Substation auxiliary control system point position topology automatic binding method, system and device

Abstract

The application belongs to the technical field of substation auxiliary control system, and specifically discloses a substation auxiliary control system point position topology automatic binding method, system and equipment, which comprises the following steps: constructing and maintaining a graph model; when it is monitored that a new device is connected, the attribute information of the new device and the initial edge weight of the new device in the graph model are obtained; the potential correlation between the new device and the existing devices in the system is predicted through a Bayesian network; the potential correlation is verified and refined by using a graph neural network, and the final topology relationship is obtained, taking the potential correlation as a constraint; according to the final topology relationship, the topology link binding between the new device and the related existing devices in the substation auxiliary control system is realized, and the graph model is synchronously updated. The application can quickly complete device connection and topology construction, shorten the system upgrade iteration cycle, and fully adapt to the complex and changeable operation monitoring and control requirements of the substation.

Description

Technical Field

[0001] This invention belongs to the field of substation auxiliary control system technology, and particularly relates to a method, system and equipment for automatic binding of point topology in substation auxiliary control system. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] The substation auxiliary monitoring system (hereinafter referred to as the substation auxiliary control system) includes various types of equipment such as security guards, fire alarms, environmental monitoring, SF6 monitoring, lighting control, and online monitoring, which can realize real-time monitoring of substation fire protection, security, equipment status, etc.

[0004] In substation auxiliary control systems, the binding of equipment locations with spatial positions and control logic is the core of intelligent management and control. Most existing technologies store data such as equipment IDs and location numbers in two-dimensional tables, and then rely on manual configuration of SQL (Structured Query Language) relational statements to establish static mappings between equipment IDs and location numbers. This method requires manual configuration of two-dimensional mapping relationships between equipment locations and physical positions. For example, manually associating sensor IDs with GIS areas and control equipment modules with monitoring area numbers one by one is not only simplistic and inefficient, but also prone to human error.

[0005] In addition, when adding new sensors or control devices, it is necessary to manually establish the binding relationship between the new devices and the existing system in terms of spatial location, control logic, and ownership. The entire process relies on manual operation and is cumbersome. It is impossible to achieve automatic binding of device locations with spatial locations and control logic, resulting in insufficient system scalability and seriously affecting the upgrade and iteration efficiency of the substation auxiliary control system. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a method, system, and equipment for automatically binding the topology of substation auxiliary control system points. This method enables automated and intelligent association of the topology of auxiliary control system equipment points, significantly improving the system's scalability and operational efficiency.

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

[0008] The first aspect of the present invention provides an automatic binding method for the topology of a substation auxiliary control system, comprising: constructing and maintaining a graph model representing the equipment, space and logical relationships within the substation auxiliary control system;

[0009] When a new device is detected to be connected, the attribute information of the new device and the initial edge weight of the new device in the graph model are obtained.

[0010] Based on the attribute information of the newly added device and the initial edge weight of the newly added device in the graph model, the potential association relationship between the newly added device and the existing devices in the system is predicted by Bayesian network.

[0011] Using the potential relationships as constraints, a subgraph containing the new device and related existing devices is constructed, and a graph neural network is used to verify and refine the potential relationships to obtain the final topological relationships;

[0012] Based on the final topology, the topology link binding between the newly added equipment and the relevant existing equipment is implemented in the substation auxiliary control system through the automated configuration interface, and the graph model is updated synchronously.

[0013] This invention constructs a relationship inference engine based on the fusion of Bayesian networks and graph neural networks. It uses Bayesian networks for initial screening of device relationships, and then utilizes graph neural networks to precisely refine these relationships. The entire process requires no manual configuration and automatically binds new device attributes, spatial relationships, logical relationships, and temporal topologies across multiple dimensions, fully meeting the real-time dynamic binding requirements of equipment location topologies in substation auxiliary control systems.

[0014] As a further solution, a graphical model representing the equipment, space, and logical relationships within the substation auxiliary control system is constructed and maintained, specifically as follows:

[0015] Using equipment nodes, spatial nodes, logical nodes, and time nodes as nodes in the graph model, and the physical connection edges, logical association edges, and spatial inclusion edges between nodes as directed edges in the graph model, a graph model of substation auxiliary control equipment is constructed. Among them, equipment nodes represent substation auxiliary control equipment and attribute information, spatial nodes represent substation physical area information, logical nodes represent control logic information, and time nodes represent equipment full life cycle status information.

[0016] This invention constructs a four-dimensional attribute graph model applicable to the field of substation auxiliary control, which can accurately depict the multi-dimensional relationships of monitoring equipment, control equipment, spatial location, control logic, etc. Through multi-dimensional fusion, it realizes a structured and integrated representation of the complex environment of the substation auxiliary control system, and fully meets the multi-dimensional correlation requirements of equipment.

[0017] As a further solution, the edge weights of the physically connected edges, logically related edges, and spatially contained edges are determined as follows:

[0018] Based on the edge type and business priority, set the basic weight value for each type of edge;

[0019] For physical connection edges, calculate the connection continuity correction coefficient, interface health correction coefficient, line length matching correction coefficient, and redundant connection number correction coefficient respectively; obtain the total weight correction coefficient of the physical connection edge by weighted summation; calculate the product of the basic weight value of the physical connection edge and the total weight correction coefficient to obtain the final edge weight of the physical connection edge.

[0020] For logically related edges, calculate the scene matching degree correction coefficient, historical linkage success rate correction coefficient, device online rate correction coefficient, and response timeliness correction coefficient respectively; obtain the total weight correction coefficient of the logically related edges by weighted summation; calculate the product of the basic weight value of the logically related edges and the total weight correction coefficient to obtain the final edge weight of the logically related edges.

