A method and system for fault diagnosis of power communication network defects based on graph matching

Abstract

This invention discloses a method and system for fault diagnosis of power communication networks based on graph matching, comprising: acquiring the topology of the power communication network to be diagnosed and converting the topology into a data graph of the power communication network; constructing a fault query graph of the power communication network using the topology and historical fault alarm information; determining the candidate set and vertex degree of each root node in the fault query graph, and determining the optimal root node in the fault query graph based on the number of nodes in the candidate set and the vertex degree; using the optimal root node as the query vertex, querying a subgraph of the power communication network data graph that is isomorphic to the subgraph of the fault query graph based on the graph matching algorithm; and determining the fault result of the power communication network to be diagnosed based on the subgraph of the power communication network data graph obtained by graph matching, which can improve the speed and accuracy of fault diagnosis.

Description

Technical Field

[0001] This invention belongs to the field of smart grid fault diagnosis technology, and particularly relates to a method and system for fault diagnosis of power communication network defects based on graph matching. Background Technology

[0002] With the continuous increase in electricity demand and the development of the social economy, the power grid scale continues to expand. The power system plays a vital role in people's livelihoods, industry, and social development, making its stable operation particularly important. As a crucial component of the power system, the power communication network undertakes the tasks of data transmission and information exchange. It not only supports monitoring and control functions in daily operations but also provides necessary communication guarantees for rapid response and fault recovery in emergencies. By monitoring and managing the status of the power network in real time, the power communication network can help the dispatch center promptly identify and address potential problems, thereby ensuring the safety and reliability of power supply. In the face of sudden faults, the power communication network can determine the type of communication network fault based on the types of original records, quickly locate it, and then take corresponding maintenance measures to ensure the smooth operation of overall services.

[0003] The core functional modules of power communication networks include fault diagnosis and service impact analysis, both indispensable for maintaining the stable operation of smart grid systems. Fault diagnosis aims to identify and classify faults by matching the current network topology with historical data in real time. This process involves continuous monitoring of the network status and comparative analysis with past fault cases to quickly determine the specific location and nature of the fault. Service impact analysis, on the other hand, assesses the scope of impact from the physical layer to the service layer when a fault occurs in a specific part (such as optical cable or equipment). This includes, but is not limited to, checking which services may be affected, thereby providing detailed guidance for rapid recovery. The existing power communication management system SG-TMS2.0 uses a relational database as its core data management technology. Although this database has good data consistency and transaction processing capabilities, it exhibits inefficiency when facing complex query requirements and rapid response scenarios. In the fault diagnosis process, the system first collects alarm information and other relevant data from the network through various sensors and stores this data in a relational database. Subsequently, based on a pre-built rule base, the system performs complex database queries to filter out datasets that meet the conditions and compares them with fault patterns in the rule base. Once a fault mode matching the current alarm is found, the system will execute the corresponding fault handling measures according to the operations specified in the rule base.

[0004] However, this approach has drawbacks. Table queries in relational databases involve numerous table joins, which consume a significant amount of time. Furthermore, whenever the network topology changes or new devices and services are introduced, these relationships need to be updated. This process involves extensive data traversal, and each re-analysis consumes considerable time, making the entire fault diagnosis process very slow. Especially in large-scale network environments, such latency can significantly extend service outage times, thereby degrading overall service quality. Therefore, solving the problem of fault diagnosis in power communication networks is particularly important. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a graph matching-based method and system for diagnosing defects in power communication networks, which improves the speed and accuracy of fault diagnosis.

[0006] The present invention adopts the following technical solution.

[0007] The first aspect of the present invention provides a method for diagnosing defects in power communication networks based on graph matching, comprising:

[0008] S1: Obtain the topology of the power communication network to be diagnosed, and convert the topology of the power communication network to be diagnosed into a data graph of the power communication network to be diagnosed;

[0009] S2: Construct a power communication network fault query map using the power communication network topology and historical defect alarm information. Each power communication network fault query map corresponds to a different type of communication network defect.

[0010] S3: Determine the candidate set and vertex degree of each root node in the power communication network fault query graph, and determine the optimal root node in the power communication network fault query graph based on the number of nodes in the candidate set and the vertex degree. Using the optimal root node as the query vertex, query the power communication network data subgraph that is isomorphic to the power communication network fault query graph subgraph in the power communication network data graph to be diagnosed based on the graph matching algorithm.

[0011] S4: Determine the defect or fault result of the power communication network to be diagnosed based on the power communication network data subgraph obtained by graph matching.

[0012] Optionally, in S2, a fault query graph of the power communication network is constructed using the power communication network topology and historical fault alarm information, including:

[0013] Analyze historical defect alarm information and filter out the nodes and links affected by the fault to determine the coverage of defect alarm information;

[0014] In the power communication network data graph to be diagnosed, the nodes within the coverage area of ​​the defect alarm information are connected to generate a power communication network fault query graph.

[0015] Optionally, in S3, the optimal root node is determined in the power communication network fault query graph based on the number of nodes in the candidate set and the vertex degree, including:

[0016] The number of nodes and vertex degree in the candidate set are normalized.

[0017] The index value of each root node in the fault query graph of the power communication network is calculated based on the normalized value of the number of nodes and the normalized value of the vertex degree in the candidate set.

