Power 5G network service quality monitoring method and system, and storage medium

Abstract

The invention relates to the technical field of electric power system communication, in particular to an electric power 5G network service quality monitoring method and system and a storage medium, and the method comprises the steps: responding to a protocol data unit session establishment request signaling, and obtaining signaling plane data and data plane performance data through a signaling plane probe and a data plane probe; associating the signaling plane data with the data plane performance data through the service quality flow identifier and the timestamp to generate a multi-dimensional spatial-temporal characteristic matrix; based on the multi-dimensional spatio-temporal characteristic matrix, a network quality prediction score is obtained through dynamic spatio-temporal causality graph network prediction; and when the network quality prediction score exceeds a preset score threshold, establishing a network resource configuration optimization model to obtain an optimal network resource configuration strategy. According to the method, the signaling plane data and the data plane performance data are subjected to correlation analysis, and the network quality is predicted by using the dynamic space-time causal graph network, so that the network resource configuration is optimized, and the service quality of the power 5G network is ensured.

Description

Technical Field

[0001] The present invention relates to the field of power system communication technology, and in particular to a method, system and storage medium for monitoring the service quality of a power 5G network. Background Art

[0002] With the rapid development of fifth-generation mobile communication technology networks, network scale and business complexity are showing an exponential growth trend. In this context, traditional network monitoring technology faces many severe challenges when dealing with the fifth-generation mobile communication technology network environment, and it is difficult to meet the high requirements of the fifth-generation mobile communication technology network for service quality monitoring.

[0003] Traditional operation, management, and maintenance mechanisms rely primarily on periodic sampling or offline log analysis. This monitoring approach struggles to perceive dynamic network changes in real time. For example, burst traffic or transient link congestion frequently occur, and traditional monitoring technologies struggle to capture these transient changes in a timely manner, making it impossible to take effective measures to address network congestion, which in turn impacts network service quality. Furthermore, traditional methods rely primarily on manual experience for troubleshooting, which is inefficient and difficult to pinpoint the root causes of cross-domain and cross-layer issues. For example, manual troubleshooting of issues with signaling and data plane coordination is difficult to quickly locate, resulting in extended troubleshooting times and reduced network service quality. Furthermore, traditional static resource allocation strategies are unable to adapt to the dynamic service demands of fifth-generation mobile communication technology networks. Services such as virtual reality, augmented reality, and the Industrial Internet of Things (IIoT) have extremely high requirements for low latency. Static resource allocation strategies cannot flexibly adjust to these changing service demands, leading to resource waste or service degradation, and failing to fully leverage the advantages of fifth-generation mobile communication technology networks.

[0004] Among fifth-generation mobile communication network quality monitoring technologies, IPv6 In-situ Flow Information Telemetry (IFIT) directly detects real-time network performance indicators by tagging service flow features, improving network monitoring capabilities to a certain extent. However, the in-depth integration of IFIT technology with 5G signaling monitoring is not yet mature. 5G network quality monitoring primarily relies on signaling monitoring, but signaling data often only reflects the status of the control plane and cannot directly characterize the quality of the service plane. The correlation analysis between 5G signaling plane anomalies and data plane issues lacks a unified framework, making it difficult to comprehensively and accurately assess network service quality.

[0005] In summary, the existing 5G network service quality monitoring technology has many shortcomings and is difficult to meet the service quality assurance needs in highly dynamic business scenarios. Therefore, a more efficient and accurate 5G network service quality monitoring method is urgently needed to improve the service quality and user experience of the 5G network. Summary of the Invention

[0006] To solve the above technical problems, the present invention provides a method, system and storage medium for monitoring the service quality of a 5G power network.

[0007] In a first aspect, the present invention provides a method for monitoring the service quality of a 5G power network, the method comprising the following steps:

[0008] In response to a protocol data unit session establishment request signaling initiated by the power terminal device, capturing signaling plane data through a signaling plane probe;

[0009] Use data plane probes to encapsulate the flow detection field at the ingress of the segment routing IPv6 tunnel to obtain data plane performance data.

[0010] The signaling plane data is associated and bound with the data plane performance data corresponding to the power service flow in the current time window through the service quality flow identifier and the timestamp to generate a multi-dimensional spatiotemporal feature matrix for the power service;

[0011] Based on the multidimensional spatiotemporal feature matrix, the power 5G network quality is predicted using a pre-constructed dynamic spatiotemporal causal graph network to obtain a network quality prediction score; the dynamic spatiotemporal causal graph network includes a dynamic causal neural graph construction layer, a spatiotemporal graph convolution layer, and a causal reasoning layer connected in sequence;

[0012] When the network quality prediction score exceeds a preset score threshold, a network resource configuration optimization model is established with minimizing the weighted sum of link delay and resource consumption as the optimization goal;

[0013] The network resource configuration optimization model is solved by the integer programming method to obtain the optimal network resource configuration strategy, and the optimal network resource configuration strategy is deployed to the power 5G network device nodes to optimize the network resource configuration.

[0014] In a further embodiment, the step of associating and binding the signaling plane data with the data plane performance data corresponding to the electric power service flow in the current time window through the quality of service flow identifier and the timestamp to generate a multidimensional spatiotemporal feature matrix for the electric power service includes:

[0015] When a signaling event is triggered, a quality of service flow identifier is allocated to the quality of service flow corresponding to the signaling event through the session management function;

[0016] When a change in the quality of service flow identifier is detected, all signaling plane data with the same quality of service flow identifier in the current time window are associated and bound with the data plane performance data to obtain initial associated data;

[0017] Based on the time synchronization protocol, the initial associated data is timestamp aligned through the synchronization message interaction process of the master and slave power 5G network device nodes to obtain timing aligned associated data;

[0018] According to the timing alignment associated data, the average link delay between the master and slave power 5G network device nodes is calculated using the message interaction timestamp in the synchronization message interaction process;

[0019] Calculate the average value of the link delay data of all power 5G network device nodes in the current time window to obtain the delay mean, and calculate the degree of deviation between the link delay data and the delay mean to obtain the delay variance;

[0020] The average link delay and the delay variance are integrated into a structured feature matrix according to the time step, network device node and feature category dimensions to generate a multidimensional spatiotemporal feature matrix for power business.

[0021] In a further embodiment, the condition for timestamp alignment is that the absolute difference between the timestamp of the data plane performance data and the timestamp of the signaling event trigger is less than a preset time window threshold.

[0022] In a further embodiment, the dynamic causal neural graph construction layer includes a physical layer graph, a cross-connection layer, and a logical layer graph;

[0023] The physical layer graph uses physical network entities in the electric power 5G network as nodes. The physical layer graph is used to dynamically calculate the edge connection weights between physical network entities based on real-time link bandwidth and link delay data, and generate a physical layer topology graph representing the status of physical resources in the electric power 5G network based on the edge connection weights;

[0024] The logical layer diagram uses logical service entities as nodes, and the logical layer diagram is used to dynamically optimize the logical service flow mapping relationship according to the guaranteed traffic bit rate compliance rate and service priority, to obtain an optimized logical service flow mapping relationship;

[0025] The cross-connection layer is used to cross-layer associate the physical network entities and logical business entities in the physical layer topology diagram through service quality flow identifiers based on the optimized logical business flow mapping relationship, forming a multi-level network topology structure including physical resources and business logic.

[0026] In a further embodiment, the spatiotemporal graph convolution layer comprises a spatial graph convolutional network and a temporal graph convolutional network in parallel;

[0027] The spatial graph convolutional network is used to aggregate the signaling plane data and data plane performance data of adjacent nodes in the multi-level network topology structure using a graph attention mechanism, and capture the network space dependencies of the multi-level network topology structure in combination with the multi-dimensional spatiotemporal feature matrix to obtain network space structure features;

[0028] The temporal graph convolutional network is used to model the temporal dependency of the data plane performance data using a gated temporal convolutional network, extract the data plane time features, and fuse the network spatial structure features with the data plane time features to output a joint representation of spatiotemporal features.

