Service network quality analysis control device based on AI

Abstract

A service network quality analysis control device based on AI comprises a network interface module which supports multi-port access and is used for flow acquisition control, a data acquisition and storage module which carries out data preprocessing and tagging and constructs an efficient storage system, and an AI analysis engine which realizes service identification, QOS demand prediction and network trend analysis. The QOS strategy decision module is used for dynamically generating a control strategy according to an AI result, the strategy execution and feedback module is used for completing strategy issuing and forming closed-loop optimization, and the management and visualization module is used for providing network state display, configuration management and alarm. Through AI intelligent analysis and dynamic strategy control, the network resource utilization rate and the operation reliability are remarkably improved, and the method is suitable for intelligent substations and other scenes with high requirements for communication stability and real-time performance; good application values and popularization prospects are realized.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of network service quality management, and in particular to an AI-based service network quality analysis and control device. BACKGROUND

[0002] The main problem faced by current intelligent substation networks is the lack of fine-grained management of the importance of different services, leading to unreasonable allocation of network resources and affecting the operation of key services. There are various types of services in substations, including real-time monitoring, control signal transmission, etc. These services have different requirements for network bandwidth, latency and reliability, but current network devices are unable to effectively distinguish between these services when processing them. Various types of services share the same network bandwidth, and in particular under high load conditions, non-critical services may seize the network resources of important services, leading to problems such as insufficient bandwidth and excessive latency for critical services, which seriously affect the stable operation of the power grid.

[0003] In actual application scenarios, control-type services in substations require extremely low latency and high reliability, and any delay or packet loss can cause serious system failure. While non-real-time data acquisition has requirements for network bandwidth, it is less sensitive to latency. However, there is currently no effective mechanism to dynamically allocate network resources according to priority for these services, leading to fluctuations in network performance and waste of resources.

[0004] In existing technologies, some network devices already have basic QOS policy management capabilities, enabling static classification of traffic and simple priority control. However, these methods mostly rely on fixed configuration rules, lack deep understanding of service content and adaptive adjustment capabilities, and are difficult to cope with the complex environment of diverse service types and dynamic network state changes in substations. In actual applications, traditional QOS policies are usually based on manually configured port, IP or protocol rules, and are unable to automatically optimize resource allocation based on real-time service behavior and network load, leading to potential performance degradation of critical services in emergency situations.

[0005] In addition, existing systems lack intelligent analysis and prediction capabilities for the running state of power services, and are unable to perform trend modeling or anomaly identification based on historical data, limiting the forward-looking and stability of network policies. At the same time, traditional network monitoring methods do not form an effective policy closed-loop feedback mechanism, resulting in the inability to use the execution effect to optimize subsequent policies, and policy adjustments lagging behind actual network changes, affecting overall response efficiency. SUMMARY

[0006] In view of the technical problems of static resource configuration, lagging policy response and insufficient intelligence level in the prior art, the technical scheme provides an AI-based service network quality analysis control device, provides an artificial intelligence-based service identification and network state analysis mechanism, realizes intelligent identification of key services, dynamic judgment of network trends and adaptive optimization control of QOS parameters, and effectively solves the above problems.

[0007] The application is implemented by the following technical scheme:

[0008] An AI-based service network quality analysis control device comprises:

[0009] A network interface module is connected with an intelligent network monitoring terminal through multiple network interfaces, collects network monitoring data in real time, collects and transmits control of various service traffics in a substation in real time, and transmits raw data to a data collection and storage module.

[0010] The data collection and storage module pre-processes, filters and labels the collected network data, and establishes a historical data storage system comprising a solid-state storage and a cache mechanism.

[0011] An AI analysis engine receives structured data, identifies service types, judges network state trends and predicts QOS requirements based on a trained machine learning model.

[0012] A QOS policy decision module dynamically formulates QOS parameters of bandwidth, priority, delay control and packet loss according to AI analysis results and current network state, and generates a control policy.

[0013] A policy execution and feedback module issues the control policy to a network interface layer or an external network device, and feeds back execution effects to the AI analysis engine, so as to realize policy closed-loop optimization.