[0018] ,

[0019] in, This represents the index value of the i-th root node in the power communication network fault query graph. This represents the normalized degree value of the i-th root node. This represents the normalized value of the number of nodes in the candidate set of the i-th root node. and These represent the weights for degree and the number of nodes in the candidate set, respectively.

[0020] Select the root node with the largest index value as the optimal root node.

[0021] Optionally, in S3, using the optimal root node as the query vertex, a graph matching algorithm is used to query a power communication network data subgraph that is isomorphic to the power communication network fault query subgraph in the power communication network data graph to be diagnosed, including:

[0022] Convert the power communication network data diagram and power communication network fault query diagram into a unified input format;

[0023] Using the optimal root node as the query vertex, the breadth-first search method is used to traverse the power communication network fault query graph starting from the query vertex according to the preset traversal order. The nodes in the power communication network fault query graph are sorted to construct the acyclic graph corresponding to the power communication network fault query graph. The matching order is generated according to the topological sorting order of the acyclic graph.

[0024] According to the matching order, the query vertex in the acyclic graph is matched with the power communication network data graph to be diagnosed in turn. During the backtracking matching process, if the next node that meets the preset matching conditions cannot be found on the current matching path to continue expanding the matching path, the current path is marked as a failed path, and the failed path is skipped to continue matching. The candidate structure is constructed based on the candidate set obtained by matching and the relationship between the nodes in the candidate set.

[0025] Based on the acyclic graph and candidate structure, it is verified whether the candidate nodes in the candidate structure meet the preset matching conditions. If they do not meet the conditions, they are deleted from the candidate structure, resulting in the final power communication network data subgraph that is isomorphic to the power communication network fault query subgraph.

[0026] Optionally, nodes and edges in the power communication network fault query graph have corresponding labels. The nodes in the power communication network fault query graph are sorted, including:

[0027] The edges in the power communication network fault query graph are sorted according to a preset hierarchy, with all edges pointing from higher-level nodes to lower-level nodes.

[0028] Vertices within the same level are grouped by label, and vertices within each group are sorted in descending order of degree.

[0029] Optionally, S3, using the optimal root node as the query vertex and employing a graph matching algorithm to query a power communication network data subgraph isomorphic to the power communication network fault query subgraph in the power communication network data graph to be diagnosed, further includes:

[0030] Check if there are non-tree edges in the power communication network fault query graph. If so, delete the corresponding node from the candidate set.

[0031] Optionally, before matching the query vertices in the acyclic graph with the power communication network data graph to be diagnosed in sequence according to the matching order, S3, based on the graph matching algorithm with the optimal root node as the query vertex, queries the power communication network data subgraph isomorphic to the power communication network fault query graph subgraph in the power communication network data graph to be diagnosed, and also includes:

[0032] A graph index is created for the fault query graph of the power communication network, and the power communication network fault query graph is traversed using the pre-built graph index.

[0033] Optionally, during graph matching, if a match is successful, the matching order is updated according to the following formula:

[0034] ,

[0035] in, To find vertex n in an acyclic graph, Let n be the priority of node n, and let n be the priority of node n. The priority of a node is set according to the node's label and degree. This represents the number of matched neighbors for node n. Let n be the path length from the starting node to node n in the acyclic graph. , , They are respectively , , The corresponding weights.

[0036] A second aspect of the present invention provides a fault diagnosis system for power communication networks based on graph matching, the system comprising:

[0037] The acquisition module is used to acquire the topology of the power communication network to be diagnosed and convert the topology of the power communication network to be diagnosed into a data graph of the power communication network to be diagnosed.

[0038] The module is used to construct a fault query graph of the power communication network using the topology of the power communication network and historical defect alarm information. Each power communication network fault query graph corresponds to a different type of communication network defect.

[0039] The graph matching module is used to determine the candidate set and vertex degree of each root node in the power communication network fault query graph, and to determine the optimal root node in the power communication network fault query graph based on the number of nodes in the candidate set and the vertex degree. Using the optimal root node as the query vertex, the module uses the graph matching algorithm to query the power communication network data subgraph that is isomorphic to the power communication network fault query graph subgraph in the power communication network data graph to be diagnosed.

[0040] The determination module is used to determine the defect and fault results of the power communication network to be diagnosed based on the power communication network data subgraph obtained by graph matching.

[0041] A third aspect of the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when loaded onto the processor, implements the above-described graph-matching-based power communication network defect fault diagnosis method.

[0042] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described graph-matching-based method for diagnosing defects in power communication networks.

[0043] Compared with the prior art, the beneficial effects of the present invention include at least the following:

[0044] This invention improves the speed and accuracy of fault diagnosis by effectively representing complex topologies in power communication networks and achieving efficient querying in the context of massive data based on graph matching. The diagnostic process, based on graph matching, eliminates the need for extensive manually labeled data, unlike deep learning, thus avoiding the time-consuming and labor-intensive nature of manual labeling. Furthermore, in environments with diverse communication devices and complex network topologies, the graph matching algorithm demonstrates greater flexibility and adaptability, especially in the dynamically changing environment of power communication networks, where it better adapts to network variations. Experimental results not only demonstrate the potential value of this method in promoting the intelligentization of power communication networks but also reveal the broad application potential of graph matching technology in future power communication network management, while reducing reliance on manually labeled data and lowering the cost and complexity of maintenance and operation.