[0021] For spatially contained edges, calculate the location matching degree correction coefficient, binding duration ratio correction coefficient, region adaptation degree correction coefficient, and movement frequency correction coefficient respectively; obtain the total weight correction coefficient of the spatially contained edge by weighted summation; calculate the product of the basic weight value of the spatially contained edge and the total weight correction coefficient to obtain the final edge weight of the spatially contained edge.

[0022] When the device location changes, the operating status changes, or the business rules are adjusted, the edge weights are automatically updated according to the changes in node attributes.

[0023] As a further solution, when a new device is detected to be connected, the attribute information of the new device is obtained, specifically:

[0024] The type of newly added device can be identified by its IP address, MAC address, or protocol type information;

[0025] By matching the unique device identifier or judging the overlap of key attributes, it is determined whether data from different data sources are the same device information. The same device information from multiple data sources is integrated and standardized to form unified device attribute information.

[0026] As a further approach, a Bayesian network is used to predict the potential relationships between the newly added device and existing devices in the system, specifically:

[0027] Based on the attribute information of the newly added device and the initial edge weight of the newly added device in the graph model, the prior probability of different relationship types between the newly added device and the existing device is predicted by Bayesian network; the relationship type with a prior probability greater than a first predetermined threshold is taken as the potential relationship type between the newly added device and the existing device.

[0028] As a further solution, the topology link binding between the newly added equipment and relevant existing equipment is implemented in the substation auxiliary control system through an automated configuration interface, specifically as follows:

[0029] A rule knowledge base is constructed and maintained, which pre-stores the topology association logic rules of different devices in the substation auxiliary control system; the topology association logic rules include: rule identifier, rule name, rule attributes, corresponding topology association relationship type, triggering condition, and execution action;

[0030] Based on the relationship type between the new equipment and the existing equipment predicted by the graph neural network, the corresponding rules are retrieved from the rule knowledge base. Based on the rules, the topology link binding between the new equipment and the relevant existing equipment is realized, thereby realizing the automatic update of the graph model of the substation auxiliary control equipment.

[0031] This invention establishes relationships between devices in a graph database by invoking topological association logic rules, and then automatically generates association edges between new nodes and other nodes on the attribute graph without manual intervention. Simultaneously, it can periodically check existing topological relationships, marking invalid or expired associations to prevent erroneous linkage.

[0032] As a further solution, the construction process of the rule knowledge base includes:

[0033] Extract the topological association logic of equipment locations in the substation auxiliary control system, transform the logic into structured rules, and define the association type, triggering conditions, triggering thresholds, execution actions, and priorities corresponding to each rule;

[0034] By simulating different scenarios using rule validation tools, the effectiveness of each rule is tested, ultimately forming a rule knowledge base.

[0035] A second aspect of the present invention provides an automatic point topology binding system for a substation auxiliary control system, comprising:

[0036] The attribute graph management module is configured to build and maintain a graph model that represents the equipment, space, and logical relationships within the substation auxiliary control system.

[0037] The device access sensing module is configured to monitor whether a new device is connected, and when a new device is detected, to obtain the attribute information of the new device and the initial edge weight of the new device in the graph model.

[0038] The association reasoning module is configured to predict the potential association between the new device and existing devices in the system using a Bayesian network based on the attribute information of the new device and the initial edge weight of the new device in the graph model; construct a subgraph containing the new device and related existing devices using the potential association as constraints; and use a graph neural network to verify and refine the potential association to obtain the final topological relationship.

[0039] The topology binding execution module is configured to bind the topology links between the newly added equipment and the relevant existing equipment in the substation auxiliary control system according to the final topology relationship through an automated configuration interface, and to synchronously update the graph model.

[0040] As a further solution, a graphical model representing the equipment, space, and logical relationships within the substation auxiliary control system is constructed and maintained, specifically as follows:

[0041] Using equipment nodes, spatial nodes, logical nodes, and time nodes as nodes in the graph model, and the physical connection edges, logical association edges, and spatial inclusion edges between nodes as directed edges in the graph model, a graph model of substation auxiliary control equipment is constructed. Among them, equipment nodes represent substation auxiliary control equipment and attribute information, spatial nodes represent substation physical area information, logical nodes represent control logic information, and time nodes represent equipment full life cycle status information.

[0042] A third aspect of the present invention provides a terminal device comprising a processor and a memory, the processor being used to implement instructions; the memory being used to store multiple instructions adapted to be loaded and executed by the processor to perform the above-described automatic binding method for point topology of a substation auxiliary control system.

[0043] A fourth aspect of the present invention provides a computer-readable storage medium storing a plurality of instructions adapted to be loaded and executed by a processor of a terminal device of the above-described automatic binding method for point topology of a substation auxiliary control system.

[0044] A fifth aspect of the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the above-described automatic binding method for point topology of a substation auxiliary control system.

[0045] Compared with the prior art, the beneficial effects of the present invention are:

[0046] (1) This invention integrates equipment node, spatial node, logical node and time node information to construct a four-dimensional attribute graph model applicable to the field of substation auxiliary control. It unifies and abstracts various heterogeneous relationships such as physical connection, logical association and spatial inclusion into directed edges with semantics and weights. It can accurately depict the multi-dimensional relationships of monitoring equipment, control equipment, spatial location, control logic, etc. Through multi-dimensional fusion, it realizes the structured and integrated representation of the complex environment of substation auxiliary control system, and fully meets the multi-dimensional association requirements of equipment.