[0014] A management and visualization module runs through the whole system process, provides a Web interface, and is used for displaying data, network state, service classification and QOS policy changes, and supports policy configuration and fault alarm, so as to improve operation and maintenance efficiency and the intelligence level of network management.

[0015] Further, the intelligent network monitoring terminal is deployed in a substation and a power supply station, is connected with a policy control center and an operation and maintenance management platform through a network, a service identification model and a network quality policy template are preset in the operation and maintenance management platform, the identification and control tasks are issued to the intelligent network monitoring terminal through the policy control center, the intelligent terminal accesses various service links in the substation through multiple network interfaces, and network data of various service links is collected in real time.

[0016] Further, the network interface module is provided with a plurality of gigabit physical network interfaces, supports mirror traffic access and bidirectional data transmission, has multi-link simultaneous access capability, and can realize concurrent collection and control of multiple types of business flows such as dispatching, monitoring, management and the like in the substation; meanwhile, supports VLAN, port isolation and rate limiting functions, and guarantees the security and stability of different business access.

[0017] The network interface module performs concurrent access and shunting processing on multiple types of business flows in the substation, and performs real-time collection, identification and mirror forwarding on the traffic of different priority businesses; the network interface module has link state detection function, can monitor the port connection state, bandwidth utilization and traffic change, and sends the original data packet into the data collection and storage module according to the preset rule as the input basis for subsequent AI analysis and strategy decision, guarantees the comprehensiveness and real-time performance of system perception.

[0018] Further, the data collection and storage module is responsible for preprocessing, filtering, feature extraction and labeling operation on the original business flow data collected by the network interface module, and classifies and stores the processed data according to the information of business type, time stamp, source / destination address, and stores the processed data in a labeled structure to provide standardized input for the AI analysis engine.

[0019] The data collection and storage module is built-in with high-speed solid-state storage unit and cache mechanism, supports high-frequency data writing and batch reading, guarantees the data processing capability of the system under high concurrency condition; meanwhile, establishes historical data storage system, provides training data support for the AI analysis engine, and realizes the record and backtracking of strategy execution effect, provides data basis for subsequent optimization and debugging.

[0020] Has the capability of real-time preprocessing, cleaning, deduplication and feature extraction on network traffic, can store the processed data in a labeled structure to provide standardized input for the AI analysis engine; meanwhile, establishes local historical data index system, realizes long-term tracking of behavior mode of key business flow and backtracking of strategy effect.

[0021] Further, the AI analysis engine deploys a model trained based on a deep learning or machine learning algorithm to intelligently analyze the preprocessed network traffic data, and automatically identifies the service flow based on the five-tuple source IP, destination IP, source port, destination port, and protocol as the basic anchor point. The algorithm is an intelligent sensing and prediction algorithm combining multi-source network data features and service behavior patterns, and its core lies in achieving high-precision identification of service types, trend modeling of network status, and dynamic deduction of service quality requirements through deep learning of network traffic in time and behavior dimensions; the algorithm first extracts multi-dimensional indexes of protocol features, transmission rate, and interaction frequency from the packet level, constructs a feature vector for clustering analysis and pattern classification to identify the service type; then uses time series modeling technology to track network link load, delay change, and congestion trend to predict the network status in the near future; finally, combined with the service features and network prediction results, a strategy mapping mechanism is used to generate real-time QOS control parameters. The entire analysis process uses an iterative optimization architecture, and the model is calibrated by introducing execution feedback to build an adaptive QOS guarantee mechanism for substation business scenarios.

[0022] Further, the trained model combines the bandwidth utilization rate, packet delay, and packet loss rate of the service flow within a time window T into a feature vector x t , obtains the service type identification result:

[0023]

[0024] wherein, is the input feature, , is the model parameter, outputs the probability of each service type, and finally obtains the service category .