[0045] This invention defines query subgraphs based on specific fault modes, making it more targeted. Algorithmically, it designs query modes based on the characteristics of power networks, improves efficiency by optimizing candidate structures and pruning failure sets, and incorporates a dynamic adaptive search algorithm. It deeply integrates graph matching algorithms with power communication fault characteristics, and solves the efficiency bottleneck of traditional methods (such as relational databases) under complex topologies through domain-specific query mode design and data verification. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0047] Figure 1 This is a schematic diagram of a graph matching-based fault diagnosis method for power communication networks provided in an embodiment of the present invention.

[0048] Figure 2 This is a schematic diagram of another graph matching-based power communication network defect diagnosis method provided in an embodiment of the present invention;

[0049] Figure 3 This is a schematic diagram of another power communication network defect fault diagnosis method based on graph matching provided in an embodiment of the present invention;

[0050] Figure 4 This is a schematic diagram of a graph matching query process provided in an embodiment of the present invention;

[0051] Figure 5 This is a schematic diagram illustrating the application of a graph matching-based fault diagnosis method for power communication networks provided in an embodiment of the present invention.

[0052] Figure 6 This is a schematic diagram of a power communication network topology provided in an embodiment of the present invention;

[0053] Figure 7 This is a schematic diagram of a query graph topology provided in an embodiment of the present invention;

[0054] Figure 8 This is a schematic diagram illustrating a comparison of query matching speeds provided in an embodiment of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0056] The technical problem this invention aims to solve is: how to effectively represent complex topologies in power communication networks and achieve efficient querying in the context of massive data, thereby improving the speed and accuracy of fault diagnosis. In conjunction with the above objective, the problem definition of graph matching is given below.

[0057] Definition 1 Graph: A graph G consists of a set of nodes V and a set of edges E, denoted as . Typically, V(G) is used to represent a finite and non-empty set of all nodes in graph G. Let |V| represent the set of relationships between different nodes in graph G. |V| is typically used to represent the total number of nodes in graph G. |E| is used to represent the total number of edges in graph G. The set of nodes in a graph must be non-empty, while the set of edges can be empty.

[0058] Definition 2: Undirected Graph: A graph G is undirected if every edge in it is undirected. Undirected edges in an undirected graph are simply called edges. (The last sentence is a common usage and doesn't need a direct translation.) or To represent edges, and nodes v and nodes w They are adjacent to each other.

[0059] Define a 3-subgraph: For two graphs and In other words, if the node set And edge set ,but It can be called Subgraphs.

[0060] Definition 4 Data Graphs and Query Graphs: A data graph can represent A query graph can be represented as ,in, and These represent the data node set and the query node set, respectively. and These represent the data edge set and the query edge set, respectively. and These represent the label set for the data graph and the label set for the query graph, respectively.

[0061] Definition 5: Subgraph isomorphism: Given a query graph P=( ) and data graph G=( A subgraph isomorphism of a query graph P in a data graph G is a subgraph from... arrive The mapping function f satisfies the following two conditions: 1. 2. .

[0062] Definition 6 Subgraph Matching: Given a query graph P and a data graph G, the goal of subgraph matching is to search and enumerate all subgraphs in G that are isomorphic to P, that is, to find all subgraphs in G that are isomorphic to P.

[0063] Definition 7 Candidate Node Set: Given a query graph P and a data graph G, a query node... The candidate node set consists of a set of data nodes, and satisfies the following: for each data node ,if It exists within a subgraph isomorphic from P to G, i.e., the query node. u With data nodes v If the matching process yields the final matching result, then... v It belongs to C(u), where C(u) represents the set of candidate nodes for query node u.

[0064] Combination Figure 1 and Figure 5 As shown, Embodiment 1 of the present invention provides a method for diagnosing defects in power communication networks based on graph matching, specifically including the following:

[0065] S1: Obtain the topology of the power communication network to be diagnosed, and convert the topology of the power communication network to be diagnosed into a data graph of the power communication network to be diagnosed.

[0066] An undirected graph can be used to generate the topology of the power communication network to be diagnosed based on the devices and relationships between them, and the topology can be converted into a data graph of the power communication network to be diagnosed according to a specified format.

[0067] Specifically, such as Figure 2 As shown, raw data such as optical routing tables, defect report tables, and defect alarm tables from the power communication network are obtained. Standardized fields (such as device type and port number) are defined using a data dictionary to ensure data consistency. A non-relational graph database is used to abstract the data into nodes and relationships, constructing a comprehensive data graph of the communication network. The topology of the power communication network to be diagnosed can be constructed by mapping optical paths to information about sites, devices, and ports, using nodes and optical path connections as edges. Defect alarm information is added as attributes to the corresponding nodes or edges in the topology of the power communication network, forming dynamic fault labels.