[0029] In a further implementation scheme, the causal reasoning layer is used to determine the predicted causal relationship weights between nodes of a multi-level network topology structure using the Granger causality analysis method, and linearly combine the joint representation of the spatiotemporal features with the predicted causal relationship weights through a fully connected layer to output a network quality prediction score for the power 5G network.

[0030] In a further embodiment, the constraints of the network resource configuration optimization model include guaranteed traffic bit rate constraints and data radio bearer constraints;

[0031] The guaranteed traffic bit rate constraint is that the actual guaranteed traffic bit rate of each quality of service flow at each time step is not less than a preset minimum traffic bit value;

[0032] The data radio bearer constraint is that the sum of the total amount of resources allocated to all data radio bearers does not exceed a preset maximum resource allocation allowable value.

[0033] In a further embodiment, the joint loss function of the dynamic spatiotemporal causal graph network is composed of a network quality loss term and a causal loss term through a weighted summation;

[0034] The network quality loss term is obtained by calculating the mean square error between the network quality prediction score and the actual network quality score of the power 5G network;

[0035] The causal loss term is obtained by calculating the cross entropy between the predicted causal relationship weights and the true causal relationship weights.

[0036] In a second aspect, the present invention provides a power 5G network service quality monitoring system, the system comprising:

[0037] A first data acquisition module is configured to capture signaling plane data through a signaling plane probe in response to a protocol data unit session establishment request signaling initiated by the power terminal device;

[0038] The second data acquisition module is configured to encapsulate a flow detection field at the ingress of the segment routing IPv6 tunnel through a data plane probe to obtain data plane performance data;

[0039] A spatiotemporal feature generation module is configured to associate and bind the signaling plane data with the data plane performance data corresponding to the power service flow in the current time window through a quality of service flow identifier and a timestamp, thereby generating a multi-dimensional spatiotemporal feature matrix for the power service;

[0040] A network quality prediction module is configured to predict the quality of the power 5G network based on the multi-dimensional spatiotemporal feature matrix using a pre-built dynamic spatiotemporal causal graph network to obtain a network quality prediction score; the dynamic spatiotemporal causal graph network includes a dynamic causal neural graph construction layer, a spatiotemporal graph convolution layer, and a causal reasoning layer connected in sequence;

[0041] A configuration model building module is used to establish a network resource configuration optimization model with minimizing latency and resource consumption as optimization goals when the network quality prediction score exceeds a preset score threshold;

[0042] The resource configuration optimization module is used to solve the network resource configuration optimization model through integer programming method, obtain the optimal network resource configuration strategy, and deploy the optimal network resource configuration strategy to the power 5G network device node to optimize the network resource configuration.

[0043] In a further embodiment, the spatiotemporal feature generation module is specifically used to:

[0044] When a signaling event is triggered, a quality of service flow identifier is assigned to the quality of service flow corresponding to the signaling event through the session management function;

[0045] When a change in the quality of service flow identifier is detected, all signaling plane data with the same quality of service flow identifier in the current time window are associated and bound with the data plane performance data to obtain initial associated data;

[0046] Based on the time synchronization protocol, the initial associated data is timestamp aligned through the synchronization message interaction process of the master and slave power 5G network device nodes to obtain timing aligned associated data;

[0047] According to the timing alignment associated data, the average link delay between the master and slave power 5G network device nodes is calculated using the message interaction timestamp in the synchronization message interaction process;

[0048] Calculate the average value of the link delay data of all power 5G network device nodes in the current time window to obtain the delay mean, and calculate the degree of deviation between the link delay data and the delay mean to obtain the delay variance;

[0049] The average link delay and the delay variance are integrated into a structured feature matrix according to the time step, network device node and feature category dimensions to generate a multidimensional spatiotemporal feature matrix for power business.

[0050] In a further embodiment, the condition for timestamp alignment is that the absolute difference between the timestamp of the data plane performance data and the timestamp of the signaling event trigger is less than a preset time window threshold.

[0051] In a further embodiment, the dynamic causal neural graph construction layer includes a physical layer graph, a cross-connection layer, and a logical layer graph;

[0052] The physical layer graph uses physical network entities in the electric power 5G network as nodes. The physical layer graph is used to dynamically calculate the edge connection weights between physical network entities based on real-time link bandwidth and link delay data, and generate a physical layer topology graph representing the status of physical resources in the electric power 5G network based on the edge connection weights;

[0053] The logical layer diagram uses logical service entities as nodes, and the logical layer diagram is used to dynamically optimize the logical service flow mapping relationship according to the guaranteed traffic bit rate compliance rate and service priority, to obtain an optimized logical service flow mapping relationship;

[0054] The cross-connection layer is used to cross-layer associate the physical network entities and logical business entities in the physical layer topology diagram through service quality flow identifiers based on the optimized logical business flow mapping relationship, forming a multi-level network topology structure including physical resources and business logic.

[0055] In a further embodiment, the spatiotemporal graph convolution layer comprises a spatial graph convolutional network and a temporal graph convolutional network in parallel;

[0056] The spatial graph convolutional network is used to aggregate the signaling plane data and data plane performance data of adjacent nodes in the multi-level network topology structure using a graph attention mechanism, and capture the network space dependencies of the multi-level network topology structure in combination with the multi-dimensional spatiotemporal feature matrix to obtain network space structure features;

[0057] The temporal graph convolutional network is used to model the temporal dependency of the data plane performance data using a gated temporal convolutional network, extract the data plane time features, and fuse the network spatial structure features with the data plane time features to output a joint representation of spatiotemporal features.

[0058] In a further implementation scheme, the causal reasoning layer is used to determine the predicted causal relationship weights between nodes of a multi-level network topology structure using the Granger causality analysis method, and linearly combine the joint representation of the spatiotemporal features with the predicted causal relationship weights through a fully connected layer to output a network quality prediction score for the power 5G network.

[0059] In a further embodiment, the constraints of the network resource configuration optimization model include guaranteed traffic bit rate constraints and data radio bearer constraints;

[0060] The guaranteed traffic bit rate constraint is that the actual guaranteed traffic bit rate of each quality of service flow at each time step is not less than a preset minimum traffic bit value;

[0061] The data radio bearer constraint is that the sum of the total amount of resources allocated to all data radio bearers does not exceed a preset maximum resource allocation allowable value.

[0062] In a further embodiment, the joint loss function of the dynamic spatiotemporal causal graph network is composed of a network quality loss term and a causal loss term through a weighted summation;

[0063] The network quality loss term is obtained by calculating the mean square error between the network quality prediction score and the actual network quality score of the power 5G network;

[0064] The causal loss term is obtained by calculating the cross entropy between the predicted causal relationship weights and the true causal relationship weights.

[0065] In a third aspect, the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0066] The present invention provides a method, system and storage medium for monitoring the service quality of a 5G power network. The method captures signaling plane data through a signaling plane probe in response to a protocol data unit session establishment request signaling initiated by a power terminal device; encapsulates a flow detection field at the entrance of a segmented routing IPv6 tunnel through a data plane probe to obtain data plane performance data; associates and binds the signaling plane data with the data plane performance data corresponding to the power business flow in the current time window through a service quality flow identifier and a timestamp to generate a multi-dimensional spatiotemporal feature matrix for the power business; based on the multi-dimensional spatiotemporal feature matrix, the power 5G network quality is predicted using a pre-constructed dynamic spatiotemporal causal graph network to obtain a network quality prediction score; when the network quality prediction score exceeds a preset score threshold, a network resource configuration optimization model is established with the optimization goal of minimizing the weighted sum of link delay and resource consumption; the network resource configuration optimization model is solved by an integer programming method to obtain an optimal network resource configuration strategy, and the optimal network resource configuration strategy is deployed to the power 5G network device nodes for network resource configuration optimization. Compared with the existing technology, this method generates a spatiotemporal feature matrix by fusing signaling plane and data plane data, and combines it with a dynamic spatiotemporal causal graph network to predict the quality of the power 5G network, thereby optimizing the network resource allocation. It realizes real-time and accurate monitoring of the service quality of the power 5G network and dynamic optimization of resources, significantly improving the service quality and resource utilization efficiency of the power 5G network. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 This is a flow chart of a method for monitoring the quality of service in a 5G power network according to an embodiment of the present invention;

[0068] Figure 2 Schematic diagram of the synchronization message interaction process based on the time synchronization protocol provided by an embodiment of the present invention;

[0069] Figure 3 This is a schematic diagram of the overall architecture of a dynamic spatiotemporal causal graph network provided by an embodiment of the present invention;

[0070] Figure 4 Schematic diagram of the spatiotemporal graph convolution feature extraction process provided by an embodiment of the present invention;

[0071] Figure 5 This is a block diagram of the power 5G network service quality monitoring system provided by an embodiment of the present invention.