[0025] After the service classification is completed, the engine further uses an LSTM time series prediction model to predict the network status:

[0026]

[0027] wherein, represents the predicted next time QOS index such as delay, bandwidth, or packet loss rate, and the input is the historical feature sequence;

[0028] Based on the identified service category and the predicted QOS requirement , the system dynamically generates strategy parameters for bandwidth, priority, and delay, and delivers them to the strategy execution module to realize closed-loop adaptive optimization of network quality.

[0029] Further, the running steps of the service flow automatic identification include:

[0030] Step A: capturing the data packets flowing through the network in real time by the network interface module, and preliminarily classifying them according to the five-tuple information (source address, destination address, source port, destination port, and protocol type);

[0031] Step B: the data acquisition and storage module performs cleaning, deduplication, and labeling processing on the captured data, and extracts key features such as traffic size, inter-packet delay, connection duration, etc., to generate a standardized feature vector;

[0032] Step C: the AI analysis engine receives the standardized feature vector and inputs it into the pre-trained machine learning model for inference, automatically identifying the type of the data packet, such as scheduling control service, monitoring video service, or management service, etc.

[0033] Step D: the identification result and the current network state data are sent to the QOS policy decision module for dynamic generation of subsequent QOS policies;

[0034] Step E: the identification result and the execution policy feedback information are recorded as samples in the data acquisition and storage module for continuous model training and identification accuracy improvement.

[0035] Further, the QOS policy decision module receives the service identification result and network state evaluation information output from the AI analysis engine, combines the current network topology structure, link resource usage, and pre-set business priority policy, and dynamically generates QOS control parameters including bandwidth allocation, priority scheduling, delay control, and packet loss limitation. The QOS policy decision module has the ability of adaptive adjustment of the policy, which can optimize the control policy in real time when the network state changes or the key service appears congestion risk, ensure that the high-priority service can still be guaranteed under the condition of limited network resources, and issue policy instructions to the policy execution and feedback module through the interface to realize dynamic and fine management of network quality.

[0036] Further, the policy execution and feedback module receives the control instructions issued by the QOS policy decision module, and issues the corresponding QOS parameters to the network interface module or external switches, routers, and other network devices for bandwidth control, priority scheduling, delay guarantee, and packet loss limitation. The policy execution and feedback module also has real-time monitoring and feedback functions, continuously tracks the policy execution effect such as traffic changes, key business performance indicators, and abnormal states, and returns the monitoring results to the AI analysis engine for optimization of model inference and policy adjustment, realizing a closed-loop adaptive control mechanism for network quality management.

[0037] Further, the management and visualization module provides a web-based graphical operation interface to display network operation status, various business traffic distribution, QOS policy change and AI analysis results, and other key indicators; the module supports policy configuration, device management, user permission control and fault alarm function, and can realize real-time visualization of policy issuing, execution effect and network anomaly; at the same time, it has the ability of log recording and historical data backtracking, which is convenient for operation and maintenance personnel to analyze trends, troubleshoot and optimize policies, improve the transparency and operability of network management, and enhance the overall intelligent operation and maintenance capability of the system.

[0038] Advantages

[0039] The AI-based business network quality analysis and control device provided by the application has the following advantages compared with the prior art:

[0040] (1) The technical solution integrates a network interface module, a data acquisition and storage module, an AI analysis engine, a QOS policy decision module, a policy execution and feedback module, and a management and visualization module, realizes real-time monitoring, dynamic optimization and intelligent management of various business traffics in the intelligent substation, improves the fine control level of network quality, enhances the network stability of substations and power supply stations, and ensures the efficient operation of key businesses.

[0041] (2) The application uses AI algorithm to intelligently identify and predict network traffic, which can analyze business demand in real time and automatically adjust QOS policy. Compared with the traditional manual scheduling method, the application has higher accuracy and efficiency, reduces the need for manual intervention, improves the dynamic allocation capability of network resources, and guarantees the priority demand of various businesses in the substation, especially for time delay sensitive businesses such as protection control and video monitoring, which has stronger network quality guarantee capability.

