Power line carrier networking method and device based on graph attention, equipment and medium

Abstract

The application discloses a power line carrier networking method and device based on graph attention, equipment and medium. The method obtains the current communication path of each communication node and the physical channel characteristics of each communication link in the power line carrier communication system, determines the abnormal communication link and the affected communication node according to the physical channel characteristics and the preset abnormal condition; determines the associated communication node, the node characteristics and the edge characteristics according to the network state information and the physical channel characteristics, updates the historical dynamic attribute graph, and obtains the real-time dynamic attribute graph; inputs the real-time dynamic attribute graph into a graph attention network model, determines the connectivity probability of each communication link, determines the standby communication path accordingly, and controls the affected communication node to switch to the standby communication path, so that the communication path reconstruction is completed before the actual failure of the abnormal link, and the continuity and reliability of the networking communication are improved.

Description

Technical Field

[0001] This application relates to the field of power line communication technology, and in particular to a method, apparatus, device and medium for power line carrier networking based on graph attention. Background Technology

[0002] With the development of smart grids, power line carrier communication has been widely used in the communication networking of distribution substations due to its ability to reuse existing power cables. However, affected by factors such as frequent switching of nonlinear loads, power line channels are prone to increased impulse noise and sudden changes in input impedance, making the communication link highly time-varying and bursty. Existing power line carrier networks mainly rely on passive reconstruction based on communication results such as packet loss and link breakage, and routing decisions cannot fully utilize the underlying physical channel change information, making it difficult for the system to adjust the communication path in time before the link actually fails.

[0003] Existing power line carrier networking methods typically trigger route reconstruction based on communication results such as acknowledgment timeouts, packet loss, or link disconnection. This makes it difficult to identify link degradation trends in advance when physical channel conditions such as sudden changes in input impedance or increased impulse noise occur. Consequently, the adjustment of communication paths lags behind actual link failures, which can easily lead to data transmission interruptions. Summary of the Invention

[0004] This application provides a graph attention-based power line carrier networking method, apparatus, device, and medium, which aims to proactively reconstruct the communication path before the actual failure of the communication link when abnormal physical channel characteristics of the link are identified, thereby reducing the risk of communication interruption caused by sudden link failure.

[0005] To achieve the above objectives, embodiments of this application provide a power line carrier networking method based on graph attention, including: The current communication path of each communication node and the physical channel characteristics of each communication link in the power line carrier communication system are obtained, wherein the power line carrier communication system includes multiple communication nodes and multiple communication links; Based on the physical channel characteristics and preset abnormal conditions, abnormal communication links are determined from multiple communication links, and the affected communication nodes corresponding to the abnormal communication links are determined from multiple communication nodes, and the network status information of the power line carrier communication system is obtained. Based on the network status information and the physical channel characteristics of each communication link, the associated communication node corresponding to the affected communication node is determined, and the node characteristics and edge characteristics of each communication link corresponding to the affected communication node and the associated communication node are determined respectively. Based on the node features and the edge features, the historical dynamic attribute map of the power line carrier communication system is updated to obtain the real-time dynamic attribute map. The real-time dynamic attribute graph is input into a preset graph attention network model to determine the connectivity probability of each communication link. Based on the connectivity probability corresponding to each of the communication links, determine the backup communication path corresponding to the affected communication node; The communication path of the affected communication node is switched from the current communication path to the backup communication path.

[0006] This application, by acquiring the current communication path of each communication node and the physical channel characteristics of each communication link in a power line carrier communication system, can clearly identify the communication path currently used by each communication node and continuously sense the physical state of each communication link, providing a basis for abnormal link identification and subsequent path adjustment. Based on the physical channel characteristics and preset abnormal conditions, abnormal communication links are identified, and the corresponding affected communication nodes and network status information are further determined. This allows for timely location of the affected objects of link anomalies and acquisition of the operating status of the power line carrier communication system at the time of the anomaly. Based on the network status information and physical channel characteristics, associated communication nodes, node characteristics, and edge characteristics are determined, enabling the identification of associated nodes participating in backup path selection and converting the operating status of communication nodes and the physical state of communication links into feature information suitable for graph structure processing. The system updates the historical dynamic attribute graph based on the node and edge characteristics, enabling the formation of a real-time dynamic attribute graph reflecting the current node and link status while preserving the original network topology. Inputting this real-time dynamic attribute graph into a graph attention network model determines the connectivity probability of each communication link, allowing for a comprehensive evaluation of the availability of each communication link by considering the relationships between communication nodes and the status of the communication links. Based on the connectivity probability, the system determines backup communication paths for affected communication nodes, enabling the selection of reliable alternative paths. Furthermore, it controls the switching of affected communication nodes from their current communication path to their backup communication path, allowing for proactive reconstruction of communication paths before further degradation or actual failure of abnormal communication links, reducing the risk of communication interruption caused by sudden link failures, and improving the continuity and reliability of power line carrier network communication.

[0007] Furthermore, the methods for constructing historical dynamic attribute graphs include: Obtain historical network status information of the power line carrier communication system and historical physical channel characteristics of each of the communication links; Based on the historical network state information and the historical physical channel characteristics of each communication link, determine the node characteristics corresponding to each communication node and the edge characteristics corresponding to each communication link; Obtain the topological connection relationship between each of the communication nodes, and map each of the communication nodes as graph nodes according to the topological connection relationship, and map each of the communication links as graph edges connecting the corresponding graph nodes; The historical dynamic attribute graph is constructed by using the node features as node attributes of the corresponding graph nodes and the edge features as edge attributes of the corresponding graph edges.

[0008] By implementing the above scheme, historical network status information of the power line carrier communication system and historical physical channel characteristics of each communication link can be obtained, thus providing a data foundation for establishing a network status benchmark. Based on the historical network status information and historical physical channel characteristics, node and edge features can be determined, converting the historical operating status of communication nodes and the historical physical status of communication links into feature information that can be processed by a graph model. By obtaining the topological connection relationships between communication nodes and mapping communication nodes as graph nodes and communication links as graph edges connecting corresponding graph nodes, the actual connection relationships of the power line carrier communication system can be preserved in the graph structure. Furthermore, by using node features as node attributes and edge features as edge attributes, a historical dynamic attribute graph that simultaneously represents historical node status, historical link status, and network topology can be constructed, providing a basis for determining the current communication path and updating the graph status after an anomaly occurs.

[0009] Furthermore, before obtaining the current communication path of each communication node and the physical channel characteristics of each communication link in the power line carrier communication system, the following steps are also included: The historical dynamic attribute graph is input into a preset graph attention network model to determine the connectivity probability of each communication link. The routing cost corresponding to each communication link is determined based on the connectivity probability of each communication link. Using the central coordinator in the power line carrier communication system as the root node, the target communication path from the central coordinator to each of the communication nodes is determined based on the routing cost corresponding to each of the communication links. A communication routing tree is constructed based on each of the target communication paths, and each of the target communication paths is determined as the current communication path corresponding to the corresponding communication node.

[0010] By inputting historical dynamic attribute graphs into a preset graph attention network model, the connectivity of each communication link can be comprehensively evaluated based on the node attributes, edge attributes, and graph connection relationships before the anomaly occurs, yielding the connectivity probability of each communication link. The routing cost is determined based on the connectivity probability, converting the link connectivity evaluation results into a unified index for path optimization. Using the central coordinator as the root node and determining the target communication path from the central coordinator to each communication node based on the routing cost of each communication link, the initial communication path with higher connectivity can be selected for each communication node. A communication routing tree is constructed based on each target communication path, forming a hierarchical communication relationship between each communication node and the central coordinator, and reducing the possibility of path loops. Determining each target communication path as the current communication path for the corresponding communication node provides a path benchmark for normal communication of communication nodes and for switching to backup paths after anomalies.

[0011] Further, based on the network state information and the physical channel characteristics of each communication link, the node characteristics corresponding to the affected communication node and its associated communication node, and the edge characteristics of each communication link are determined, including: The network status information includes node operation information and link status information; Based on the network status information and the physical channel characteristics of each communication link, the associated communication node corresponding to the affected communication node is determined; Obtain the node operation information corresponding to the affected communication node and its associated communication node from the network status information, and determine the node characteristics corresponding to the affected communication node and its associated communication node; For each communication link, the link state information of the communication link is obtained from the network state information. The physical channel features and link state information of the communication link are time-synchronized and standardized. The processed physical channel features and link state information are then fused to obtain the edge features corresponding to the communication link.