[0068] Combination Figure 6 As shown, ={Substation A, Substation B, Substation C, Equipment A, Equipment B, Equipment C, Equipment D, A(1), A(2), B(1), B(2), C(1), C(2), C(3), C(4), C(5)} ={(Substation A, Equipment A),(Equipment A, A(1)),(Equipment A, A(2)),(Substation A, Substation B),(Substation B, Equipment B),(Equipment B, B(1)),(Equipment B, B(2)),(Substation B, Substation C),(Substation C, Equipment C),(Equipment C, C(1)),(Equipment C, C(2)),(Equipment C, C(3)),(Substation C, Equipment D),(Equipment D, C(4)),(Equipment D, C(5)),(A(1), B(1)),(B(2), C(1)),(A(2), C(5))}, which represents the equipment and its connection method in the power communication network through the set of nodes and edges.

[0069] S2: Construct a power communication network fault query graph using the power communication network topology and historical defect alarm information. Each power communication network fault query graph corresponds to a different type of communication network defect.

[0070] Specifically, by analyzing defect and fault information, communication network faults are obtained, including transmission optical path interruption, optical cable interruption, equipment power supply interruption, equipment board fault, and equipment power supply fault.

[0071] Understandably, these faults correspond to different power communication network fault query diagrams, and different features are used to query the fault category.

[0072] Combination Figure 7 As shown, Figure 7 The corresponding query graph is shown in Table 1.

[0073] Table 1

[0074]

[0075] In Table 1, the basic information of the first module is (t, number of nodes, number of edges); the basic information of the second module is (v, ID identifier, node label, node degree). v indicates that this row contains node information. Taking v 0 0 1 as an example, the first digit 0 represents the node's unique identifier ID. The second digit 0 represents the node's label, indicating that vertex 0's attribute is a substation, used to describe the node's attribute or type. In power communication fault detection, this is simplified to three categories: station (i.e., substation) is represented by 0, equipment is 1, and its ports are 2. The third digit represents the node's degree, indicating the number of edges connected to the node; here, it indicates there is one edge connected to the node. The basic information of the third module is (e, starting node ID, target node ID). If it starts with 'e', ​​it means this row of data contains edge data. Taking e 0 3 as an example, 0 represents the starting node ID of the edge, i.e., the source node, and 3 represents the target node ID of the edge, i.e., the target node.

[0076] S2 utilizes the power communication network topology and historical fault alarm information to construct a power communication network fault query graph, including:

[0077] Analyze historical defect alarm information and filter out the nodes and links affected by the fault to determine the coverage of defect alarm information;

[0078] In the power communication network data graph to be diagnosed, the nodes within the coverage area of ​​the defect alarm information are connected to generate a power communication network fault query graph.

[0079] Historical defect alarm information includes defect ticket alarm information and defect ticket information. Defect ticket alarm information records the alarm details when the fault was triggered, while defect ticket information stores the fault report content, such as time information, location information, and fault alarm type information.

[0080] S3: Determine the candidate set and vertex degree of each root node in the power communication network fault query graph, and determine the optimal root node in the power communication network fault query graph based on the number of nodes in the candidate set and the vertex degree. Using the optimal root node as the query vertex, query the power communication network data subgraph that is isomorphic to the subgraph of the power communication network fault query graph in the power communication network data graph to be diagnosed based on the graph matching algorithm.

[0081] Optional, combined Figure 3 As shown, in S3, the optimal root node is determined in the fault query graph of the power communication network based on the number of nodes in the candidate set and the vertex degree, including:

[0082] Normalizing the number of nodes and vertex degree in the candidate set can normalize them to the range of (0,1).

[0083] The index value of each root node in the fault query graph of the power communication network is calculated based on the normalized value of the number of nodes and the normalized value of the vertex degree in the candidate set.

[0084] ,

[0085] in, This represents the index value of the i-th root node in the power communication network fault query graph. This represents the normalized degree value of the i-th root node. This represents the normalized value of the number of nodes in the candidate set of the i-th root node. and These represent the weights for degree and the number of nodes in the candidate set, respectively.

[0086] Select the root node with the largest index value as the optimal root node.

[0087] In this way, selecting nodes with fewer candidate nodes and higher degrees as query vertices facilitates better pruning operations and improves matching efficiency.

[0088] Optionally, in S3, using the optimal root node as the query vertex, a graph matching algorithm is used to query a power communication network data subgraph that is isomorphic to the power communication network fault query subgraph in the power communication network data graph to be diagnosed, including:

[0089] S3.1: Convert the power communication network data diagram and power communication network fault query diagram to be diagnosed into a unified input format.

[0090] The unified input format conforms to the input format of the graph matching algorithm.

[0091] S3.2: Using the optimal root node as the query vertex, the breadth-first search method is used to traverse the power communication network fault query graph starting from the query vertex according to the preset traversal order. The nodes in the power communication network fault query graph are sorted to construct the acyclic graph corresponding to the power communication network fault query graph. The matching order is generated according to the topological sorting order of the acyclic graph.

[0092] It is understandable that sorting the nodes in the fault query graph of the power communication network is the traversal order, which is the node access order when constructing an acyclic graph.

[0093] Optionally, nodes and edges in the power communication network fault query graph have corresponding labels. S3.2 sorts the nodes in the power communication network fault query graph, including:

[0094] The edges in the power communication network fault query graph are sorted according to a preset hierarchy, with all edges pointing from higher-level nodes to lower-level nodes.

[0095] Vertices within the same level are grouped by label, and vertices within each group are sorted in descending order of degree.