[0042] (3) The AI analysis engine of the application supports edge deployment, which can quickly analyze data and adjust policies on site, reduces the dependence on central servers, and improves the system response speed and stability. Compared with the traditional centralized computing method, edge deployment not only optimizes resource utilization, but also improves the processing capacity of devices, and reduces the delay.

[0043] (4) The application realizes the closed-loop adaptive optimization of network management and control through modular design, which makes the collaborative work between modules more efficient. The management and visualization module provides real-time data display and alarm function, helps operation and maintenance personnel to quickly master network status, timely adjust policy, and trace historical data for analysis, which improves the convenience and operability of operation and maintenance management.

[0044] (5) The strategy execution and feedback module of the present invention has a real-time feedback mechanism, which can quickly respond to changes in network status and fluctuations in business demand, ensure that the QoS strategy is effectively executed, and optimize the prediction model of the AI ​​analysis engine through continuous feedback, making network quality management more intelligent and refined, and improving the adaptability and flexibility of the entire system. Attached Figure Description

[0045] Fig. 1 This is a schematic diagram of the overall architecture of the present invention.

[0046] Fig. 2 This is the internal circuit diagram of the present invention.

[0047] Fig. 3 This is a connection diagram of the device of the present invention. Detailed Implementation

[0048] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. The described embodiments are merely some embodiments of the present invention, and not all embodiments. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the design concept of the present invention should fall within the protection scope of the present invention.

[0049] Example 1:

[0050] like Figs. 1-2 As shown, an AI-based business network quality analysis and control device includes: a network interface module, a data acquisition and storage module, an AI analysis engine, a QoS policy decision-making module, a policy execution and feedback module, and a management and visualization module.

[0051] Network interface module: Connects to the intelligent network monitoring terminal through a multi-port network interface, collects network monitoring data in real time, performs real-time collection and transmission control of various business traffic in the substation, and transmits the raw data to the data acquisition and storage module.

[0052] As the core data communication component of the portable test terminal, the network interface module is equipped with two Gigabit Ethernet interfaces, one for connecting the network under test and the other for connecting the test benchmark server / application server. It supports DHCP and static IP configuration modes, automatically adapting to different types of LAN or WAN environments. During network quality testing, the network interface module is responsible for accurately and losslessly sending the generated test traffic to the target network and transmitting the received return traffic data to the processor for subsequent analysis. It features automatic link status detection, real-time interface traffic monitoring, and abnormal link disconnection alerts, ensuring stable and reliable network connections and providing a high-quality data path for the entire test system.

[0053] Data collection and storage module: preprocess, filter and label the collected network data, and establish a historical data storage system containing solid-state storage and cache mechanism.

[0054] The data collection and storage module is deployed inside the test reference server and is responsible for receiving and storing network test data from portable test terminals and other monitoring devices; real-time collection of link state parameters such as end-to-end delay, bandwidth utilization, link packet loss rate, packet loss position, available bandwidth and other key performance indicators; built-in cache mechanism can temporarily save collected data locally when the network is unstable to avoid data loss; the collected data is first processed by the preprocessing program for verification, deduplication and formatting before being written into the database; at the same time, the module works with the AI analysis engine to synchronize part of the feature data to the model analysis interface; supports multi-thread concurrent processing to improve collection efficiency and is suitable for large-scale terminal synchronous testing scenarios.

[0055] AI analysis engine: receives structured data, identifies business types, judges network state trends and predicts QOS requirements based on trained machine learning models.

[0056] The AI analysis engine provides the device with core intelligent analysis capabilities, automatically extracts features, classifies and clusters based on collected network index data, identifies key business flow features and abnormal traffic; the engine integrates a lightweight deep neural network model that can run locally on edge devices to quickly complete traffic pattern recognition, QOS demand prediction and congestion trend evaluation; based on real-time and historical test data, the AI engine automatically identifies network performance bottlenecks, determines whether bandwidth, packet loss or delay causes quality degradation, and pushes optimization recommendations to the policy decision module; with self-learning function, continuously optimizes the analysis model as the data volume increases, improving the accuracy of network anomaly identification and early warning.

