Television transmission network management system and operation and maintenance method thereof, electronic device and storage medium

Abstract

The application relates to a television transmission network management system and a maintenance method thereof, electronic equipment and a storage medium, and relates to the technical field of communication network management. The system comprises a data acquisition module, which is used for collecting multi-source heterogeneous data in a television transmission network in real time; a data processing module, which is used for performing space-time alignment, data management and feature engineering processing on the collected data to form network state data flow; an intelligent analysis module, which is used for performing collaborative analysis to generate at least one analysis result from predictive alarms, fault root cause positioning results, resource optimization strategies and quantitative correlation between network performance and user experience; and an instruction execution module, which is used for generating network control instructions based on the analysis results, and issuing the network control instructions to target equipment after security verification to realize predictive maintenance, fault repair and resource dynamic scheduling of the television transmission network. The application can significantly improve network reliability, resource utilization efficiency and user experience quality.

Description

Technical Field

[0001] This invention relates to the field of communication network management technology, and in particular to a television transmission network management system and its operation and maintenance method, electronic equipment and storage medium. Background Technology

[0002] With the deepening of the convergence of telecommunications, broadcasting, and the internet, and the rapid development of video services such as ultra-high-definition (4K / 8K), interactive television, and streaming media, modern television transmission networks have evolved into an extremely complex and heterogeneous mega-system. This system integrates multiple transmission media, including satellite, fiber optic cables, and internet distribution, as well as a large number of network elements such as encoders, multiplexers, switches, and optical transceivers. This dramatic increase in scale and complexity poses unprecedented challenges to the capabilities of network management systems.

[0003] Currently, television transmission network management systems generally adopt a passive operation and maintenance model of "monitoring-alarm-manual handling". Specifically, firstly, performance and status data are collected from the devices through protocols such as Simple Network Management Protocol (SNMP) and User Terminal Equipment Wide Area Network Management Protocol (TR069); secondly, the administrator sets fixed thresholds for various parameters, and the system generates an alarm when the data exceeds the limit; finally, the operation and maintenance personnel rely on their personal experience, combined with alarms and topology diagrams, to manually analyze and locate faults, and manually perform recovery operations such as switching and configuration.

[0004] The aforementioned traditional passive operation and maintenance model exposes many shortcomings when facing modern television transmission networks: 1) Alarm storms and false alarms / missed alarms: A single root cause fault can trigger a chain of alarms, drowning out the real problem; fixed thresholds cannot adapt to dynamic network changes, resulting in a large number of invalid alarms or missed alarms for slow degradation; 2) Low efficiency in fault location: The troubleshooting process heavily relies on the personal experience of experts, with long average repair times, and experience is difficult to solidify and pass on; 3) Passive response rather than proactive prevention: The system only alarms after a fault occurs, and cannot predict and intervene before service interruption. For critical services such as live streaming, post-incident remediation is often too late; 4) Rigid resource scheduling: Bandwidth, routing, and other configuration strategies are mostly static and cannot be dynamically optimized according to real-time service load and network conditions, resulting in low resource utilization or service quality degradation during peak hours; 5) Data silos: Data from subsystems such as transmission network management, monitoring systems, and bitstream analysis are independent, lacking correlation analysis, making it difficult to directly link network signal degradation with the user's actual viewing experience (QoE) problem. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a television transmission network management system and its operation and maintenance method, electronic equipment and storage medium.

[0006] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: In a first aspect, the present invention provides a television transmission network management system, which adopts the following technical solution: A television transmission network management system, comprising: The data acquisition module is used to acquire multi-source heterogeneous data in the television transmission network in real time; The data processing module is used to perform spatiotemporal alignment, data governance, and feature engineering on the collected multi-source heterogeneous data to form a network state data stream. The intelligent analysis module is used to perform collaborative analysis on the network status data stream through multiple interconnected artificial intelligence models, thereby generating at least one of the following analysis results: predictive alarms, fault root cause location results, resource optimization strategies, and quantitative correlation between network performance and user experience. The instruction execution module is used to generate network control instructions based on the analysis results and send the network control instructions to the target device in the television transmission network to drive the target device to perform corresponding operations, thereby realizing at least one of predictive maintenance, fault repair and dynamic resource scheduling of the television transmission network; wherein at least some of the network control instructions are security verified before being sent.

