Automatic planning system for communication routing path of transformer substation

Abstract

The invention relates to the technical field of electric power communication, and particularly discloses an automatic planning system for a substation communication routing path, which is used for solving the problem that the traditional substation communication routing mainly depends on a static routing strategy and is difficult to adapt to dynamically changing network environments such as line faults, bandwidth change and equipment damage. The problem that a shortest path algorithm and a large model technology are combined does not exist; according to the method, network topology structure data, node attribute data and link attribute data are extracted from a substation communication network, the shortest path between a source node and a target node is calculated by using a shortest path algorithm, a result is stored in a distance table, and a preliminary path is calculated through the parallel shortest path; and training a large model by using the marked data set, inputting the preprocessed data into the trained large model, obtaining a network state sensing result, and combining a shortest path calculation result and the network state sensing result to improve the intelligence of the substation communication network.

Description

Technical Field

[0001] The present invention relates to the technical field of power communication, and more specifically, to a substation communication routing path automatic planning system. Background Art

[0002] With the development of smart grid technology, the complexity and data volume of substation communication networks have increased rapidly. The modern power grid has higher and higher requirements for communication routing planning. The traditional routing selection method based on manual experience has been difficult to meet the optimal routing requirements in a large-scale and complex network environment. The existing literature (Xu Wei, Liu Zhenyu. Research on Low-Latency Communication Routing Technology in Power Internet of Things [J]. Industrial Innovation Research, 2024, (16): 99-101.) analyzed and discussed the implementation path of low-latency communication routing technology in the power Internet of Things, proposed to use an SDN controller to allocate bandwidth resources, apply network slicing technology to improve the reliability of data transmission, rely on routing selection optimization to reduce communication transmission latency, and gave Figure 2 as shown in the schematic diagram of the multi-path low-latency communication routing system model.

[0003] In order to improve communication efficiency and optimize network resource allocation, it is particularly important to study automated and intelligent communication path planning technologies. In the prior art, the shortest path algorithm is a classic problem in graph theory. Its goal is to find the route with the smallest total weight of the path from the source node to the target node in the network topology, and it has good computational efficiency in path optimization. Substation internal and external communications usually require low latency and high reliability. In the event of emergencies such as equipment failures and link interruptions, the shortest path algorithm can quickly recalculate available paths to ensure the continuity of the communication system; while the large model is a complex model constructed based on technologies such as deep learning, big data analysis, and graph neural network (GNN). Its core advantage is that it can automatically extract features, capture potential patterns from massive data, and perform accurate prediction and decision support, and can better explore the potential patterns of the network topology and adapt to the dynamically changing network state. In an environment where the network state is constantly changing, the large model technology can be used to perform real-time modeling and prediction of the current network state, and then dynamically adjust the routing strategy. For example, by learning the current topology through GNN and then combining the short-term prediction of traffic by Transformer, the optimal routing can be planned in advance; traditional substation communication routing mainly relies on static routing strategies and is difficult to adapt to the dynamically changing network environment, such as line failures, bandwidth changes, equipment damage, etc., and does not combine the shortest path algorithm and the large model technology to solve the above problems. To solve the above problems, a technical solution is provided herein. Summary of the Invention

[0004] To overcome the above-mentioned defects of the prior art, the present invention provides an automatic planning system for substation communication routing paths, which combines the shortest path algorithm and large model technology to solve the problem that traditional substation communication routing mainly relies on static routing strategies and is difficult to adapt to dynamic network environments such as line failures, bandwidth changes, equipment damage, etc., and does not combine the shortest path algorithm and large model technology, so as to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] An automatic planning system for substation communication routing paths includes a data preprocessing module, a shortest path calculation module, a large model training module, a network status perception module, a routing planning module, and a result output module; the data preprocessing module is used to extract network topology structure data, node attribute data, and link attribute data from the substation communication network, and perform cleaning, conversion, and annotation on the data; the shortest path calculation module is used to calculate the shortest path between the source node and the target node using the shortest path algorithm, and save the result to the distance table, and calculate the preliminary path through parallel shortest path calculation; the shortest path calculation module includes a shortest path update unit; the shortest path update unit is used to select the node u with the smallest current path distance using a priority queue, traverse all neighbor nodes v of node u, and update to find a shorter path through node u. If the path length from the current node v to node u is less than the known path length d[v] from the origin to node v, then update d[v] to the path length from node u to node v. The formula for updating the shorter path found through node u is:

[0007] if: d[u]+w(u,v)<d[v];

[0008] then: d[v] = d[u]+w(u,v);

[0009] In the formula: d[u] is the current known shortest path length from the source point to node u, w(u,v) is the weight of the edge from node u to node v, and d[v] is the current known shortest path length from the source point to node v;

[0010] Repeat the above steps until all nodes are visited.

[0011] As a further solution of the present invention, the data preprocessing module is used to extract network topology structure data, node attribute data, and link attribute data from the substation communication network, and perform cleaning, conversion, and annotation on the data;

[0012] The large model training module is used to train the large model using the labeled dataset;

[0013] The network status perception module is used to input the preprocessed data into the trained large model to obtain the perception result of the network status;

[0014] The routing planning module is used to combine the shortest path calculation result and the network status perception result to perform routing planning according to business rules and constraint conditions;

[0015] The result output module is used to output the planning result in a visual or text form.

[0016] As a further solution of the present invention, the data preprocessing module includes a data collection unit, a data cleaning unit, and a data conversion unit;

[0017] The data collection unit is used to collect the topology data of the substation communication network, the node attribute data (including substation nodes, control center nodes, communication nodes), and the link (communication link) attribute data (including link delay, link stability, and link load);

[0018] The data cleaning unit is used to filter out the abnormal and invalid data to ensure the accuracy and consistency of the data;

[0019] The data conversion unit is used to uniformly convert the data into a graph data structure and store it using the graph data structure, giving the adjacency matrix A and the weight matrix W; among them, the adjacency matrix A is expressed as:

[0020]

[0021] In the formula: A ij is the adjacency matrix of the edge from node i to node j;

[0022] The weight matrix W is expressed as:

[0023]

[0024] In the formula: W ij is the weight of the edge from node i to node j.

[0025] As a further solution of the present invention, the shortest path calculation module further includes a path initialization unit, a parallel shortest path calculation unit, and a preliminary path result output unit; the path initialization unit is used to set the initial distance from the source node to all nodes to infinity, and the distance from the source node to itself to 0; the parallel shortest path calculation unit is used to perform hierarchical processing by dividing the path distance set and update the paths in the distance set through parallel calculation; the preliminary path result output unit is used to store the shortest path length from the source node to the target node and record the shortest path length.

[0026] As a further solution of the present invention, the parallel shortest path calculation unit performs hierarchical processing by dividing the path distance set, and updates the paths in the distance set in combination with parallel calculation. The specific steps are as follows:

[0027] Divide the path distance set: Set the step size parameter Δ, and divide all nodes into the path distance set according to the current shortest path distance through the path division model. The formula of the path division model is:

[0028] B[p] = {v ∈ Q | pΔ ≤ d[v] < (p + 1)Δ};

[0029] In the formula: B[p] is the p-th path distance set, pΔ is the minimum value in the p-th path distance set, (p + 1)Δ is the maximum value in the p-th path distance set, and p is the index in the path distance set, which is determined by the shortest node path;

[0030] Set initialization: Set the initial node s. At this time, the minimum node distance d[s] = 0, the minimum distance d[c] between the initial node and the remaining nodes c is d[c] = ∞, and At this time, the initial node is divided into the first path distance set B[0];

[0031] Process the nodes in the set: Select the path distance set B[p] with the current minimum index, and perform the following operations in parallel for all nodes in the path distance set B[p]:

[0032] d[c] = min(d[c], d[a] + w[c, a]);

[0033] In the formula: d[c] is the minimum distance between the initial node and the remaining node c, d[a] is the minimum distance between the initial node and node c, w[c, a] is the weight of the edge from node c to node a, and min is to take the smaller of the two values;

[0034] If d[c] is updated, the node is reallocated to a new path distance set, and the processed path distance set is removed. If all current path distance sets are non-empty, continue to select the path distance set with the minimum path distance until the shortest path distances of all nodes have been updated.