[0047] (2) When a new device is detected, the present invention constructs a relationship inference engine based on the fusion of Bayesian network and graph neural network. The Bayesian network is used to initially screen the relationship type between the new device and other devices, and then the graph neural network is used to accurately correct the relationship type, so as to realize the automatic prediction of the relationship type between the new device and the corresponding area control device.

[0048] Meanwhile, by pre-building a rule knowledge base, the system automatically calls the corresponding rules based on the predicted relationship types between device locations, thereby achieving automatic association of device relationships at the database level.

[0049] This invention, through its mechanisms of automatic discovery of new devices, automatic prediction of relationship types, and dynamic binding of point topology, can quickly complete device access and topology construction, shorten the system upgrade and iteration cycle, and fully adapt to the complex and ever-changing operation monitoring and control needs of substations. Attached Figure Description

[0050] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0051] Figure 1 This is a flowchart of the automatic binding method for substation auxiliary control system point topology in an embodiment of the present invention;

[0052] Figure 2 This is a schematic diagram of the inference process of fusing Bayesian network and graph neural network in an embodiment of the present invention. Detailed Implementation

[0053] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0054] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0055] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0056] Example 1

[0057] In one or more embodiments, a method for automatic binding of point topology in a substation auxiliary control system is disclosed, combining... Figure 1 Specifically, it includes the following process:

[0058] S101: Obtain multi-source data information of substation auxiliary control equipment and construct a graphical model of substation auxiliary control equipment;

[0059] Substation auxiliary control equipment mainly includes: monitoring sensors used to collect data on environment and equipment status, including sulfur hexafluoride sensors, temperature and humidity sensors, etc.; execution control equipment used to respond to monitoring data and perform control operations, including explosion-proof fans (such as explosion-proof axial flow fans), other linkage control modules (such as control modules M001, M002, etc.); management equipment such as sulfur hexafluoride monitoring host and fire protection host, as well as related auxiliary equipment.

[0060] This embodiment obtains multi-source device information from different data sources such as network protocols, industrial protocols, management system APIs (application programming interfaces), and message queues, and then performs a data cleaning process.

[0061] The data cleaning process mainly includes:

[0062] (1) Data validation: For different data sources (network protocols, industrial protocols, APIs (application programming interfaces), message queues), validate data format (such as the validity of IP and MAC addresses), integrity (such as the device location and type fields being non-empty), and timeliness (such as removing offline data that has not been updated for more than 24 hours);

[0063] (2) Data deduplication: Integrate cross-source data to form unified device attribute information; the specific method is as follows:

[0064] Using the device's unique identifier as an anchor point, data from different sources is linked, while retaining the latest version. Core identifiers include device ID, MAC address, IP address, or sensor SN (Serial Number), etc. These identifiers are unique in each data source and can be directly used as the basis for association. For example, if the device MAC address obtained by the LLDP protocol matches the MAC address in the device file entered by the API, it can be determined that the data belongs to the same device across different sources.

[0065] When there is no unique identifier, key attributes are selected. In this embodiment, device type, installation location, and detection range are selected as key attributes. If at least two key attributes highly overlap, the data are determined to be related. For example, sensor device data collected by the industrial protocol with "detection range 5m, installation location GIS room 201 West Zone" can be determined to be related to alarm data in the message queue with "type SF6 sensor, location R201-West Zone" because of the matching key attributes.

[0066] Alternatively, data attribution can be determined by combining scenario information such as equipment functions, linkage relationships, and data generation timing. For example, wind turbine start-up command data (from the control protocol) and SF6 concentration exceeding the limit data (from the sensor protocol) in a certain area can be determined to be related due to the business logic of "gas exceeding the limit - wind turbine start-up".

[0067] (3) Data standardization: Map the names of equipment types with different protocols (such as “SF6 sensor”, “sulfur hexafluoride sensor”) to standard terms, and standardize the location information (such as “GIS Room 201 East Zone”, “201GIS East”) to a unified format (such as “R201-East Zone”).

[0068] Encode structured data into graph node feature vectors, such as encoding device types and normalizing location coordinates.

[0069] (4) Abnormal data processing: use interpolation to repair missing equipment operating parameters (such as sensor concentration values) and use threshold filtering to remove abnormal values ​​(such as invalid data that exceeds the detection range).

[0070] This embodiment uses devices, space, logic, and time as core nodes, and various heterogeneous relationships such as physical connections, logical associations, and spatial inclusion as edges to construct a four-dimensional graph model suitable for the field of substation auxiliary control. This four-dimensional graph model is stored in a graph database and can reflect the dynamic topology of the substation auxiliary control system.

[0071] Specifically, each type of node carries unique dynamic attributes, and the node information mainly includes:

[0072] (1) Equipment nodes: covering monitoring sensors (such as SF6 sensors), control equipment (such as explosion-proof fans), etc., and attributes include equipment ID, IP address, detection accuracy, operating status and control module number, etc.

[0073] (2) Spatial nodes: represent the physical area of ​​the substation (such as GIS (Gas Insulated Metal Enclosed Switchgear) room, switch room), and the attributes include GIS coordinates, floor elevation, spatial type and area number, etc.

[0074] (3) Logical nodes: store control policies and association rules, with attributes such as Cypher statements (declarative query language) (e.g., “anomaly monitoring - control device startup”), trigger conditions, and priorities.

[0075] (4) Time nodes: Record the status of the equipment throughout its entire life cycle. Attributes include commissioning time, maintenance time, scrapping time, data update timestamp, etc.