[0007] The beneficial effects of this invention are as follows: The data acquisition module collects multi-source heterogeneous data in the television transmission network in real time; the data processing module performs spatiotemporal alignment, data governance, and feature engineering on the collected multi-source heterogeneous data to form a high-quality, correlated network status data stream, realizing deep integration and unified governance of cross-domain data; the intelligent analysis module uses multiple interconnected artificial intelligence models to perform collaborative analysis on the network status data stream, realizing accurate early warning of network faults, automated location of fault root causes, dynamic optimization and scheduling of network resources, and quantitative correlation between network performance and user experience, thereby transforming the traditional passive response and experience-driven operation and maintenance mode into an active prevention and intelligent decision-making operation and maintenance mode; the instruction execution module generates network control instructions based on the analysis results, and after verifying the instructions that need to be verified (such as critical or high-risk instructions), it issues them together with the instructions that do not need to be verified, thereby achieving at least one of predictive maintenance, rapid fault repair, and dynamic resource scheduling for the television transmission network while ensuring operational safety. Through the closed-loop collaboration of the aforementioned modules, the system can proactively identify and address potential risks, significantly reducing unplanned service interruptions. Its intelligent scheduling strategy enables on-demand, real-time allocation of network resources, avoiding resource idleness or congestion under static configuration. Simultaneously, user-experience-oriented quantitative correlation and root cause analysis ensure that any optimization and repair actions directly and accurately target user-perceived quality shortcomings. Ultimately, the synergistic effect of these technologies significantly improves the operational reliability, resource utilization efficiency, and user experience quality of the television transmission network.

[0008] Based on the above technical solution, the present invention can be further improved as follows.

[0009] Furthermore, the system also includes: The model optimization module is used to collect the full-link data corresponding to each operation and maintenance event, from collaborative analysis to network control command execution, after each operation and maintenance event is completed, and to iteratively optimize the multiple artificial intelligence models based on the full-link data. The operation and maintenance event refers to a complete network control operation process triggered by the intelligent analysis module and completed by the instruction execution module.

[0010] The beneficial effects of adopting the above-mentioned further solutions are: by collecting full-link data after each operation and maintenance event and using iterative training on multiple artificial intelligence models, the entire system has the ability to continuously self-optimize based on real operation feedback, thereby continuously adapting to changes in the network environment, learning new operation and maintenance scenarios, and autonomously improving its decision-making accuracy and adaptability in prediction, diagnosis, scheduling, and correlation analysis.

[0011] Furthermore, the data acquisition module includes at least two of the following units: The network data acquisition unit is used to collect operating status data, performance index data, and configuration parameter data of various network element devices in the television transmission network. The service data acquisition unit is used to capture the service data stream flowing through the television transmission network by probes deployed at multiple key nodes of the television transmission network, and to collect network transport layer indicators and multi-level error monitoring information conforming to the TR101-290 standard from the service data stream through deep packet inspection or dedicated protocol analysis. The user experience acquisition unit is used to collect at least one type of user experience quality data from the logs of the technical monitoring equipment and in conjunction with the quality evaluation of the bitstream analyzer, including signal loss, cyclic redundancy check error, black screen, still frame, and mute. An external data acquisition unit is used to connect to an external data source through an external data interface to acquire external data related to the operation of the television transmission network.

[0012] The beneficial effects of adopting the above-mentioned further solutions are: by using at least two of the network data acquisition units, service data acquisition units, user experience acquisition units, and external data acquisition units, it is possible to achieve comprehensive and multi-dimensional unified acquisition of multi-source heterogeneous data of the television transmission network, which can break down data silos and lay the foundation for building a complete and interconnected network data view.

[0013] Furthermore, the data processing module includes: The spatiotemporal alignment unit is used to add a unified timestamp, device / service identifier, and topology location information to the multi-source heterogeneous data collected by the data acquisition module, forming a raw data stream with spatiotemporal tags; The spatiotemporal data lake adopts a hybrid storage structure of time-series database and graph database to receive and store the raw data stream with spatiotemporal tags; wherein, the time-series database is used to store time-series sequence data of performance indicators and status data, and the graph database is used to store network topology relationship data representing the relationships between devices, ports and links. The data governance unit is used to perform data cleaning, standardization, normalization and data alignment operations on the data stored in the spatiotemporal data lake; The feature engineering unit is used to extract high-level temporal features from data-governed time-series data, extract graph structure features from data-governed network topology data, and fuse the high-level temporal features and the graph structure features to form the network state data stream.