[0057] QOS policy decision module: dynamically formulates QOS parameters such as bandwidth, priority, delay control and packet loss based on AI analysis results and current network state to generate control strategies.

[0058] The QOS policy decision module receives prediction results and optimization suggestions from the AI analysis engine, combines pre-set policy templates and current network state to dynamically generate QOS configuration instructions that adapt to different business types (such as voice, video, control instructions, etc.); supports multi-dimensional strategy generation, including priority adjustment, bandwidth reservation, forwarding path optimization, etc.; has rule learning and strategy evolution function, which can continuously adjust and improve control strategies according to test results; after the strategy is generated, it is immediately pushed to the policy execution and feedback module to complete the specific strategy landing and validation process, ensuring that network resources are reasonably scheduled and allocated during business peak periods.

[0059] Policy execution and feedback module: The control policy is issued to the network interface layer or external network equipment, and the execution effect is fed back to the AI analysis engine to realize the closed-loop optimization of the policy.

[0060] The policy execution and feedback module is deployed in the middle layer between the edge test terminal and the policy control server, receives and analyzes the control commands issued by the QOS policy decision module, calls the terminal local system control interface or switch device interface to realize policy execution, such as adjusting bandwidth limit, port priority, VLAN label, etc.; The module monitors the network state changes after policy execution in real time and returns feedback data, which is used to judge the policy effect and whether further adjustment is needed; It has a policy conflict detection and rollback mechanism. Once it is found that the implementation of the policy leads to a decrease in network performance, the default state can be quickly restored to ensure the continuity and stability of critical business; It supports linkage with mainstream SDN controllers to expand policy execution capabilities.

[0061] Management and visualization module: Throughout the whole process of the system, a Web interface is provided to show data, network status, business classification and QOS policy changes; and it supports policy configuration and fault alarm to improve operation and maintenance efficiency and the intelligent level of network management.

[0062] The management and visualization module is deployed in the application server to provide a graphical operation interface and state monitoring screen for operation and maintenance personnel; It supports task configuration, device state monitoring, policy management, log query, and abnormality early warning functions; Through real-time visual charts, it displays core indicators such as network delay, packet loss, available bandwidth, and QOS policy execution of each terminal; The module interacts with the database to query historical test records and device operation logs, and generates network quality analysis reports in PDF format that can be exported; It integrates a multi-role permission management system to distinguish the function permissions of operation and maintenance personnel, management personnel, and ordinary viewing users; It supports remote access and Web management, and is suitable for centralized management needs of a large number of variable substations.

[0063] The automatic identification task of business flow runs, and the running steps include:

[0064] Step A: The network interface module captures the data packets flowing through the network in real time, and classifies them according to the five-tuple information (source address, destination address, source port, destination port, protocol type);

[0065] Step B: The data acquisition and storage module cleans, de-duplicates, and labels the captured data, and extracts key features such as traffic size, inter-packet delay, connection duration, etc., to generate standardized feature vectors;

[0066] Step C: The AI analysis engine receives the standardized feature vector and inputs it into a pre-trained machine learning model for inference, automatically identifying the type of service to which the data packet belongs, such as dispatch control services, monitoring video services, or management services, etc.

[0067] Step D: The identification results are sent together with the current network state data to the QOS policy decision module for the dynamic generation of subsequent QOS policies.

[0068] Step E: The identification results and execution strategy feedback information are recorded as samples in the data collection and storage module for continuous model training and identification accuracy improvement.

[0069] As shown in Fig. 3 , the intelligent network monitoring terminal is deployed in substations and power supply stations and connected with the policy control center and the operation and maintenance platform through the network. The operation and maintenance platform is preconfigured with a business identification model and a network quality policy template, and the identification and control tasks are issued to the intelligent network monitoring terminal through the policy control center. The intelligent terminal accesses various business links in the substation through multiple network interfaces and collects network data of various business links in real time.

