Deterministic network traffic management methods, systems, devices, and media

Abstract

The application relates to the technical field of network management, and discloses a deterministic network traffic management method, system, device and medium, fuzzy processing is performed on acquired multi-element network data to obtain a traffic health state fuzzy set; the traffic health state fuzzy set is matched according to a fuzzy membership degree rule base to obtain an abnormal traffic probability value; a second graph neural network structure constructed based on a graph attention and a gated recurrent unit is used to perform spatiotemporal feature extraction on a first graph neural network structure obtained according to feature representation of the traffic health state fuzzy set; after linear transformation is performed on the obtained spatiotemporal state vector, a health state result is obtained according to the abnormal traffic probability value and a dynamic probability threshold value obtained; and according to the dynamic probability threshold value, the health state result and a deterministic network priority mapping rule, traffic scheduling is performed on each multi-element network data. The method improves the dynamic adaptability, accuracy and time delay controllability of power communication network traffic management.

Description

Technical Field

[0001] This invention relates to the field of network management technology, and in particular to a method, system, device, and medium for traffic management in deterministic networks. Background Technology

[0002] With the rapid development of smart grid technology, smart substations, as core nodes of the power system, are continuously improving their digitalization and informatization levels, propelling power communication networks into a new stage of high business integration. The types of services operating in smart substations are becoming increasingly diverse, covering real-time control services such as telemetry, remote control, and remote signaling, as well as non-real-time services such as event logging, status monitoring, and video. These various services are transmitted in the network using multiple protocols, including IEC 61850, MMS (Manufacturing Message Specification), and TCP / IP, forming a dynamic hybrid communication scenario with multiple protocols coexisting and multiple services intertwined. Against this backdrop, the service traffic of power communication networks is experiencing explosive growth. The large-scale access of smart devices and high-frequency data acquisition are leading to a sharp increase in network bandwidth demand. On the other hand, critical services are placing higher demands on real-time performance and reliability.

[0003] However, existing power communication network traffic management methods have the following significant drawbacks: insufficient dynamic adaptability, as existing methods mostly rely on fixed window flow limiting algorithms or static rule matching, which are difficult to adapt to the dynamic mixed communication scenarios of multiple protocols and services in substations, resulting in a high misjudgment rate for normal services and frequent failure of static rules; lack of latency controllability, as network devices only rely on simple priority queues or bandwidth allocation strategies for traffic scheduling, which cannot guarantee the deterministic latency of critical service flows.

[0004] Therefore, improving the dynamic adaptability, accuracy, and latency controllability of power communication network traffic management has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] This invention provides a method, system, device, and medium for traffic management in deterministic networks, aiming to solve the technical problem of how to improve the dynamic adaptability, accuracy, and latency controllability of traffic management in power communication networks, thereby achieving the effect of improving the dynamic adaptability, accuracy, and latency controllability of traffic management in power communication networks.

[0006] In a first aspect, the present invention provides a traffic management method for a deterministic network, the method comprising:

[0007] The multivariate network dataset of each network element node in the acquired deterministic network to be processed is fuzzified to obtain a fuzzy set of traffic health status.

[0008] The fuzzy set of traffic health status of each network element node is matched according to the preset fuzzy membership rule base to obtain the abnormal traffic probability value of the multivariate network dataset corresponding to each network element node.

[0009] Based on the feature representation of each of the multivariate network datasets, a first graph neural network structure connecting each of the network element nodes is obtained. A second graph neural network structure based on graph attention and gated recurrent units is used to extract spatiotemporal features from the first graph neural network structure to obtain the spatiotemporal state vector of each of the network element nodes at the next moment. The spatiotemporal state vector is then subjected to linear transformation, error calculation and optimization adjustment in sequence to obtain the dynamic probability threshold of each of the network element nodes.

[0010] The abnormal traffic probability value corresponding to each network element node is compared and analyzed with the dynamic probability threshold to obtain the health status result of each network element node;

[0011] Based on the dynamic probability threshold, the health status result, and the pre-set deterministic network priority mapping rules, traffic scheduling is performed on the multi-dimensional network data corresponding to each network element node.

[0012] Preferably, the step of fuzzifying the multivariate network dataset of each network element node in the acquired deterministic network to obtain a fuzzy set of traffic health status includes:

[0013] Based on the preset data acquisition time interval and preset acquisition time window length, acquire the measurement time series data, traffic index data and event data of each network element node in the deterministic network to be processed;

[0014] According to the pre-set fuzzification rules, the traffic indicator data is mapped to the pre-set fuzzification rules to obtain the first health status warning value set;

[0015] The measurement time series data is fuzzified using the deviation standardization method to obtain the second health status warning value set;

[0016] The event-type data is numerically quantified to obtain event-quantified data. The event-quantified data is then fuzzified using the deviation standardization method to obtain a third health status warning value set.

[0017] A fuzzy set of traffic health status is constructed based on the first health status warning value set, the second health status warning value set, and the third health status warning value set.

[0018] Preferably, obtaining the first graph neural network structure connecting each of the network element nodes based on the feature representation of each of the multivariate network datasets includes:

[0019] A spatiotemporal state vector preprocessing method based on a sliding window is used to perform feature representation on the multivariate network dataset of each network element node, thereby obtaining the feature representation vector of each multivariate network dataset.

[0020] A graph embedding method is used to calculate the similarity value between the feature representation vector of each network element node and the feature representation vectors of the other network element nodes;

[0021] Each similarity value is binarized to obtain the binarized connection relationship between each network element node and the other network element nodes;

[0022] Based on the feature representation vector and the binary connection relationship, a first graph neural network structure is constructed.

[0023] Preferably, the step of using a second graph neural network structure based on graph attention and gated recurrent units to extract spatiotemporal features from the first graph neural network structure to obtain the spatiotemporal state vector of each network element node at the next moment includes:

[0024] The reset gate of the gated loop unit is embedded with a reset gate graph attention mechanism, the update gate of the gated loop unit is embedded with an update gate graph attention mechanism, and the hidden state graph attention mechanism is embedded in the hidden state calculation unit of the gated loop unit to obtain the second graph neural network structure.