[0072] Explanation of the accompanying symbols: 101, first data acquisition module; 102, second data acquisition module; 103, spatiotemporal feature generation module; 104, network quality prediction module; 105, configuration model construction module; 106, resource configuration optimization module. DETAILED DESCRIPTION

[0073] The following describes the embodiments of the present invention in detail with reference to the accompanying drawings. The embodiments are provided for illustrative purposes only and are not to be construed as limiting the present invention. The accompanying drawings are provided for reference and illustration only and do not constitute a limitation on the scope of protection of the present invention. Many changes may be made to the present invention without departing from the spirit and scope of the present invention.

[0074] refer to Figure 1 , an embodiment of the present invention provides a method for monitoring the service quality of a power 5G network, such as Figure 1 As shown, the method includes the following steps:

[0075] S1. In response to a protocol data unit session establishment request signaling initiated by the power terminal device, signaling plane data is captured through a signaling plane probe.

[0076] S2. Use a data plane probe to encapsulate the flow detection field at the ingress of the segment routing IPv6 tunnel to obtain data plane performance data.

[0077] In this embodiment, a signaling plane probe is used to embed a lightweight agent in the Access and Mobility Management Function (AMF) and Session Management Function (SMF) nodes. The lightweight agent can capture the Protocol Data Unit (PDU) Session Resource Setup Request signaling in the Next Generation Application Protocol (NGAP) in real time. When the PDU session establishment request signaling is initiated by the power terminal equipment (such as a smart meter or a substation monitoring terminal), the signaling plane probe is used to capture the signaling event (Protocol Data Unit Session Establishment Request signaling), extract key parameters such as the 5G Quality of Service Identifier (5G QoS Identifier, 5QI), Allocation and Retention Priority (ARP), and Guaranteed Flow Bit Rate (GFBR), and form signaling plane data. At the same time, this embodiment uses the data plane probe to generate signaling plane data between the User Plane Function (UPF) and the next generation node. At the entrance of the Segment Routing IPv6 (SRv6) tunnel of the gNB (B), data packets are encapsulated and an in-band Operations, Administration, and Maintenance (IOAM) field is added. This IOAM field records information such as the hop-by-hop latency, packet loss rate, and instantaneous queue length of the power service flow, forming data plane performance data.

[0078] S3. The signaling plane data is associated and bound with the data plane performance data corresponding to the electric power service flow in the current time window through the service quality flow identifier and the timestamp to generate a multi-dimensional spatiotemporal feature matrix for the electric power service.

[0079] In some embodiments, the step of associating and binding the signaling plane data with the data plane performance data corresponding to the electric power service flow in the current time window by using the quality of service flow identifier and the timestamp to generate a multidimensional spatiotemporal feature matrix for the electric power service includes:

[0080] When a signaling event is triggered, a quality of service flow identifier is assigned to the quality of service flow corresponding to the signaling event through the session management function;

[0081] When a change in the quality of service flow identifier is detected, all signaling plane data with the same quality of service flow identifier in the current time window are associated and bound with the data plane performance data to obtain initial associated data;

[0082] Based on the time synchronization protocol, the initial associated data is timestamp aligned through the synchronization message interaction process of the master and slave power 5G network device nodes to obtain timing aligned associated data;

[0083] According to the timing alignment associated data, the average link delay between the master and slave power 5G network device nodes is calculated using the message interaction timestamp in the synchronization message interaction process;

[0084] Calculate the average value of the link delay data of all power 5G network device nodes in the current time window to obtain the delay mean, and calculate the degree of deviation between the link delay data and the delay mean to obtain the delay variance;

[0085] The average link delay and the delay variance are integrated into a structured feature matrix according to the time step, network device node and feature category dimensions to generate a multidimensional spatiotemporal feature matrix for power business.

[0086] In 5G power networks, signaling events (such as session establishment and resource allocation) and data plane performance indicator data (such as latency and packet loss) are typically collected independently. For example, when a user equipment (UE) initiates a protocol data unit (PDU) session establishment request, a corresponding quality of service (QoS) flow (QoS Flow) begins transmitting data on the data plane. To comprehensively analyze the network status, it is necessary to correlate signaling events with flow-based operations and maintenance (IOAM) data, two different layers of data. Therefore, this embodiment uses a QoS Flow Identifier (QFI) and a timestamp to bind signaling events with IOAM data. Each QoS flow has a unique QoS Flow Identifier (QFI). Based on this feature, this embodiment can associate signaling events with corresponding data plane IOAM data. The timestamp can accurately record the time when the event occurred, thereby ensuring data alignment. The specific steps for binding signaling events with data plane IOAM data using the QoS Flow Identifier and timestamp are as follows:

[0087] In this embodiment, when a signaling event (PDU session establishment request) is triggered by a power terminal device (such as a smart meter or a substation monitoring terminal), the session management function will assign a unique service quality flow identifier QFI to the service quality flow (QoS Flow) corresponding to the signaling event, and monitor the dynamic changes of the service quality flow identifier QFI in real time during the operation of the power 5G network (such as QFI switching or resource reallocation caused by business priority adjustment). When the service quality flow identifier in the power 5G network changes instantly, it is necessary to associate the signaling time with the data plane IOAM data in the current time window, and accurately align the timestamps within a limited time to ensure the accuracy of network quality monitoring, that is, all the services with the corresponding service quality flow identifier (QFI) and timestamp in the current time window are associated with the data plane IOAM data in the current time window. The signaling events of the same QFI (such as PDU session establishment delay) are bound to the data plane performance data (such as SRv6 tunnel delay), and all IOAM data belonging to the same QFI (such as delay, packet loss rate, queue depth) are merged into the corresponding signaling event to form an end-to-end service quality monitoring record, ensuring that the control plane and user plane data of the same service correspond one-to-one, thereby obtaining the initial associated data bound to the QFI and the time window. In some embodiments, the condition for timestamp alignment is that the absolute difference between the timestamp of the data plane performance data and the timestamp triggering the signaling event is less than the preset time window threshold.

[0088] It should be noted that to ensure that the signaling event and the IOAM data are strictly matched in timing, this embodiment uses the Precision Time Protocol (PTP) to perform nanosecond clock synchronization through the PTP synchronization message interaction between the master and slave power 5G network nodes (such as the master clock is the substation control center and the slave clock is the base station), calibrate the clock deviation, and thus synchronize the signaling trigger time with the IOAM recording time. The time window can be set to 1 millisecond. In this embodiment, when the timestamp of the data plane IOAM data is within the time window of the signaling event, the IOAM data is considered to be associated with the signaling event. The timestamp alignment condition indicates that the timestamp of the data plane IOAM data must fall within the time window of the signaling event to be considered associated with the signaling event. The timestamp alignment condition can be expressed as:

[0089] |t event -t ioam |<Δt

[0090] Where, t event The signaling trigger timestamp is the time when the signaling event occurs; t ioamThe timestamp for recording IOAM data may be the hop-by-hop delay recording time of a segment routing tunnel based on IPv6; Δt is the time window, which is dynamically adjusted by the network management policy. For example, in this embodiment, it may be set to 1ms.