[0012] Implementing the above scheme, by dividing network status information into node operation information and link status information, the current operating status of the power line carrier communication system can be described from both the communication node and communication link levels. Based on the network status information and the physical channel characteristics of each communication link, the associated communication nodes corresponding to the affected communication nodes can be determined, clarifying the range of nodes that are communicationally associated with the affected communication nodes and can participate in subsequent path reconstruction. Obtaining the node operation information of the affected and associated communication nodes from the network status information and determining node characteristics allows the node characteristics to reflect the current operating and data processing status of the relevant communication nodes. Obtaining the link status information for each communication link and synchronizing the physical channel characteristics and link status information in time reduces status deviations caused by different acquisition times. Standardization processing reduces the impact of differences in dimensions and numerical ranges between different types of features. Further feature fusion of the processed physical channel characteristics and link status information forms side features that simultaneously reflect the underlying physical state and actual communication state of the communication link, improving the completeness of the side features' representation of the current link state.

[0013] Further, based on the node features and the edge features, the historical dynamic attribute map of the power line carrier communication system is updated to obtain a real-time dynamic attribute map, including: From the historical dynamic attribute graph, determine the graph nodes corresponding to the affected communication nodes and the associated communication nodes, as well as the graph edges corresponding to each communication link; Update the node attributes of the corresponding graph nodes based on the node characteristics of the affected communication nodes and the associated communication nodes; Based on the edge characteristics corresponding to each communication link, the edge attributes of the corresponding graph edges are updated to obtain the real-time dynamic attribute graph.

[0014] By implementing the above scheme, the graph nodes corresponding to the affected and associated communication nodes can be determined from the historical dynamic attribute graph, enabling accurate location of graph structure objects that need to be updated based on the current node status. By determining the graph edges corresponding to each communication link, the correspondence between the current edge features of each communication link and the corresponding graph edges in the historical dynamic attribute graph can be established. Updating the node attributes of the corresponding graph nodes based on the node features of the affected and associated communication nodes allows the graph nodes to reflect the current operating status of the relevant communication nodes when the anomaly occurs. Updating the edge attributes of the corresponding graph edges based on the edge features of each communication link allows the graph edges to reflect the current physical and communication status of each communication link. The resulting real-time dynamic attribute graph can reflect the node and link status after the anomaly occurs while retaining the original network topology, providing a graph model input that is more consistent with the current network status for recalculating the link connectivity probability.

[0015] Further, the real-time dynamic attribute graph is input into a preset graph attention network model to determine the connectivity probability corresponding to each communication link, including: For any given graph node, determine each associated graph node connected to that graph node, and the graph edges between that graph node and each of the associated graph nodes; For any graph node, the attention relevance of the graph node to each of the associated graph nodes is determined based on the node attributes of the graph node, the node attributes of each of the associated graph nodes, and the edge attributes of the graph edges between the graph node and each of the associated graph nodes. For any of the graph nodes, the attention relevance of the graph node to each of the associated graph nodes is normalized to obtain the attention weight of the graph node to each of the associated graph nodes. For any of the graph nodes, based on the attention weight of the graph node to each of the associated graph nodes, the node attributes of each of the associated graph nodes are aggregated to obtain the node aggregated features corresponding to the graph node. For any of the graph edges, the connectivity probability of the communication link corresponding to the graph edge is determined based on the node aggregation features of the graph nodes at both ends of the graph edge and the edge attributes of the graph edge.

[0016] By implementing the above scheme, by identifying all associated graph nodes connected to any given graph node and the graph edges between them, the neighboring nodes and corresponding link ranges involved in graph attention calculation can be clearly defined. Attention relevance is determined based on the node attributes of the current graph node, the node attributes of each associated graph node, and the edge attributes of the corresponding graph edges. This allows for the simultaneous consideration of the states of the nodes at both ends of the link and the state of the link itself when evaluating the degree of communication association between graph nodes. Normalization of each attention relevance results in comparable attention weights, distinguishing the degree of influence of different associated graph nodes and corresponding links on the current graph node. Feature aggregation of the node attributes of each associated graph node based on the attention weights allows the fusion of the state information of neighboring nodes into the current graph node according to their importance, obtaining node aggregation features reflecting the local network environment. Furthermore, the connectivity probability of the communication link is determined based on the node aggregation features of the graph nodes at both ends of the graph edge and the edge attributes of that graph edge. This allows for a comprehensive quantitative evaluation of the availability of the communication link by considering both the local network states at both ends of the link and the state of the link itself.

[0017] Further, based on the connectivity probability corresponding to each of the communication links, a backup communication path corresponding to the affected communication node is determined, including: For each of the communication links, the routing cost corresponding to the communication link is determined based on the connectivity probability of the communication link. The central coordinator in the power line carrier communication system is determined as the root node, and an alternative communication path from the central coordinator to the affected communication node is determined based on the routing cost corresponding to each communication link.

[0018] By implementing the above scheme, the link connectivity evaluation results output by the graph attention network can be converted into cost indicators that can be directly used in the path optimization process by determining the corresponding routing cost based on the connectivity probability of each communication link. This allows different communication links to participate in path comparison according to their connectivity reliability. By determining the central coordinator as the root node, the communication convergence target of the backup path can be unified. By determining the backup communication path from the central coordinator to the affected communication node based on the routing cost of each communication link, communication links with low connectivity probability or abnormal risks can be avoided, and a more reliable alternative communication path can be determined for the affected communication node, thus providing a path basis for it to switch from the current communication path to the backup communication path.

[0019] Secondly, this application provides a power line carrier networking device based on graph attention, characterized in that it includes: a feature acquisition module, an anomaly judgment module, a feature determination module, a graph update module, a probability calculation module, a path determination module, and a path switching module; The feature acquisition module is used to acquire the current communication path of each communication node and the physical channel characteristics of each communication link in the power line carrier communication system, wherein the power line carrier communication system includes multiple communication nodes and multiple communication links. The anomaly detection module is used to determine the abnormal communication link from the multiple communication links based on the physical channel characteristics and preset anomaly conditions, and to determine the affected communication node corresponding to the abnormal communication link from the multiple communication nodes and obtain the network status information of the power line carrier communication system. The feature determination module is used to determine the associated communication node corresponding to the affected communication node based on the network status information and the physical channel characteristics of each communication link, and to determine the node characteristics and edge characteristics of each communication link corresponding to the affected communication node and the associated communication node, respectively. The graph update module is used to update the historical dynamic attribute graph of the power line carrier communication system according to the node features and the edge features, so as to obtain the real-time dynamic attribute graph. The probability calculation module is used to input the real-time dynamic attribute graph into a preset graph attention network model to determine the connectivity probability corresponding to each communication link. The path determination module is used to determine the backup communication path corresponding to the affected communication node based on the connectivity probability corresponding to each communication link. The path switching module is used to control the communication path of the affected communication node to switch from the current communication path to the backup communication path.

[0020] Thirdly, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the graph attention-based power line carrier networking method described in the above application.

[0021] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the graph attention-based power line carrier networking method described in the above application. Attached Figure Description

[0022] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. The drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating a power line carrier networking method based on graph attention, provided in an embodiment of this application. Figure 2 This is a schematic diagram of the operation process of a power line carrier network according to an embodiment of this application; Figure 3 This is a schematic diagram of the graph attention link evaluation and topology reconstruction execution process provided in an embodiment of this application; Figure 4 This is a schematic diagram of a power line carrier networking device based on graph attention, provided in another embodiment of this application. Detailed Implementation

[0024] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The purpose of providing these embodiments is to make the disclosure of this application more thorough and comprehensive. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0025] In the description of this application, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first," "second," "third," etc., may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0026] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to the internal communication between two components. The terms "vertical," "horizontal," "left," "right," "upper," "lower," and similar expressions used herein are for illustrative purposes only and do not indicate or imply that the device or component referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as limiting this application. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0027] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used in this specification is for the purpose of describing specific embodiments only and is not intended to limit this application. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0028] To address the communication interruption caused by sudden link failures when the physical channel characteristics of the link are abnormal in existing technologies, such as... Figure 1 As shown in the figure, this application provides a schematic flowchart of a power line carrier networking method based on graph attention. The power line carrier networking method based on graph attention in this embodiment includes steps S101 to S107, which are detailed below: Step S101: Obtain the current communication path of each communication node and the physical channel characteristics of each communication link in the power line carrier communication system, wherein the power line carrier communication system includes multiple communication nodes and multiple communication links.