[0025] Using the second graph neural network structure, graph topology learning is performed on the first graph neural network structure to obtain the spatiotemporal state vector of each network element node at the next moment.

[0026] Preferably, the step of using the second graph neural network structure to perform graph topology learning on the first graph neural network structure to obtain the spatiotemporal state vector of each network element node at the next moment includes:

[0027] The feature representation vector of the first graph neural network structure is operated on using the updated gate graph attention mechanism to obtain the output data of the first graph attention mechanism. The feature representation vector is operated on using the reset gate graph attention mechanism to obtain the output data of the second graph attention mechanism. The feature representation vector is operated on using the hidden state graph attention mechanism to obtain the output data of the third graph attention mechanism.

[0028] The update gate is used to operate on the output data of the first graph attention mechanism and the feature representation vector to obtain the update gate output data. The reset gate is used to operate on the output data of the second graph attention mechanism and the feature representation vector to obtain the reset gate output data. The hidden state calculation unit is used to operate on the output data of the third graph attention mechanism, the feature representation vector and the reset gate output data to obtain the hidden state data.

[0029] Based on the update gate output data and the hidden state data, the feature representation vector is iteratively calculated to obtain the spatiotemporal state vector of each network element node at the next moment.

[0030] Preferably, the step of sequentially performing linear transformation, error calculation, and optimization adjustment on the spatiotemporal state vector to obtain the dynamic probability threshold for each network element node includes:

[0031] A linear transformation is performed on the spatiotemporal state vector of each network element node to obtain the predicted value of each network element node.

[0032] The abnormal error of each network element node is obtained by the absolute value of the difference between the predicted value and the actual value corresponding to each network element node.

[0033] The abnormal errors are standardized to obtain the health score of each network element node;

[0034] The health score is adjusted based on the service type and the sensitivity of the service type to obtain the dynamic probability threshold for each network element node.

[0035] Preferably, the step of comparing and analyzing the abnormal traffic probability value corresponding to each network element node with the dynamic probability threshold to obtain the health status result of each network element node includes:

[0036] The abnormal traffic probability value corresponding to each network element node is compared with the dynamic probability threshold. If the abnormal traffic probability value is greater than the dynamic probability threshold, the multi-dimensional network data corresponding to the network element node is determined to be in an abnormal health state.

[0037] If the abnormal traffic probability value is not greater than the dynamic probability threshold, then the multi-dimensional network data corresponding to the network element node is determined to be in a healthy and normal state.

[0038] Secondly, the present invention also provides a traffic management system for deterministic networks, which implements the traffic management method for deterministic networks described above. The system includes: a data processing module, a data fuzzy matching module, a data graph structure processing module, a comparison analysis module, and a traffic scheduling module.

[0039] The data processing module is used to perform fuzzification processing on the multivariate network dataset of each network element node of the acquired deterministic network to be processed, so as to obtain a fuzzy set of traffic health status.

[0040] The data fuzzy matching module is used to match the fuzzy set of traffic health status of each network element node according to the preset fuzzy membership rule library, so as to obtain the abnormal traffic probability value of the multivariate network dataset corresponding to each network element node.

[0041] The data graph structure processing module is used to obtain a first graph neural network structure connecting each of the network element nodes based on the feature representation of each of the multivariate network datasets, and to extract spatiotemporal features of the first graph neural network structure using a second graph neural network structure constructed based on graph attention and gated recurrent units, so as to obtain the spatiotemporal state vector of each of the network element nodes at the next moment, and to perform linear transformation, error calculation and optimization adjustment on the spatiotemporal state vector in sequence to obtain the dynamic probability threshold of each of the network element nodes;

[0042] The comparison and analysis module is used to compare and analyze the abnormal traffic probability value and the dynamic probability threshold corresponding to each network element node to obtain the health status result of each network element node.

[0043] The traffic scheduling module is used to perform traffic scheduling on the multi-dimensional network data corresponding to each network element node according to the dynamic probability threshold, the health status result and the pre-set deterministic network priority mapping rules.

[0044] Thirdly, the present invention also provides a computer device, the computer device including a memory, a processor and a transceiver, which are connected to each other via a bus; the memory is used to store a set of computer program instructions and data, and to transmit the stored data to the processor, the processor executing the computer program instructions stored in the memory to perform the aforementioned deterministic network traffic management method.

[0045] Fourthly, the present invention also provides a computer-readable storage medium storing a computer program that, when executed, implements the traffic management method for deterministic networks described above.

[0046] This application provides a method, system, device, and medium for traffic management in deterministic networks. Compared with the prior art, the beneficial effects of the embodiments of this application are as follows:

[0047] To address the multi-source heterogeneous traffic in substations, a distributed fuzzy inference approach is designed to independently process single-dimensional features, avoid rule conflicts, and dynamically adapt to the dynamic mixed scenarios of multiple protocols and services in substations, reducing the misjudgment rate of normal services. Using the network element nodes where monitoring equipment is located as graph nodes and the service flow interaction relationships as edges, an asymmetric dependent first graph neural network structure is constructed. Graph attention and gated recurrent units capture the spatiotemporal coupling characteristics of multi-source data, forming a directed graph topology network, resulting in a second graph neural network structure. The current dynamic probability threshold is obtained based on the second graph neural network structure. Both the first and second graph neural network structures can follow the temporal drift of multi-source network data, allowing the dynamic probability threshold to dynamically adjust to adapt to long-term, slow network changes. The higher the bandwidth utilization, the lower the dynamic probability threshold, achieving adaptive sensitivity adjustment, ensuring deterministic latency of critical service flows, improving the accuracy of network data health status judgment, enhancing the perception of network topology changes, and realizing dynamic matching between traffic management strategies and network status. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the steps of a traffic management method for a deterministic network provided in a preferred embodiment of the present invention;