[0096] Specifically, in the breadth-first search process, the root node is defined as the highest level (e.g., level 0), its direct neighbors are the next level (level 1), and so on. Hierarchical orientation facilitates subsequent matching according to topological order, reducing redundant computation and optimizing the backtracking process. Labels are attribute information of nodes or edges; node labels are device types, and edge labels are connection types. The sorting operation targets the query graph, and the search goal is to find all subgraphs in the data graph that match the query graph structure—that is, subgraphs whose topological structure is completely identical to the query graph, and whose node and edge labels completely match the query graph. Thus, optimizing label grouping and sorting further improves matching efficiency.

[0097] S3.3: According to the matching order, the query vertex in the acyclic graph is matched with the power communication network data graph to be diagnosed in turn. During the backtracking matching process, if the next node that meets the preset matching conditions cannot be found on the current matching path to continue expanding the matching path, the current path is marked as a failed path, and the failed path is skipped to continue matching. The candidate structure is constructed based on the candidate set obtained by matching and the relationship between the nodes in the candidate set.

[0098] Specifically, a candidate structure is built starting from the query vertex. For each query vertex, a set of candidate vertices in the data graph is generated, i.e., a candidate set. This candidate set is part of the candidate structure and represents the data graph nodes that each query vertex might match. The candidate structure also includes the matching relationships between nodes in the candidate set. An acyclic graph is used to build the candidate structure, converting all embeddings in the power communication network fault query graph into embeddings in the candidate structure. Based on the already constructed acyclic graph (i.e., the query graph) and the candidate structure, a backtracking process is performed to find all possible embeddings.

[0099] Optionally, during graph matching, if a match is successful, the matching order is updated according to the following formula:

[0100] ,

[0101] in, To find vertex n in an acyclic graph, Let n be the priority of node n, and let n be the priority of node n. The priority of a node is set according to the node's label and degree. This represents the number of matched neighbors for node n. Let n be the path length from the starting node to node n in the acyclic graph. , , They are respectively , , The corresponding weights.

[0102] Thus, the matching status is recalculated. and Updating the matching order prioritizes matching nodes with more connections to already matched nodes, and prioritizes matching nodes adjacent to already matched nodes. This reduces the search space, allows for faster construction of the matching subgraph, and improves matching quality by prioritizing closely connected nodes. It also better preserves the topology of the query graph. Furthermore, dynamically adjusting the matching order avoids unnecessary matching attempts and improves overall search efficiency.

[0103] Optionally, a subset of the query graph that is related to the partial embedding is found in the search graph, the subset is identified as the failure set, and all partial embeddings containing the failure set are removed from the search space during the graph matching process to determine the final candidate set.

[0104] The failure set shown is a subset of the query graph associated with partial embeddings. A partial embedding occurs when, during graph matching, some nodes in the query graph have been matched to nodes in the data graph, but the matching of the entire query graph is not yet complete. The subset of the query graph and the subgraph in the data graph are topologically identical, meaning the connection relationships between nodes and edges are the same. The failure set indicates that, based on the current partial embedding, no node in the data graph can be found to expand the current partial embedding. The search space can be understood as the matching paths from query graph nodes to data graph nodes. During the matching process, if no next node satisfying the preset matching conditions can be found on the current matching path to continue expanding the matching path, the current path is marked as a failure path, and the failure path is skipped to continue matching. Unmatched query nodes are added to the failure set. All partial embeddings, including those in the failure set, are removed from the search space corresponding to the query graph. This avoids wasting computational resources on invalid paths, thereby improving search efficiency.

[0105] Optional, combined Figure 4 As shown, before matching the query vertices in the acyclic graph with the power communication network data graph to be diagnosed in sequence according to the matching order, S3, using the optimal root node as the query vertex, queries the power communication network data subgraph isomorphic to the power communication network fault query graph subgraph in the power communication network data graph to be diagnosed based on the graph matching algorithm. This also includes:

[0106] A graph index is created for the fault query graph of the power communication network, and the power communication network fault query graph is traversed using the pre-built graph index.

[0107] In this embodiment, the index entries in the graph index may include a unique identifier (node ​​ID) for each node, and the node's attributes or feature vectors (type or degree). Label indexes or degree distribution indexes can be built to quickly filter the candidate set of nodes in the data graph that may match the query. The candidate range can also be further narrowed down through local constraints, such as adjacent edge types.

[0108] In this way, by building a graph index and avoiding direct manipulation of the graph data, we can enhance subsequent pruning capabilities and improve the matching rate.

[0109] Furthermore, the constructed graph index can be compressed, which reduces the space occupied by the graph index, thereby improving the speed of constructing the graph index and further improving the query speed.

[0110] Furthermore, candidate structures are pruned by alternating the use of acyclic graphs in the query graph, specifically as follows:

[0111] Construct a directed acyclic graph based on the query graph. qD and reverse graph qD -1, about to qD Reverse the direction of all edges in;

[0112] For each node u in the query graph, construct an initial candidate structure: initialize the candidate set C(u) as the set of all nodes with the same label in the data graph; if there exists any u such that... If the condition is not met, then no result will be returned directly.

[0113] By alternating the use of directed acyclic graphs qD and reverse graph qD -1 Perform bottom-up optimization on the initial candidate structure:

[0114] Perform simple acyclic graph dynamic programming on the initial candidate structures, based on the inverse graph. qD The topological order of -1 checks whether the nodes in the candidate set meet the matching conditions from the leaf nodes to the root node.

