knowledge graph, the candidate cause-effect chain {P i} is constructed, and the weight W(P i ) is calculated; the main cause-effect chain is determined by comparing the weight, and the possible root cause is inferred;

[0060] The abnormal node V a and the abnormal node association relationship E a are constructed into an abnormal subgraph G anomaly ,

[0061] G anomaly =(V a ,E a );

[0062] Comparing the preset fault mode template M k , the matching degree S(M k ,G anomaly ) is calculated:

[0063] ;

[0064] When the matching degree S(M k ,G anomaly ) exceeds the threshold θ m , it is considered that the anomaly is highly matched with the mode M k , thereby identifying the fault type;V k , E k are the abnormal nodes and abnormal node association relationship of the mode M k , respectively;

[0065] Comprehensive causal chain analysis and pattern matching results, output abnormal diagnosis report, including fault root cause inference, fault type and the corresponding suggestion.

[0066] Further, based on the topology structure of the knowledge graph, the propagation path and node characteristics of the abnormal area are dynamically displayed, and by clicking the device node, the abnormal reason of the node and the influence on the whole system are queried, as follows:

[0067] From the time series database, graph database and knowledge graph, the latest abnormal detection results, causal chain and propagation path are extracted to form a graphical data structure;

[0068] WebSocket is used to realize real-time communication between the front end and the back end, to ensure that the visualization interface reflects the state changes in time;

[0069] Based on the graph database, a graphical display engine is used to render the real-time topology graph, and the abnormal area, propagation path and node state are dynamically labeled;

[0070] After the user clicks the device node, the device node details are queried through the REST API and the detailed results are displayed in the pop-up window or side bar.

[0071] Further, a graphical display engine is used to render the real-time topology graph, and the abnormal area, propagation path and node state are dynamically labeled, as follows:

[0072] Each device node displays its normal / abnormal state through different colors, sizes or shapes, and the abnormal device is identified by red color, and the abnormal time period or abnormal level information is prompted on the node;

[0073] According to the graph database data, the connection relationship of each device is drawn, and the thickness or transparency of the edge reflects the edge weight and transmission cost;

[0074] When the abnormal propagation path is detected, the connection of the path is highlighted in different colors, indicating the abnormal propagation direction and path weight;

[0075] Using power knowledge graph data, the upstream and downstream causal chain related to the anomaly is embedded between nodes, or the key relationship is displayed in a floating label manner on the graph, and the fault root cause and its propagation mechanism are intuitively displayed.

[0076] A synchronous line loss intelligent diagnosis analysis method based on a power knowledge graph, comprising the following steps:

[0077] S1: Obtain power grid node data, and perform dynamic synchronization on the power grid node data through a real-time streaming data analysis framework Apache Kafka;

[0078] S2: Store the time series data in a dynamic database, and store the topology and semantic relationship of the power grid in the graph database based on the power network knowledge graph;

[0079] S3: Extract dynamic data from the time series database, identify abnormal nodes and time periods, and extract the topology structure from the graph database, and analyze the propagation path of the anomaly;

[0080] S4: Based on the detection result, build a causal chain, infer the root cause of the anomaly, and match the abnormal mode to identify the possible fault type;

[0081] S5: Based on the topology structure of the knowledge graph, dynamically display the propagation path of the abnormal area and the node characteristics, and by clicking the device node, query the abnormal reason of the node and the influence on the entire system.

[0082] The present application has the following beneficial effects:

[0083] 1. The present application effectively solves the synchronous line loss diagnosis problem of large-scale power grid, and effectively improves the power operation efficiency and economic benefit;

[0084] 2. The present application acquires each node data of the power grid in real time through the data acquisition layer, and realizes real-time streaming synchronization of data in the data access layer by using Apache Kafka, so that the system can timely capture the power grid operation state and abnormal dynamics, and based on the time series database and the graph database, the system can work cooperatively, so that the historical data of the device can be recorded, and the real-time topology and semantic relationship of the power grid can be reflected, thereby providing comprehensive data support for subsequent analysis;