[0014] The beneficial effects of adopting the above-mentioned further solutions are as follows: By assigning unified timestamps, device / service identifiers, and topological location information to multi-source heterogeneous data through the spatiotemporal alignment unit, the data fusion obstacles caused by dispersed data sources, inconsistent formats, and misaligned time sequences are effectively solved, thus ensuring data consistency and relevance at the data entry point. By introducing a spatiotemporal data lake with a hybrid storage architecture of time-series and graph databases, efficient and unified storage and management of time-series data and topological relationship data in the original data stream with spatiotemporal labels are achieved, providing a unified data foundation with both time and network structure dimensions for subsequent analysis. The data governance unit effectively improves data quality and consistency through cleaning, standardization, normalization, and alignment operations on the stored data, eliminating the interference of noise and anomalies on the artificial intelligence analysis model. The feature engineering unit further extracts and integrates advanced time-series features and graph structure features from the governed time-series and topological data, constructing a network state data stream that simultaneously reflects network dynamic behavior and topological dependencies, providing deep, structured, and high-information-density feature inputs for accurate prediction, diagnosis, and optimization of subsequent collaborative analysis modules.

[0015] Furthermore, the intelligent analysis module includes: The correlation analysis unit is used to establish and update the quantitative correlation between network performance and user experience based on the network key performance index matrix, user experience quality index matrix and the full-link data in the network state data stream, using causal discovery algorithms and attribution analysis models. The prediction unit is used to employ a model that combines a spatiotemporal convolutional network with an attention mechanism, and to analyze the historical alarm time-series characteristics, historical performance time-series characteristics, and graph structure characteristics in the network state data stream based on the current quantized correlation, so as to generate predictive alarms for network device or link performance degradation. The diagnostic unit is used to employ a heterogeneous graph neural network and, based on the predictive alarm and the current quantized correlation, to perform fault propagation analysis on the real-time alarm timing characteristics, real-time performance timing characteristics, and graph structure characteristics in the network state data stream, so as to generate diagnostic results including the fault root cause location results, location confidence, and inference path. The scheduling optimization unit is used to employ a multi-agent deep deterministic policy gradient algorithm, and based on the diagnostic results and the current quantized correlation, analyze the temporal characteristics of the entire network state in the network state data stream based on the current network state, so as to generate at least one resource optimization strategy among dynamic routing strategy, adaptive coding strategy, and protection switching pre-decision.

[0016] The beneficial effects of adopting the above-mentioned further scheme are as follows: By constructing a deep collaborative and closed-loop iterative system of correlation analysis unit, prediction unit, diagnosis unit, and scheduling optimization unit, a continuously evolving intelligent analysis system is formed. The correlation analysis unit not only establishes a quantitative correlation between network performance and user experience, but also dynamically updates its causal knowledge graph based on full-link data, providing the system with an actionable dynamic cognitive foundation. Under this cognitive guidance, the prediction unit integrates historical multi-dimensional features to achieve accurate early warning; the diagnosis unit combines real-time features and topological relationships to achieve interpretable root cause localization; and the scheduling optimization unit, based on the specific root cause localization results output by the diagnosis unit, achieves precise focusing and target alignment when analyzing the temporal characteristics of the entire network status, thereby generating dynamic routing and coding optimization strategies that can avoid known fault points and prioritize key user experience resources. This closed-loop mechanism of "dynamic cognition guiding precise perception, perception results driving optimization decisions, and decision effect feedback reinforcing cognition" enables the system to achieve a synergistic enhancement effect in prediction accuracy, diagnostic precision, scheduling applicability, and quality correlation, ultimately achieving a fundamental leap from passive response to proactive, adaptive intelligent operation and maintenance.

[0017] Furthermore, the instruction execution module includes: The strategy management unit is used to verify and merge the analysis results and generate the network control commands. A digital twin sandbox is used to simulate and deduce the network control command in a digital twin model when the network control command is a preset command, so as to verify the security of the network control command. The intelligent control unit is used to issue network control commands to the target device through a standardized northbound interface or script; wherein the issued network control commands include network control commands that do not require verification and / or verified network control commands.

[0018] The beneficial effects of adopting the above-mentioned further solutions are as follows: The strategy management unit comprehensively verifies and intelligently integrates the analysis results to generate accurate and executable network control commands, ensuring the rationality and consistency of the commands from the source. The digital twin sandbox performs full-process simulation and deduction of some network control commands (such as preset commands such as critical or high-risk commands) in the simulation environment, pre-identifying and avoiding potential execution risks and business impacts, realizing a safe and controllable mechanism of "simulation verification first, then real execution". The intelligent control unit issues commands uniformly and automatically through standardized interfaces, realizing the reliable issuance of verified commands and commands that do not require verification. While improving the speed of operation and maintenance response and the level of automation, it can significantly reduce the risk of misoperation and secondary failures, thereby ensuring the operational safety, service continuity, and execution determinism of the television transmission network during the implementation of predictive maintenance, fault repair, and resource scheduling.