[0115] When performing subgraph matching, extended acyclic graph dynamic programming is used to further refine the candidate set by considering the necessary conditions for embedding (such as neighbor safety and edge matching).

[0116] The candidate structures processed in the second step are then subjected to an inverse graph again. qD -1 is used for extended cycle graph dynamic programming to further reduce invalid candidates.

[0117] Understandably, in each acyclic graph dynamic programming process, the candidate set C(u) is updated based on the calculation results. If a certain node... vIf a candidate does not meet the necessary conditions for embedding, it is removed from C(u). If, after pruning, a C(u) becomes an empty set, then the query graph termination graph matching algorithm cannot be embedded in the data graph. The pruning process only needs to be performed three times to filter out most invalid candidates. When the candidate set no longer changes, the pruning process converges, yielding the final candidate structure.

[0118] This reduces the amount of data that needs further inspection and allows for the rapid elimination of results that do not meet the conditions in a low-cost manner, thus narrowing the candidate set and improving query efficiency and performance.

[0119] S3.4: Based on the acyclic graph and candidate structure, verify whether the candidate nodes in the candidate structure meet the preset matching conditions. If they do not meet the conditions, delete them from the candidate structure to obtain the final power communication network data subgraph that is isomorphic to the power communication network fault query subgraph.

[0120] The accuracy of graph matching results can be further improved by validating the candidate set in the candidate structure.

[0121] Optionally, S3, using the optimal root node as the query vertex and employing a graph matching algorithm to query a power communication network data subgraph isomorphic to the power communication network fault query subgraph in the power communication network data graph to be diagnosed, further includes:

[0122] Before generating the candidate set, check if there are non-tree edges in the power communication network fault query graph. If so, delete the corresponding node in the candidate set.

[0123] In this embodiment, an adaptive matching order framework based on DAG (Directed Acyclic Graph) sorting is employed, where the matching order is the topological order of the acyclic graph constructed based on the query graph. This allows all non-tree edges to be checked as early as possible, thereby reducing redundant Cartesian products.

[0124] S4: Determine the defect or fault result of the power communication network to be diagnosed based on the power communication network data subgraph obtained by graph matching.

[0125] It should be noted that if the subgraph of the power communication network data retrieved is a subgraph with only one vertex and / or a subgraph with more than 2 vertices and a degree of 1 for each node, then there are no defects or faults in the power communication network.

[0126] In this embodiment, the methods for determining equipment board faults and equipment power supply faults in communication network defects are quite unique. Equipment board faults are characterized by originating from the same board or different boards within the same device. Equipment power supply faults, on the other hand, originate from the board itself; typically, a substation board issues a "POWER ABNORMAL" error, or a substation protection network issues a "POWER FAIL" error. These faults are both board-level faults, with relatively singular origins. Therefore, these two types of easily identifiable faults can be directly queried, or the methods provided in this embodiment can be used for fault querying. The main difference between optical path interruption and optical cable interruption lies at the substation site level. If there is no connection between substations, it is an optical cable interruption; conversely, it is an optical path interruption.

[0127] Therefore, considering the above situations and the application background of power communication network defect diagnosis, the embodiments of the present invention exclude some special cases. For data graph subgraphs that meet any of the following conditions, the embodiments of the present invention will not consider them as having defects or faults: subgraphs with only one vertex, no connecting edges connected to it, subgraphs with more than 2 vertices and nodes with a degree of 1 and no alarms.

[0128] Embodiment 2 of the present invention provides a power communication network defect fault diagnosis system based on graph matching, which operates the power communication network defect fault diagnosis method based on graph matching as described in Embodiment 1. The system includes:

[0129] The acquisition module is used to acquire the topology of the power communication network to be diagnosed and convert the topology of the power communication network to be diagnosed into a data graph of the power communication network to be diagnosed.

[0130] The module is used to construct a fault query graph of the power communication network using the topology of the power communication network and historical defect alarm information. Each power communication network fault query graph corresponds to a different type of communication network defect.

[0131] The graph matching module is used to determine the candidate set and vertex degree of each root node in the power communication network fault query graph, and to determine the optimal root node in the power communication network fault query graph based on the number of nodes in the candidate set and the vertex degree. Using the optimal root node as the query vertex, the module uses the graph matching algorithm to query the power communication network data subgraph that is isomorphic to the power communication network fault query graph subgraph in the power communication network data graph to be diagnosed.

[0132] The determination module is used to determine the defect and fault results of the power communication network to be diagnosed based on the power communication network data subgraph obtained by graph matching.

[0133] Embodiment 3 of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded onto the processor, it implements the graph matching-based power communication network defect diagnosis method described in Embodiment 1.

[0134] Embodiment 4 of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the graph matching-based power communication network defect diagnosis method according to Embodiment 1.

[0135] Combination Figure 8 As shown, Embodiment 5 of the present invention provides a verification embodiment.