[0085] 3、The detection module of the application can accurately identify abnormal nodes and abnormal time periods from time series data, and in combination with the power grid topology information in the graph database, deeply analyzes the propagation path of the abnormal event in the network, helps the operation and maintenance personnel to timely discover potential risks, and uses the power network knowledge graph to construct a causal chain, infers the root cause of the abnormality based on the abnormality detection result, not only improves the accuracy of fault diagnosis, but also provides a basis for subsequent decision and processing. BRIEF DESCRIPTION OF DRAWINGS

[0086] Figure 1 The system architecture of the application is shown in the figure.

[0087] Figure 2 The method flowchart in an embodiment of the application is shown in the figure. DETAILED DESCRIPTION

[0088] The application will be further described in detail below in combination with the drawings and specific embodiments:

[0089] Reference Figure 1 In this embodiment, a synchronous line loss intelligent diagnosis and analysis system based on power knowledge graph is provided, which includes a data acquisition layer, a data access layer, a data storage layer, a data analysis layer and a visualization layer.

[0090] The data acquisition layer obtains power grid node data.

[0091] The data access layer establishes a real-time streaming data analysis framework Apache Kafka to dynamically synchronize the power grid node data.

[0092] The data storage layer includes a time series database and a graph database, the time series data is stored in the time series database, and the graph database stores the topology and semantic relationship of the power grid based on the power network knowledge graph.

[0093] The data analysis layer includes a detection module and an inference engine, the detection module extracts dynamic data from the time series database, identifies abnormal nodes and time periods, and extracts the topology structure from the graph database to analyze the propagation path of the abnormality; the inference engine combines the power network knowledge graph, constructs a causal chain based on the detection result, infers the root cause of the abnormality, and matches the abnormal pattern to identify the possible fault type.

[0094] The visualization layer dynamically displays the propagation path of the abnormal area and the node characteristics based on the topology structure of the knowledge graph, and queries the abnormal reason of the node and the influence on the entire system by clicking the device node.

[0095] In this embodiment, the power grid node data includes device sensor data (such as high-frequency current, voltage fluctuation), industrial protocol data, smart meter data, and power grid operation history data; the industrial protocol data (such as IEC 104, Modbus) is accessed through a protocol adapter, and the device sensor data (such as MQTT) is accessed through a message middleware; the smart meter data is accessed through an AMI (advanced metering infrastructure) system; and the power grid operation history data is imported in batches through a Rest API or an ETL tool.

[0096] In this embodiment, a real-time streaming data analysis framework Apache Kafka is established to dynamically synchronize the power grid node data, as follows:

[0097] A distributed MQTT message broker is used to decode real-time data into a standardized format; a KafkaProducer is used to convert the data into a message unit in a Topic message queue for subsequent consumption; each type of data (such as current, voltage, user behavior data, and operation history data) is mapped to a different Kafka Topic to realize separation and management of real-time data streams; multiple Topics are divided according to data sources and task requirements, topic-node-data is responsible for real-time node data of the power grid, including voltage, current, and load fluctuation; topic-user-behavior is responsible for user load and electricity behavior anomaly related data; topic-ops-data is responsible for power grid device operation history records; and Kafka partitions are designed according to data size and load. For example, topic-node-data is partitioned according to substation ID or transformer ID to support high concurrency processing.

[0098] In this embodiment, a graph database stores the topology and semantic relationship of the power grid based on a power network knowledge graph, as follows:

[0099] The power network knowledge graph G=(V,E,A) includes entities V, relationships E, and semantic attributes A, as follows:

[0100] The entities V include power grid devices (such as transformers, meters, and lines), regional divisions, and user nodes;

[0101] The relationships E include physical connections between devices, regional attributions, and electricity behaviors;

[0102] The semantic attributes A describe the characteristics of each entity or relationship (such as device capacity, topology type, and operation records);

[0103] Entities and relationships:

[0104] Device set V e={v1,v2,…,v i ,…,v n}, such as transformers, switches; region set V r ={r1,r2,…,r m}, such as station area, regional division; connection relationship set , indicating the physical connection between devices:

[0105] E c ={(v i ,v j ) | device v i With device v j There is a connection}

[0106] Attribute Modeling:

[0107] For any v i ∈V e , define its attribute as A(v i )={a1,a2,…,a k}, such as device number, power capacity, and status;

[0108] Computing device i and device v j The shortest connection path L(i,j):

[0109]

[0110] Where d(e) is the weight of edge e (such as transmission distance); P is the set of connection relationships E in the graph. c A subset of represents all possible connection paths between node i and node j; path represents the actual specific path;

[0111] Simulation of upstream propagation impacts:

[0112] For a certain device failure, determine its impact on the upstream node V upstream Impact:

[0113] ;

[0114] in, Represents the slave node v i There is a path that can reach node f;

[0115] Divide load subnets based on regions and connection relationships , for partition analysis:

[0116] ;

[0117] in, denotes that the node v is directly connected to the nodes in the set V through the edge e in the graph; e

[0118] and associate the measurement data di(t) in the time series database and the node attribute A(vi) in the graph database.

[0119] In this embodiment, the detection module extracts dynamic data from the time series database, identifies abnormal nodes and time periods, and extracts topological structure from the graph database to analyze the propagation path of the anomaly, as follows:

[0120] Extract dynamic data from the time series database, identify abnormal nodes and time periods;

[0121] Extract time series data of each power grid node from the time series database ;

[0122] Apply threshold detection, sliding window statistics and prediction error method to determine whether each node is abnormal at each time;

[0123] Define an anomaly indicator function :

[0124]

[0125] Collect all nodes existing at any time t : :

[0126] ;

[0127] wherein, is the value interval of time t;

[0128] For each collected node , determine the abnormal time period set according to consecutive time segments :

[0129] ;

[0130] Obtain the abnormal node V abnormal and the abnormal time period T abnormal ;

[0131] According to the abnormal node, according to the power network knowledge graph, calculate the shortest path from the abnormal node to other nodes; analyze to obtain the propagation influence range of upstream or downstream.

[0132] ​In this embodiment, based on the abnormal node and the power network knowledge graph, the shortest path from the abnormal node to other nodes is calculated; the upstream or downstream propagation impact range is analyzed, as follows:

[0133] From the abnormal node v k ∈V abnormal Start from the abnormal node and calculate the distance from the abnormal node to other nodes v j The shortest path:

[0134] ;

[0135] If we set an abnormal node v f is the fault point, then all the f Node v i Considered to be affected by the upstream of the fault point, the upstream propagation range is obtained:

[0136] ;

[0137] Downstream spread:

[0138] ;

[0139] For the abnormal node v f To any affected node v∈V affected , calculate the total cost of the propagation path C ( P f,v ):

[0140]

[0141] Among them, P f,v v f The shortest path to v.

[0142] In this embodiment, combined with the power network knowledge graph, a causal chain is constructed based on the detection results to infer the root cause of the anomaly, match the anomaly pattern, and identify the possible fault type, as follows:

[0143] For each detected abnormal node v f , using the power network knowledge graph to query upstream related entities and relationships, and construct candidate causal chains {P i} and calculate its weight W(P i ); Determine the main causal chain by comparing the weights and infer the possible root cause;

[0144] The abnormal node V a and abnormal node association relationship E a Construct abnormal subgraph G anomaly ,

[0145] G anomaly a ,E a );

[0146] Comparing the preset fault mode template M k , the matching degree S(M k ,G anomaly ) is calculated:

[0147] ;

[0148] When the matching degree S(M k ,G anomaly ) exceeds the threshold value θ m , it is considered that the anomaly is highly matched with the mode M k , thereby identifying the fault type; V k , E k are the abnormal nodes and abnormal node association of the mode M k , respectively;

[0149] Comprehensive causal chain analysis and pattern matching results, output anomaly diagnosis report, including fault root cause inference (based on causal chain), fault type (based on pattern matching) and corresponding suggestions.