[0049] Figure 2 This is a schematic diagram of the structure of a traffic management system for a deterministic network provided in a preferred embodiment of the present invention;

[0050] Figure 3 This is an internal structural diagram of the computer device in an embodiment of the present invention;

[0051] Figure label:

[0052] 1-Data processing module, 2-Data fuzzy matching module, 3-Data graph structure processing module, 4-Comparison and analysis module, 5-Traffic scheduling module. Detailed Implementation

[0053] The embodiments of the present invention are described in detail below with reference to the accompanying drawings. The embodiments are provided for illustrative purposes only and should not be construed as limiting the invention. The accompanying drawings are for reference and illustration only and do not constitute a limitation on the scope of protection of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of this invention. In the description of this invention, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first," "second," "third," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0054] In the description of this invention, it should be noted that the term "and / or" as used herein includes any and all combinations of one or more of the associated listed items. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0055] In the description of this invention, it should be noted that, unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this specification is for the purpose of describing specific embodiments only and is not intended to limit the invention. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0056] Please see Figure 1 The diagram illustrates the steps of a traffic management method for a deterministic network. In an embodiment of the present invention, a traffic management method for a deterministic network is provided, the method comprising:

[0057] S1. The multivariate network dataset of each network element node in the acquired deterministic network to be processed is fuzzified to obtain a fuzzy set of traffic health status. In the preferred embodiment of this application, a deterministic network refers to a network communication environment that can provide predictable and guaranteed transmission performance (such as latency, jitter, and packet loss rate) for specific services. Its core feature is to eliminate or control the uncertainty in network transmission through technical means to ensure that key services can reliably complete data interaction within a specified time. Deterministic networks use time-sensitive scheduling (such as the time triggering mechanism of TSN time-sensitive networks), resource reservation (such as bandwidth and cache pre-allocation), priority mapping, and other technologies to pre-plan transmission paths and time windows for key services, ensuring controllable latency, minimal jitter, and guaranteed reliability. In substation networks, a deterministic network specifically refers to a communication environment that carries high-priority real-time services. These services are directly related to the safe and stable operation of the power system and must rely on deterministic network transmission guarantees, such as SV streams (sampled value streams), GOOSE streams (general object-oriented substation event streams), and remote control / remote adjustment commands. In the target substation network, network traffic monitoring tools are used to acquire time-series traffic data, traffic indicator data, and event type data of the deterministic network to be processed within the current acquisition period at each network element node, based on preset data acquisition intervals and preset acquisition time window lengths. The measurement time-series data includes time-series data representing traffic type characteristics at each network element node, such as the measured and reported values ​​corresponding to voltage and current. Traffic indicator data includes indicators such as network protocol type, frame length, arrival time, IP address, port number, and bandwidth occupancy. Event type data includes power system event data used for exchange or reporting during communication, such as equipment switch status values, alarm events, equipment status change events, and fault events. In practical applications, the preset data acquisition interval is generally the shortest reporting cycle of the communication network message, but can also be set in milliseconds, seconds, or minutes according to application requirements. The preset acquisition time window length is set according to actual needs, generally tens to hundreds of times the acquisition interval.

[0058] Furthermore, the measurement time-series data, traffic indicator data, and event data of each network element node in the deterministic network are preprocessed. The core step of the preprocessing is to fuzzify the collected measurement time-series data, traffic indicator data, and event data. For traffic indicator data, the fuzzification method is based on pre-defined fuzzification rules, mapping the precise values ​​or types of each traffic indicator data to predefined fuzzification rules to obtain a first set of health status warning values. The pre-defined fuzzification rules include at least the following: for frame length, if it exceeds twice the historical average of the frame length over a preset time, it is mapped to a "high" health status warning value; if it exceeds 1.5 times but does not exceed twice the historical average of the frame length over a preset time, it is mapped to a "medium" health status warning value; if it does not exceed 1.5 times the historical average of the frame length over a preset time, it is mapped to a "low" health status warning value; for arrival time, if it exceeds twice the historical average of the arrival time over a preset time, it is mapped to... The "High" health status warning value is mapped to a "Medium" health status warning value if it exceeds 1.5 times but does not exceed 2 times the historical average arrival time of the preset time. If it does not exceed 1.5 times the historical average arrival time of the preset time, it is mapped to a "Low" health status warning value. For protocol types, HTTP is mapped to a "Low" health status warning value, FTP and SMTP are mapped to a "Medium" health status warning value, and SSH and Telnet are mapped to a "High" health status warning value. For port numbers, port 80 is mapped to a "Low" health status warning value, port 21 or 25 is mapped to a "Medium" health status warning value, and port 22 or 23 is mapped to a "High" health status warning value. For IP addresses, IP addresses in the preset blacklist are mapped to a "High" health status warning value, IP addresses in the preset whitelist are mapped to a "Low" health status warning value, and IP addresses that are neither in the blacklist nor the whitelist are mapped to a "Medium" health status warning value. The fuzzy rules for protocol type and port number are mainly based on the application scenario of power deterministic network. They are set according to the priority level and business risk level of protocol type and port number. For example, if there is a significant risk of SSH protocol execution lag, its health status warning value is set to a higher level. If the Telnet protocol has a high priority level, its health status warning value is set to a higher level.

[0059] The above only shows the fuzzification rules for some frequently used traffic indicator data. For traffic indicator data not covered by the above fuzzification rules, fuzzification rules can be set according to the extent to which they deviate from the normal value. The specific setting method will not be elaborated here.

[0060] For measurement time series data, the values ​​usually do not follow a Gaussian distribution. In view of this, the deviation standardization method is used to standardize the measurement time series data to achieve fuzzification, mapping all measurement time series data to the range of [0,1] and converting them to the same dimension to eliminate the influence between different dimensions.