[0136] This embodiment analyzes a relatively large-scale local power communication network in China, which has 208 nodes and 297 computational edges. The operating environment for this embodiment is an Intel(R) Core i9-13900HX CPU with a clock speed of 5.40 GHz, 16GB of memory, and Ubuntu 18.04 operating system. Graph matching is achieved by constructing a query graph, establishing candidate structures, backtracking, and pruning using the failure set. In constructing the query graph, an acyclic graph is built starting from query graph 𝑞. In this process, the first step is to select a query vertex as the root node. When selecting this root node, both the candidate set and the vertex's degree need to be considered. To facilitate better pruning, the vertex should satisfy both the conditions of having a small number of candidates and a high degree. Then, a breadth-first search (BFS) method is used to traverse the query graph starting from the root node, orienting all edges from high to low level. Vertices within the same level are grouped and sorted by label, so that vertices with less common labels are placed at the front, and vertices within each group are sorted in descending order of degree. The second step uses an acyclic graph (AGraph) to build a candidate structure. The goal here is to convert all embeddings of the query graph into embeddings in the candidate structure. In the third step, during backtracking, based on the constructed query graph and candidate structure, a backtracking process is performed to find all possible embeddings. This process employs an adaptive matching order framework based on DAG sorting, where the matching order always follows the topological order of the acyclic graph. This allows for early checking of all non-tree edges, thus reducing redundant Cartesian products. The fourth step introduces the concept of a failure set to prune part of the search space. The failure set is a subset of the query graph N associated with partial embeddings, where no embedding in N extends the partial embeddings. Once the failure set is found, some partial embeddings in the search space can be safely removed. During enumeration, a backtracking-based framework is used to recursively explore all possible matches. Matching is attempted starting with the first node in the matching order; if all query nodes can be successfully matched, the match is added to the result set N. If the recursion termination condition has not been met, the algorithm continues to select the next query node, further narrowing down its candidate node set, traversing all possible candidate nodes, updating the matching state, and recursively entering the matching process for the next query node.

[0137] To investigate the impact of graph matching technology on the target communication network, this paper compares and analyzes it with the relational database currently used by the power grid. In relational databases, the join operation can be used to combine rows from two or more tables in SQL operations, which is very useful when processing data with complex relationships. However, it can become time-consuming on large datasets, especially when multiple tables are joined, potentially leading to performance bottlenecks. For large-scale communication networks, graph matching technology offers better scalability and performance advantages compared to traditional join queries. The evaluated defect order data contains duplicate orders, thus requiring a reduction in duplicate defect orders. This phenomenon stems from:

[0138] In real-world operating environments, defects sometimes fail to be detected and fixed in a timely manner, causing the same problem to trigger alarms again after a period of time. Due to the long time interval between the two alarms, the system often treats multiple alarms caused by the same root cause as different independent events, thus generating multiple defect reports.

[0139] Furthermore, when multiple geographically dispersed sites share the same line or the same equipment, if the line or equipment malfunctions, the diagnostic system, based on manual judgment or preset rules, may mistakenly identify it as being caused by multiple independent defects, and thus create multiple independent defect records.

[0140] Experimental results show that traditional relational databases are significantly slower than graph matching algorithms when dealing with various fault types. Among the newer graph matching algorithms, the algorithm in this embodiment is more efficient in terms of time processing than CECI's algorithm. In the query for these three fault types, the query efficiency for power supply equipment terminals is higher than for the other two fault types, optical path interruption and optical cable interruption, because its characteristics are more obvious compared to the other two types of graph matching patterns. Experimental comparisons demonstrate that graph matching is more effective for fault diagnosis.

[0141] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0142] This disclosure can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of this disclosure.

[0143] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination of the foregoing. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0144] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0145] Computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing the status information of the computer-readable program instructions to implement various aspects of this disclosure.

[0146] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for diagnosing defects of a power communication network based on graph matching, characterized in that, Comprise: S1: taking the station, device, port as the node, taking the optical path connection as the edge, constructing the topology structure of the power communication network to be diagnosed through the mapping of the optical path and the station, device, port information, adding the defect alarm information to the node or edge corresponding to the topology structure of the power communication network, forming a dynamic fault label and converting the topology structure of the power communication network to be diagnosed into a power communication network data graph to be diagnosed; S2: constructing a power communication network fault query graph using the power communication network topology structure and historical defect alarm information, each power communication network fault query graph corresponding to a different type of communication network defect fault; S3: determining the candidate set and vertex degree of each root node in the power communication network fault query graph, and determining the optimal root node in the power communication network fault query graph according to the number of nodes in the candidate set and the vertex degree, and querying the power communication network data subgraph isomorphic to the power communication network fault query graph subgraph in the power communication network data graph to be diagnosed based on the graph matching algorithm with the optimal root node as the query vertex; wherein, the candidate structure is constructed from the power communication network fault query vertex, the candidate structure includes the candidate set and the matching relationship between the nodes in the candidate set, and the acyclic graph of the power communication network fault query graph is alternately used to prune the candidate structure; S4: determining the defect fault result of the power communication network to be diagnosed according to the power communication network data subgraph obtained by graph matching; In S3, the optimal root node in the power communication network fault query graph is determined according to the number of nodes in the candidate set and the vertex degree, comprising: Normalizing the number of nodes in the candidate set and the vertex degree; According to the normalized value of the number of nodes in the candidate set and the normalized value of the vertex degree, the index value of each root node in the power communication network fault query graph is calculated; ， wherein, denotes the i-th root node indicator value of the power communication network fault query graph, denotes the degree normalization value of the i-th root node, denotes the candidate set node number normalization value of the i-th root node, and denote the weights of the degree and the candidate set node number, respectively; Select the root node with the maximum index value as the optimal root node.