[0029] In this embodiment, the PLC node is the communication node, and the physical connectivity branch is the communication link; the time-domain impedance gradient, the frequency-domain impulse noise peak power, and the channel state information are all physical channel characteristics. The following description will continue to use terms such as PLC node and physical connectivity branch.

[0030] In this embodiment, the process of acquiring the physical channel characteristics of each communication link and subsequently forming the edge characteristics of each communication link is collaboratively completed by the underlying physical layer interface and the media access control layer of each PLC node. Specifically, the underlying physical layer interface is used to collect port voltage, port current, background noise, channel state information, and received signal strength indication. The media access control layer is used to collect data packet delivery rate based on the data frame transmission results, reception results, acknowledgment results, and retransmission status. The underlying physical layer interface and the media access control layer respectively provide the acquired data to the feature processing unit to form physical channel characteristics and link state information characterizing the communication link status.

[0031] Specifically, the time-domain impedance transition gradient is extracted. For PLC nodes... With PLC node The physical connection between the branches allows the PLC node to collect the port voltage of its transmitting port in real time. and port current And according to the port voltage and port current Determine the current carrier frequency instantaneous input impedance The instantaneous input impedance The specific calculation method is as follows: in, Indicates at time The voltage at the transmitting port was collected; Indicates at time The current at the transmitting port is collected; Indicates the time of signal acquisition; Indicates the carrier frequency for power line carrier communication; Indicates at time and carrier frequency The instantaneous input impedance.

[0032] Furthermore, based on the change of the instantaneous input impedance over time, the PLC nodes are determined. With PLC node Temporal impedance jump gradient of the physically connected components The specific calculation formula is as follows: in, and These represent the PLC nodes at both ends of the physically connected branch; PLC node With PLC node The time-domain impedance jump gradient of the physically connected component is used to characterize the degree to which the input impedance of the physically connected component changes over time.

[0033] It is understandable that when a high-power load is switched on or off in a power line carrier communication system, the impedance of the power line network may change significantly within a short period of time. Therefore, the time-domain impedance transition gradient can be used to characterize the physical channel environment anomalies caused by load switching.

[0034] Furthermore, the peak power of broadband impulse noise in the frequency domain is estimated. The PLC node collects the channel background noise spectrum of the physically connected branch within a preset idle time slot, and determines the peak power of the impulse noise of that physically connected branch based on the received signal, transmitted signal, and channel frequency domain response. The specific calculation method is as follows: in, Indicates the receiver is at frequency The received signal obtained at the location; Indicates the frequency of the transmitting end. The transmitted signal was sent from the location; Indicates the physical connectivity component at frequency Channel frequency domain response at the location; This indicates the frequency or subcarrier frequency in an orthogonal frequency division multiplexing system.

[0035] Furthermore, channel state information is acquired. Based on the signal transmission status on each orthogonal frequency division multiplexing subcarrier, the PLC node acquires the channel state information of the corresponding physically connected branch. ,in, It can include channel amplitude, channel phase or channel response corresponding to multiple subcarriers, used to characterize the signal transmission status of the physical connectivity branch on different subcarriers.

[0036] Therefore, PLC node With PLC node The physical channel characteristics of the physically connected branches include the time-domain impedance gradient. Peak power of impulse noise and channel state information .

[0037] Step S102: Based on the physical channel characteristics and preset abnormal conditions, determine the abnormal communication link from the multiple communication links, determine the affected communication node corresponding to the abnormal communication link from the multiple communication nodes, and obtain the network status information of the power line carrier communication system.

[0038] Step S103: Based on the network status information and the physical channel characteristics of each communication link, determine the associated communication node corresponding to the affected communication node, and determine the node characteristics and edge characteristics of each communication link corresponding to the affected communication node and the associated communication node, respectively.

[0039] As a preferred embodiment, based on the network state information and the physical channel characteristics of each communication link, the node characteristics corresponding to the affected communication node and its associated communication node, and the edge characteristics of each communication link are determined, including: The network status information includes node operation information and link status information; Obtain the node operation information corresponding to the affected communication node and its associated communication node from the network status information, and determine the node characteristics corresponding to the affected communication node and its associated communication node; For each communication link, the link state information of the communication link is obtained from the network state information. The physical channel features and link state information of the communication link are time-synchronized and standardized. The processed physical channel features and link state information are then fused to obtain the edge features corresponding to the communication link.

[0040] Implementing the above-mentioned preferred scheme, by dividing network status information into node operation information and link status information, the current operating status of the power line carrier communication system can be described from both the communication node and communication link levels. Based on the network status information and the physical channel characteristics of each communication link, the associated communication nodes corresponding to the affected communication nodes can be determined, clarifying the range of nodes that are communicationally associated with the affected communication nodes and can participate in subsequent path reconstruction. Obtaining the node operation information of the affected and associated communication nodes from the network status information and determining node characteristics enables the node characteristics to reflect the current operating and data processing status of the relevant communication nodes. Obtaining the link status information for each communication link and synchronizing the physical channel characteristics and link status information in time reduces status deviations caused by different acquisition times. Standardization processing reduces the impact of differences in dimensions and numerical ranges between different types of features. Further feature fusion of the processed physical channel characteristics and link status information forms side features that simultaneously reflect the underlying physical state and actual communication state of the communication link, improving the completeness of the side features' representation of the current link state.

[0041] Step S104: Update the historical dynamic attribute map of the power line carrier communication system according to the node features and the edge features to obtain the real-time dynamic attribute map.

[0042] As a preferred embodiment, the pre-constructed dynamic attribute graph of the entire network is updated based on the node features and the edge features to obtain the updated dynamic attribute graph, including: From the dynamic attribute graph of the entire network, determine the graph nodes corresponding to the affected communication nodes and their associated communication nodes, as well as the graph edges corresponding to each communication link; Based on the node characteristics corresponding to the affected communication nodes and their associated communication nodes, update the node attributes of the corresponding graph nodes; Based on the edge characteristics corresponding to each communication link, the edge attributes of the corresponding graph edges are updated to obtain the updated dynamic attribute graph.

[0043] In this embodiment, the received signal strength indication and historical data packet delivery rate are the link state information; the initial edge feature vector is the edge feature.

[0044] Continuing with the above embodiments, for PLC nodes With PLC node For any physically connected branch, the time-domain impedance transition gradient of that physically connected branch is obtained from the underlying physical layer interface of the PLC node. Peak power of impulse noise and channel state information The received signal strength indication of the physical connectivity branch is obtained by the media access control layer of the PLC node. and historical data packet delivery rate Furthermore, the time-domain impedance transition gradient, impulse noise peak power, channel state information, received signal strength indication, and historical data packet delivery rate are time-aligned so that each feature corresponds to the same or matching feature acquisition time window.

[0045] To reduce the differences in dimensions and numerical ranges among different types of features, the time-domain impedance transition gradient, impulse noise peak power, and received signal strength indication are subjected to maximum-min normalization. The normalized physical channel features are then concatenated with link state information to obtain the physical connectivity branch. Corresponding initial edge feature vector The specific calculation formula is as follows: in, Represents physically connected components The initial edge feature vector; This represents the time-domain impedance jump gradient of this branch; Indicates the peak power of the frequency domain impulse noise; Represents physical layer channel state information; Indicates received signal strength; Indicates the historical data packet delivery rate; , and These are the maximum and minimum normalized constants corresponding to impedance transition, noise power, and received signal strength, respectively.

[0046] Furthermore, during continuous communication in the power line carrier communication system, each PLC node continuously monitors the physical channel characteristics of each active communication branch in the current communication path through the underlying physical layer interface. A set of security thresholds corresponding to the physical channel characteristics is pre-set in the central coordinator or the corresponding communication node.

[0047] In this embodiment, the set of security thresholds includes an impedance jump gradient threshold and an impulse noise power threshold. When the time-domain impedance jump gradient or impulse noise peak power of any active communication branch meets the corresponding anomaly judgment condition, the active communication branch is determined to be an abnormal communication link.

[0048] In this embodiment, the anomaly detection condition is specifically as follows: in, For communication nodes and communication nodes The time-domain impedance jump gradient of the communication link between them; For communication nodes and communication nodes Peak impulse noise power of the communication link between them; This indicates the preset impedance transition gradient threshold; This indicates the preset impulse noise power threshold.