[0070] The intelligent network monitoring terminal serves as an edge device for data collection and preliminary processing. The terminal includes an embedded processor unit, a solid-state storage unit, two gigabit Ethernet interfaces, a power interface, a network interface, and a remote management communication module. The processor unit uses a low-power, high-performance processor that can efficiently run AI models and network analysis algorithms to ensure real-time processing of network data and generation of optimization suggestions. The network interface is equipped with two Ethernet interfaces that support multiple network environment access and can flexibly schedule between different business streams to ensure real-time data transmission. The solid-state storage unit is equipped with an SSD solid-state storage unit to ensure fast data reading and storage, and the monitoring terminal stores key business data and test results in real time. The power interface and communication module provide stable power through the power system, and the remote management communication module supports real-time data exchange with the remote operation and maintenance platform.

[0071] The policy control center is deployed in the core server and is responsible for analyzing, predicting, and optimizing network traffic based on real-time data collected from the intelligent network monitoring terminal and AI analysis engine, and generating specific QOS control strategies. The data collection module collects real-time data from the intelligent terminal, including network performance indicators such as bandwidth, latency, and packet loss rate, and transmits the data to the policy control center. The AI analysis engine automatically identifies and predicts real-time traffic based on a deep learning model, generating QOS parameters such as service priority, bandwidth allocation, and latency control.

[0072] The operation and maintenance management platform provides a unified monitoring and management interface for users, displays network traffic status, real-time QOS execution of various businesses, alarm information and historical data through a Web interface, and provides decision support for operation and maintenance personnel. Visualization function: support real-time chart display of network traffic and QOS policy execution effect, help operation and maintenance personnel quickly identify potential problems and adjust the policy. Fault alarm and policy adjustment: support custom fault alarm and policy adjustment function, facilitate timely response to network anomalies and optimize network configuration.

[0073] In order to verify the feasibility and superiority of the present scheme, the inventors provide specific practical operation cases, and the specific case content is:

[0074] Case 1:

[0075] System deployment

[0076] Device preparation: deploy a portable test terminal in a certain 110kV substation, ensure that it accesses the business data network and management network of the substation through the dual network port, use UPS to ensure stable power supply, and set the device to 7x24 hour operation state.

[0077] Remote configuration: the test reference server and the application server are deployed in the city company data center, and a communication channel is established with the portable test terminal through a secure VPN link. Data channel encrypted transmission ensures the security of test data.

[0078] Environment building: the operation and maintenance personnel log in to the application server system console, create a test task, and configure the test parameters as follows: packet sending rate 300Mbps, packet size 512KB, test time 10 minutes, target for evaluating the stability and QOS support capability of the business link of a remote site.

[0079] After the task setting is completed, the test task is issued to the portable test terminal through the system task scheduling module, and the AI analysis strategy template is set synchronously to enable intelligent decision analysis.

[0080] Test execution

[0081] Test start: the operation and maintenance personnel click "one-key start" through the application server Web interface, the portable test terminal automatically connects to the reference server, and starts packet sending test. The test traffic is sent to the remote target node along the business path, and the RTT, bandwidth utilization rate, packet loss rate and other key performance indicators are recorded.

[0082] Data collection: the data collection module of the reference server collects network state data generated by the test traffic in real time, and automatically archives it into the database; at the same time, the temperature sensor of the portable terminal monitors the device environment temperature and synchronously uploads, preventing measurement error caused by overheating environment.

[0083] III. Data Analysis and Feedback

[0084] AI Analysis: During the test, the AI analysis engine learns and detects anomalies from the real-time collected bandwidth and latency data, identifying network congestion points. The system determines that the current service delay peak is 175ms, exceeding the maximum 150ms threshold allowed for VoIP services.

[0085] Dynamic Adjustment: The data analysis module returns the anomaly determination result to the application server, and the system automatically issues a new policy to reduce the packet sending rate to 250Mbps, reducing congestion impact and retesting until the delay value is reduced to 144ms, meeting the call quality standard.

[0086] Final Analysis Module Output: The "Link QOS Capability Report" contains the maximum available bandwidth prediction value of the current link as 862Mbps, and the RTT fluctuation range as ±12ms, and marks in the report that there is a high peak congestion risk on the current link.