2. The power communication network defect fault diagnosis method based on graph matching according to claim 1, wherein: In S2, the power communication network fault query graph is constructed using the power communication network topology structure and historical defect alarm information, comprising: Analyzing the historical defect alarm information and screening out the nodes and links affected by the fault to determine the coverage range of the defect alarm information; In the power communication network data graph to be diagnosed, the nodes in the coverage range of the defect alarm information are connected to generate the power communication network fault query graph.

3. The power communication network defect fault diagnosis method based on graph matching according to claim 1, wherein: In S3, the power communication network data subgraph isomorphic to the power communication network fault query graph subgraph in the power communication network data graph to be diagnosed is queried based on the graph matching algorithm with the optimal root node as the query vertex, comprising: Convert the power communication network data graph to be diagnosed and the power communication network fault query graph into a unified input format; The optimal root node is taken as a query vertex, a breadth-first search method is used to traverse the power communication network fault query graph according to a preset traversal order from the query vertex, nodes in the power communication network fault query graph are sorted, a loop-free graph corresponding to the power communication network fault query graph is constructed, and a matching order is generated according to a topological order of the loop-free graph; According to the matching order, the query vertex in the loop-free graph is matched with the power communication network data graph to be diagnosed in sequence, and in the backtracking matching process, if the next node that meets the preset matching condition cannot be found on the current matching path to continue to expand the matching path, the current path is marked as a failed path, and the matching is continued by skipping the failed path, and a candidate structure is constructed based on a candidate set obtained by matching and a relationship between nodes in the candidate set; Based on the loop-free graph and the candidate structure, whether the candidate nodes in the candidate structure meet the preset matching condition is verified, and if not, the candidate nodes are deleted in the candidate structure, to obtain a power communication network data subgraph that is isomorphic to the power communication network fault query subgraph.

4. The power communication network defect fault diagnosis method based on graph matching according to claim 3, wherein: The nodes and edges in the power communication network fault query graph have corresponding labels, and the nodes in the power communication network fault query graph are sorted, including: The edges in the power communication network fault query graph are sorted according to a preset level, and the directions of all edges are from high-level nodes to low-level nodes; The vertices in the same level are grouped according to labels, and the vertices in each group are arranged in descending order according to the degrees of the vertices.

5. The power communication network defect fault diagnosis method based on graph matching according to claim 3, wherein: In S3, the optimal root node is taken as a query vertex, and a graph matching algorithm is used to query a power communication network data subgraph that is isomorphic to a power communication network fault query subgraph in the power communication network data graph to be diagnosed, and further comprising: Checking whether there is a non-tree edge in the power communication network fault query graph, and if there is, deleting the corresponding node in the candidate set.

6. The power communication network defect fault diagnosis method based on graph matching according to claim 3, wherein: Before sequentially matching the query vertex in the loop-free graph with the power communication network data graph to be diagnosed according to the matching order, in S3, the optimal root node is taken as a query vertex, and a graph matching algorithm is used to query a power communication network data subgraph that is isomorphic to a power communication network fault query subgraph in the power communication network data graph to be diagnosed, and further comprising: Establishing a graph index for the power communication network fault query graph, and traversing the power communication network fault query graph by using the pre-constructed graph index.

7. The power communication network defect fault diagnosis method based on graph matching according to claim 3, wherein: In the graph matching process, if the matching is successful, the matching order is updated according to the following formula: ， wherein, is the priority of node n in the acyclic graph, is the priority of node n, which is set according to the label and degree of the node, is the number of matched neighbors of node n, is the path length from the start node to node n in the acyclic graph, , , are the respective weights, , , corresponding weights.

8. A graph matching-based power communication network defect fault diagnosis system, which operates the graph matching-based power communication network defect fault diagnosis method according to any one of claims 1 to 7, characterized by, The system comprises: An acquisition module configured to acquire a topology structure of a power communication network to be diagnosed, and convert the topology structure of the power communication network to be diagnosed into a power communication network data graph to be diagnosed; The constructing module is configured to construct power communication network fault query graphs by using the power communication network topology and historical defect alarm information, each power communication network fault query graph corresponding to a different type of communication network defect fault; The graph matching module is configured to determine a candidate set and a vertex degree of each root node in the power communication network fault query graph, determine an optimal root node in the power communication network fault query graph according to the number of nodes in the candidate set and the vertex degree, and query a power communication network data subgraph isomorphic to a subgraph of the power communication network fault query graph in the power communication network data graph to be diagnosed based on the graph matching algorithm with the optimal root node as a query vertex. The determining module is configured to determine a defect fault result of the power communication network to be diagnosed according to the power communication network data subgraph obtained by the graph matching. 9.An electronic device comprising a processor and a storage medium; characterized in that: The storage medium is configured to store instructions; The processor is configured to operate according to the instructions to perform the steps of the power communication network defect fault diagnosis method based on graph matching according to any one of claims 1-7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the steps of the power communication network defect fault diagnosis method based on graph matching according to any one of claims 1-7.

Images