[0076] Edges are abstractions of heterogeneous relationships between nodes, and are all directed edges with semantic meaning. They mainly include three types:

[0077] Physical connection edge: reflects the hardware connection relationship between devices, such as the control circuit connection between sensors and control equipment.

[0078] Logical association edges: Represent the linkage relationship at the business level, such as the "linkage control" between sensors and control equipment, or the "triggering" relationship where a sensor triggers the start of a fan.

[0079] Spatial containment edges: describe the relationship between devices and physical areas, such as the "LOCATED_IN" relationship of a sensor "located" in a certain GIS room. LOCATED_IN is a declarative query language in graph databases that represents the "located" relationship.

[0080] Various edge weights are dynamically determined through a combination of business rule assignment and data feature correction, specifically:

[0081] First, based on substation operation and maintenance experience, and combined with edge type and business priority, set the basic weight value for each type of edge; for example, for physical connection edges, if the connection is to core equipment, the basic weight value is set to 0.9, and if the connection is to ordinary equipment, the basic weight value is set to 0.7.

[0082] Then, combining device attributes and operational data, for the three types of edges, multi-dimensional correction coefficients are calculated from different correction dimensions, and the total correction coefficient is obtained by weighted summation; the details are as follows:

[0083] (1) For physical connection edges, the correction coefficients are calculated from the aspects of connection continuity rate, interface health, line length matching degree and number of redundant connections respectively;

[0084] Connection continuity correction factor: The higher the connection uptime, the more stable the hardware link, and the more positive the weight correction.

[0085] Interface health correction factor: The fewer interface errors, the higher the reliability of data transmission, and the closer the physical connection weight is to the base value.

[0086] Line length matching correction factor: The closer the line length is to the standard design value, the lower the signal loss and the stronger the connection stability.

[0087] Redundancy correction factor: As an example, the maximum redundancy of the core connection of the substation is 2 lines; the more redundancy, the stronger the fault tolerance of the connection (when the redundancy is 0, K4=0, and the weight will be greatly reduced).

[0088] The total correction factor for physical connection edges is: ;

[0089] in, , , and These are preset weightings for connection uptime, interface health, line length matching, and redundant connections, which can be adjusted as needed. .

[0090] The final weight of the physical connection edge is: ; The base weights of the physically connected edges.

[0091] (2) For logically related edges, the correction coefficients are calculated from the aspects of scene matching degree, historical linkage success rate, device online rate and response timeliness;

[0092] Scene matching correction coefficient: Coverage area = π × detection radius²; The more comprehensive the monitoring range, the higher the matching degree of the linked scenarios, and the more positive the weight correction.

[0093] Historical success rate correction factor: The more reliable the coordinated execution, the higher the effectiveness of the business logic, and the closer the weight is to the base value.

[0094] Equipment online rate correction factor: The more stable the devices on both sides of the linkage, the lower the risk of false linkage or no response, and the more positive the weight correction.

[0095] Response timeliness correction factor: The faster the response, the more positive the weight correction will be, meeting the substation's "second-level processing" requirements.

[0096] The total correction factor for logically related edges is: ;

[0097] in, , , and These are preset weightings for scene matching accuracy, historical linkage success rate, device online rate, and response timeliness, which can be adjusted as needed. .

[0098] The final weight of the logically related edges is: ; The basic weights of logically related edges.

[0099] (3) For spatially contained edges, the correction coefficients are calculated from the aspects of position matching degree, binding duration ratio, regional adaptability and movement frequency;

[0100] Position matching correction factor: The coordinates are latitude and longitude; the closer the device is to the center of the region, the more accurate the attribution relationship and the more positive the weight correction.

[0101] Binding duration percentage correction factor: The longer the binding lasts, the more stable the affiliation becomes, and the closer the weight is to the base value.

[0102] Region adaptation correction factor: The equipment function and regional needs matching score is between 1 and 10 points, which is an experience score; the better the equipment function matches the region, the higher the rationality of the attribution, and the more positive the weight adjustment.

[0103] Mobile frequency correction factor: As an example, the maximum number of moves allowed is 5; the fewer moves, the more stable the affiliation, and the closer the weight is to the base value.

[0104] The total correction factor for the space containing edges is: ;

[0105] in, , , and These are preset weightings for scene matching accuracy, historical linkage success rate, device online rate, and response timeliness, which can be adjusted as needed. .

[0106] The final weight of the edge contained in the space is: ; The basic weights of the edges contained in the space.

[0107] Substation auxiliary control equipment is subject to dynamic scenarios such as relocation, expansion, and failure. A single-dimensional weight is difficult to adapt to these changes. In this embodiment, the weight of each type of edge is calculated from multiple dimensions. When the equipment location changes, the operating status changes, or the business rules are adjusted, the edge weight is automatically iterated with the node attribute update, without the need for manual intervention.

[0108] The above node information and directed edges are stored in a graph database, ultimately resulting in a four-dimensional graph model applicable to the field of substation auxiliary control. This embodiment provides a solid data foundation for various monitoring and emergency linkages through the fusion of multi-dimensional relationships, realizing a structured and integrated representation of the complex environment of the substation auxiliary control system.

[0109] S102: When a new device is detected to be connected, the attribute information of the new device is obtained through multi-source data fusion.

[0110] In this embodiment, data related to the auxiliary control equipment is collected in real time from multiple channels to monitor the newly added equipment; the data sources mainly include:

[0111] (1) Network protocol probe: Listen to protocol messages such as LLDP (Link Layer Discovery Protocol), CDP (Cisco Discovery Protocol), and mDNS (Multicast DNS Protocol) in the local area network to discover newly added monitoring sensors, control devices and their connected neighbor information.