[0091] This embodiment integrates discrete signaling events (such as RRC connection success rate) and continuous data plane performance data (such as SRv6 tunnel delay average and GFBR compliance rate) into a structured feature matrix to provide high-dimensional input data for network quality prediction and fault diagnosis. The aligned IOAM data needs to be aggregated into feature vectors, such as Figure 2 As shown, this embodiment takes the delay indicator as an example. After the master and slave devices complete the synchronization message interaction process based on PTP, the master and slave device message interaction timestamp T in the PTP synchronization process is 1i ~T 6i , calculate the average value of the two-way transmission delay, assuming that the network is symmetrical and the master-slave frequency ratio is γ, which is used to compensate for clock drift. The average link delay is meanLinkDelay i The calculation formula for the clock offset is:

[0092]

[0093] Offset=T 2i -T 1i -meanLinkDelay i

[0094] Where, meanLinkDelay i is the average link delay, which is the average delay from request to completion; Offset is the clock deviation, which is the time difference between the master and slave clocks and is used for timestamp correction; T 6i T is the response time of receiving from the clock; 3i The delay request time sent by the slave clock; T 5i Send response time for the master clock; T 4i The master clock receives the delay request time; T 2i The time when the slave clock receives the synchronization request; T 1i Send synchronization request time to the master clock.

[0095] After using the PTP protocol for nanosecond-level clock synchronization, the timestamps of the IOAM data are corrected based on the clock offset to ensure consistency with the signaling event time base. The aligned IOAM data needs to be aggregated into a structured feature matrix. Statistics are calculated for the IOAM data associated with each QFI. The statistics include the mean delay and the variance of the delay. The calculation formulas for the mean delay and the variance of the delay are:

[0096]

[0097] Where, is the mean delay, which is the average of all link delays in time window i; N is the number of data samples in the time window; is the delay variance, which is the degree of dispersion of the delay data.

[0098] In this embodiment, the above features are integrated into a three-dimensional tensor according to the three dimensions of time step, network node, and feature category. The final structured feature matrix is ​​a multidimensional array. Each row represents a data record after association, and each column represents a feature. Its spatiotemporal feature matrix E = R T×S×N is a three-dimensional tensor, and its mathematical expression is as follows:

[0099]

[0100] Where, E is the spatiotemporal feature matrix; η RRC is the RRC connection success rate, which is the ratio of successful establishment of radio resource control connection; GFBR is the guaranteed flow bit rate; θ cell-load is the cell load, which is the current cell radio resource utilization rate; S is the number of network nodes, which includes the total number of physical nodes (such as gNB, UPF) and logical nodes (such as QoS Flow); T is the number of time steps. This embodiment considers three feature category indicators, namely signaling indicators, data plane indicators, and environmental variables. Among them, signaling indicators include PDU session establishment delay and Radio Resource Control (RRC) connection success rate; data plane indicators include SRv6 tunnel end-to-end delay mean, delay variance, and packet loss rate; power environment variables include cell load, thereby generating a multi-dimensional spatiotemporal feature matrix for power services.

[0101] In summary, this embodiment solves the problem of data separation between the signaling plane and the data plane through the dual constraints of QFI and PTP timestamps, ensures the accurate association of multi-source data in time series, and combines statistical aggregation to generate structured features, providing high-quality input data for the spatiotemporal graph convolution layer of the subsequent spatiotemporal causal graph neural network model.

[0102] S4. Based on the multi-dimensional spatiotemporal feature matrix, the power 5G network quality is predicted using a pre-built dynamic spatiotemporal causal graph network to obtain a network quality prediction score.

[0103] This embodiment is based on the Dynamic Spatio-Temporal Causal Graph Neural Network (DSCGN), with network elements as nodes. Edge weights are dynamically calculated through signaling dependencies and link states. At the same time, the TinyML model is deployed at the edge nodes to achieve real-time prediction of microburst traffic. In some implementations, such as Figure 3 、 Figure 4 As shown, the dynamic spatiotemporal causal graph network includes a dynamic causal neural graph construction layer, a spatiotemporal graph convolution layer and a causal reasoning layer connected in sequence. In this embodiment, the dynamic causal neural graph construction layer includes a physical layer graph, a cross-connection layer and a logical layer graph; the physical layer graph uses the physical network entities in the power 5G network as nodes, and the physical layer graph is used to dynamically calculate the edge connection weights between the physical network entities based on the real-time link bandwidth and link delay data, and generate a physical layer topology graph representing the physical resource status of the power 5G network based on the edge connection weights; the logical layer graph uses the logical business entities as nodes, and the logical layer graph is used to dynamically optimize the logical business flow mapping relationship based on the guaranteed traffic bit rate compliance rate and business priority, and obtain an optimized logical business flow mapping relationship; the cross-connection layer is used to cross-layer associate the physical network entities and the logical business entities in the physical layer topology graph through the service quality flow identifier according to the optimized logical business flow mapping relationship, so as to form a multi-level network topology structure containing physical resources and business logic.

[0104] The spatiotemporal graph convolution layer includes a parallel spatial graph convolution network and a temporal graph convolution network; the spatial graph convolution network is used to utilize the graph attention mechanism to aggregate the signaling surface data and data surface performance data of adjacent nodes in the multi-level network topology structure, and combine the multi-dimensional spatiotemporal feature matrix to capture the network space dependency of the multi-level network topology structure to obtain the network space structure features; the temporal graph convolution network is used to use a gated temporal convolution network to model the temporal dependency of the data surface performance data, extract the data surface time features, and fuse the network space structure features and the data surface time features to output a joint representation of spatiotemporal features.

[0105] The causal inference layer is used to determine the predicted causal relationship weights between nodes in a multi-level network topology structure using the Granger causality analysis method, and linearly combine the joint representation of the spatiotemporal features with the predicted causal relationship weights through a fully connected layer to output a network quality prediction score for the power 5G network.

[0106] In a specific embodiment, the input of the dynamic causal neural graph construction layer is real-time network topology, signaling events, and in-band operation, management and maintenance data IOAM. These input data are processed by the physical layer graph, cross-connection layer and logical layer graph respectively, and finally output a dynamically updated multi-level network topology structure. The multi-level network topology structure is a dynamically updated multi-level physical layer and logical layer graph structure. In this embodiment, the physical layer graph uses the physical network entities in the power 5G network as nodes. The physical network entities include the next generation base station (Generation NodeB, gNB), user plane function entity (UserPlane Function, UPF), access and mobility management function entity (Access and Mobility Management Function Entity Management Function (AMF). In this embodiment, real-time link bandwidth and latency data (such as the transmission performance between a base station and a substation) are input into the physical layer graph, so that the physical layer graph dynamically calculates the edge connection weights between physical network entities based on the real-time link bandwidth and latency, thereby obtaining a physical layer topology graph representing the physical resource status of the power 5G network. In this embodiment, the edge connection weights can be calculated based on the link weight between the base station and the UPF by weighting the bandwidth ratio and the inverse of the latency. The logical layer graph uses logical service entities as nodes, and the logical service entities include quality of service flows (QoS flows) and data radio bearers (DRBs). In this embodiment, the guaranteed traffic bit rate (such as the bandwidth guarantee rate of smart meter data flows) and service priorities (such as the urgency of relay protection instructions) are input into the logical layer graph, so that the logical layer graph dynamically adjusts the edge weights between logical service entities based on the guaranteed traffic bit rate and service priorities (such as prioritizing mapping high-priority service flows to low-latency DRBs), thereby obtaining an optimized logical service flow mapping relationship. The quality of service flow identifier (QoS flow) is used across the connection layer to identify the traffic flow. The QFI (Qualified Field Identifier) ​​cross-layer associates the physical network entities in the physical layer diagram with the logical business entities in the logical layer diagram, and outputs a multi-level network topology that integrates physical resources and business logic.