[0049] Optionally, the anomaly determination result can be converted into a software trigger event or a hardware interrupt signal to trigger the central coordinator to perform the path reconstruction process corresponding to the abnormal communication link.

[0050] By implementing the above-mentioned preferred scheme, the graph nodes corresponding to the affected and associated communication nodes can be determined from the historical dynamic attribute graph, enabling accurate location of graph structure objects that need to be updated based on the current node status. By determining the graph edges corresponding to each communication link, the correspondence between the current edge features of each communication link and the corresponding graph edges in the historical dynamic attribute graph can be established. Updating the node attributes of the corresponding graph nodes based on the node features of the affected and associated communication nodes allows the graph nodes to reflect the current operating status of the relevant communication nodes when the anomaly occurs. Updating the edge attributes of the corresponding graph edges based on the edge features of each communication link allows the graph edges to reflect the current physical and communication status of each communication link. The resulting real-time dynamic attribute graph can reflect the node and link status after the anomaly occurs while retaining the original network topology, providing a graph model input that is more consistent with the current network status for recalculating the link connectivity probability.

[0051] Step S105: Input the real-time dynamic attribute graph into a preset graph attention network model to determine the connectivity probability corresponding to each communication link.

[0052] As a preferred embodiment, the updated dynamic attribute graph is input into a preset graph attention network model to determine the connectivity probability corresponding to each communication link, including: For any given graph node, the attention relevance of the graph node to each of the associated graph nodes is determined based on the node attributes of the graph node, the node attributes of each associated graph node connected to the graph node, and the edge attributes of the graph edges between the graph node and each of the associated graph nodes. For any of the graph nodes, the attention relevance of the graph node to each of the associated graph nodes is normalized to obtain the attention weight of the graph node to each of the associated graph nodes. For any graph node, based on the attention weight of the graph node to each associated graph node, the node attributes of each associated graph node are aggregated to obtain the node aggregated features corresponding to the graph node. For any of the graph edges, the connectivity probability of the communication link corresponding to the graph edge is determined based on the node aggregation features of the graph nodes at both ends of the graph edge and the edge attributes of the graph edge.

[0053] In this embodiment, the original feature vector of a node is the node attribute of the graph node; the edge attribute matrix is ​​the edge attribute of the graph edge; the first-order physical neighbor is the associated graph node; the non-normalized attention score is the attention relevance; the final attention weight is the attention weight; and the updated representation of the node is the node aggregation feature.

[0054] It should be noted that the central coordinator corresponds to the central coordinator used in the power line carrier communication system to perform network control and path calculation, and the edge computing unit of the central coordinator is used to run the preset graph attention network model.

[0055] In this embodiment, after the real-time dynamic attribute graph is input into the preset graph attention network model, the preset graph attention network model performs node feature transformation, attention relevance calculation, attention weight normalization, and neighbor feature aggregation in sequence according to the node attributes of each graph node, the edge attributes of each graph edge, and the connection relationship between graph nodes, to obtain the node aggregation features corresponding to each graph node.

[0056] For any graph node in the real-time dynamic attribute graph Obtain the original feature vector of the node in the graph. The original feature vectors of the nodes are transformed by linear mapping to obtain graph nodes. The corresponding high-dimensional hidden state vector Among them, the high-dimensional hidden state vector It can be determined according to the following formula: in, Represents a node The high-dimensional hidden state (hidden feature) vector generated after linear mapping; This represents the shared, learnable weight matrix used for node feature dimension transformation; Represents a node The original state feature vector.

[0057] In the same way, for graph nodes Linear mapping is performed on all connected graph nodes to obtain the high-dimensional hidden state vector corresponding to each graph node. By applying the same linear mapping to the original feature vectors of each graph node, the node attributes of different graph nodes can be transformed into the same latent feature space, providing a unified data dimension for subsequent calculation of the attention correlation between graph nodes.

[0058] For any graph node i, based on the graph edge connections in the real-time dynamic attribute graph, determine all associated graph nodes connected to graph node i, and form a set of first-order physical neighbors for graph node i. .

[0059] For a first-order physical neighbor set For any associated graph node j, obtain the high-dimensional hidden state vector of graph node i. The high-dimensional hidden state vector of the associated graph node j, and the edge attribute matrix of the graph edges between graph node i and associated graph node j. .

[0060] Furthermore, the high-dimensional hidden state vector High-dimensional hidden state vector and edge attribute matrix Feature concatenation is performed, and the non-normalized attention score of graph node i to associated graph node j is determined through a shared attention mechanism. The nonnormalized attention score It can be determined according to the following formula: Where 'a' represents the learnable parameter vector in the shared attention mechanism; symbol This represents the concatenation operation of vectors or matrices; LeakyReLU represents the linear rectified activation function with leakage, used to perform nonlinear processing on the calculation results after feature concatenation.

[0061] Due to the edge attribute matrix The graph contains physical channel characteristics and link state information of the corresponding communication link. Therefore, the non-normalized attention score is affected not only by the state of the graph nodes at both ends of the graph edge, but also by the current state of the communication link corresponding to that edge. When the noise of the communication link increases, the degree of impedance change increases, or other link states deteriorate, the edge attribute matrix of the corresponding graph edge changes, thereby reducing the attention relevance of the communication link.

[0062] Furthermore, for any graph node i, determine the set of its first-order physical neighbors. The non-normalized attention scores of each associated graph node are calculated, and the non-normalized attention scores are normalized to obtain the attention weights of graph node i to each associated graph node.

[0063] The attention weight of graph node i to associated graph node j can be determined by the following formula: in, This represents the attention weight of graph node i to its associated graph node j, which is used to characterize the degree of influence of associated graph node j and the communication link between them on the feature aggregation of graph node i. This represents the nonnormalized attention score of graph node i to its associated graph node j; Let represent the nonnormalized attention score of graph node i to its associated graph node k.

[0064] The normalization process ensures that the attention weights corresponding to graph node i are within a uniform numerical range, making the attention weights of associated graph nodes comparable. Associated graph nodes with better communication link states receive higher attention weights, while those with poorer communication link states receive lower attention weights.

[0065] Furthermore, for any graph node i, based on the attention weights of this graph node to each associated graph node, the original feature vectors of each associated graph node are weighted and aggregated to obtain the node aggregated feature corresponding to graph node i. The node aggregation feature It can be determined according to the following formula: Where i represents the graph node currently undergoing feature aggregation; j represents the set of first-order physical neighbors of graph node i. Any node in the associated graph; Represents the set of first-order physical neighbors of graph node i; Represents graph nodes Attention weights for node j in the association graph; This represents a shared, learnable weight matrix used for feature transformation of the original feature vectors of nodes in the association graph; This represents a shared learnable weight matrix used for feature transformation of the original feature vectors of nodes in the association graph.

[0066] It is understandable that by weighting and aggregating the original feature vectors of each associated graph node according to attention weights, the aggregated feature of graph node i can simultaneously include the operating status of all associated graph nodes around graph node i and the status influence of the corresponding communication links. Associated graph nodes with higher attention weights have a greater impact on the aggregated feature, while associated graph nodes with lower attention weights have a smaller impact. Based on the above method, the aggregated feature of each graph node in the real-time dynamic attribute graph is determined.

[0067] For any graph edge, obtain the node aggregation features corresponding to the two graph nodes at the ends of the edge and the edge attributes of the edge. Then, through fully connected processing and probability mapping, determine the connectivity probability of the communication link corresponding to the graph edge. Specifically, for a graph edge connecting graph node i and graph node j, the connectivity probability of its corresponding communication link can be determined according to the following formula: in, It represents the connectivity probability of the communication link between graph node i and graph node j, and is used to characterize the degree to which the communication link maintains effective communication under the current node state and link state; This represents a learnable weight matrix used to determine connectivity probabilities based on node aggregation features and edge attributes; This represents the node aggregation feature corresponding to node i in the graph; This represents the node aggregation feature corresponding to node j in the graph; The edge attribute represents the graph edge between graph node i and graph node j, used to characterize the physical channel state and link state of the communication link; Sigmoid represents the probability mapping function, used to map the fully connected processing result to a preset probability interval.