[0019] Furthermore, the system also includes: The human-computer interaction module is used to provide a human-computer interaction interface and to visualize and alarm the intermediate process data and output results of the intelligent analysis module and the instruction execution module.

[0020] The beneficial effects of adopting the above-mentioned further solutions are as follows: Through the visual interface provided by the human-computer interaction module, maintenance personnel can intuitively and in real time grasp the network status, intelligent analysis process and execution results, which greatly enhances the system transparency and user-friendliness. At the same time, the module ensures that key information can be delivered in a timely manner through multi-dimensional alarm notifications, providing maintenance personnel with an effective entry point for auxiliary decision-making and manual intervention. While giving full play to the system's automation and intelligence capabilities, it retains the necessary human-computer collaboration and final decision-making control, thereby achieving an efficient combination of intelligent operation and maintenance and expert experience, and improving the reliability of the overall operation and maintenance process and the efficiency of human-computer collaboration.

[0021] Furthermore, the human-computer interaction module includes at least one of the following functional units: The global situation map unit is used to map the real-time health, load and alarm level of network devices or links in a three-dimensional network topology view by using the color, size and dynamic effects of the network devices or links, and to scroll and display the predictive alarms and the root cause location results of the faults. The fault cause-effect chain unit is used to visualize the propagation chain, root cause node, and reasoning path of a fault event in a way that combines timeline and topology graph. The predictive maintenance dashboard unit is used to display a list of risky devices or links, a predicted performance degradation trend curve, and a recommended maintenance plan for a specified future period. The model operation and maintenance center unit is used to provide performance monitoring views, version management interfaces, sample annotation tools, and retraining trigger entry points for the multiple artificial intelligence models.

[0022] The beneficial effects of adopting the above-mentioned further solutions are as follows: Through the global situation map unit, maintenance personnel can intuitively grasp the real-time operating status and potential risks of the entire network in a three-dimensional topology form, achieving "a complete picture from one map"; through the fault causal chain unit, the complex fault propagation process and root cause location results can be presented in a visual and traceable causal chain, greatly improving the intuitiveness and interpretability of fault diagnosis; through the predictive maintenance dashboard unit, predictive analysis results can be transformed into specific and executable maintenance plans, promoting the transformation of the maintenance model from post-processing to pre-planning; through the model maintenance center unit, a unified visualization tool platform is provided for the performance monitoring, iterative optimization, and lifecycle management of multiple artificial intelligence models, effectively reducing the threshold for model maintenance and ensuring the continuity and reliability of the system's intelligent analysis capabilities.

[0023] Secondly, this invention provides an operation and maintenance method for a television transmission network management system, employing the following technical solution: A method for operating and maintaining a television transmission network management system includes: Real-time acquisition of multi-source heterogeneous data in television transmission networks; The collected multi-source heterogeneous data is subjected to spatiotemporal alignment, data governance, and feature engineering to form a network state data stream; The network status data stream is analyzed collaboratively by multiple interconnected artificial intelligence models to generate at least one of the following analysis results: predictive alarms, root cause location results, resource optimization strategies, and quantitative correlation between network performance and user experience. Based on the analysis results, network control commands are generated and sent to target devices in the television transmission network to drive the target devices to perform corresponding operations, thereby achieving at least one of predictive maintenance, fault repair, and dynamic resource scheduling of the television transmission network; wherein at least some of the network control commands are security verified before being sent.

[0024] Thirdly, the present invention provides an electronic device that adopts the following technical solution: An electronic device includes a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the operation and maintenance method of the television transmission network management system as described in the second aspect.

[0025] Fourthly, the present invention provides a computer-readable storage medium, which adopts the following technical solution: A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the operation and maintenance method of the television transmission network management system as described in the second aspect.