[0035] As a further solution of the present invention, the network state perception module is used to input the preprocessed data into the trained large model to obtain the perception result of the network state. The specific steps include:

[0036] Data preparation: Construct a training data set, including the input network topology structure, node and link attributes, and corresponding label data (including target paths, link states);

[0037] Model design: Learn the state embedding vector h for each node through the graph neural network modelv ∈R s , where R s is an s-dimensional real vector space, and h v is the state vector of the node and can generate the output vector o v ;

[0038] h v = f(x v , x co[v] , h ne[v] , x ne[v] );

[0039] o v = g(h v , x v );

[0040] In the formula: x v is the feature vector of node v, x co[v] is the feature vector of the edge associated with node v, h ne[v] is the state vector of the neighbor node of node v, x ne[v] is the feature vector of the neighbor node of node v, f() is the local transfer function, and g() is the local output function;

[0041] Calculate state parameters based on the feature vector, state vector, output vector of the node, and the feature vector of the neighbor node.

[0042] The technical effects and advantages of an automatic substation communication routing path planning system of the present invention: By extracting network topology structure data, node attribute data, and link attribute data from the substation communication network, calculating the shortest path between the source node and the target node using the shortest path algorithm, and saving the result in the distance table, calculating the preliminary path through parallel shortest path calculation, training a large model using the labeled dataset, inputting the preprocessed data into the trained large model, obtaining the perception result of the network state, and combining the shortest path calculation result and the network state perception result, the intelligence of the substation communication network is improved, communication congestion is avoided, and network stability is guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 is a schematic structural diagram of an automatic substation communication routing path planning system provided by the present invention;

[0044] Figure 2 is a schematic diagram of a multi-path low-latency communication routing system model in the prior art;

[0045] Figure 3 is a flowchart of the operation of an automatic substation communication routing path planning system provided by the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0046] Next, the technical solutions in the present invention will be clearly and completely described in conjunction with the accompanying drawings in the present invention. Obviously, the described technical solutions are only a part of the present invention, rather than all of it. Based on the technical solutions in the present invention, all other technical solutions obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0047] A substation communication routing path automatic planning system includes a data preprocessing module, a shortest path calculation module, a large model training module, a network status perception module, a routing planning module, and a result output module; the data preprocessing module is connected to the shortest path calculation module, the shortest path calculation module is connected to the large model training module, the large model training module is connected to the network status perception module, the network status perception module is connected to the routing planning module, and the routing planning module is connected to the result output module.

[0048] The data preprocessing module is used to extract network topology structure data, node attribute data, and link attribute data from the substation communication network, and perform data cleaning, conversion, and annotation on the data.

[0049] The shortest path calculation module is used to calculate the shortest path between the source node and the target node using the shortest path algorithm, save the result in the distance table, and calculate the preliminary path through parallel shortest path calculation.

[0050] The large model training module is used to train the large model using the labeled dataset.

[0051] The network status perception module is used to input the preprocessed data into the trained large model to obtain the perception result of the network status.

[0052] The routing planning module is used to combine the shortest path calculation result and the network status perception result, and perform routing planning according to service rules and constraint conditions.

[0053] The result output module is used to output the planning result in a visual or text form.

[0054] Specifically, the data preprocessing module includes a data collection unit, a data cleaning unit, and a data conversion unit.

[0055] The data collection unit is used to collect the topology data of the substation communication network, node attribute data (including substation nodes, control center nodes, and communication nodes), and link (communication link) attribute data (including link delay, link stability, and link load).

[0056] The data cleaning unit is used to filter out abnormal and invalid data to ensure the accuracy and consistency of the data.