[0112] (2) Southbound interface gateway: Through industrial protocols such as Modbus (application layer message transmission protocol), IEC104, IEC61850, etc., it polls or subscribes to monitoring data (such as gas concentration) and control equipment operating status (such as start-stop status, speed) and other change information.

[0113] (3) Management System API: Receives instructions (such as sensor calibration, control equipment location change) from the upper management system or manually triggered by maintenance personnel.

[0114] (4) Message queue: Subscribe to topics such as Kafka (Apache Kafka Message Queues) and MQTT (Message Queues Telemetry Transport) to obtain various system events such as monitoring data exceeding limits alarms.

[0115] When a new device is detected, the device type (such as SF6 sensor, wind turbine, etc.) can be accurately identified based on the collected information such as IP address, MAC address, protocol type, and manufacturer.

[0116] By matching the unique identifier of the device or judging the overlap of key attributes, it is determined whether data from different data sources are the same device information. The same device information from multiple data sources is integrated and standardized to form unified device attribute information. As a specific example, device attribute information may include: device type, installation location, detection range, control radius, and control module number, etc.

[0117] The specific methods for cross-source data integration are the same as those described above, and will not be detailed here again.

[0118] Meanwhile, when a new device is added, based on the four-dimensional graph model constructed in step S101, the average edge weights of the physical connection edges, logical association edges, and spatial inclusion edges between the new device and other devices of the same type are calculated respectively, and used as the initial edge weights of each type of edge between the new device and other devices in the graph model.

[0119] S103: Based on the attribute information of the new device and the initial edge weight of the new device in the graph model, predict the potential association between the new device and the existing devices in the system through a Bayesian network;

[0120] S104: Using potential relationships as constraints, construct a subgraph containing new devices and related existing devices, and use a graph neural network to verify and refine the potential relationships to obtain the final topological relationships.

[0121] This embodiment constructs a relationship inference engine based on the fusion of Bayesian networks and graph neural networks:

[0122] The first stage uses Bayesian networks to perform preliminary screening of device relationships. Bayesian networks have limited ability to fuse multi-source data (such as protocol messages, operation and maintenance records, and real-time monitoring data), but they are fast in computation. Therefore, the input of Bayesian networks is structured attribute data and the initial edge weights of new devices and existing devices. These data have relatively clear relationships, and Bayesian networks can quickly make these sparse dependencies explicit, thus achieving preliminary screening of device relationships.

[0123] The second stage, based on the initially screened relationship types, utilizes graph neural networks to precisely refine the relationship types. Graph convolutional neural networks can fuse unstructured data (such as equipment response time-series data, text-based maintenance records, etc.) with structured attributes for inference, fully uncovering deep-seated correlation features between devices, correcting the coarse screening results of Bayesian networks, and improving the accuracy of correlations. This ensures both data processing speed and accuracy, meeting the real-time and accuracy requirements of automatic topology binding of substation auxiliary control equipment locations.

[0124] Specifically, in combination Figure 2 Define device set D ={ d 1, d 2,..., d n (Including monitoring sensors, control equipment, etc.) d 1, d 2,..., d n These represent the device name and the set of device relationship types, respectively. R ={ r 1, r 2,..., r m (such as monitoring coverage, linkage control, spatial attribution, etc.) r 1, r 2,..., r m These represent the relation type and the set of device attributes, respectively. A ={ a 1, a 2,..., a k}, a 1,a 2,..., a k These represent device attributes; the initial set of edge weights W between the newly added device and other devices is { w 1, w 2,..., w t}, w 1, w 2,..., w t These represent the initial edge weights between devices.

[0125] The device set includes existing devices and newly detected devices. The attribute information of the existing and newly detected devices has already been obtained in the preceding steps. The device relationship type is a fixed category predefined based on the substation auxiliary control scenario and does not change with the addition of new devices.

[0126] For any two devices , ∈ D (such as adding monitoring sensors) With control equipment Relationship to be inferred r ∈ R A relationship between two devices can be determined based on a Bayesian network. r Prior probability:

[0127] ;

[0128] in, Indicates equipment Attribute information, Indicates equipment Attribute information;

[0129] P ( r ) is a relation r The prior probability is derived from statistical experience in substation deployment;

[0130] A relationship is known to exist. r Under these conditions, equipment and The likelihood probability of the current attribute combination is obtained through training on historical data;

[0131] For equipment and The marginal probability of attribute combinations is obtained by reasoning using the initial edge weights, and satisfies... .

[0132] Bayesian networks can provide the prior probability of every type of relationship existing between two devices; a probability threshold can be set. (like =0.6), when the calculated prior probability At that time, ( , , As a potential relationship, they enter the second stage.

[0133] The second stage uses a graph neural network (in this embodiment, a graph convolutional neural network) to build a relationship prediction model. Device attributes, historical interaction data, and spatial and logical association information between nodes are integrated into the graph structure for deep feature mining, so as to achieve accurate verification and probability correction of potential relationships.

[0134] This embodiment uses existing and newly added equipment in the substation auxiliary control system as nodes, maintaining the relationship types between existing equipment and constructing edges between them. Based on the potential relationship types between the newly added equipment and other equipment obtained in the first stage, initial edges for the newly added equipment are constructed, and a subgraph containing the newly added equipment and related existing equipment, i.e., the equipment association graph, is built. G =( V , E ); where V represents a node and E represents an edge; the device association graph is used for prediction in graph convolutional neural networks.