[0107] The spatiotemporal graph convolution layer is divided into two processes: spatial convolution and temporal convolution. Among them, spatial convolution aggregates the information of neighbor nodes of the current network element based on the graph attention mechanism to capture topological dependencies; temporal convolution uses a gated temporal convolutional network to model temporal changes. In this embodiment, the multi-level network topology structure and the multi-dimensional spatiotemporal feature matrix are input into the spatial graph convolution network. The spatial graph convolution network uses the graph attention mechanism (GAT) to aggregate the signaling surface data (such as RRC connection success rate) and data surface performance data (such as SRv6 tunnel delay) of adjacent nodes (such as substations, base stations and UPFs) to capture network spatial dependencies and obtain feature vectors that characterize the network spatial structure. In this embodiment, the data surface performance data is input into the temporal graph convolution network, and the temporal graph convolution network uses a gated temporal convolutional network (TemporalConvolutional The time series dependency of data plane performance data (such as periodic load fluctuation) is modeled by using a TCN (Traffic Flow Network) to extract time dimension features and obtain feature vectors that characterize the time changes of the data plane. Finally, this embodiment concatenates and weights the spatial structure features and the time features to generate a joint representation of spatiotemporal features. In a specific embodiment, this embodiment assumes that A physical ∈{0,1} S×S Represents the adjacency matrix of a node v in the network topology G = (ν, ε) composed of network elements. The spatiotemporal convolutional layer jointly models the temporal and spatial features and extracts the feature E through the spatiotemporal encoder. (l) ,have:

[0108] H=Encoder(E)

[0109] And H∈R N , then the feature extracted by the (l+1)th layer of the spatiotemporal graph convolution is:

[0110]

[0111] Where A physical is the physical layer adjacency matrix, which is used to describe the connection relationship between physical network elements such as gNB / UPF / AMF (1 indicates direct connection); G is the network topology; v is the set of physical / logical nodes, such as gNB and QoS Flow; ε is the edge set; H is the spatiotemporal coding feature, the low-dimensional feature vector output by the encoder, which is used to characterize the network status; Encoder(*) is the spatiotemporal encoder; E (l) is the spatiotemporal feature matrix of the lth layer of spatiotemporal graph convolution; R is a set of real numbers; H (l) is the feature matrix of the lth layer of spatiotemporal graph convolution; σ(*) is the sigmoid activation function; It is the matrix concatenation operation; is the normalized degree matrix, Symmetric normalization of ; is an adjacency matrix with self-loops; W (l) is the learnable parameter matrix.

[0112] The 5G network system trains the basic model locally, then aggregates global parameters and collects optimized network data in real time to update the model parameters. The reasoning process ensures that the base station obtains the latest signaling events, IOAM data, and network topology status at regular intervals. The physical and logical layer graph structures are then reconstructed based on the current network status. The future network quality level and failure probability are calculated using a dynamic spatiotemporal causal graph neural network. In this embodiment, the input of the causal reasoning layer is spatiotemporal features and historical fault records. The main methods for obtaining historical fault records include Next Generation Application Protocol (NGAP) / S1 Application Protocol (S1AP) signaling event record timestamps, QFIs, and failure cause codes collected by network elements such as access and mobility management functions and session management functions; hop-by-hop packet loss rate and latency distribution of IPv6-based segment routing tunnels; device alarms obtained through Simple Network Management Protocol (SNMP) traps or system logs; and key performance indicators (KPIs) in base station performance management counters (PM counters). The indicator mainly includes the radio resource utilization of gNB, the throughput of UPF, and the signaling processing delay of the core network node.

[0113] In this embodiment, the spatiotemporal features are jointly represented and input into the causal inference layer. The causal inference layer constructs a causal graph and uses Granger Causality Analysis to quantify the causal influence strength between network elements (such as gNBs and QoS flows), forming weighted directed edges to determine the causal relationship weights between nodes. The spatiotemporal features and causal relationship weights are linearly combined in the fully connected layer and mapped to a network quality prediction score. In this embodiment, the graph attention mechanism (GAT) is used to predict the network quality prediction score in the future time window. The formula for mapping the fully connected layer to the score is:

[0114] y r =σ(HW p +b p )

[0115] Where y rScore the network quality prediction in the future time window; W p is the weight matrix, the initial value of the weight matrix is ​​usually set by random initialization and optimized by the back propagation algorithm during the training process; b p is a bias vector, which is used to adjust the output of the node. The initial value is also randomly set and adjusted during the training process.

[0116] In order to train the spatiotemporal causal graph neural network so that it can accurately predict network quality through the probability of delay exceeding the standard and identify the causal path, the joint loss function of the dynamic spatiotemporal causal graph network is composed of a network quality loss term and a causal loss term through weighted summation. The network quality loss term is obtained by calculating the mean square error between the network quality prediction score of the power 5G network and the actual network quality score. The causal loss term is obtained by calculating the cross entropy between the predicted causal relationship weight and the actual causal relationship weight. In this embodiment, the joint loss function of the dynamic spatiotemporal causal graph network is specifically:

[0117] L = α·MSE(y r ,y true )+β·CrossEntropy(p causal ,p true )

[0118] In the formula, MSE(y r ,y true ) is the mean square error term of the prediction task; CrossEntropy(p causal ,p true ) is the cross entropy term of the causal path identification task; y true is the true value of the network quality score; p true is the true causal relationship weight; p causal is the weight of the predicted causal relationship; α is the prediction task weight coefficient; β is the causal analysis weight coefficient. The prediction task weight coefficient α controls the weight of the mean square error term. When α is increased, the model pays more attention to accurately predicting network quality; the causal analysis weight coefficient β controls the weight of the cross entropy term. When β is increased, the model pays more attention to the interpretability of the causal relationship. The calculation process formula of the weight coefficients α and β is as follows:

[0119]

[0120] Where, α t and β t are the prediction task weight coefficient and causal analysis weight at the current moment respectively; α t+1 and β t+1 are the prediction task weight coefficient and causal analysis weight at time (t+1) respectively; λ is the learning rate coefficient, which is used to prevent the denominator from being zero and adjust the speed of weight update.

[0121] In terms of intelligent prediction and evaluation, this embodiment uses a dynamic spatiotemporal causal graph neural network to abstract physical and logical network elements into graph nodes. Causal relationships between nodes are defined based on signaling transmission processes. The weights between nodes are determined based on in-band operation, management, and maintenance data from IPv6 in-band detection technology. This data includes link bandwidth, latency, and packet loss rate. For example, the weight between the access and mobility management function and the next-generation base station is determined by the frequency of next-generation application protocol signaling interactions. The weights of the number of hops and end-to-end latency in IPv6-based segment routing tunnels are calculated based on the hop-by-hop latency in the IOAM field. This embodiment uses graph convolution and attention mechanisms to capture nonlinear interactions between network elements, explicitly modeling spatial and causal dependencies between them. It also dynamically updates the graph structure, supporting millisecond-level reasoning and significantly improving prediction accuracy. In summary, this embodiment uses dynamic graph construction, spatiotemporal feature fusion, and causal reasoning to accurately characterize the physical resource status and service logic relationships of the power 5G network and output a network quality score, which closely meets the requirements of high-reliability, low-latency scenarios such as smart grids.

[0122] S5. When the network quality prediction score exceeds a preset score threshold, a network resource configuration optimization model is established with minimizing the weighted sum of link delay and resource consumption as the optimization goal.

[0123] S6. Solve the network resource configuration optimization model through integer programming method to obtain the optimal network resource configuration strategy, and deploy the optimal network resource configuration strategy to the power 5G network device node to optimize the network resource configuration.