[0150] In this embodiment, based on the topology structure of the knowledge graph, the propagation path and the node characteristics of the abnormal area are dynamically displayed, and by clicking the device node, the abnormal reason of the node and the influence on the entire system are queried, as follows:

[0151] The latest anomaly detection results, causal chains and propagation paths are extracted from the time series database, graph database and knowledge graph to form a graphical data structure;

[0152] Real-time communication between the front end and the back end is realized by using WebSocket to ensure that the visual interface reflects state changes in a timely manner;

[0153] Based on the graph database, a graphical display engine (such as D3.js, Cytoscape.js or Sigma.js) is used to render a real-time topology graph, and abnormal areas, propagation paths and node states are dynamically labeled;

[0154] After the user clicks the device node, the device node details (including abnormal reasons, abnormal time period, causal chain information and propagation impact) are queried through the REST API and the detailed results are displayed in a pop-up window or side bar.

[0155] In this embodiment, a graphical display engine is used to render a real-time topology graph, and abnormal areas, propagation paths and node states are dynamically labeled, as follows: ​

[0156] Each device node displays its normal / abnormal state by different colors, sizes or shapes, and the abnormal device is marked with red color and prompts the abnormal time period or abnormal level information on the node;

[0157] The connection relationship of each device is drawn according to the graph database data, and the thickness or transparency of the edge reflects the edge weight and transmission cost;

[0158] When the abnormal propagation path is detected, the connection of the path is highlighted with different colors, indicating the abnormal propagation direction and path weight;

[0159] Using power knowledge graph data, the upstream and downstream causal chain related to the abnormality is embedded between the nodes, or the key relationship is displayed on the graph in the form of floating annotation, to intuitively show the fault root cause and its propagation mechanism.

[0160] Reference Figure 2 In the embodiment, a synchronous line loss intelligent diagnosis and analysis method based on a power knowledge graph is provided, including the following steps:

[0161] S1: Obtain power grid node data, and perform dynamic synchronization on the power grid node data through a real-time streaming data analysis framework Apache Kafka;

[0162] S2: Store the time series data in a dynamic database, and store the topology and semantic relationship of the power grid in the graph database based on the power network knowledge graph;

[0163] S3: Extract dynamic data from the time series database, identify abnormal nodes and time periods, and extract the topology structure from the graph database to analyze the propagation path of the abnormality;

[0164] S4: Based on the detection result, construct a causal chain, infer the root cause of the abnormality, and match the abnormal pattern to identify the possible fault type;

[0165] S5: Based on the topology structure of the knowledge graph, dynamically display the propagation path of the abnormal area and the node characteristics, and by clicking the device node, query the abnormal reason of the node and the influence on the entire system.

[0166] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0167] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0168] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0169] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0170] The above descriptions are only preferred embodiments of the present application, and are not intended to limit the present application to other forms. Any person skilled in the art can make modifications or improvements on the basis of the above disclosed technical content without departing from the technical scope of the present application. Any simple modifications, equivalent changes and improvements made on the basis of the above embodiments without departing from the technical scope of the present application shall fall within the scope of the present application.