[0057] The data conversion unit is used to uniformly convert data into a graph data structure and store it using the graph data structure, giving the adjacency matrix A and the weight matrix W. Among them, the adjacency matrix A is expressed as:

[0058]

[0059] In the formula: A ij is the adjacency matrix of the edge from node i to node j;

[0060] The weight matrix W is expressed as:

[0061]

[0062] In the formula: W ij is the weight of the edge from node i to node j.

[0063] The data collection unit focuses on collecting various types of data, including communication network topologies, node attributes (substations, control centers, communication nodes), and link attributes (delay, stability, load), etc., ensuring rich and comprehensive data sources; the data cleaning unit is responsible for removing abnormal and invalid data, ensuring data accuracy and consistency, and providing a reliable basis for subsequent analysis; the data conversion unit uniformly converts the original data into a graph data structure (by constructing the adjacency matrix A and the weight matrix W), facilitating subsequent processing and analysis using graph algorithms; through the filtering of abnormal data and invalid data by the data cleaning unit, data noise and errors can be significantly reduced, improving the overall quality and reliability of the data, thereby providing more accurate support for decision-making; after converting the data into a graph data structure, various algorithms in graph theory (such as shortest path, connectivity analysis, network optimization, etc.) can be used to efficiently analyze and schedule the substation communication network; the adjacency matrix A clearly expresses the direct connection relationship between nodes, while the weight matrix W provides specific weight information for the edges (such as link weights or set to infinity to indicate disconnection), enabling graph algorithms to quickly obtain the relationship between nodes and transmission costs; the data preprocessing module decomposes and organizes the data processing process, enabling each unit to be optimized for specific tasks, improving the overall data processing efficiency, and at the same time adopting a standardized data format (such as a graph data structure), making the interfaces of subsequent systems or algorithms more unified and reducing the complexity of system integration and maintenance.

[0064] Specifically, the shortest path calculation module includes a path initialization unit, a shortest path update unit, a parallel shortest path calculation unit, and a preliminary path result output unit; the path initialization unit is connected to the shortest path update unit, the shortest path update unit is connected to the parallel shortest path calculation unit, and the parallel shortest path calculation unit is connected to the preliminary path result output unit;

[0065] The path initialization unit is used to set the initial distance from the source node to all nodes to infinity and the distance from the source node to itself to 0;

[0066] The shortest path update unit is used to select the node u with the minimum current path distance using a priority queue. The formula for selecting the node u with the minimum current path distance is:

[0067]

[0068] where: u is the node with the minimum current path distance, v is the node v, d[v] is the currently known shortest path length from the source point to the node v, q is the set of nodes, is to filter the node with the minimum current distance;

[0069] Traverse all neighbor nodes v of the node u and update to find a shorter path through the node u. If the path length from the current node v to the node u is less than the known path length d[v] from the origin to the node v, then update d[v] to the path length from the node u to the node v. The formula for updating the shorter path found through the node u is:

[0070] if: d[u] + w(u, v) < d[v];

[0071] then: d[v] = d[u] + w(u, v);

[0072] where: d[u] is the currently known shortest path length from the source point to the node u, w(u, v) is the weight of the edge from the node u to the node v, and d[v] is the currently known shortest path length from the source point to the node v;

[0073] Repeat the above steps until all nodes have been visited;

[0074] The parallel shortest path calculation unit is used to perform hierarchical processing by partitioning the path distance set and update the paths in the distance set in combination with parallel computing;

[0075] The preliminary path result output unit is used to store the shortest path length from the source node to the target node and record the shortest path length.