[0124] During the service quality monitoring process of the power 5G network, when the network quality prediction score (delay exceedance probability) is greater than the 90% threshold, the system automatically triggers the data radio bearer (DRB) remapping and transmission path switching operation to re-establish the mapping relationship between the service quality flow and the data radio bearer. The adaptive decision execution phase aims to generate the optimal resource allocation strategy. This strategy needs to balance the delay cost and resource overhead under the conditions of satisfying the guaranteed flow bit rate constraint and the data radio bearer constraint to minimize future delay and resource consumption. In this embodiment, the constraints of the network resource configuration optimization model include guaranteed flow bit rate constraint and data radio bearer constraint; the guaranteed flow bit rate constraint is that the actual guaranteed flow bit rate of each service quality flow at each time step is not less than the preset flow bit minimum value; the data radio bearer constraint is that the sum of the total resources allocated to all data radio bearers does not exceed the preset maximum resource allocation allowable value to avoid excessive resource allocation. Under the premise of ensuring service quality, the optimization problem can be modeled as follows:

[0125]

[0126]

[0127] Where π is the resource allocation strategy; F is the expected value, which is used to measure the expected value of future delay and resource consumption; d t (π) is the delay cost at time t under resource allocation strategy π; c t (π) is the resource cost at time t under resource allocation strategy π; GFBR t is the actual guaranteed flow bit rate at time t; GFBR min is the minimum value of traffic bits; DBR j The total amount of resources allocated to the jth data radio bearer; J is the number of data radio bearers; DBR max is the maximum allowed value for resource allocation; τ is the time window length, which represents the number of future time steps considered in the optimization problem; in this embodiment, in order to ensure a global trade-off between prediction performance and resource efficiency, the weights are kept consistent with the weights in the loss function to ensure a global trade-off between prediction performance and resource efficiency.

[0128] The DRB allocation problem can be further modeled as a 0-1 integer programming problem:

[0129]

[0130] Where x i,j is a binary decision variable indicating whether the quality of service flow is mapped to the data radio bearer; u i,j is the utility function, which comprehensively considers latency improvement and resource utilization.

[0131] Adaptive decision-making is achieved through a lightweight machine learning (TinyML) model deployed on edge nodes. This model can dynamically adjust resource allocation based on the real-time network status and optimization goals, thereby ensuring that the network can still maintain efficient and stable service quality when facing micro-burst traffic and dynamic topology changes, thereby guaranteeing the service quality of the power 5G network. Through this optimization mechanism, this embodiment can rationally utilize resources and reduce operating costs while ensuring network performance, while improving the overall efficiency and reliability of the power 5G network.

[0132] In summary, considering that in the field of power 5G network service quality monitoring, traditional operation, management, and maintenance methods rely on periodic polling or offline log analysis, making it difficult to capture sudden conditions such as microbursts caused by high-concurrency services in real time, which in turn leads to network congestion and affects user experience. To solve the above problems, this embodiment uses signaling plane probes and data plane probes to collect 5G signaling, completes in-band operation, management, and maintenance field encapsulation, associates signaling events with IOAM data in the format of service quality flow identifiers and timestamps, generates a multi-dimensional spatiotemporal feature matrix, and uses an improved graph neural network to construct a causal topology graph to achieve real-time prediction of microburst traffic, thereby adjusting the mapping rules of service quality flows to data radio bearers based on the prediction results. Therefore, this embodiment, through the coordinated operation of signaling plane and data plane probes, combined with the improved graph neural network model, not only achieves real-time monitoring and prediction of network quality, but also significantly improves monitoring accuracy and resource utilization through adaptive resource allocation optimization, effectively addressing the problems caused by microburst traffic and dynamic topology changes.

[0133] An embodiment of the present invention provides a method for monitoring the service quality of a 5G power network. The method comprises: capturing signaling plane data via a signaling plane probe in response to a protocol data unit session establishment request signaling initiated by a power terminal device; encapsulating a flow detection field at a segmented routing IPv6 tunnel entrance via a data plane probe to obtain data plane performance data; associating and binding the signaling plane data with data plane performance data corresponding to a power service flow in a current time window via a quality of service flow identifier and a timestamp to generate a multidimensional spatiotemporal feature matrix for power services; predicting the quality of the 5G power network based on the multidimensional spatiotemporal feature matrix using a pre-constructed dynamic spatiotemporal causal graph network to obtain a network quality prediction score; the dynamic spatiotemporal causal graph network includes a dynamic causal neural graph construction layer, a spatiotemporal graph convolution layer, and a causal inference layer connected in sequence; establishing a network resource configuration optimization model with the optimization objective of minimizing the weighted sum of link delay and resource consumption when the network quality prediction score exceeds a preset score threshold; solving the network resource configuration optimization model via an integer programming method to obtain an optimal network resource configuration strategy, and deploying the optimal network resource configuration strategy to the power 5G network device nodes for network resource configuration optimization. Compared with the existing technology, this method generates a spatiotemporal feature matrix by fusing signaling plane and data plane data, and combines it with a dynamic spatiotemporal causal graph network to predict the quality of the power 5G network, thereby optimizing the network resource allocation. It realizes real-time and accurate monitoring of the service quality of the power 5G network and dynamic optimization of resources, significantly improving the service quality and resource utilization efficiency of the power 5G network.

[0134] It should be noted that the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of this application.

[0135] In one embodiment, Figure 5 As shown, an embodiment of the present invention provides a power 5G network service quality monitoring system, the system comprising:

[0136] The first data acquisition module 101 is configured to capture signaling plane data through a signaling plane probe in response to a protocol data unit session establishment request signaling initiated by the power terminal device;

[0137] The second data acquisition module 102 is configured to encapsulate a flow detection field at the ingress of the segment routing IPv6 tunnel through a data plane probe to acquire data plane performance data.

[0138] The spatiotemporal feature generation module 103 is configured to associate and bind the signaling plane data with the data plane performance data corresponding to the power service flow in the current time window through the quality of service flow identifier and the timestamp, thereby generating a multi-dimensional spatiotemporal feature matrix for the power service;

[0139] The network quality prediction module 104 is configured to predict the quality of the power 5G network based on the multi-dimensional spatiotemporal feature matrix using a pre-constructed dynamic spatiotemporal causal graph network to obtain a network quality prediction score; the dynamic spatiotemporal causal graph network includes a dynamic causal neural graph construction layer, a spatiotemporal graph convolution layer, and a causal reasoning layer connected in sequence;

[0140] A configuration model building module 105 is configured to establish a network resource configuration optimization model with minimizing latency and resource consumption as optimization goals when the network quality prediction score exceeds a preset score threshold;

[0141] The resource configuration optimization module 106 is used to solve the network resource configuration optimization model through the integer programming method, obtain the optimal network resource configuration strategy, and deploy the optimal network resource configuration strategy to the power 5G network device node to optimize the network resource configuration.

[0142] In this embodiment, the spatiotemporal feature generation module is specifically used to:

[0143] When a signaling event is triggered, a quality of service flow identifier is allocated to the quality of service flow corresponding to the signaling event through the session management function;

[0144] When a change in the quality of service flow identifier is detected, all signaling plane data with the same quality of service flow identifier in the current time window are associated and bound with the data plane performance data to obtain initial associated data;

[0145] Based on the time synchronization protocol, the initial associated data is timestamp aligned through the synchronization message interaction process of the master and slave power 5G network device nodes to obtain timing aligned associated data;

[0146] According to the timing alignment associated data, the average link delay between the master and slave power 5G network device nodes is calculated using the message interaction timestamp in the synchronization message interaction process;

[0147] Calculate the average value of the link delay data of all power 5G network device nodes in the current time window to obtain the delay mean, and calculate the degree of deviation between the link delay data and the delay mean to obtain the delay variance;

[0148] The average link delay and the delay variance are integrated into a structured feature matrix according to the time step, network device node and feature category dimensions to generate a multidimensional spatiotemporal feature matrix for power business.

[0149] In this embodiment, the condition for timestamp alignment is that the absolute difference between the timestamp of the data plane performance data and the timestamp of the signaling event trigger is less than a preset time window threshold.