[0087] IV. Operation and Maintenance Decision and Troubleshooting

[0088] View Results: The operation and maintenance personnel view the test details and AI analysis results through the application server and find that the RTT in the path from the substation to a certain business data center has shown an abnormal upward trend in the recent three tests.

[0089] Problem Identification: The operation and maintenance personnel confirm that no new business load has been added between the substation and the data center by comparing historical data, suspecting abnormal traffic. Subsequently, they log in to the switch system and find that a non-legal business VLAN port occupies a large amount of bandwidth, suspected of unauthorized access.

[0090] Remote Processing: Through the policy execution module, remote policy instructions are issued to isolate the unauthorized VLAN port and enable ACL to limit abnormal traffic.

[0091] Re-test Confirmation: Re-execute the test task, the network delay returns to the baseline value of 145ms, the link available bandwidth increases to 880Mbps, the AI analysis engine judges that the risk is removed, and the status evaluation is "green and stable".

[0092] Through this actual operation case, the system realizes the whole process application from test task creation, traffic sending, data collection, intelligent analysis, policy optimization, fault identification to repair closed loop, not only improves the network quality guarantee capability of the substation, but also provides a replicable reference for subsequent deployment of other sites.

Claims

1. An AI-based business network quality analysis and control device, characterized in that: include: Network interface module: Connects to the intelligent network monitoring terminal through a multi-port network interface, collects network monitoring data in real time, performs real-time collection and transmission control of various business traffic in the substation, and transmits the raw data to the data acquisition and storage module; Data acquisition and storage module: preprocesses, filters, and tags the acquired network data, and establishes a historical data storage system that includes solid-state storage and caching mechanisms; AI analytics engine: Receives structured data, and based on a trained machine learning model, identifies business types, judges network status trends, and predicts QoS requirements. QoS policy decision module: Dynamically formulates QoS parameters such as bandwidth, priority, latency control, and packet loss based on AI analysis results and current network status, and generates control policies; The strategy execution and feedback module distributes control policies to the network interface layer or external network devices and feeds back the execution results to the AI ​​analysis engine to achieve closed-loop optimization of the policies. Management and visualization module: It runs through the entire system process, providing a web interface to display data, network status, service categories and QoS policy changes; and supports policy configuration and fault alarms to improve operation and maintenance efficiency and network management intelligence.

2. The AI-based service network quality analysis and control device according to claim 1, characterized in that: The intelligent network monitoring terminal is deployed in substations and power supply stations, and is connected to the strategy control center and operation and maintenance management platform through the network. The operation and maintenance management platform presets a service identification model and a network quality strategy template, and the strategy control center sends the identification and control tasks to the intelligent network monitoring terminal. The intelligent terminal accesses various business links within the substation through multiple network interfaces and collects network data of these business links in real time.

3. The AI-based service network quality analysis and control device according to claim 2, characterized in that: The network interface module is equipped with multiple gigabit-level physical network interfaces, supporting mirrored traffic access and bidirectional data transmission. It performs concurrent access and traffic distribution processing for various types of business traffic within the substation, and collects, identifies, and mirrors traffic for different priority services in real time. The network interface module has a link status detection function, which can monitor port connection status, bandwidth utilization, and traffic changes, and send the raw data packets to the data acquisition and storage module according to preset rules as the input basis for subsequent AI analysis and strategy decision-making.

4. The AI-based service network quality analysis and control device according to claim 1, characterized in that: The data acquisition and storage module is responsible for preprocessing, filtering, feature extraction, and tagging the raw business traffic data collected by the network interface module, and classifying and storing it according to business type, timestamp, and source / destination address information; the data acquisition and storage module has a built-in high-speed solid-state storage unit and caching mechanism.