[0068] Based on the above method, the connectivity probability of the communication link corresponding to each edge in the real-time dynamic attribute graph is determined, resulting in the connectivity probability set for each communication link in the power line carrier communication system. It can be understood that this calculation process can simultaneously utilize the node states around the nodes at both ends of the communication link, as well as the physical channel state and link state of the communication link itself. Therefore, when a communication link experiences noise enhancement, impedance abrupt changes, or deterioration of its communication state, its corresponding connectivity probability can be reduced accordingly.

[0069] By implementing the above-mentioned preferred scheme, by identifying all associated graph nodes connected to any given graph node and the graph edges between them, the neighboring nodes and corresponding link ranges involved in graph attention calculation can be clearly defined. Attention relevance is determined based on the node attributes of the current graph node, the node attributes of each associated graph node, and the edge attributes of the corresponding graph edges. This allows for the simultaneous consideration of the states of the nodes at both ends of the link and the state of the link itself when evaluating the degree of communication association between graph nodes. Normalization of each attention relevance results in comparable attention weights, distinguishing the degree of influence of different associated graph nodes and corresponding links on the current graph node. Feature aggregation of the node attributes of each associated graph node based on the attention weights allows the fusion of the state information of neighboring nodes into the current graph node according to their importance, obtaining node aggregation features reflecting the local network environment. Furthermore, the connectivity probability of the communication link is determined based on the node aggregation features of the graph nodes at both ends of the graph edge and the edge attributes of that graph edge. This allows for a comprehensive quantitative evaluation of the availability of the communication link by considering both the local network states at both ends of the link and the state of the link itself.

[0070] Step S106: Determine the backup communication path corresponding to the affected communication node based on the connectivity probability corresponding to each communication link.

[0071] As a preferred embodiment, determining the backup communication path corresponding to the affected communication node based on the connectivity probability of each communication link includes: For each of the communication links, the routing cost corresponding to the communication link is determined based on the connectivity probability of the communication link. The central coordinator in the power line carrier communication system is determined as the root node, and an alternative communication path from the central coordinator to the affected communication node is determined based on the routing cost corresponding to each communication link.

[0072] In this embodiment, the globally optimal routing tree is the communication routing tree determined based on the routing cost of each communication link; in the abnormal reconstruction scenario, the target path from the central coordinator to the affected communication node in the communication routing tree is the backup communication path; the optimal backup relay path is the backup communication path.

[0073] It should be noted that for each communication link, its corresponding routing cost is determined based on the link's connectivity probability. Specifically, a monotonically decreasing cost mapping relationship is used, so that the higher the connectivity probability of a communication link, the lower its routing cost; and the lower the connectivity probability of a communication link, the higher its routing cost.

[0074] In this embodiment, the communication link Routing cost Determined according to the following formula: Furthermore, after determining the routing cost corresponding to each communication link, the central coordinator is used as the root node. Based on the topological connection relationship between each communication node and the routing cost corresponding to each communication link, a path optimization algorithm is used to determine multiple feasible communication paths from the central coordinator to the affected communication nodes, and the sum of the routing costs of each communication link included in each feasible communication path is calculated.

[0075] Preferably, a path optimization algorithm is used to achieve the globally optimal route. Minimize the total routing cost by selecting the communication path that satisfies the preset optimization conditions from multiple feasible communication paths, and use it as the backup communication path for the affected communication node.

[0076] Optionally, the path optimization algorithm can be a shortest path algorithm. By converting connectivity probability into routing cost and performing path optimization based on the routing cost of each communication link, communication links with higher connectivity probability can have higher priority in the selection of backup communication paths.

[0077] By implementing the above-mentioned preferred scheme, the link connectivity evaluation results output by the graph attention network can be converted into cost indicators that can be directly used in the path optimization process by determining the corresponding routing cost based on the connectivity probability of each communication link. This allows different communication links to participate in path comparison according to their connectivity reliability. By determining the central coordinator as the root node, the communication convergence target of the backup path can be unified. By determining the backup communication path from the central coordinator to the affected communication node based on the routing cost of each communication link, communication links with low connectivity probability or abnormal risks can be avoided, and a more reliable alternative communication path can be determined for the affected communication node, thus providing a path basis for it to switch from the current communication path to the backup communication path.

[0078] Step S107: Control the communication path of the affected communication node to switch from the current communication path to the backup communication path.

[0079] Specifically, after determining the backup communication path, the central coordinator identifies the target parent node or target relay node corresponding to the affected communication node based on the backup communication path, generates a path switching instruction, and sends the path switching instruction to the affected communication node. The affected communication node, according to the path switching instruction, updates its current parent node or current next-hop communication node in its MAC forwarding table to the target parent node or target relay node, thereby switching from the current communication path to the backup communication path. The MAC forwarding table records the correspondence between the parent node, next-hop communication node, or target MAC address used by the communication node when forwarding data frames. Updating the MAC forwarding table is a specific implementation method for controlling the switching of communication paths by affected communication nodes.

[0080] By using the physical channel characteristics meeting preset abnormal conditions as the trigger for path reconstruction, the system can initiate real-time dynamic attribute graph updates, connectivity probability recalculation, and backup communication path determination processes without waiting for the media access control layer to experience acknowledgment timeouts, continuous data retransmissions, or actual data packet loss. This allows the system to complete communication path switching before the abnormal communication link deteriorates further.

[0081] As a preferred embodiment, the method for constructing the historical dynamic attribute graph includes: Obtain historical network status information of the power line carrier communication system and historical physical channel characteristics of each of the communication links; Based on the historical network state information and the historical physical channel characteristics of each communication link, determine the node characteristics corresponding to each communication node and the edge characteristics corresponding to each communication link; Obtain the topological connection relationship between each of the communication nodes, and map each of the communication nodes as graph nodes according to the topological connection relationship, and map each of the communication links as graph edges connecting the corresponding graph nodes; The node features are used as node attributes of the corresponding graph nodes, and the edge features are used as edge attributes of the corresponding graph edges to construct the dynamic attribute graph of the entire network.

[0082] It should be noted that the current communication path is determined based on the historical dynamic attribute map, which is a dynamic attribute map constructed based on the network status information before the abnormal communication link was identified; the real-time dynamic attribute map is a dynamic attribute map obtained by updating the historical dynamic attribute map based on the network status information and physical channel characteristics obtained after the abnormal communication link was identified.

[0083] In this embodiment, the dynamic weighted graph is the historical dynamic attribute graph; vertices are the graph nodes; edges are the graph edges; state feature vectors are the node features; initial edge feature vectors are the edge features; and the edge attribute matrix is ​​the edge attributes. CCO represents the Central Coordinator, PCO represents the Agent Coordinator, and STA represents the Site Node. Both PCO and STA are communication nodes.

[0084] Specifically, the central coordinator constructs a historical dynamic attribute graph based on historical network status information, historical physical channel characteristics, and topological connections between communication nodes obtained before the abnormal communication link is identified. This provides a data foundation for determining the current communication path and updating the graph after the anomaly occurs.

[0085] Furthermore, the central coordinator abstracts the power line carrier communication system into a historical dynamic attribute graph. The historical dynamic attribute graph includes a set of vertices. Sum of edges .in, This represents a historical dynamic attribute map constructed based on historical network state information and historical physical channel characteristics; A set of vertices representing a historical dynamic attribute graph, wherein each vertex in the set corresponds to the central coordinator, the agent coordinator, and the site node in the power line carrier communication system. The set of edges represents the historical dynamic attribute graph, where each edge in the set corresponds to a communication link between different communication nodes.

[0086] The central coordinator acquires the topological connectivity between communication nodes through periodic beacon probes, and generates a topological adjacency matrix based on these connectivity relationships. When the communication node With communication nodes When there is a direct electric field line connection between two vertices, an edge is created between the corresponding vertices; when there is no direct electric field line connection, no edge is created between the corresponding vertices. The formula for determining the topological adjacency matrix is ​​as follows: Where A represents the topological adjacency matrix corresponding to the power line carrier communication system; Represents the first element in the topological adjacency matrix. Line 1 The matrix elements of the columns are used to represent communication nodes. With communication nodes Is there a direct power line connection between them? and These represent the node indices of two different communication nodes; when communication nodes With communication nodes When there is a direct power line connection between them Take the first preset value; when the communication node With communication nodes When there is no direct power line connection between them, Take the second preset value.