[0150] In this embodiment, the dynamic causal neural graph construction layer includes a physical layer graph, a cross-connection layer, and a logical layer graph;

[0151] The physical layer graph uses physical network entities in the electric power 5G network as nodes. The physical layer graph is used to dynamically calculate the edge connection weights between physical network entities based on real-time link bandwidth and link delay data, and generate a physical layer topology graph representing the status of physical resources in the electric power 5G network based on the edge connection weights;

[0152] The logical layer diagram uses logical service entities as nodes, and the logical layer diagram is used to dynamically optimize the logical service flow mapping relationship according to the guaranteed traffic bit rate compliance rate and service priority, to obtain an optimized logical service flow mapping relationship;

[0153] The cross-connection layer is used to cross-layer associate the physical network entities and logical business entities in the physical layer topology diagram through service quality flow identifiers based on the optimized logical business flow mapping relationship, forming a multi-level network topology structure including physical resources and business logic.

[0154] In this embodiment, the spatiotemporal graph convolution layer includes a spatial graph convolution network and a temporal graph convolution network in parallel;

[0155] The spatial graph convolutional network is used to aggregate the signaling plane data and data plane performance data of adjacent nodes in the multi-level network topology structure using a graph attention mechanism, and capture the network space dependencies of the multi-level network topology structure in combination with the multi-dimensional spatiotemporal feature matrix to obtain network space structure features;

[0156] The temporal graph convolutional network is used to model the temporal dependency of the data plane performance data using a gated temporal convolutional network, extract the data plane time features, and fuse the network spatial structure features with the data plane time features to output a joint representation of spatiotemporal features.

[0157] In this embodiment, the causal reasoning layer is used to determine the predicted causal relationship weights between nodes in a multi-level network topology structure using the Granger causality analysis method, and linearly combine the joint representation of the spatiotemporal features with the predicted causal relationship weights through a fully connected layer to output a network quality prediction score for the power 5G network.

[0158] In this embodiment, the constraints of the network resource configuration optimization model include guaranteed traffic bit rate constraints and data radio bearer constraints;

[0159] The guaranteed traffic bit rate constraint is that the actual guaranteed traffic bit rate of each quality of service flow at each time step is not less than a preset minimum traffic bit value;

[0160] The data radio bearer constraint is that the sum of the total amount of resources allocated to all data radio bearers does not exceed a preset maximum resource allocation allowable value.

[0161] In this embodiment, the joint loss function of the dynamic spatiotemporal causal graph network is composed of a network quality loss term and a causal loss term through weighted summation;

[0162] The network quality loss term is obtained by calculating the mean square error between the network quality prediction score and the actual network quality score of the power 5G network;

[0163] The causal loss term is obtained by calculating the cross entropy between the predicted causal relationship weights and the true causal relationship weights.

[0164] For the specific definition of a power 5G network service quality monitoring system, please refer to the above-mentioned definition of a power 5G network service quality monitoring method, which will not be repeated here. Those of ordinary skill in the art will appreciate that the various modules and steps described in conjunction with the embodiments disclosed in this application can be implemented in hardware, software, or a combination of both. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0165] An embodiment of the present invention provides a power 5G network service quality monitoring system, wherein the system responds to the protocol data unit session establishment request signaling initiated by the power terminal device through a first data acquisition module and captures signaling plane data through a signaling plane probe; a second data acquisition module encapsulates the flow detection field at the segment routing IPv6 tunnel entrance through a data plane probe to obtain data plane performance data; a spatiotemporal feature generation module associates and binds the signaling plane data with the data plane performance data corresponding to the power business flow in the current time window through a service quality flow identifier and a timestamp to generate a multi-dimensional spatiotemporal feature matrix for the power business; a network quality prediction module is based on the multi-dimensional spatiotemporal feature matrix, The system uses a pre-built dynamic spatiotemporal causal graph network to predict the quality of the electric power 5G network and obtain a network quality prediction score. The dynamic spatiotemporal causal graph network consists of a sequentially connected dynamic causal neural graph construction layer, a spatiotemporal graph convolution layer, and a causal inference layer. When the network quality prediction score exceeds a preset score threshold, the configuration model construction module establishes a network resource configuration optimization model with the optimization objective of minimizing the weighted sum of link latency and resource consumption. The resource configuration optimization module solves the network resource configuration optimization model using integer programming methods to obtain the optimal network resource configuration strategy, which is then deployed to the electric power 5G network device nodes for network resource configuration optimization. Compared with existing technologies, this system generates a spatiotemporal feature matrix by fusing signaling and data plane data, and uses the dynamic spatiotemporal causal graph network to predict the quality of the electric power 5G network, thereby optimizing network resource configuration. This system achieves real-time and accurate monitoring of the service quality of the electric power 5G network and dynamic resource optimization, significantly improving the service quality and resource utilization efficiency of the electric power 5G network.

[0166] In one embodiment, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the above method are implemented.

[0167] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., an SSD).

[0168] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing related hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods.

[0169] The above-described embodiments merely represent several preferred implementations of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent. It should be noted that a person skilled in the art could make several improvements and substitutions without departing from the technical principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be based on the scope of protection of the claims.

Claims

1. A method for monitoring the service quality of a power 5G network, characterized in that: The following steps are involved: In response to a protocol data unit session establishment request signaling initiated by the power terminal device, capturing signaling plane data through a signaling plane probe; Use data plane probes to encapsulate the flow detection field at the ingress of the segment routing IPv6 tunnel to obtain data plane performance data. The signaling plane data is associated and bound with the data plane performance data corresponding to the power service flow in the current time window through the service quality flow identifier and the timestamp to generate a multi-dimensional spatiotemporal feature matrix for the power service; Based on the multidimensional spatiotemporal feature matrix, the power 5G network quality is predicted using a pre-constructed dynamic spatiotemporal causal graph network to obtain a network quality prediction score; the dynamic spatiotemporal causal graph network includes a dynamic causal neural graph construction layer, a spatiotemporal graph convolution layer, and a causal reasoning layer connected in sequence; When the network quality prediction score exceeds a preset score threshold, a network resource configuration optimization model is established with minimizing the weighted sum of link delay and resource consumption as the optimization goal; The network resource configuration optimization model is solved by the integer programming method to obtain the optimal network resource configuration strategy, and the optimal network resource configuration strategy is deployed to the power 5G network device nodes to optimize the network resource configuration.

2. A method for monitoring the service quality of a power 5G network according to claim 1, characterized in that: The step of associating and binding the signaling plane data with the data plane performance data corresponding to the electric power service flow in the current time window through the quality of service flow identifier and the timestamp to generate a multi-dimensional spatiotemporal feature matrix for the electric power service comprises: When a signaling event is triggered, a quality of service flow identifier is assigned to the quality of service flow corresponding to the signaling event through the session management function; When a change in the quality of service flow identifier is detected, all signaling plane data with the same quality of service flow identifier in the current time window are associated and bound with the data plane performance data to obtain initial associated data; Based on the time synchronization protocol, the initial associated data is timestamp aligned through the synchronization message interaction process of the master and slave power 5G network device nodes to obtain timing aligned associated data; According to the timing alignment associated data, the average link delay between the master and slave power 5G network device nodes is calculated using the message interaction timestamp in the synchronization message interaction process; Calculate the average value of the link delay data of all power 5G network device nodes in the current time window to obtain the delay mean, and calculate the degree of deviation between the link delay data and the delay mean to obtain the delay variance; The average link delay and the delay variance are integrated into a structured feature matrix according to the time step, network device node and feature category dimensions to generate a multidimensional spatiotemporal feature matrix for power business.

3. A method for monitoring the service quality of a power 5G network according to claim 2, characterized in that: The condition for timestamp alignment is that the absolute difference between the timestamp of the data plane performance data and the timestamp of the signaling event trigger is less than a preset time window threshold.

4. A method for monitoring the service quality of a power 5G network according to claim 1, characterized in that: The dynamic causal neural graph construction layer includes a physical layer graph, a cross-connection layer and a logic layer graph; The physical layer graph uses physical network entities in the electric power 5G network as nodes. The physical layer graph is used to dynamically calculate the edge connection weights between physical network entities based on real-time link bandwidth and link delay data, and generate a physical layer topology graph representing the status of physical resources in the electric power 5G network based on the edge connection weights; The logical layer diagram uses logical service entities as nodes, and the logical layer diagram is used to dynamically optimize the logical service flow mapping relationship according to the guaranteed traffic bit rate compliance rate and service priority, to obtain an optimized logical service flow mapping relationship; The cross-connection layer is used to cross-layer associate the physical network entities and logical business entities in the physical layer topology diagram through service quality flow identifiers based on the optimized logical business flow mapping relationship, forming a multi-level network topology structure including physical resources and business logic.