[0135] Data standardization and encoding are performed on information such as device type code, normalized installation location coordinates, detection range, protocol type, and initial signal strength to generate node feature vectors. =[ ], Indicates specific characteristics, l For feature dimensions.

[0136] The device association graph, node feature vectors, historical interaction data between nodes (such as past linkage execution records of existing devices, data transmission latency, response success rate, etc.), and spatial and logical association information (such as the GIS region to which the device belongs, the control module affiliation, and related parameters of preset linkage rules, etc.) are used as inputs to the graph convolutional neural network.

[0137] This embodiment uses unstructured data, such as historical interaction data between nodes and spatial and logical association information, along with structured feature vector data, as input to the graph convolutional neural network. This allows the network to learn simultaneously from both the structured features and unstructured associations, resulting in a more accurate representation of relationships. These unstructured data contain rich association information. The graph convolutional neural network automatically learns meaningful implicit features such as "relationship strength" and "interaction type" from this unstructured data and uses them as edge features or weights influencing message passing. Simultaneously, the network can directly aggregate information from the most likely related neighbor nodes, resulting in faster convergence.

[0138] Graph convolutional neural networks aggregate features from neighboring nodes through multiple rounds of graph convolution operations to update node embedding representations.

[0139] ;

[0140] in, For nodes In the t The embedding vector of the layer, N( i ) is a node In the device association diagram G The set of neighboring nodes in; and The first t The weight matrix and bias vector of the convolutional layer; σ (·) represents the activation function; in this embodiment, the ReLU function is used. σ ( x )=max(0, x ); For nodes In the t+ Embedding vectors at layer 1.

[0141] The final node is embedded into the vector through the output layer. and ( T The total number of convolutional layers is mapped to the probability of the existence of a relation:

[0142] ( r | , =Sigmoid(MLP( ⊕ ));

[0143] Where ⊕ represents the vector concatenation operation; MLP (·) is a multilayer perceptron used to fit the mapping relationship between node embeddings and relation probabilities; Sigmoid (·) is the activation function that maps the output to the interval [0,1], i.e., Sigmoid(x) = .

[0144] Similarly, graph convolutional neural networks can be used to obtain the probability of each potential relationship type between a new device and other devices; a probability threshold can then be set. (for example =0.8), when ( r | , )≥ At that time, confirm the equipment and Existence Relationship r It outputs high-confidence relationship results.

[0145] Ultimately, the types of relationships between the newly added device and other devices were obtained, leading to the final topological relationships.

[0146] This embodiment utilizes a relationship inference engine that integrates Bayesian networks and graph convolutional neural networks. It automatically binds new devices to multi-dimensional topologies, including attributes, spatial relationships, logical relationships, and time, without requiring manual configuration. This meets the real-time dynamic binding requirements of equipment location topologies in substation auxiliary control systems.

[0147] S105: Based on the final topology, the topology link binding between the new equipment and the relevant existing equipment is realized in the substation auxiliary control system through the automatic configuration interface, and the graph model is updated synchronously.

[0148] In this embodiment, in order to achieve automatic association of the topology relationship of newly added equipment, a rule knowledge base is constructed in the graph database. The rule knowledge base pre-stores the topology association logic rules of different equipment in the substation auxiliary control system. The topology association logic rules specifically include: rule identifier (ID), rule name (such as "SF6 sensor - explosion-proof fan linkage rule"), rule attributes (such as trigger threshold, priority, etc.), corresponding topology association relationship type, trigger conditions (clearly specifying the equipment type, spatial range, attribute matching conditions, etc. to which the rule applies), and execution actions (defining the specific actions after the conditions are met, such as creating associated edges, updating relationship attributes, marking invalid relationships, etc.).

[0149] By invoking topology association logic rules, the system can automatically establish relationships between devices in the graph database, and then automatically generate association edges between new nodes and other nodes on the attribute graph, without manual intervention. Simultaneously, it can periodically check existing topology relationships, marking invalid or expired associations to prevent erroneous linkages (such as retaining linkage relationships even after sensors have gone offline).

[0150] In this embodiment, based on the relationship type between the newly added equipment and other equipment predicted by the graph convolutional neural network, the corresponding rules are retrieved from the rule knowledge base, and the corresponding association relationship between the newly added equipment and other equipment is established based on the rules, thereby realizing the automatic updating of the graph model of the substation auxiliary control equipment.

[0151] As a specific implementation method, the process of generating the rule knowledge base is as follows:

[0152] (1) Organize the core business of substation auxiliary control scenarios (such as gas leakage linkage and temperature and humidity control) and extract the topological association logic (such as "linkage between sensors and control equipment in the same space").

[0153] (2) Based on the Cypher language, the logic is transformed into structured rules, and the triggering conditions, association types and attributes are defined;

[0154] In this embodiment, the rule knowledge base mainly stores two categories of rules: relationship discovery rules and relationship verification rules. Relationship discovery rules are used to guide the automatic creation of new device topology edges, while relationship verification rules are used to verify the timeliness and validity of existing topology relationships.

[0155] (3) Test the effectiveness of rules by simulating scenarios using rule validation tools to form an initial rule base;

[0156] (4) It supports manually adding and modifying rules through the management system API or visual interface. After the rules are updated, they are synchronized to the graph database in real time without restarting the system.

[0157] The execution process after the rule is invoked is as follows:

[0158] (1) Based on the final relationship type between the newly added device and other devices, match the corresponding scenario rules (such as "SF6 sensor and explosion-proof fan linkage rule in the same area") from the rule knowledge base.