5. The AI-based service network quality analysis and control device according to claim 4, characterized in that: The AI ​​analysis engine is deployed based on models trained using deep learning or machine learning algorithms. It intelligently analyzes preprocessed network traffic data, using the source IP, destination IP, source port, destination port, and protocol of a five-tuple as basic anchor points to automatically identify service flows. The algorithm is an intelligent perception and prediction algorithm that combines multi-source network data features with service behavior patterns. Its core lies in achieving high-precision identification of service types, trend modeling of network status, and dynamic inference of service quality requirements through deep learning of network traffic in time and behavior dimensions. The algorithm first extracts multi-dimensional indicators such as protocol features, transmission rate, and interaction frequency from the data packet level, constructs feature vectors for cluster analysis and pattern classification to identify service types. Then, it uses time series modeling technology to track network link load, latency changes, and congestion trends to predict network status trends in the near future. Finally, by combining business characteristics and network prediction results, a policy mapping mechanism is used to generate real-time QoS control parameters.

6. The AI-based service network quality analysis and control device according to claim 5, characterized in that: The trained model combines the features of bandwidth utilization, inter-packet latency, and packet loss rate of the service flow within the time window T into a feature vector x. t ; The business type identification result is obtained: ； in, As input features, , For model parameters, Output the probability of each business type to obtain the category to which the business belongs. ; After the business classification is completed, the engine further uses the LSTM time series prediction model to predict the network state: ； in, This represents the predicted QoS metric for the next time step, such as latency, bandwidth, or packet loss rate. The input is a historical feature sequence. Based on the identified business category and predicted QoS requirements The system dynamically generates policy parameters for bandwidth, priority, and latency, and sends them to the policy execution module to achieve closed-loop adaptive optimization of network quality.

7. The AI-based service network quality analysis and control device according to claim 5, characterized in that: The steps for automatic business flow identification include: Step A: The network interface module captures data packets flowing through the network in real time and performs preliminary classification based on the five-tuple information (source address, destination address, source port, destination port, protocol type); Step B: The data acquisition and storage module cleans, deduplicatizes, and tags the captured data, and extracts key features such as traffic volume, packet latency, and connection duration to generate standardized feature vectors. Step C: The AI ​​analysis engine receives standardized feature vectors and inputs them into a pre-trained machine learning model for inference, automatically identifying the business type of the data packet, such as scheduling control business, surveillance video business, or management business, etc. Step D: Send the identification results and current network status data into the QoS policy decision module for dynamic generation of subsequent QoS policies; Step E: The recognition results and the feedback information from the execution strategy are recorded as samples in the data acquisition and storage module for continuous model training and improvement of recognition accuracy.

8. The AI-based service network quality analysis and control device according to claim 1, characterized in that: The QoS policy decision module receives service identification results and network status assessment information from the AI ​​analysis engine, and dynamically generates QoS control parameters, including bandwidth allocation, priority scheduling, latency control, and packet loss limitation, by combining the current network topology, link resource usage, and preset service priority policies.

9. The AI-based service network quality analysis and control device according to claim 5, characterized in that: The policy execution and feedback module receives control commands from the QoS policy decision module and sends corresponding QoS parameters to the network interface module or external network devices such as switches and routers. It performs bandwidth control, priority scheduling, latency protection, and packet loss limitation for different service flows. The policy execution and feedback module also has real-time monitoring and feedback functions, continuously tracking the policy execution effect, such as traffic changes, key service performance indicators, and abnormal states. It then sends the monitoring results back to the AI ​​analysis engine to optimize model inference and policy adjustment, thereby realizing a closed-loop adaptive control mechanism for network quality management.

10. The AI-based service network quality analysis and control device according to claim 5, characterized in that: The management and visualization module provides a web-based graphical user interface that displays key indicators such as network operating status, traffic distribution of various services, QoS policy changes, and AI analysis results. This module supports policy configuration, device management, user access control, and fault alarm functions, and can provide real-time visualization of policy issuance, execution effects, and network anomalies. It also features log recording and historical data backtracking capabilities, facilitating trend analysis, fault diagnosis, and policy optimization by operations and maintenance personnel, thereby improving the transparency and operability of network management and enhancing the overall intelligent operation and maintenance capabilities of the system.

Images