[0087] Based on the topological adjacency matrix, each communication node is mapped to a vertex in the historical dynamic attribute graph, and each communication link is mapped to an edge connecting the corresponding vertex, thereby preserving the actual topological connection relationship of the power line carrier communication system in the historical dynamic attribute graph.

[0088] Furthermore, the central coordinator obtains the node operation information corresponding to each communication node from the historical network state information, and constructs the state feature vector of the corresponding communication node based on the node operation information. .

[0089] In this embodiment, the node operation information includes the communication node's CPU computing load rate, service buffer queue length, and current transmit power. (Communication node) The corresponding state feature vector can be determined according to the following formula: in, Represents communication node The corresponding state feature vector, i.e., the communication node Corresponding node features; Represents communication node The CPU computing load rate is used to characterize the current computing resource utilization of the communication node; Represents communication node The length of the service buffer queue is used to characterize the degree of data backlog for services currently pending processing or forwarding at the communication node. This represents the current transmit power of communication node i, used to characterize the power state used by the communication node when transmitting power line carrier signals.

[0090] Furthermore, the state feature vectors corresponding to each communication node are determined, and each state feature vector is used as the node attribute of the corresponding vertex, thereby enabling the vertices in the historical dynamic attribute graph to represent the historical operating state of the corresponding communication node.

[0091] It should be noted that the CPU computing load rate, service buffer queue length, and current transmit power are merely examples of node operating information. In other implementations, node characteristics can be constructed based on the communication node's storage resource usage, service priority, or other information that can characterize the communication node's operating capabilities.

[0092] For PLC nodes With PLC node For any physically connected branch, the time-domain impedance transition gradient of that physically connected branch is obtained from the underlying physical layer interface of the PLC node. Peak power of impulse noise and channel state information The received signal strength indication of the physical connectivity branch is obtained by the media access control layer of the PLC node. and historical data packet delivery rate .

[0093] Furthermore, the time-domain impedance transition gradient, impulse noise peak power, channel state information, received signal strength indication, and historical data packet delivery rate are time-aligned so that each feature corresponds to the same or matching feature acquisition time window.

[0094] To reduce the differences in dimensions and numerical ranges among different types of features, the time-domain impedance transition gradient, impulse noise peak power, and received signal strength indication are subjected to maximum-min normalization. The normalized physical channel features are then concatenated with link state information to obtain the physical connectivity branch. Corresponding initial edge feature vector The specific calculation formula is as follows: in, Represents physically connected components The initial edge feature vector; This represents the time-domain impedance jump gradient of this branch; Indicates the peak power of the frequency domain impulse noise; Represents physical layer channel state information; Indicates received signal strength; Indicates the historical data packet delivery rate; , and These are the maximum and minimum normalized constants corresponding to impedance transition, noise power, and received signal strength, respectively.

[0095] Furthermore, to match the dimensionality of the initial edge feature vector with the node features in the graph attention network, the initial edge feature vector is input into a multilayer perceptron for high-dimensional mapping to obtain the communication node. With communication nodes The edge attribute matrix corresponding to the graph edges between them The formula for determining the edge attribute matrix is ​​as follows: in, This represents the learnable weight matrix used in a multilayer perceptron to perform feature transformation on the initial edge feature vectors; The term represents the bias term corresponding to the learnable weight matrix; ReLU represents the linear rectified activation function.

[0096] By setting the output dimension of the learnable weight matrix, the edge attribute matrix can be optimized. The feature dimensions are matched with the feature dimensions of the nodes, which makes it easier for the subsequent graph attention network to calculate the attention relevance based on the node attributes of the nodes at both ends of the graph edge and the edge attributes of the graph edge.

[0097] The edge attribute matrix corresponding to each communication link is determined in the above manner, and each edge attribute matrix is ​​used as the edge attribute of the corresponding graph edge, thereby constructing a historical dynamic attribute graph that simultaneously includes topological connection relationships, node operating status, and communication link status.

[0098] By implementing the above preferred scheme, historical network state information of the power line carrier communication system and historical physical channel characteristics of each communication link can be obtained, thus providing a data foundation for establishing a network state benchmark. Based on the historical network state information and historical physical channel characteristics, node and edge features can be determined, converting the historical operating states of communication nodes and the historical physical states of communication links into feature information that can be processed by a graph model. By obtaining the topological connection relationships between communication nodes and mapping communication nodes as graph nodes and communication links as graph edges connecting corresponding graph nodes, the actual connection relationships of the power line carrier communication system can be preserved in the graph structure. Furthermore, by using node features as node attributes and edge features as edge attributes, a historical dynamic attribute graph that simultaneously represents historical node states, historical link states, and network topology can be constructed, providing a basis for determining the current communication path and updating the graph state after an anomaly occurs.

[0099] As a preferred embodiment, before controlling the multiple communication nodes to communicate according to their respective current communication paths, the current communication path is first determined. The specific determination method includes: The dynamic attribute graph of the entire network is input into the preset graph attention network model to determine the connectivity probability of each communication link. The routing cost corresponding to each communication link is determined based on the connectivity probability of each communication link. Using the central coordinator in the power line carrier communication system as the root node, the target communication path from the central coordinator to each of the communication nodes is determined based on the routing cost corresponding to each of the communication links. A communication routing tree is constructed based on each of the target communication paths, and each of the target communication paths is determined as the current communication path corresponding to the corresponding communication node.

[0100] By implementing the above preferred scheme, and inputting the historical dynamic attribute graph into a preset graph attention network model, the connectivity of each communication link can be comprehensively evaluated based on the node attributes, edge attributes, and graph connection relationships before the anomaly occurs, thus obtaining the connectivity probability corresponding to each communication link. The routing cost is determined based on the connectivity probability, converting the link connectivity evaluation results into a unified index usable for path optimization. Using the central coordinator as the root node and determining the target communication path from the central coordinator to each communication node based on the routing cost of each communication link, the initial communication path with higher connectivity can be selected for each communication node. A communication routing tree is constructed based on each target communication path, forming a hierarchical communication relationship between each communication node and the central coordinator, and reducing the possibility of path loops. Determining each target communication path as the current communication path of the corresponding communication node provides a path benchmark for normal communication of communication nodes and for switching to backup paths after anomalies.

[0101] See Figure 2 To more clearly illustrate the execution order and interrelationships of the various technical processes in this embodiment, the following is a combination of... Figure 2 The overall process of the graph attention-based power line carrier networking method is explained. For example... Figure 2 As shown, after the power line carrier communication system is powered on and completes network initialization, it sequentially performs multi-dimensional physical channel feature perception and extraction, communication graph network modeling based on physical priors, graph attention network link reasoning and prediction, and topology adaptive reconstruction and optimal route distribution, thereby generating a routing table and guiding each terminal node to communicate.

[0102] After the system is powered on, the power line carrier network is initialized first. Specifically, basic information about each communication node and each communication link in the power line carrier communication system is acquired, the topological connection relationship between each communication node is determined, and the correspondence between communication nodes and communication links is established, providing a foundation for subsequent construction of dynamic attribute graphs and determination of communication paths.

[0103] In step S1, multi-dimensional physical channel feature sensing and extraction are performed. During communication, each communication node monitors the corresponding communication link through the underlying physical layer interface, extracting physical channel features such as time-domain impedance transitions, frequency-domain impulse noise, and channel state information (CSI). Specifically, the time-domain impedance transitions characterize changes in the communication link's input impedance caused by factors such as load switching; the frequency-domain impulse noise characterizes the degree of non-stationary noise interference experienced by the communication link; and the channel state information (CSI) characterizes the signal transmission state of the communication link on different subcarriers. This yields multi-dimensional physical channel features describing the underlying physical state of each communication link.

[0104] In step S2, physical prior-based communication graph modeling is performed. Communication nodes in the power line carrier communication system are mapped to graph nodes, and each communication link is mapped to a graph edge connecting the corresponding graph node. Node features are determined based on the node operation information of each communication node, and edge features are determined based on the physical channel characteristics and link state information of each communication link. These node features and edge features are then used as node attributes of the corresponding graph node and edge attributes of the graph edge, respectively. By mapping the physical channel characteristics to physical weights, a dynamic attribute graph G is constructed, which simultaneously represents the topological connections, communication node operation states, and communication link physical states of the power line carrier communication system.

[0105] During the system initialization phase, the dynamic attribute map G can be a historical dynamic attribute map constructed based on historical network status information and historical physical channel characteristics. After an abnormal communication link is detected, the historical dynamic attribute map can be updated based on the currently acquired network status information and physical channel characteristics to obtain a real-time dynamic attribute map that reflects the current network status.