[0076] The running code of the shortest path update unit is as follows:

[0077]

[0078]

[0079]

[0080]

[0081] The path initialization unit uniformly sets the initial distances from the source node to each node, ensuring the consistency of the starting state, and sets the distance of the source node itself to 0 to form a clear benchmark; the shortest path update unit uses a priority queue to select the node with the smallest current path distance, thus ensuring that the path information is updated in the most favorable direction in each iteration; the parallel shortest path calculation unit parallelizes the calculation tasks by hierarchical processing of the path distance set, making full use of multi-core or distributed computing resources to improve the overall computing efficiency; the preliminary path result output unit timely records and outputs the shortest path results for convenient subsequent analysis and scheduling; using the priority queue mechanism to select the node with the smallest current distance can quickly narrow the search scope, reduce unnecessary calculations, traverse all neighbors of the node and update the path length in real time, ensuring that each update moves towards the optimal solution, which conforms to the basic principle of the greedy strategy; by partitioning the path distance set and adopting parallel computing, the shortest path calculation task for a large-scale graph is decomposed into multiple smaller subtasks, thereby reducing the overall computing time; hierarchical processing of the path distance set can not only effectively utilize computing resources, but also improve the system's response speed and scalability, and is applicable to scenarios with large-scale networks or high real-time requirements; through careful functional partitioning and optimized design of the shortest path calculation module, not only the computing efficiency and accuracy are improved, but also the system's expansion ability and adaptability are enhanced, and it can provide an efficient and reliable solution for the shortest path problem in complex networks.

[0082] Specifically, the parallel shortest path calculation unit performs hierarchical processing by partitioning the path distance set, and combines parallel computing to update the paths in the distance set. The specific steps are as follows:

[0083] Partition the path distance set: Set the step size parameter Δ, and divide all nodes into the path distance set according to the current shortest path distance through the path partitioning model. The formula of the path partitioning model is:

[0084] B[p] = {v ∈ Q | pΔ ≤ d[v] < (p + 1)Δ};

[0085] In the formula: B[p] is the p-th path distance set, pΔ is the minimum value in the p-th path distance set, (p + 1)Δ is the maximum value in the p-th path distance set, and p is the index within the path distance set, which is determined by the shortest node path;

[0086] Set initialization: Set the initial node s. At this time, the minimum node distance d[s] = 0, the minimum distance d[c] between the initial node and the remaining nodes c = ∞, and At this time, the initial node is divided into the first path distance set B[0];

[0087] Processing of set nodes: Select the path distance set B[p] with the current minimum index, and perform the following operations in parallel on all nodes within the path distance set B[p]:

[0088] d[c] = min(d[c], d[a] + w[c, a]);

[0089] Where: d[c] is the minimum distance from the initial node to the remaining node c, d[a] is the minimum distance from the initial node to node c, w[c, a] is the weight of the edge from node c to node a, and min is to take the smaller of the two values;

[0090] If d[c] is updated, the node is reallocated to the new path distance set, and the processed path distance set is removed. If the current path distance set is not empty, continue to select the minimum path distance set until the shortest path distances of all nodes have been updated.

[0091] By dividing the nodes into different sets according to the current shortest path distance, the shortest path calculation problem in a large-scale graph can be decomposed into several sub-problems. Within each path distance set, the distances between nodes can be updated simultaneously, thus making full use of parallel computing resources and significantly accelerating the update speed; after the nodes are divided into different distance sets, only the set with the current minimum index needs to be processed each time, reducing the frequency of global node traversal and the unnecessary computational amount. When the shortest path distance of a node is updated, it is reallocated to the corresponding path distance set, which can quickly spread the update result and avoid repeated calculations; the path distance set and the parallel computing module are independent of each other, facilitating expansion or adjustment in scenarios with large-scale networks and high real-time requirements. By setting different step parameters, the granularity of set division can be flexibly adjusted to adapt to network situations of different scales and complexities; always selecting the path distance set with the current minimum index for processing can quickly discover and update the shortest path, so that the algorithm converges faster as a whole. In local parallel updates, by considering multiple paths simultaneously, shorter paths can be found more efficiently, promoting the rapid iterative update of the overall shortest path calculation; through the hierarchical management and parallel update strategy of the path distance set, both the calculation process is optimized and the calculation speed and efficiency are significantly improved, which is applicable to the real-time solution of the shortest path problem in large-scale complex networks.