[0061] For event-type data, which is usually non-numerical or discrete, non-numerical data includes event type and status description, while discrete data includes event level and occurrence frequency. Therefore, event-type data needs to be preprocessed to be converted into quantifiable values ​​to obtain event quantification data. Then, the event quantification data is mapped to the range [0,1] using deviation standardization. For example, for event types, they are converted into values ​​through label encoding or one-hot encoding, and "remote signaling change", "protection action", and "communication interruption" are mapped to integers such as 0, 1, and 2, respectively. Finally, deviation standardization is used to map the data and eliminate dimensional differences.

[0062] In a preferred embodiment of this application, the multi-dimensional spatiotemporal state data of each network node are fuzzed to eliminate the influence between dimensions and achieve the fusion of multi-dimensional spatiotemporal state vectors.

[0063] S2. Match the fuzzy set of traffic health status for each network element node according to the preset fuzzy membership rule base to obtain the abnormal traffic probability value of the multivariate network dataset corresponding to each network element node; In the preferred embodiment of this application, for the first health status warning value set, a fuzzy membership rule base is constructed based on expert experience. For example, the membership degree corresponding to the "high" health status warning value is 0.7, the membership degree corresponding to the "medium" health status warning value is 0.2, and the membership degree corresponding to the "low" health status warning value is 0.1. For the second and third health status warning value sets, they are themselves mapping values ​​in the range of [0,1], which can be directly used. For each network element node, the fuzzy set of traffic health status is matched according to the above fuzzy membership rule base to obtain the abnormal traffic probability value of each network element node. For example, if the frame length of a voltage measurement report is 4 times the average, the arrival time is 2.5 times the average, the protocol type is SSH, the port number is 22, and the IP address is an address in the blacklist, the resulting first health status warning value set is [high, high, high, high, high]. Further, based on the fuzzy membership rule base, the above first health status warning value set is subjected to distributed inference matching, and then a weighted calculation of the abnormal traffic probability value is obtained. The weight values ​​are determined according to the importance of each traffic indicator data, and the sum of all weight values ​​is 1. This abnormal traffic probability value is compared with a dynamic probability threshold to determine the health status of the multi-dimensional network data corresponding to each network node, obtaining the health status result of the multi-dimensional network data corresponding to each network element node, and traffic scheduling is performed on the multi-dimensional network data based on the health status result. For details on determining the dynamic probability threshold, see step S3.

[0064] S3. Based on the feature representation of each of the multivariate network datasets, a first graph neural network structure connecting each of the network element nodes is obtained. A second graph neural network structure based on graph attention and gated recurrent units is used to extract spatiotemporal features from the first graph neural network structure, obtaining the spatiotemporal state vector of each network element node at the next time step. The spatiotemporal state vector is then subjected to linear transformation, error calculation, and optimization adjustment in sequence to obtain the dynamic probability threshold of each network element node. In a preferred embodiment of this application, a sliding window-based spatiotemporal state vector preprocessing method is used to represent the features of the multivariate network dataset for each network element node. A fixed-length "window" is extracted from the continuous time series data, transforming the time series data into feature segments that can be used as model input. Specifically, features of a specific timestamp are first extracted from the multivariate network dataset. The multivariate network dataset is represented as follows:

[0065]

[0066] in, Represents a multivariate network dataset. This represents the number of timestamps in a multivariate network dataset. Indicates the number of network element nodes. It represents the set of real numbers.

[0067] To better represent the state of the power grid system at each timestamp, at each timestamp Above, using each timestamp The size of the top is A sliding window is used, and the time-series data within the sliding window is selected as the feature representation of each network element node at that timestamp. Specifically, it is represented as follows:

[0068]

[0069] in, Indicates the timestamp of a multivariate network dataset. Time-related characteristics Indicates size is The first feature within the sliding window, Indicates size is The second feature within the sliding window, Indicates size is The third feature within the sliding window, Indicates size is The first sliding window One characteristic.

[0070] The concatenation of all timestamps forms the feature representation of the multivariate network dataset. The feature representation vector of the multivariate network dataset is:

[0071]

[0072] in, This indicates that the multivariate network dataset is represented by the number of timestamps. The number of network element nodes is The sliding window size is The feature representation vector, Indicates the timestamp of the first network element node. Time-related characteristics This indicates the second network element node at the timestamp. Time-related characteristics Indicates the first Each network element node at the timestamp The characteristics of time.

[0073] In a preferred embodiment of this application, the network element nodes for traffic monitoring are used as graph nodes in a graph neural network, and the relationships between these nodes are represented by edges. Therefore, the influence of one network element node on another can be represented by the edges of the graph neural network. Considering that the dependencies between network element nodes are not necessarily symmetrical, a directed graph is used to represent the inherent structure of the data. Accurate graph structure information not only greatly improves the accuracy of the graph neural network but also helps to understand the internal connections of the power network system and detect anomalies.

[0074] In a preferred embodiment of this application, two graph structure learning methods are employed. For the first method, a graph embedding method is used to perform data similarity analysis on the feature representation vector of each network element node in the case of unknown topology, calculating the edge weight between any two network element nodes. Specifically, the similarity between the feature representation vector of the network element node itself and the feature representation vectors of all other network element nodes is calculated to obtain a neighborhood similarity matrix. The similarity between the network element node itself and each of the other network element nodes is used as the edge weight between corresponding two network element nodes. The edge weights are further binarized to obtain the binarized connection relationship. The processing procedure is as follows:

[0075]

[0076] in, Represents network element nodes and network element nodes Similarity between them Represents network element nodes Feature representation vector within the timestamp window Represents network element nodes Feature representation vector within the timestamp window Represents network element nodes and network element nodes pass The weight selection yields the binarized connection relationships between nodes. This indicates the highest similarity value. Each network element node.