5. A method for monitoring the service quality of a power 5G network according to claim 4, characterized in that: The spatiotemporal graph convolution layer includes a spatial graph convolution network and a temporal graph convolution network connected in parallel; The spatial graph convolutional network is used to aggregate the signaling plane data and data plane performance data of adjacent nodes in the multi-level network topology structure using a graph attention mechanism, and capture the network space dependencies of the multi-level network topology structure in combination with the multi-dimensional spatiotemporal feature matrix to obtain network space structure features; The temporal graph convolutional network is used to model the temporal dependency of the data plane performance data using a gated temporal convolutional network, extract the data plane time features, and fuse the network spatial structure features with the data plane time features to output a joint representation of spatiotemporal features.

6. A method for monitoring the service quality of a power 5G network according to claim 5, characterized in that: The causal reasoning layer is used to determine the predicted causal relationship weights between nodes in a multi-level network topology structure using the Granger causality analysis method, and linearly combine the joint representation of the spatiotemporal features with the predicted causal relationship weights through a fully connected layer to output a network quality prediction score for the power 5G network.

7. A method for monitoring the service quality of a power 5G network according to claim 1, characterized in that: The constraints of the network resource configuration optimization model include guaranteed traffic bit rate constraints and data radio bearer constraints; The guaranteed traffic bit rate constraint is that the actual guaranteed traffic bit rate of each quality of service flow at each time step is not less than a preset minimum traffic bit value; The data radio bearer constraint is that the sum of the total amount of resources allocated to all data radio bearers does not exceed a preset maximum resource allocation allowed value.

8. A method for monitoring the service quality of a power 5G network according to claim 6, characterized in that: The joint loss function of the dynamic spatiotemporal causal graph network is composed of a network quality loss term and a causal loss term through weighted summation; The network quality loss term is obtained by calculating the mean square error between the network quality prediction score and the actual network quality score of the power 5G network; The causal loss term is obtained by calculating the cross entropy between the predicted causal relationship weights and the true causal relationship weights.

9. A power 5G network service quality monitoring system, characterized in that: The system comprises: A first data acquisition module is configured to capture signaling plane data through a signaling plane probe in response to a protocol data unit session establishment request signaling initiated by the power terminal device; The second data acquisition module is configured to encapsulate a flow detection field at the ingress of the segment routing IPv6 tunnel through a data plane probe to obtain data plane performance data; A spatiotemporal feature generation module is configured to associate and bind the signaling plane data with the data plane performance data corresponding to the power service flow in the current time window through a quality of service flow identifier and a timestamp, thereby generating a multi-dimensional spatiotemporal feature matrix for the power service; A network quality prediction module is configured to predict the quality of the power 5G network based on the multi-dimensional spatiotemporal feature matrix using a pre-built dynamic spatiotemporal causal graph network to obtain a network quality prediction score; the dynamic spatiotemporal causal graph network includes a dynamic causal neural graph construction layer, a spatiotemporal graph convolution layer, and a causal reasoning layer connected in sequence; A configuration model building module is used to establish a network resource configuration optimization model with minimizing latency and resource consumption as optimization goals when the network quality prediction score exceeds a preset score threshold; The resource configuration optimization module is used to solve the network resource configuration optimization model through integer programming method, obtain the optimal network resource configuration strategy, and deploy the optimal network resource configuration strategy to the power 5G network device node to optimize the network resource configuration.

10. A power 5G network service quality monitoring system according to claim 9, characterized in that: The spatiotemporal feature generation module is specifically used to: When a signaling event is triggered, a quality of service flow identifier is assigned to the quality of service flow corresponding to the signaling event through the session management function; When a change in the quality of service flow identifier is detected, all signaling plane data with the same quality of service flow identifier in the current time window are associated and bound with the data plane performance data to obtain initial associated data; Based on the time synchronization protocol, the initial associated data is timestamp aligned through the synchronization message interaction process of the master and slave power 5G network device nodes to obtain timing aligned associated data; According to the timing alignment associated data, the average link delay between the master and slave power 5G network device nodes is calculated using the message interaction timestamp in the synchronization message interaction process; Calculate the average value of the link delay data of all power 5G network device nodes in the current time window to obtain the delay mean, and calculate the degree of deviation between the link delay data and the delay mean to obtain the delay variance; The average link delay and the delay variance are integrated into a structured feature matrix according to the time step, network device node and feature category dimensions to generate a multidimensional spatiotemporal feature matrix for power business.

11. A power 5G network service quality monitoring system according to claim 10, characterized in that: The condition for timestamp alignment is that the absolute difference between the timestamp of the data plane performance data and the timestamp of the signaling event trigger is less than a preset time window threshold.

12. The power 5G network service quality monitoring system according to claim 9, characterized in that: The dynamic causal neural graph construction layer includes a physical layer graph, a cross-connection layer and a logic layer graph; The physical layer graph uses physical network entities in the electric power 5G network as nodes. The physical layer graph is used to dynamically calculate the edge connection weights between physical network entities based on real-time link bandwidth and link delay data, and generate a physical layer topology graph representing the status of physical resources in the electric power 5G network based on the edge connection weights; The logical layer diagram uses logical service entities as nodes, and the logical layer diagram is used to dynamically optimize the logical service flow mapping relationship according to the guaranteed traffic bit rate compliance rate and service priority, to obtain an optimized logical service flow mapping relationship; The cross-connection layer is used to cross-layer associate the physical network entities and logical business entities in the physical layer topology diagram through service quality flow identifiers based on the optimized logical business flow mapping relationship, forming a multi-level network topology structure including physical resources and business logic.

13. A power 5G network service quality monitoring system according to claim 12, characterized in that: The spatiotemporal graph convolution layer includes a spatial graph convolution network and a temporal graph convolution network connected in parallel; The spatial graph convolutional network is used to aggregate the signaling plane data and data plane performance data of adjacent nodes in the multi-level network topology structure using a graph attention mechanism, and capture the network space dependencies of the multi-level network topology structure in combination with the multi-dimensional spatiotemporal feature matrix to obtain network space structure features; The temporal graph convolutional network is used to model the temporal dependency of the data plane performance data using a gated temporal convolutional network, extract the data plane time features, and fuse the network spatial structure features with the data plane time features to output a joint representation of spatiotemporal features.

14. A power 5G network service quality monitoring system according to claim 13, characterized in that: The causal reasoning layer is used to determine the predicted causal relationship weights between nodes in a multi-level network topology structure using the Granger causality analysis method, and linearly combine the joint representation of the spatiotemporal features with the predicted causal relationship weights through a fully connected layer to output a network quality prediction score for the power 5G network.

15. The power 5G network service quality monitoring system according to claim 9, characterized in that: The constraints of the network resource configuration optimization model include guaranteed traffic bit rate constraints and data radio bearer constraints; The guaranteed traffic bit rate constraint is that the actual guaranteed traffic bit rate of each quality of service flow at each time step is not less than a preset minimum traffic bit value; The data radio bearer constraint is that the sum of the total amount of resources allocated to all data radio bearers does not exceed a preset maximum resource allocation allowed value.

16. The power 5G network service quality monitoring system according to claim 14, characterized in that: The joint loss function of the dynamic spatiotemporal causal graph network is composed of a network quality loss term and a causal loss term through weighted summation; The network quality loss term is obtained by calculating the mean square error between the network quality prediction score and the actual network quality score of the power 5G network; The causal loss term is obtained by calculating the cross entropy between the predicted causal relationship weights and the true causal relationship weights.

17. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed, the method according to any one of claims 1 to 8 is implemented.

Images