[0092] Specifically, the network state perception module is used to input the preprocessed data into the trained large model to obtain the perception result of the network state. The specific steps include:

[0093] Data preparation: Construct a training data set, including the input network topology structure, node and link attributes, and corresponding label data (including target paths, link states);

[0094] Model design: Use a graph neural network model to learn the state embedding vector h for each node v ∈R s, where R s is an s-dimensional real vector space, h v is the state vector of the node, which can generate the output vector o v ;

[0095] h v = f(x v , x co[v] , h ne[v] , x ne[v] );

[0096] o v = g(h v , x v );

[0097] In the formula: x v is the feature vector of node v, x co[v] is the feature vector of the edge associated with node v, h ne[v] is the state vector of the neighbor node of node v, x ne[v] is the feature vector of the neighbor node of node v, f() is the local transfer function, and g() is the local output function;

[0098] Calculate the state parameters based on the feature vector, state vector, output vector of the node, and the feature vector of the neighbor node;

[0099] Model training: Use the training dataset to train the model, and optimize the performance of the model by continuously adjusting the parameters so that it can accurately predict the network state and path selection;

[0100] Model evaluation: Use the validation dataset to evaluate the accuracy and generalization ability of the model to ensure that the model can effectively predict the network state.

[0101] By inputting various data such as network topology structure, node features, and link attributes, it can comprehensively reflect the actual situation of the network. Using the local transfer function and local output function of the graph neural network, it can not only capture the local features of the node itself and its neighbors, but also transmit the global network state information through the state embedding vector. After sufficient training and evaluation, the model can accurately predict the network state and link state, providing a reliable basis for subsequent path selection and network scheduling decisions, and obtaining and updating the node state in real time, which helps to timely detect network failures or bottlenecks, and then adjust the network path and resource allocation to improve the overall operation efficiency of the network. The constructed training dataset contains rich network instances and label information, enabling the model to have good generalization ability when facing different network topologies and state changes. Evaluating the model through the validation dataset and continuously optimizing the model parameters can effectively prevent overfitting problems and improve the robustness of the model in practical applications. By utilizing the powerful feature extraction ability of the graph neural network, it realizes the fine capture and accurate prediction of the network state, not only improving the network operation efficiency, but also providing a solid data support and intelligent analysis basis for subsequent path selection and network scheduling decisions.

[0102] Specifically, the network state perception module inputs the preprocessed data into the trained large model to obtain the perception result of the network state. The specific steps are as follows:

[0103] Data input: Input the real-time collected network data (including node state and link state) into the trained graph neural network model;

[0104] State perception: Through the inference process of the model, obtain the overall situation of the network and the real-time states of each node and link;

[0105] Result analysis: Analyze the perception result output by the model to understand the current network health status, potential problems, and optimization opportunities.

[0106] Specifically, the routing planning module is used to combine the shortest path calculation result and the network state perception result to perform routing planning according to business rules and constraint conditions;

[0107] Initial path planning: Based on the shortest path calculation result, initially determine the path plan;

[0108] Path optimization: Combine the network state perception result to optimize and adjust the initial path plan;

[0109] Evaluation and verification: Evaluate the optimized path plan to ensure that it meets the business rules and constraint conditions;

[0110] Finally determine the path: Output the finally determined routing path plan.

[0111] By inputting network data (node status, link status) into a pre-trained graph neural network model in real time, the overall health status of the network and the real-time status of each component can be reflected in a timely manner. The perception results output by the model help to quickly identify potential problems and provide a basis for subsequent routing adjustment and maintenance decisions; using the shortest path calculation results, a scientific initial plan for routing planning is provided to ensure selection within the optimal path range. Combining the network status perception results, the preliminary plan is optimized and adjusted according to the actual situation of the current network to achieve a more intelligent and flexible routing planning; in the process of path planning, business rules and network constraints are fully considered to ensure that the final plan is not only optimal mathematically but also meets the safety and reliability requirements of practical applications. It can consider traditional indicators such as path length and delay, and can also integrate real-time status information such as network load and link stability to achieve a comprehensive balance of multiple indicators; continuously adjust the path planning according to the real-time network status to make the allocation of network resources (such as bandwidth, load) more reasonable, thereby improving the overall operation efficiency. The optimized path plan can quickly respond to network anomalies or traffic fluctuations and reduce the risk of service interruption caused by congestion or failures; by integrating real-time network status perception and intelligent routing planning, not only accurate monitoring and dynamic adjustment of the network are achieved, but also an efficient and reliable routing solution can be provided under business rules and network constraints, thereby greatly enhancing the intelligence and adaptability of network operation.