[0077] Based on the binarized connection relationships, the various network element nodes are connected to obtain the first graph neural network structure. However, this first graph neural network structure is too simple and may lose important intrinsic spatiotemporal correlation features. Therefore, this application introduces a graph attention mechanism and a gated recurrent unit (ROU). The gated RNU is a variant of recurrent neural networks commonly used for processing time-series data. It dynamically adjusts the transmission and forgetting of information through two "gates," avoiding information loss caused by long sequences in traditional RNUs. The two "gates" are the reset gate and the update gate. The reset gate determines whether to forget historical information, controlling the impact of past hidden states on the current computation. The update gate determines whether to update the current hidden state, balancing the weights of historical and current information. The gated RNU has significant advantages in time series learning tasks, capturing long-term dependencies in time series and better representing the temporal characteristics of the sequence data. The graph attention mechanism focuses more on the spatial dependencies between different time series data, assigning different attention coefficients to different neighboring nodes to capture more detailed spatial characteristics, and exhibiting good performance in modeling the spatial dependencies of multivariate time series. The multi-source spatiotemporal state data of the multivariate network dataset in this application also have strong intrinsic connections in space, that is, they have spatiotemporal features. Therefore, based on the gated recurrent unit, the graph attention mechanism is embedded into the gated recurrent unit, so that the gated recurrent unit can update the weight value and reset the hidden state based on the graph attention mechanism, which can simultaneously capture the underlying spatial and temporal dependency characteristics.

[0078] Specifically, for any feature representation vector in a multivariate network dataset, the graph attention mechanism operation is represented as:

[0079]

[0080] in, Representation of graph attention mechanism in timestamps Output at time This represents a graph attention network.

[0081] For each node, the graph attention mechanism operation is represented as:

[0082]

[0083] in, Representation of graph attention mechanism in timestamps At the time of the first The output of each network element node Indicates the timestamp Time The feature representation vector of each network element node, where ReLU represents the rectified linear unit activation function. Represents network element nodes The self-attention coefficient, Represents network element nodes and network element nodes Attention coefficient between them This represents a trainable weight matrix used for linear transformations. , Represents a node The set of neighboring nodes.

[0084] For nodes and nodes The formula for calculating the attention coefficient between them is:

[0085]

[0086] in, Indicates series connection. Let represent the learning vector for the attention mechanism, and LeakyReLU represent the rectified linear unit activation function with leakage. Indicates the first Each network element node at the timestamp Time-related characteristics Indicates the first Each network element node at the timestamp Time-related characteristics Indicates the first The neighboring nodes at the timestamp The characteristics of time.

[0087] A graph attention mechanism is used to replace the dot product operation of the gated recurrent unit. The graph attention mechanism is embedded in the gated recurrent unit to obtain the second graph neural network structure. Specifically, a reset gate graph attention mechanism is embedded in the reset gate of the gated recurrent unit, an update gate graph attention mechanism is embedded in the update gate of the gated recurrent unit, and a hidden state graph attention mechanism is embedded in the hidden state calculation unit of the gated recurrent unit. The second graph neural network structure is represented as follows:

[0088]

[0089] in, Represents timestamp The reset gate outputs data in the second neural network structure. Represents timestamp The update gate outputs data in the second neural network structure. Represents timestamp The hidden state data of the neural network structure in the second graph. This represents the reset gate graph attention mechanism for reset gate embedding. This represents the update gate graph attention mechanism for updating gate embeddings. The hidden state graph attention mechanism represents the embedding of intermediate states. Let denot be any activation function, and let tanh denote the hyperbolic tangent function. , , , , and All of these represent network weights, which are learnable network parameters. Represents the Hadamard product. This represents the graph embedding vector of the first graph neural network. Both represent network bias. This indicates the neural network structure in the second graph at the timestamp. Output at time This indicates the neural network structure in the second graph at the timestamp. The output at the time, that is, at the timestamp The spatiotemporal state vector at time.

[0090] The node feature representations of the first neural network structure are input into the second neural network structure to obtain the spatiotemporal state vector of each network element node at the next time step. Furthermore, a fully connected layer is used to perform a linear transformation on the spatiotemporal state vector of each network element node to obtain the predicted value of each network element node at the next time step. The formula for calculating the predicted value is as follows:

[0091]

[0092] in, Indicates the timestamp Time The predicted value of each network element node. This represents a linear transformation of a fully connected layer. This indicates the neural network structure in the second graph at the timestamp. At the time of the first The output of each network element node.

[0093] In a preferred embodiment of this application, after obtaining the predicted value of each network element node, the difference between the predicted value and the actual value corresponding to each network element node is calculated, and this difference is taken as the abnormal error of that network element node. The formula for calculating the abnormal error is as follows:

[0094]

[0095] in, Indicates the first Each network element node at the timestamp Abnormal errors at that time Indicates the first Each network element node at the timestamp The true value of a moment, when time arrives The actual measurement was obtained at that time.

[0096] Since the abnormal errors of network element nodes may vary significantly in numerical value, it is necessary to standardize the abnormal errors of each network element node for easier observation. In the preferred embodiment of this application, the median and interquartile range are used for standardization to obtain a health score. The health score calculation formula is as follows:

[0097]

[0098] in, Indicates the first Each network element node at the timestamp Health score at that time express The median over time, express Interquartile spacing in the time dimension.

[0099] Furthermore, the health score is optimized to obtain a dynamic probability threshold. The formula for calculating the dynamic probability threshold is:

[0100]

[0101] in, Indicates the first Each network element node at the timestamp The dynamic probability threshold at time, and the initial value of the dynamic probability threshold. To determine the health status warning value by matching the health status data according to the fuzzy membership rule base during the aforementioned fuzzy inference process, based on the initial probability threshold set according to the application scenario; This represents the weighting coefficient. The value is set according to the business type and its sensitivity in the actual application environment, and can generally be set to 3-5.