[0159] (2) Verify the consistency between the triggering conditions of the verification rules and the device attributes and spatial information;

[0160] (3) If the conditions are met, execute the topology edge creation or update operation defined in the rules;

[0161] (4) After execution, record the rule execution log (including execution time, involved devices, and relationship type) for subsequent operation and maintenance audit and rule optimization.

[0162] This embodiment uses the Cypher graph query language to write rule fragments, directly connecting to the four-dimensional graph model, supporting out-of-the-box writing and dynamic updating of rules. By constructing a rule knowledge base based on a declarative query language, it stores the topology binding and business reasoning rules for auxiliary control scenarios, enabling flexible configuration and rapid invocation of rules, and providing a logical basis for the automatic creation of topology-related edges.

[0163] This embodiment uses a relationship inference engine that fuses Bayesian networks and graph convolutional neural networks to automatically identify the relationship types between new devices and other devices. By calling matching rules, it automatically establishes the association between devices at the database level. From the automatic identification of associations to the automatic binding of associations between devices, it truly realizes the "plug and play, connect and bind" of new devices, and completes the automatic creation of topological association edges without manual intervention.

[0164] Example 2

[0165] In one or more embodiments, an automatic point topology binding system for a substation auxiliary control system is disclosed, specifically including:

[0166] The attribute graph management module is configured to build and maintain a graph model that represents the equipment, space, and logical relationships within the substation auxiliary control system.

[0167] The device access sensing module is configured to monitor whether a new device is connected, and when a new device is detected, to obtain the attribute information of the new device and the initial edge weight of the new device in the graph model.

[0168] The association reasoning module is configured to predict the potential association between the new device and existing devices in the system using a Bayesian network based on the attribute information of the new device and the initial edge weight of the new device in the graph model; construct a subgraph containing the new device and related existing devices using the potential association as constraints; and use a graph neural network to verify and refine the potential association to obtain the final topological relationship.

[0169] The topology binding execution module is configured to bind the topology links between the newly added equipment and the relevant existing equipment in the substation auxiliary control system according to the final topology relationship through an automated configuration interface, and to synchronously update the graph model.

[0170] Furthermore, a graphical model representing the equipment, space, and logical relationships within the substation auxiliary control system is constructed and maintained, specifically as follows:

[0171] Using equipment nodes, spatial nodes, logical nodes, and time nodes as nodes in the graph model, and the physical connection edges, logical association edges, and spatial inclusion edges between nodes as directed edges in the graph model, a graph model of substation auxiliary control equipment is constructed. Among them, equipment nodes represent substation auxiliary control equipment and attribute information, spatial nodes represent substation physical area information, logical nodes represent control logic information, and time nodes represent equipment full life cycle status information.

[0172] It should be noted that the specific implementation methods of the above modules are the same as those in Example 1, and will not be described in detail again.

[0173] Example 3

[0174] In one or more embodiments, a terminal device is disclosed, comprising a processor and a memory, wherein the processor is used to implement instructions; and the memory is used to store multiple instructions adapted to be loaded by the processor and executed by the processor to automatically bind the substation auxiliary control system point topology method of Embodiment 1.

[0175] Example 4

[0176] In one or more embodiments, a computer-readable storage medium is disclosed, wherein a plurality of instructions are stored, the instructions being adapted to be loaded by a processor of a terminal device and executed by the automatic binding method for substation auxiliary control system point topology in Embodiment 1.

[0177] Example 5

[0178] In one or more embodiments, a computer program product is disclosed, including a computer program that, when executed by a processor, implements the automatic binding method for substation auxiliary control system point topology in Embodiment 1.

[0179] In such an embodiment, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by the central processing unit, it performs the various functions defined in the steps of Embodiment 1. Alternatively, the computer program instructions can also be stored in a computer-readable storage medium capable of directing a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium implement the operational process of Embodiment 1.

[0180] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for automatic binding of point topology in a substation auxiliary control system, characterized in that, include: Construct and maintain a graphical model that represents the equipment, space, and logical relationships within the substation auxiliary control system; Specifically, the graph model of substation auxiliary control equipment is constructed by using equipment nodes, spatial nodes, logical nodes, and time nodes as nodes in the graph model, and using the physical connection edges, logical association edges, and spatial inclusion edges between nodes as directed edges in the graph model. When a new device is detected to be connected, the attribute information of the new device and the initial edge weight of the new device in the graph model are obtained. Based on the attribute information of the newly added device and the initial edge weight of the newly added device in the graph model, the potential association relationship between the newly added device and the existing devices in the system is predicted by Bayesian network. Using the potential relationships as constraints, a subgraph containing the new device and related existing devices is constructed, and a graph neural network is used to verify and refine the potential relationships to obtain the final topological relationships; Based on the final topology, the topology link binding between the newly added equipment and the relevant existing equipment is implemented in the substation auxiliary control system through the automated configuration interface, and the graph model is updated synchronously. Specifically, this involves constructing and maintaining a rule knowledge base, which pre-stores topology association logic rules for different devices in the substation auxiliary control system; the topology association logic rules include: rule identifier, rule name, rule attributes, corresponding topology association type, triggering condition, and execution action; Based on the relationship type between the new equipment and the existing equipment predicted by the graph neural network, the corresponding rules are retrieved from the rule knowledge base. Based on the rules, the topology link binding between the new equipment and the relevant existing equipment is realized, thereby realizing the automatic update of the graph model of the substation auxiliary control equipment.