Claims

1. The intelligent diagnosis and analysis system for synchronous line loss based on the power knowledge graph is characterized by: Including data acquisition layer, data access layer, data storage layer, data analysis layer and visualization layer; The data acquisition layer obtains power grid node data; The data access layer establishes a real-time streaming data analysis framework Apache Kafka to dynamically synchronize power grid node data; The data storage layer includes a time series database and a graph database. Time series data is stored in the time series database. The graph database stores the topology and semantic relationships of the power grid based on the power network knowledge graph. The power network knowledge graph G = (V, E, A) includes entity V, relationship E and semantic attribute A, specifically as follows: entity V includes power grid equipment, regional division, and user node; The data analysis layer includes a detection module and an inference engine. The detection module extracts dynamic data from the time series database, identifies abnormal nodes and time periods, and extracts topology structures from the graph database to analyze the propagation path of abnormalities. The details are as follows: Extract dynamic data from the time series database and identify abnormal nodes and time periods; Extract the time series data of each power grid node from the time series database ; Apply threshold detection, sliding window statistics and prediction error methods to determine whether each node is abnormal at each moment; Define exception indicator function : All existing at any time t Node Collect : ； in, is the value interval of time t; For each node collected , based on consecutive time segments , determine the abnormal time period set : ； Get the marked abnormal node V abnormal and abnormal time period T abnormal ; Based on the abnormal node and the power network knowledge graph, the shortest path from the abnormal node to other nodes is calculated; the upstream or downstream propagation impact range is analyzed, as follows: From the abnormal node v k ∈V abnormal Start from the abnormal node and calculate the distance from the abnormal node to other nodes v j The shortest path: ； If we set an abnormal node v f is the fault point, then all the f Node v i Considered to be affected by the upstream of the fault point, the upstream propagation range is obtained: ； Downstream propagation range: ； For the abnormal node v f To any affected node v∈V affected , calculate the total cost of the propagation path C ( P f,v ): ； Among them, P f,v v f The shortest path to v, d(e) is the weight of edge e; The inference engine combines the power network knowledge graph to build a causal chain based on the detection results, infer the root cause of the anomaly, and match the anomaly pattern to identify the possible fault type; The visualization layer, based on the topological structure of the knowledge graph, dynamically displays the propagation path and node characteristics of the abnormal area. By clicking on the device node, you can query the cause of the abnormality of the node and its impact on the entire system.

2. The intelligent diagnosis and analysis system for synchronous line loss based on electric power knowledge graph according to claim 1 is characterized in that: The grid node data includes device sensor data, industrial protocol data, smart meter data and grid operation and maintenance historical data; the industrial protocol data is accessed through a protocol adapter, and the device sensor data is accessed through a message middleware; smart meter data is accessed through the AMI system; and grid operation and maintenance historical data is imported in batches through a Rest API or ETL tool.

3. The intelligent diagnosis and analysis system for synchronous line loss based on electric power knowledge graph according to claim 2 is characterized in that: The real-time streaming data analysis framework Apache Kafka is established to dynamically synchronize power grid node data, as follows: A distributed MQTT message broker is used to decode real-time data into a standardized format. Kafka Producer is used to convert data into message units in a Topic message queue for subsequent consumption. Each type of data is mapped to a different Kafka Topic to separate and manage real-time data streams. Multiple Topics are divided according to data sources and task requirements. Topic-node-data is responsible for real-time node data of the power grid, including voltage, current, and load fluctuations. Topic-user-behavior is responsible for data related to user load and abnormal electricity consumption behavior. Topic-ops-data is responsible for the historical operation and maintenance records of power grid equipment. Kafka partitions are designed based on data scale and load.

4. The intelligent diagnosis and analysis system for synchronous line loss based on electric power knowledge graph according to claim 1 is characterized in that: The graph database stores the topology and semantic relationships of the power grid based on the power network knowledge graph, as follows: The power network knowledge graph G = (V, E, A) includes entity V, relationship E and semantic attribute A, as follows: Entity V includes power grid equipment, regional divisions, and user nodes; Relationship E includes the physical connection between devices, regional affiliation, and power usage behavior; Semantic attribute A, which describes the characteristics of each entity or relationship; Entities and relationships: Device Collection V e ={v1,v2,…,v i ,…,v n }; Region set V r ={r1,r2,…,r m }; Connection relationship set , indicating the physical connection between devices: E c ={(v i ,v j ) | device v i With device v j There is a connection} Attribute Modeling: For any v i ∈V e , define its attribute as A(v i )={a1,a2,…,a k }; Computing device i and device v j The shortest connection path L(i,j): ； Where d(e) is the weight of edge e; P is the set of connection relations E in the graph. c A subset of represents all possible connection paths between node i and node j; path represents the actual specific path; Simulation of upstream propagation impacts: For a certain device failure, determine its impact on the upstream node V upstream Impact: ； in, Represents the slave node v i There is a path that can reach node f; Divide load subnets based on regions and connection relationships , for partition analysis: ； in, Represents node v and set V e The nodes in are directly connected through the edges e in the graph; And associate the measurement data di(t) in the time series database with the node attributes A(vi) in the graph database.