[0106] In step S3, graph attention network link inference and prediction are performed. The dynamic attribute graph G is input into a preset graph attention network model. For any graph node, the attention score of the graph node to each associated graph node is calculated based on the node attributes of the graph node, the node attributes of each associated graph node connected to the graph node, and the edge attributes of the corresponding graph edges. The attention scores are normalized to obtain the corresponding attention weights, and the node attributes of each associated graph node are aggregated based on the attention weights to obtain the node aggregation features of each graph node. Further, based on the node aggregation features of the graph nodes at both ends of each graph edge and the edge attributes of the graph edge, the connectivity probability of the communication link corresponding to the graph edge is determined, and a connectivity probability matrix composed of the connectivity probabilities of each communication link is output.

[0107] In step S4, topology adaptive reconstruction and optimal route distribution are performed. Based on the connectivity probability matrix, the connectivity probability of each communication link is converted into a corresponding routing cost. Using the central coordinator as the root node, a path optimization algorithm is used to solve the spanning tree, obtaining the communication paths from the central coordinator to each communication node. Since communication links subject to high noise interference or significant impedance jumps have lower connectivity probabilities and higher routing costs, high-noise links can be avoided in advance when solving the spanning tree, and communication links with higher connectivity probabilities are prioritized to form communication paths.

[0108] During normal networking, the current communication path corresponding to each communication node is determined based on the spanning tree. When abnormal communication links are determined based on physical channel characteristics such as time-domain impedance jumps, frequency-domain impulse noise, and channel state information (CSI), the dynamic attribute graph G is updated and the connectivity probability matrix is ​​recalculated. Based on the re-determined connectivity probability and routing cost, the backup communication path corresponding to the affected communication node is determined to achieve topology adaptive reconstruction.

[0109] Finally, a routing table is generated based on the spanning tree or backup communication path, and the routing table or corresponding path switching instructions are sent to the corresponding communication nodes to guide terminal node communication. Each terminal node determines the next-hop communication node corresponding to the data frame based on the routing table; affected communication nodes update their parent node, next-hop communication node, or MAC forwarding relationship according to the path switching instructions, switching from the current communication path to the backup communication path. Thus, communication path adjustment can be completed before the abnormal communication link deteriorates further or is actually disconnected, achieving proactive network reconfiguration of the power line carrier network.

[0110] See Figure 3 , Figure 3This diagram illustrates the graph attention link evaluation and topology reconstruction process provided in this embodiment. To more clearly explain the specific process by which the graph attention network evaluates the connectivity of communication links and performs periodic route updates or immediate path reconstruction based on the anomaly judgment results of physical channel characteristics, the following is a detailed explanation. Figure 3 Please provide an explanation.

[0111] like Figure 3 As shown, the node state feature matrix and physical edge feature matrix of the power line carrier communication system are first obtained. The node state feature matrix consists of node operation information such as the load rate and service buffer queue length of each communication node, while the physical edge feature matrix consists of physical channel features such as the impedance gradient and impulse noise of each communication link. The node state feature matrix and physical edge feature matrix are input into a preset graph attention network model, and cross-layer feature linear mapping and hidden state generation, physical feature-driven attention score calculation, and nonlinear aggregation and low-weight suppression based on the Softmax function are performed sequentially. Specifically, by mapping the node attributes of graph nodes and the edge attributes of graph edges to the hidden feature space, corresponding node hidden states and edge attributes are generated. Attention scores are calculated based on the node hidden states of adjacent graph nodes and the edge attributes of the graph edges between them, so that communication links with large impedance changes or strong impulse noise receive lower attention weights. The attention scores corresponding to each associated graph node are normalized, and feature aggregation is performed based on the normalized attention weights to obtain the node aggregation features corresponding to each graph node.

[0112] Subsequently, based on the node aggregation features corresponding to the two endpoints of each graph edge and the edge attributes of that graph edge, the connectivity probability of the corresponding communication link is determined using the Sigmoid function, and a connectivity probability matrix composed of the connectivity probabilities of each communication link is output. Further, it is determined whether the physical channel characteristics of the communication link are abnormal, specifically whether the impedance jump gradient of the communication link exceeds a preset impedance threshold, or whether the peak power of the impulse noise exceeds a preset noise threshold.

[0113] When the physical channel characteristics of a communication link become abnormal, the connectivity probability of the current abnormal communication link is reduced, and the graph attention network model is triggered to perform real-time recalculation. Based on the redefined attention weights and connectivity probabilities, candidate relay nodes with higher attention weights are searched from the associated communication nodes of the affected communication node to determine an alternative communication path. Before actual data packet loss occurs on the abnormal communication link, an alternative path switching command is generated and sent to the affected communication node, causing the affected communication node to switch from the current communication path to the alternative communication path.

[0114] When the physical channel characteristics of the communication links are not abnormal, the connectivity probability of each communication link is converted into the corresponding routing cost weight. An improved Dijkstra algorithm is used to optimize the path of each communication link, and the optimal global routing tree with the central coordinator as the root node is calculated. The routing forwarding table is generated, updated and distributed according to a preset period to guide each communication node to forward data.

[0115] Therefore, whether it is to perform global route updates according to a preset period or to perform immediate backup path switching when physical channel characteristics are abnormal, communication path adjustment can be completed based on the communication link connectivity probability output by the graph attention network, thereby realizing seamless topology reconstruction of the power line carrier communication system.

[0116] By implementing the embodiments of this application, and by acquiring the current communication path of each communication node and the physical channel characteristics of each communication link in the power line carrier communication system, it is possible to clearly identify the communication path currently used by each communication node and continuously sense the physical state of each communication link, providing a basis for abnormal link identification and subsequent path adjustment. Based on the physical channel characteristics and preset abnormal conditions, abnormal communication links are determined, and the corresponding affected communication nodes and network status information are further identified. This allows for timely location of the affected objects of link anomalies and acquisition of the operating status of the power line carrier communication system at the time of the anomaly. Based on the network status information and physical channel characteristics, associated communication nodes, node characteristics, and edge characteristics are determined, enabling the identification of associated nodes participating in backup path selection and converting the operating status of communication nodes and the physical state of communication links into features suitable for graph structure processing. The system generates a real-time dynamic attribute graph that reflects the current node and link status while preserving the original network topology. This graph is then input into a graph attention network model to determine the connectivity probability of each communication link. This allows for a comprehensive evaluation of the availability of each communication link by considering the relationships between communication nodes and the status of the communication links. Based on the connectivity probability, backup communication paths for affected communication nodes are determined, enabling the selection of reliable alternative paths. Furthermore, the system controls the switching of affected communication nodes from their current communication path to their backup communication path. This proactively reconstructs communication paths before the abnormal communication links deteriorate further or fail, reducing the risk of communication interruption caused by sudden link failures and improving the continuity and reliability of power line carrier network communication.

[0117] Based on the above method embodiments, this application provides corresponding apparatus embodiments.

[0118] like Figure 4As shown, another embodiment of this application provides a power line carrier networking device based on graph attention, including: a feature acquisition module 100, an anomaly judgment module 200, a feature determination module 300, a graph update module 400, a probability calculation module 500, a path determination module 600, and a path switching module 700. The feature acquisition module 100 is used to acquire the current communication path of each communication node and the physical channel characteristics of each communication link in the power line carrier communication system, wherein the power line carrier communication system includes multiple communication nodes and multiple communication links. The anomaly detection module 200 is used to determine the abnormal communication link from the multiple communication links according to the physical channel characteristics and preset anomaly conditions, and to determine the affected communication node corresponding to the abnormal communication link from the multiple communication nodes and obtain the network status information of the power line carrier communication system. The feature determination module 300 is used to determine the associated communication node corresponding to the affected communication node based on the network status information and the physical channel characteristics of each communication link, and to determine the node characteristics and edge characteristics of each communication link corresponding to the affected communication node and the associated communication node, respectively. The graph update module 400 is used to update the historical dynamic attribute graph of the power line carrier communication system according to the node features and the edge features, so as to obtain the real-time dynamic attribute graph. The probability calculation module 500 is used to input the updated dynamic attribute graph into a preset graph attention network model to determine the connectivity probability of each communication link. The path determination module 600 is used to determine the backup communication path corresponding to the affected communication node based on the connectivity probability corresponding to each communication link. The path switching module 700 is used to control the communication path of the affected communication node to switch from the current communication path to the backup communication path.