2. The automatic binding method for point topology of a substation auxiliary control system as described in claim 1, characterized in that, Equipment nodes represent the auxiliary control equipment and attribute information of the substation, spatial nodes represent the physical area information of the substation, logical nodes represent the control logic information, and time nodes represent the status information of the equipment throughout its entire life cycle.

3. The automatic binding method for point topology of a substation auxiliary control system as described in claim 1, characterized in that, The specific method for determining the edge weights of the physical connection edges, logical association edges, and spatially contained edges is as follows: Based on the edge type and business priority, set the basic weight value for each type of edge; For physical connection edges, calculate the connection continuity correction coefficient, interface health correction coefficient, line length matching correction coefficient, and redundant connection number correction coefficient respectively; obtain the total weight correction coefficient of the physical connection edge by weighted summation; calculate the product of the basic weight value of the physical connection edge and the total weight correction coefficient to obtain the final edge weight of the physical connection edge. For logically related edges, calculate the scene matching degree correction coefficient, historical linkage success rate correction coefficient, device online rate correction coefficient, and response timeliness correction coefficient respectively; The total weight correction coefficient of logically related edges is obtained by weighted summation; The final edge weight of the logically related edge is obtained by multiplying the base weight value of the logically related edge by the total weight correction coefficient. For spatially contained edges, calculate the location matching degree correction coefficient, binding duration ratio correction coefficient, region adaptation degree correction coefficient, and movement frequency correction coefficient respectively; obtain the total weight correction coefficient of the spatially contained edges by weighted summation; The final edge weights of the spatially contained edges are obtained by multiplying the basic weight values ​​of the edges with the total weight correction coefficient. When the device location changes, the operating status changes, or the business rules are adjusted, the edge weights are automatically updated according to the changes in node attributes.

4. The automatic binding method for point topology of a substation auxiliary control system as described in claim 1, characterized in that, When a new device is detected to be connected, the attribute information of the new device is obtained, specifically: The type of newly added device can be identified by its IP address, MAC address, or protocol type information; By matching the unique device identifier or judging the overlap of key attributes, it is determined whether data from different data sources are the same device information. The same device information from multiple data sources is integrated and standardized to form unified device attribute information.

5. The automatic binding method for point topology of a substation auxiliary control system as described in claim 1, characterized in that, The potential relationships between the new device and existing devices in the system are predicted using a Bayesian network, specifically as follows: Based on the attribute information of the newly added device and the initial edge weight of the newly added device in the graph model, the prior probability of different relationship types between the newly added device and the existing device is predicted by Bayesian network; the relationship type with a prior probability greater than a first predetermined threshold is taken as the potential relationship type between the newly added device and the existing device.

6. The automatic binding method for point topology of a substation auxiliary control system as described in claim 1, characterized in that, The construction process of the rule knowledge base includes: Extract the topological association logic of equipment locations in the substation auxiliary control system, transform the logic into structured rules, and define the association type, triggering conditions, triggering thresholds, execution actions, and priorities corresponding to each rule; By simulating different scenarios using rule validation tools, the effectiveness of each rule is tested, ultimately forming a rule knowledge base.

7. An automatic point topology binding system for a substation auxiliary control system, characterized in that, include: The attribute graph management module is configured to build and maintain a graph model that represents the equipment, space, and logical relationships within the substation auxiliary control system. Specifically, the graph model of substation auxiliary control equipment is constructed by using equipment nodes, spatial nodes, logical nodes, and time nodes as nodes in the graph model, and using the physical connection edges, logical association edges, and spatial inclusion edges between nodes as directed edges in the graph model. The device access sensing module is configured to monitor whether a new device is connected, and when a new device is detected, to obtain the attribute information of the new device and the initial edge weight of the new device in the graph model. The association reasoning module is configured to predict the potential association between the new device and existing devices in the system using a Bayesian network based on the attribute information of the new device and the initial edge weight of the new device in the graph model; construct a subgraph containing the new device and related existing devices using the potential association as constraints; and use a graph neural network to verify and refine the potential association to obtain the final topological relationship. The topology binding execution module is configured to bind the topology links between the newly added equipment and the relevant existing equipment in the substation auxiliary control system according to the final topology relationship through an automated configuration interface, and to synchronously update the graph model. Specifically, this involves constructing and maintaining a rule knowledge base, which pre-stores topology association logic rules for different devices in the substation auxiliary control system; the topology association logic rules include: rule identifier, rule name, rule attributes, corresponding topology association type, triggering condition, and execution action; Based on the relationship type between the new equipment and the existing equipment predicted by the graph neural network, the corresponding rules are retrieved from the rule knowledge base. Based on the rules, the topology link binding between the new equipment and the relevant existing equipment is realized, thereby realizing the automatic update of the graph model of the substation auxiliary control equipment.

8. The automatic point topology binding system for a substation auxiliary control system as described in claim 7, characterized in that, Equipment nodes represent the auxiliary control equipment and attribute information of the substation, spatial nodes represent the physical area information of the substation, logical nodes represent the control logic information, and time nodes represent the status information of the equipment throughout its entire life cycle.

9. A terminal device comprising a processor and a memory, the processor for implementing instructions; the memory for storing multiple instructions, characterized in that, The instructions are adapted to be loaded by a processor and executed by the automatic binding method for substation auxiliary control system point topology as described in any one of claims 1-6.

10. A computer-readable storage medium storing a plurality of instructions, characterized in that, The instructions are adapted to be loaded and executed by the processor of the terminal device using the automatic binding method for substation auxiliary control system point topology as described in any one of claims 1-6.

11. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the automatic binding method for point topology of the substation auxiliary control system as described in any one of claims 1-6.

Images