[0102] From the above formula for calculating the dynamic probability threshold, it can be seen that... This reflects the change in the current state of the deterministic network system relative to its inertial state. The dynamic probability threshold, achieved through the first and second neural network structures, dynamically adapts to the fluctuations within the normal range of the deterministic network system. When the fluctuations are strong, the dynamic probability threshold is appropriately increased; when the fluctuations are weak, it is appropriately decreased. In a stable state, the threshold returns to near the initial probability threshold. Therefore, the dynamic probability threshold designed in this application increases the stability and accuracy of traffic health status inference for deterministic networks compared to a fixed probability threshold. Furthermore, in practical applications, if the dynamic probability threshold deviates significantly from the initial threshold or exceeds the normal range, a health status alarm will be directly triggered. Specifically, a deviation threshold is set, and the deviation between the dynamic probability threshold and the initial probability threshold is calculated. If the deviation exceeds the deviation threshold, a health status warning is triggered. The deviation threshold is selected based on the specific application scenario requirements, such as 0.2.

[0103] S4. Compare and analyze the abnormal traffic probability value corresponding to each network element node with the dynamic probability threshold to obtain the health status result of each network element node; in the preferred embodiment of this application, the abnormal traffic probability value corresponding to each network element node and the dynamic probability threshold are compared. If the abnormal traffic probability value is greater than the dynamic probability threshold, the multi-dimensional network data of the network element node is determined to be in an abnormal health status; if the abnormal traffic probability value is not greater than the dynamic probability threshold, the multi-dimensional network data of the network element node is determined to be in a normal health status.

[0104] In this application, both the first and second graph neural network structures can follow the time drift of multivariate network data, thereby enabling the dynamic probability threshold to be dynamically adjusted to adapt to the long-term slow changes in the network, thus improving the accuracy of network data health status determination.

[0105] S5. Based on the dynamic probability threshold, the health status result, and the pre-set deterministic network priority mapping rule, traffic scheduling is performed on the multi-dimensional network data corresponding to each network element node. In a preferred embodiment of this application, based on the health status result, early warning and processing are performed on multi-dimensional network data with abnormal health status. For multi-dimensional network data with normal health status, a deterministic network priority mapping rule is constructed according to the real-time requirements of deterministic networks, as shown in Table 1 as an example of a deterministic network priority mapping rule.

[0106] Table 1

[0107]

[0108] As shown in Table 1, the priority and latency requirements of healthy multi-dimensional network data can be determined based on the dynamic probability threshold. The deterministic network assigns the priority and latency requirements to the network controller, thereby completing the multi-dimensional network data traffic scheduling process based on priority and latency requirements.

[0109] In a preferred embodiment of the present invention, the multivariate network dataset of each network element node in the acquired deterministic network to be processed is fuzzified to obtain a fuzzy set of traffic health status; the fuzzy set of traffic health status of each network element node is matched according to a preset fuzzy membership rule base to obtain the abnormal traffic probability value of the multivariate network dataset corresponding to each network element node; based on the feature representation of each multivariate network dataset, a first graph neural network structure connecting each network element node is obtained, and a second graph neural network structure based on graph attention and gated recurrent units is used to extract spatiotemporal features from the first graph neural network structure to obtain the spatiotemporal state vector of each network element node at the next moment, and the spatiotemporal state vector is sequentially subjected to linear transformation, error calculation and optimization adjustment to obtain the dynamic probability threshold of each network element node; the abnormal traffic probability value corresponding to each network element node is compared and analyzed with the dynamic probability threshold to obtain the health status result of each network element node; according to the dynamic probability threshold, the health status result and the preset deterministic network priority mapping rule, traffic scheduling is performed on the multivariate network data corresponding to each network element node. The deterministic network traffic management method disclosed in this application, targeting multi-source heterogeneous traffic in substations, designs a distributed fuzzy inference approach to independently process single-dimensional features, avoid rule conflicts, and dynamically adapt to the dynamic mixed scenarios of multiple protocols and services in substations, reducing the misjudgment rate of normal services. Using the network element nodes where monitoring equipment is located as graph nodes and the service flow interaction relationships as edges, an asymmetric dependent first graph neural network structure is constructed. Graph attention and gated recurrent units capture the spatiotemporal coupling characteristics of multi-source data, forming a directed graph topology network, resulting in a second graph neural network structure. The current dynamic probability threshold is obtained based on the second graph neural network structure. Both the first and second graph neural network structures can follow the time drift of multi-source network data, allowing the dynamic probability threshold to dynamically adjust to adapt to long-term, slow network changes. The higher the bandwidth utilization, the lower the dynamic probability threshold, achieving adaptive sensitivity adjustment, ensuring deterministic latency of critical service flows, improving the accuracy of network data health status judgment, enhancing the perception of network topology changes, and realizing dynamic matching between traffic management strategies and network status.

[0110] Accordingly, such as Figure 2The diagram shows the structure of a traffic management system for a deterministic network. Based on a traffic management method for a deterministic network, this embodiment of the invention also provides a traffic management system for a deterministic network that implements the traffic management method for a deterministic network disclosed in this embodiment. The system includes: a data processing module 1, a data fuzzy matching module 2, a data graph structure processing module 3, a comparison and analysis module 4, and a traffic scheduling module 5.

[0111] The data processing module 1 is used to perform fuzzification processing on the multivariate network dataset of each network element node of the acquired deterministic network to be processed, so as to obtain a fuzzy set of traffic health status.

[0112] The data fuzzy matching module 2 is used to match the fuzzy set of traffic health status of each network element node according to the preset fuzzy membership rule library, so as to obtain the abnormal traffic probability value of the multivariate network dataset corresponding to each network element node.

[0113] The data graph structure processing module 3 is used to obtain a first graph neural network structure connecting each of the network element nodes based on the feature representation of each of the multivariate network datasets, and to extract spatiotemporal features of the first graph neural network structure using a second graph neural network structure constructed based on graph attention and gated recurrent units, so as to obtain the spatiotemporal state vector of each of the network element nodes at the next moment, and to perform linear transformation, error calculation and optimization adjustment on the spatiotemporal state vector in sequence to obtain the dynamic probability threshold of each of the network element nodes;

[0114] The comparison and analysis module 4 is used to compare and analyze the abnormal traffic probability value and the dynamic probability threshold corresponding to each network element node to obtain the health status result of each network element node.