[0119] It is understood that the above-described device embodiments correspond to the method embodiments of this application, and can implement the graph attention-based power line carrier networking method provided by any of the above-described method embodiments of this application.

[0120] It should be noted that the device embodiments described above are merely illustrative, and some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided in this application, the connection relationships between modules indicate that they have communication connections, which can specifically be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0121] Based on the above embodiments of the power line carrier networking method based on graph attention, another embodiment of this application provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the power line carrier networking method based on graph attention of any embodiment of this application.

[0122] For example, in this embodiment, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete this application. The one or more module units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the terminal device.

[0123] The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0124] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting all parts of the terminal device via various interfaces and lines.

[0125] Based on the above-described method embodiments, another embodiment of this application provides a computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to execute the graph attention-based power line carrier networking method described in any of the above-described method embodiments of this application.

[0126] The modules / units integrated in the device / terminal equipment, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0127] Based on the above-described method embodiments, another embodiment of this application provides a computer program product, including a computer program or instructions, which, when executed by a communication device, implements a power line carrier networking method based on graph attention.

[0128] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this application. It should be understood that the above descriptions are merely specific embodiments of this application and are not intended to limit the scope of protection of this application. In particular, it should be noted that any modifications, equivalent substitutions, or improvements made by those skilled in the art within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A power line carrier networking method based on graph attention, characterized in that, include: The current communication path of each communication node and the physical channel characteristics of each communication link in the power line carrier communication system are obtained, wherein the power line carrier communication system includes multiple communication nodes and multiple communication links; Based on the physical channel characteristics and preset abnormal conditions, abnormal communication links are determined from multiple communication links, and the affected communication nodes corresponding to the abnormal communication links are determined from multiple communication nodes, and the network status information of the power line carrier communication system is obtained. Based on the network status information and the physical channel characteristics of each communication link, the associated communication node corresponding to the affected communication node is determined, and the node characteristics and edge characteristics of each communication link corresponding to the affected communication node and the associated communication node are determined respectively. Based on the node features and the edge features, the historical dynamic attribute map of the power line carrier communication system is updated to obtain the real-time dynamic attribute map. The real-time dynamic attribute graph is input into a preset graph attention network model to determine the connectivity probability of each communication link. Based on the connectivity probability corresponding to each of the communication links, determine the backup communication path corresponding to the affected communication node; Control the affected communication node to switch its communication path from the current communication path to the backup communication path.

2. The power line carrier networking method based on graph attention according to claim 1, characterized in that, The method for constructing the historical dynamic attribute graph includes: Obtain historical network status information of the power line carrier communication system and historical physical channel characteristics of each of the communication links; Based on the historical network state information and the historical physical channel characteristics of each communication link, determine the node characteristics corresponding to each communication node and the edge characteristics corresponding to each communication link; Obtain the topological connection relationship between each of the communication nodes, and map each of the communication nodes as graph nodes according to the topological connection relationship, and map each of the communication links as graph edges connecting the corresponding graph nodes; The historical dynamic attribute graph is constructed by using the node features as node attributes of the corresponding graph nodes and the edge features as edge attributes of the corresponding graph edges.

3. The power line carrier networking method based on graph attention according to claim 2, characterized in that, Before obtaining the current communication path of each communication node and the physical channel characteristics of each communication link in the power line carrier communication system, the following steps are also included: The historical dynamic attribute graph is input into a preset graph attention network model to determine the connectivity probability of each communication link. The routing cost corresponding to each communication link is determined based on the connectivity probability of each communication link. Using the central coordinator in the power line carrier communication system as the root node, the target communication path from the central coordinator to each of the communication nodes is determined based on the routing cost corresponding to each of the communication links. A communication routing tree is constructed based on each of the target communication paths, and each of the target communication paths is determined as the current communication path corresponding to the corresponding communication node.

4. The power line carrier networking method based on graph attention according to claim 1, characterized in that, The step of determining the associated communication node corresponding to the affected communication node based on the network state information and the physical channel characteristics of each communication link, and determining the node characteristics and edge characteristics of each communication link corresponding to the affected communication node and the associated communication node respectively, includes: The network status information includes node operation information and link status information; Based on the network status information and the physical channel characteristics of each communication link, the associated communication node corresponding to the affected communication node is determined; Obtain the node operation information corresponding to the affected communication node and the associated communication node from the network status information, and determine the node characteristics corresponding to the affected communication node and the associated communication node; For each communication link, the link state information of the communication link is obtained from the network state information. The physical channel features and link state information of the communication link are time-synchronized and standardized. The processed physical channel features and link state information are then fused to obtain the edge features corresponding to the communication link.

5. The power line carrier networking method based on graph attention according to claim 2, characterized in that, The step of updating the historical dynamic attribute map of the power line carrier communication system based on the node features and the edge features to obtain the real-time dynamic attribute map includes: From the historical dynamic attribute graph, determine the graph nodes corresponding to the affected communication nodes and the associated communication nodes, as well as the graph edges corresponding to each communication link; Update the node attributes of the corresponding graph nodes based on the node characteristics of the affected communication nodes and the associated communication nodes; Based on the edge characteristics corresponding to each communication link, the edge attributes of the corresponding graph edges are updated to obtain the real-time dynamic attribute graph.

6. The power line carrier networking method based on graph attention according to claim 2, characterized in that, The step of inputting the real-time dynamic attribute graph into a preset graph attention network model to determine the connectivity probability corresponding to each communication link includes: For any given graph node, determine each associated graph node connected to that graph node, and the graph edges between that graph node and each of the associated graph nodes; For any graph node, the attention relevance of the graph node to each of the associated graph nodes is determined based on the node attributes of the graph node, the node attributes of each of the associated graph nodes, and the edge attributes of the graph edges between the graph node and each of the associated graph nodes. For any of the graph nodes, the attention relevance of the graph node to each of the associated graph nodes is normalized to obtain the attention weight of the graph node to each of the associated graph nodes. For any of the graph nodes, based on the attention weight of the graph node to each of the associated graph nodes, the node attributes of each of the associated graph nodes are aggregated to obtain the node aggregated features corresponding to the graph node. For any of the graph edges, the connectivity probability of the communication link corresponding to the graph edge is determined based on the node aggregation features of the graph nodes at both ends of the graph edge and the edge attributes of the graph edge.

7. The power line carrier networking method based on graph attention according to claim 1, characterized in that, The step of determining the backup communication path corresponding to the affected communication node based on the connectivity probability corresponding to each communication link includes: For each of the communication links, the routing cost corresponding to the communication link is determined based on the connectivity probability of the communication link. The central coordinator in the power line carrier communication system is determined as the root node, and an alternative communication path from the central coordinator to the affected communication node is determined based on the routing cost corresponding to each communication link.

8. A power line carrier networking device based on graph attention, characterized in that, include: The module includes a feature acquisition module, an anomaly detection module, a feature determination module, a graph update module, a probability calculation module, a path determination module, and a path switching module. The feature acquisition module is used to acquire the current communication path of each communication node and the physical channel characteristics of each communication link in the power line carrier communication system, wherein the power line carrier communication system includes multiple communication nodes and multiple communication links. The anomaly detection module is used to determine the abnormal communication link from the multiple communication links based on the physical channel characteristics and preset anomaly conditions, and to determine the affected communication node corresponding to the abnormal communication link from the multiple communication nodes and obtain the network status information of the power line carrier communication system. The feature determination module is used to determine the associated communication node corresponding to the affected communication node based on the network status information and the physical channel characteristics of each communication link, and to determine the node characteristics and edge characteristics of each communication link corresponding to the affected communication node and the associated communication node, respectively. The graph update module is used to update the historical dynamic attribute graph of the power line carrier communication system according to the node features and the edge features, so as to obtain the real-time dynamic attribute graph. The probability calculation module is used to input the real-time dynamic attribute graph into a preset graph attention network model to determine the connectivity probability corresponding to each communication link. The path determination module is used to determine the backup communication path corresponding to the affected communication node based on the connectivity probability corresponding to each communication link. The path switching module is used to control the communication path of the affected communication node to switch from the current communication path to the backup communication path.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the power line carrier networking method based on graph attention as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the power line carrier networking method based on graph attention as described in any one of claims 1 to 7.

Images