[0112] Embodiment 2

[0113] As Figure 3 shown in the operation flowchart of an automatic planning system for substation communication routing paths, the operation steps of an automatic planning system for substation communication routing paths are as follows:

[0114] Data preprocessing: In a substation communication network, extract network topology data, including the connection relationships of each substation, control center, and communication node. During the preprocessing process, clean the data and convert it into a standard format, label the node types and link attributes, and provide a data basis for subsequent large model training;

[0115] Shortest path calculation: Calculate the shortest path from the source substation to the target substation. Select the node with the smallest current distance through a priority queue and gradually update the distances of neighbor nodes to finally determine the shortest path;

[0116] Large model training: Use the collected network topology and attribute data to train a graph neural network model. The model design includes an input layer, a hidden layer, and an output layer. During the training process, repeatedly adjust the parameters to optimize the prediction ability of the model;

[0117] Network status perception: During the network operation, real-time network data is input into the trained large model to obtain the perception results of the current network status, including the status of each node and the availability information of the links;

[0118] Routing planning: Combining the shortest path calculation results and the network status perception results, path planning is carried out according to service requirements. The initial path plan is based on the shortest path algorithm and is optimized and adjusted in combination with the real-time network status to finally determine the optimal path plan;

[0119] Result output: The finally determined routing path plan is displayed to the user through a graphical interface, including the detailed information of the nodes and links on the path, path performance indicators, etc. The user can intuitively view the path planning results through the interface for further analysis and optimization.

[0120] In the embodiment of the present invention, by extracting network topology structure data, node attribute data, and link attribute data from the substation communication network, using the shortest path algorithm to calculate the shortest path between the source node and the target node, and saving the results in the distance table, the initial path is calculated by parallel shortest path calculation, the large model is trained using the labeled dataset, the preprocessed data is input into the trained large model to obtain the perception results of the network status, and combining the shortest path calculation results and the network status perception results, the intelligence of the substation communication network is improved, communication congestion is avoided, and network stability is ensured.

[0121] As mentioned above, it is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed in this application can easily think of changes or substitutions, which should all be covered within the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

[0122] Finally: The above is only the preferred solution of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An automatic planning system for communication routing paths in a substation, characterized in that, It includes a data preprocessing module, a shortest path calculation module, a large model training module, a network status perception module, a routing planning module, and a result output module; the data preprocessing module is used to extract network topology structure data, node attribute data, and link attribute data from the substation communication network, and perform data cleaning, transformation, and annotation; the shortest path calculation module is used to calculate the shortest path between the source node and the target node using the shortest path algorithm, and save the result to the distance table, and calculate the preliminary path through parallel shortest path calculation; the shortest path calculation module includes a shortest path update unit; the shortest path update unit is used to select the node u with the smallest current path distance using the priority queue, traverse all neighbor nodes v of the node u, and update to find a shorter path through the node u. If the path length from the current node v to the node u is less than the known path length d[v] from the origin to the node v, at this time, update d[v] to the path length from the node u to the node v, and the formula for updating the shorter path found through the node u is: if: d[u] + w(u, v) < d[v]; then: d[v] = d[u] + w(u, v); In the formula: d[u] is the current known shortest path length from the source point to the node u, w(u, v) is the weight of the edge from the node u to the node v, and d[v] is the current known shortest path length from the source point to the node v; Repeat the above steps until all nodes are visited.

2. The automatic planning system for substation communication routing paths according to claim 1, wherein The data preprocessing module is used to extract network topology structure data, node attribute data, and link attribute data from the substation communication network, and perform data cleaning, transformation, and annotation; The large model training module is used to train the large model using the labeled dataset; The network status perception module is used to input the preprocessed data into the trained large model to obtain the perception result of the network status; The routing planning module is used to combine the shortest path calculation result and the network status perception result, and perform routing planning according to business rules and constraints; The result output module is used to output the planning result in a visual or text form.