[0115] The traffic scheduling module 5 is used to perform traffic scheduling on the multi-dimensional network data corresponding to each network element node according to the dynamic probability threshold, the health status result and the pre-set deterministic network priority mapping rule.

[0116] For specific limitations regarding a traffic management system for a deterministic network, please refer to the above-described limitations regarding a traffic management method for a deterministic network, which will not be repeated here. Those skilled in the art will recognize that the various modules and steps described in conjunction with the embodiments disclosed in this invention can be implemented in hardware, software, or a combination of both. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0117] like Figure 3The diagram shows the internal structure of a computer device. An embodiment of the present invention provides a computer device including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the steps described in the deterministic network traffic management method embodiment above, for example... Figure 1 Steps S1 to S5 as described above.

[0118] Those skilled in the art will understand that the illustrations Figure 3 This is merely an example of a computer device and does not constitute a limitation on the computer device. It may include more or fewer components than shown, or combine certain components, or different components. For example, the computer device may also include input / output devices, network access devices, buses, etc.

[0119] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the computer device, connecting various parts of the computer device via various interfaces and lines.

[0120] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the computer device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0121] If the modules integrated into the computer device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0122] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0123] Accordingly, embodiments of the present invention provide a computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the steps described in the embodiments of the deterministic network traffic management method as described above, for example... Figure 1 Steps S1 to S5 as described above.

[0124] In summary, the present application provides a method, system, device, and medium for traffic management in deterministic networks, addressing the technical problem of improving the dynamic adaptability, accuracy, and latency controllability of traffic management in power communication networks. The method includes: fuzzifying the acquired multivariate network dataset of each network element node in the deterministic network to be processed, obtaining a fuzzy set of traffic health status; matching the fuzzy set of traffic health status of each network element node according to a preset fuzzy membership rule base, obtaining the abnormal traffic probability value of the multivariate network dataset corresponding to each network element node; and obtaining a first graph connecting the various network element nodes based on the feature representation of each multivariate network dataset. A neural network structure is constructed, and a second graph neural network structure based on graph attention and gated recurrent units is used to extract spatiotemporal features from the first graph neural network structure to obtain the spatiotemporal state vector of each network element node at the next time step. The spatiotemporal state vector is then subjected to linear transformation, error calculation, and optimization adjustment to obtain the dynamic probability threshold of each network element node. The abnormal traffic probability value corresponding to each network element node is compared and analyzed with the dynamic probability threshold to obtain the health status result of each network element node. Based on the dynamic probability threshold, health status result, and pre-set deterministic network priority mapping rules, traffic scheduling is performed on the multi-dimensional network data corresponding to each network element node. The deterministic network traffic management method disclosed in this application, targeting multi-source heterogeneous traffic in substations, designs a distributed fuzzy inference approach to independently process single-dimensional features, avoid rule conflicts, and dynamically adapt to the dynamic mixed scenarios of multiple protocols and services in substations, reducing the misjudgment rate of normal services. Using the network element nodes where monitoring equipment is located as graph nodes and the service flow interaction relationships as edges, an asymmetric dependent first graph neural network structure is constructed. Graph attention and gated recurrent units capture the spatiotemporal coupling characteristics of multi-source data, forming a directed graph topology network, resulting in a second graph neural network structure. The current dynamic probability threshold is obtained based on the second graph neural network structure. Both the first and second graph neural network structures can follow the time drift of multi-source network data, allowing the dynamic probability threshold to dynamically adjust to adapt to long-term, slow network changes. The higher the bandwidth utilization, the lower the dynamic probability threshold, achieving adaptive sensitivity adjustment, ensuring deterministic latency of critical service flows, improving the accuracy of network data health status judgment, enhancing the perception of network topology changes, and realizing dynamic matching between traffic management strategies and network status.

[0125] The various embodiments in this specification are described in a progressive manner. For directly identical or similar parts of the embodiments, refer to each other. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. It should be noted that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0126] The embodiments described above are merely preferred embodiments of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various improvements and substitutions without departing from the technical principles of this application, and these improvements and substitutions should also be considered within the scope of protection of this application. Therefore, the scope of protection of this patent application should be determined by the scope of the claims.

Claims

1. A traffic management method for a deterministic network, characterized in that, The method includes: The multivariate network dataset of each network element node in the acquired deterministic network to be processed is fuzzified to obtain a fuzzy set of traffic health status. The fuzzy set of traffic health status of each network element node is matched according to the preset fuzzy membership rule base to obtain the abnormal traffic probability value of the multivariate network dataset corresponding to each network element node. Based on the feature representation of each of the multivariate network datasets, a first graph neural network structure connecting each of the network element nodes is obtained. A second graph neural network structure based on graph attention and gated recurrent units is used to extract spatiotemporal features from the first graph neural network structure to obtain the spatiotemporal state vector of each of the network element nodes at the next moment. The spatiotemporal state vector is then subjected to linear transformation, error calculation and optimization adjustment in sequence to obtain the dynamic probability threshold of each of the network element nodes. The abnormal traffic probability value corresponding to each network element node is compared and analyzed with the dynamic probability threshold to obtain the health status result of each network element node; Based on the dynamic probability threshold, the health status result, and the pre-set deterministic network priority mapping rules, traffic scheduling is performed on the multi-dimensional network data corresponding to each network element node; The step of sequentially performing linear transformation, error calculation, and optimization adjustment on the spatiotemporal state vector to obtain the dynamic probability threshold for each network element node includes: A linear transformation is performed on the spatiotemporal state vector of each network element node to obtain the predicted value of each network element node. The abnormal error of each network element node is obtained by the absolute value of the difference between the predicted value and the actual value corresponding to each network element node. The abnormal errors are standardized to obtain the health score of each network element node; The health score is adjusted based on the service type and the sensitivity of the service type to obtain the dynamic probability threshold for each network element node.