5. The intelligent diagnosis and analysis system for synchronous line loss based on electric power knowledge graph according to claim 1 is characterized in that: The above method combines the power network knowledge graph, builds a causal chain based on the detection results, infers the root cause of the anomaly, matches the anomaly pattern, and identifies the possible fault type, as follows: For each detected abnormal node v f , using the power network knowledge graph to query upstream related entities and relationships, and construct candidate causal chains {P i } and calculate its weight W(P i ); Determine the main causal chain by comparing the weights and infer the possible root cause; The abnormal node V a and abnormal node association relationship E a Construct abnormal subgraph G anomaly , G anomaly =(V a ,E a ); Compare with the preset failure mode template M k , calculate the matching degree S(M k ,G anomaly ): ； When the matching degree S(M k ,G anomaly ) exceeds the threshold θ m When the anomaly is considered to be related to the mode M k Highly matched, thus identifying the fault type; V k 、E k Mode M k Abnormal nodes and abnormal node association relationships; Based on the causal chain analysis and pattern matching results, an abnormal diagnosis report is output, including the root cause inference, fault type and corresponding suggestions.

6. The intelligent diagnosis and analysis system for synchronous line loss based on electric power knowledge graph according to claim 1 is characterized in that: The knowledge graph-based topology dynamically displays the propagation path and node characteristics of the abnormal area. By clicking on a device node, you can query the cause of the abnormality and its impact on the entire system, as follows: Extract the latest anomaly detection results, causal chains, and propagation paths from time series databases, graph databases, and knowledge graphs to form a graphical data structure; Use WebSocket to achieve real-time communication between the front-end and back-end, ensuring that the visual interface reflects status changes in a timely manner; Based on the graph database, a graphical display engine is used to render real-time topology maps, and dynamically mark abnormal areas, propagation paths, and node status; After the user clicks a device node, the device node details are queried through the REST API and the detailed results are displayed in a pop-up window or sidebar.

7. The intelligent diagnosis and analysis system for synchronous line loss based on electric power knowledge graph according to claim 6 is characterized in that: The graphical display engine is used to render the real-time topology map and dynamically mark abnormal areas, propagation paths, and node status, as follows: Each device node displays its normal / abnormal status through different colors, sizes, or shapes. Abnormal devices are marked in red, and the abnormal time period or abnormality level information is displayed on the node; Draw the connection relationship between each device based on the graph database data. The thickness or transparency of the edge reflects the edge weight and transmission cost. When an abnormal propagation path is detected, the path lines are highlighted in different colors to indicate the abnormal propagation direction and path weight; By utilizing the power knowledge graph data, the upstream and downstream causal chains related to the anomaly are embedded between nodes, or key relationships are displayed on the graph in the form of suspended annotations to intuitively show the root cause of the fault and its propagation mechanism.

8. An analysis method based on the intelligent diagnosis and analysis system for synchronous line loss based on the power knowledge graph according to any one of claims 1 to 7, comprising the following steps: S1: Obtain grid node data and dynamically synchronize it through the real-time streaming data analysis framework Apache Kafka; S2: The time series data is stored in a dynamic database. The graph database stores the topology and semantic relationships of the power grid based on the power network knowledge graph. S3: Extract dynamic data from the time series database, identify abnormal nodes and time periods, and extract the topology from the graph database to analyze the propagation path of the anomaly; S4: Build a causal chain based on the detection results, infer the root cause of the anomaly, and match the anomaly pattern to identify the possible fault type; S5: Based on the topological structure of the knowledge graph, the propagation path and node characteristics of the abnormal area are dynamically displayed. By clicking on the device node, the cause of the abnormality of the node and its impact on the entire system can be queried.

Images