3. The automatic planning system for substation communication routing paths according to claim 1, wherein, The data preprocessing module includes a data collection unit, a data cleaning unit, and a data conversion unit; The data collection unit is used to collect the topology data of the substation communication network, node attribute data (including substation nodes, control center nodes, communication nodes), and link (communication link) attribute data (including link delay, link stability, and link load); The data cleaning unit is used to filter out abnormal and invalid data to ensure the accuracy and consistency of the data; The data conversion unit is used to uniformly convert the data into a graph data structure, store it using the graph data structure, and give the adjacency matrix A and the weight matrix W; among them, the adjacency matrix A is expressed as: Where: A ij is the adjacency matrix of the edge from node i to node j; The weight matrix W is expressed as: Where: W ij is the weight of the edge from node i to node j.

4. The automatic planning system for substation communication routing paths according to claim 3, characterized in that, The shortest path calculation module also includes a path initialization unit, a parallel shortest path calculation unit, and a preliminary path result output unit; the path initialization unit is used to set the initial distance from the source node to all nodes to infinity, and the distance from the source node to itself to 0; The parallel shortest path calculation unit is used to perform hierarchical processing by dividing the path distance set and update the paths in the distance set through parallel computing; the preliminary path result output unit is used to store the shortest path length from the source node to the target node and record the shortest path length.

5. The automatic planning system for substation communication routing paths according to claim 4, wherein The parallel shortest path calculation unit performs hierarchical processing by dividing the path distance set and updates the paths in the distance set through parallel computing. The specific steps are as follows: Dividing the path distance set: Set the step size parameter Δ, and divide all nodes into the path distance set according to the current shortest path distance through the path division model. The formula of the path division model is: B[p] = {v ∈ Q | pΔ ≤ d[v] < (p + 1)Δ}; In the formula: B[p] is the p-th path distance set, pΔ is the minimum value in the p-th path distance set, (p + 1)Δ is the maximum value in the p-th path distance set, p is the index in the path distance set, which is determined by the shortest node path; Set initialization: Set the initial node s. At this time, the minimum node distance d[s]=0, the minimum distance d[c] between the initial node and the remaining nodes c = ∞, and At this time, the initial node is divided into the first path distance set B[0]; Processing the set nodes: Select the path distance set B[p] with the current minimum index, and perform the following operations on all nodes in the path distance set B[p] in parallel: d[c] = min(d[c], d[a] + w[c,a]); In the formula: d[c] is the minimum distance between the initial node and the remaining node c, d[a] is the minimum distance between the initial node and node c, w[c,a] is the weight of the edge from node c to node a, and min is to take the smaller of the two values; If d[c] is updated, reassign the node to a new path distance set, remove the processed path distance set. If all current path distance sets are non-empty, continue to select the path distance set with the minimum path distance until the shortest path distances of all nodes have been updated.

6. The automatic planning system for substation communication routing paths according to claim 1, wherein The network state perception module is used to input the preprocessed data into the trained large model to obtain the perception result of the network state. The specific steps include: Data preparation: Construct a training data set, including the input network topology structure, node and link attributes, and corresponding label data (including the target path, link state); Model Design: Learn the state embedding vector h for each node through a graph neural network model v ∈R s , where R s is an s-dimensional real vector space, and h v is the state vector of the node and can generate the output vector o v ; h v = f(x v , x co[v] , h ne[v] , x ne[v] ); o v = g(h v , x v ); where: x v is the feature vector of node v, x co[v] is the feature vector of the edge associated with node v, h ne[v] is the state vector of the neighbor nodes of node v, x ne[v] is the feature vector of the neighbor nodes of node v, f() is the local transition function, and g() is the local output function; Calculate the state parameters based on the feature vector, state vector, output vector of the node and the feature vector of the neighbor nodes.

Images