2. The traffic management method for deterministic networks as described in claim 1, characterized in that, The process of fuzzifying the multivariate network dataset of each network element node in the acquired deterministic network to obtain a fuzzy set of traffic health status includes: Based on the preset data acquisition time interval and preset acquisition time window length, acquire the measurement time series data, traffic index data and event data of each network element node in the deterministic network to be processed; According to the pre-set fuzzification rules, the traffic indicator data is mapped to the pre-set fuzzification rules to obtain the first health status warning value set; The measurement time series data is fuzzified using the deviation standardization method to obtain the second health status warning value set; The event data is numerically quantified to obtain event quantified data. The event quantified data is then fuzzified using the deviation standardization method to obtain a third health status warning value set. A fuzzy set of traffic health status is constructed based on the first health status warning value set, the second health status warning value set, and the third health status warning value set.

3. The traffic management method for deterministic networks as described in claim 1, characterized in that, The step of obtaining the first graph neural network structure connecting each of the network element nodes based on the feature representation of each of the multivariate network datasets includes: A spatiotemporal state vector preprocessing method based on a sliding window is used to perform feature representation on the multivariate network dataset of each network element node, thereby obtaining the feature representation vector of each multivariate network dataset. A graph embedding method is used to calculate the similarity value between the feature representation vector of each network element node and the feature representation vectors of the other network element nodes; Each similarity value is binarized to obtain the binarized connection relationship between each network element node and the other network element nodes; Based on the feature representation vector and the binary connection relationship, a first graph neural network structure is constructed.

4. The traffic management method for deterministic networks as described in claim 3, characterized in that, The second graph neural network structure, constructed based on graph attention and gated recurrent units, is used to extract spatiotemporal features from the first graph neural network structure to obtain the spatiotemporal state vector of each network element node at the next time step, including: The reset gate of the gated loop unit is embedded with a reset gate graph attention mechanism, the update gate of the gated loop unit is embedded with an update gate graph attention mechanism, and the hidden state graph attention mechanism is embedded in the hidden state calculation unit of the gated loop unit to obtain the second graph neural network structure. Using the second graph neural network structure, graph topology learning is performed on the first graph neural network structure to obtain the spatiotemporal state vector of each network element node at the next moment.

5. The traffic management method for deterministic networks as described in claim 4, characterized in that, The second graph neural network structure is used to learn the graph topology of the first graph neural network structure to obtain the spatiotemporal state vector of each network element node at the next moment, including: The feature representation vector of the first graph neural network structure is operated on using the updated gate graph attention mechanism to obtain the output data of the first graph attention mechanism. The feature representation vector is operated on using the reset gate graph attention mechanism to obtain the output data of the second graph attention mechanism. The feature representation vector is operated on using the hidden state graph attention mechanism to obtain the output data of the third graph attention mechanism. The update gate is used to operate on the output data of the first graph attention mechanism and the feature representation vector to obtain the update gate output data. The reset gate is used to operate on the output data of the second graph attention mechanism and the feature representation vector to obtain the reset gate output data. The hidden state calculation unit is used to operate on the output data of the third graph attention mechanism, the feature representation vector and the reset gate output data to obtain the hidden state data. Based on the update gate output data and the hidden state data, the feature representation vector is iteratively calculated to obtain the spatiotemporal state vector of each network element node at the next moment.

6. The traffic management method for deterministic networks as described in claim 1, characterized in that, The step of comparing and analyzing the abnormal traffic probability value corresponding to each network element node with the dynamic probability threshold to obtain the health status result of each network element node includes: The abnormal traffic probability value corresponding to each network element node is compared with the dynamic probability threshold. If the abnormal traffic probability value is greater than the dynamic probability threshold, the multi-dimensional network data corresponding to the network element node is determined to be in an abnormal health state. If the abnormal traffic probability value is not greater than the dynamic probability threshold, then the multi-dimensional network data corresponding to the network element node is determined to be in a healthy and normal state.

7. A traffic management system for deterministic networks, used to implement the traffic management method for deterministic networks according to any one of claims 1-6, characterized in that, The system includes: a data processing module, a data fuzzy matching module, a data graph structure processing module, a comparison and analysis module, and a traffic scheduling module; The data processing module is used to perform fuzzification processing on the multivariate network dataset of each network element node of the acquired deterministic network to be processed, so as to obtain a fuzzy set of traffic health status. The data fuzzy matching module is used to match the fuzzy set of traffic health status of each network element node according to the preset fuzzy membership rule library, so as to obtain the abnormal traffic probability value of the multivariate network dataset corresponding to each network element node. The data graph structure processing module is used to obtain a first graph neural network structure connecting each of the network element nodes based on the feature representation of each of the multivariate network datasets, and to extract spatiotemporal features of the first graph neural network structure using a second graph neural network structure constructed based on graph attention and gated recurrent units, so as to obtain the spatiotemporal state vector of each of the network element nodes at the next moment, and to perform linear transformation, error calculation and optimization adjustment on the spatiotemporal state vector in sequence to obtain the dynamic probability threshold of each of the network element nodes; The comparison and analysis module is used to compare and analyze the abnormal traffic probability value and the dynamic probability threshold corresponding to each network element node to obtain the health status result of each network element node. The traffic scheduling module is used to perform traffic scheduling on the multi-dimensional network data corresponding to each network element node according to the dynamic probability threshold, the health status result and the pre-set deterministic network priority mapping rules.

8. A computer device, characterized in that: The computer device includes a memory, a processor, and a transceiver connected to each other via a bus; the memory stores a set of computer program instructions and data, and transmits the stored data to the processor, which executes the computer program instructions stored in the memory to perform a traffic management method for a deterministic network as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program that, when executed, implements the traffic management method for a deterministic network as described in any one of claims 1 to 6.

Images