[0026] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and will become apparent from the description or may be learned by practice of the invention. Attached Figure Description

[0027] Figure 1 A schematic diagram of the architecture of the television transmission network management system provided by the present invention; Figure 2 A schematic diagram illustrating the working principle of the intelligent analysis module provided by this invention; Figure 3 A flowchart illustrating the operation and maintenance method of the television transmission network management system provided by the present invention; Figure 4A flowchart illustrating the operation and maintenance method for handling "insufficient optical power alarms and potential service interruption risks" in the television transmission network management system provided by this invention. Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0028] The principles and features of the present invention are described below. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0029] Please refer to Figure 1 , Figure 1 This is a schematic diagram of the architecture of the television transmission network management system provided by the present invention. Figure 1 As shown, the system adopts a layered, decoupled, and intelligently embedded architecture, which mainly includes: data acquisition module 1, data processing module 2, intelligent analysis module 3, and instruction execution module 4.

[0030] Data acquisition module 1 is used to acquire multi-source heterogeneous data in the television transmission network in real time.

[0031] In one embodiment, the data acquisition module 1 includes at least two of the following units: network data acquisition unit, service data acquisition unit, user experience acquisition unit, and external data acquisition unit.

[0032] To fully cover the four major data dimensions of television transmission networks: operation, services, experience, and external environment, such as... Figure 1 As shown, the data acquisition module 1 can be configured simultaneously with network data acquisition unit 11, business data acquisition unit 12, user experience acquisition unit 13, and external data acquisition unit 14. These units constitute the data acquisition layer, acting as the "sensory nerves" of the system, responsible for the real-time acquisition of multi-source heterogeneous data.

[0033] It is understood that in practical applications, the above-mentioned acquisition units can be flexibly combined and selected according to the specific network scale, monitoring needs and cost constraints. For example, including only any two or three of them is also within the scope of feasible embodiments of the present invention.

[0034] The four acquisition units of data acquisition module 1 will be described in detail below.

[0035] The network data acquisition unit 11 is used to collect the operating status data, performance index data and configuration parameter data of various network element devices in the television transmission network.

[0036] For example, the network data acquisition unit 11 collects operational status data (such as device online status, CPU / memory usage, port status, etc.), performance index data (such as bit error rate, signal-to-noise ratio, transmit / receive power, delay jitter, etc.), and configuration parameter data (such as encoding parameters, modulation method, output power, frequency point settings, etc.) of network element devices such as encoders, multiplexers, optical transceivers, switches, and modulators in the television transmission network through standard management protocols such as Simple Network Management Protocol (SNMP), Technical Report 069 (TR-069), and Network Configuration Protocol (NetConf), as well as a command-line interface (CLI) accessed through Secure Shell (SSH) or remote login (Telnet) protocols.

[0037] The service data acquisition unit 12 is used to capture the service data stream flowing through the television transmission network through probes deployed at multiple key nodes of the television transmission network, and to collect network transport layer indicators and multi-level error monitoring information conforming to the TR101-290 standard from the service data stream through deep packet inspection or dedicated protocol analysis.

[0038] For example, the service data acquisition unit 12 captures the full or sampled service data stream flowing through the television transmission network in a mirroring or splitting manner using dedicated hardware probes or virtualized software probes deployed at the core layer, aggregation layer, access layer, and content delivery network edge nodes of the television transmission network. Then, it uses deep packet inspection technology or dedicated audio and video protocol analysis technology to parse the service data stream to collect network transport layer indicators (such as service traffic size, latency jitter, bit rate and its fluctuation range) and multi-level error monitoring information conforming to the TR101-290 standard (such as program clock reference (PCR) jitter, consecutive packet loss (CC) count, transport stream error interval, etc.).

[0039] User experience acquisition unit 13 is used to collect at least one type of user experience quality data from the logs of the technical monitoring device and in conjunction with the quality evaluation of the bitstream analyzer, including signal loss, cyclic redundancy check error, black screen, still frame and mute.

[0040] For example, the user experience acquisition unit 13 first collects the device operation logs from technical monitoring devices such as set-top boxes, smart TVs, and streaming media terminals deployed in the user's home network, and extracts underlying error events including signal loss, cyclic redundancy check errors, decryption failure, channel switching failure, and application crashes. At the same time, the user experience acquisition unit 13 connects to the bitstream analyzer deployed on the network side or the service platform side, obtains the quality evaluation report generated after the analyzer performs real-time analysis of the program stream, and extracts the quality detection results and timestamps of presentation layer problems such as black screen, still frame, mute, mosaic, audio-visual asynchrony, and color abnormality.

[0041] The external data acquisition unit 14 is used to connect to external data sources through an external data interface to acquire external data related to the operation of the television transmission network.

[0042] For example, the external data acquisition unit 14 establishes a connection with various external data sources and acquires external data that has a potential or explicit relationship with the operation of the television transmission network through a standardized or customized external data interface. External data includes, but is not limited to, the following categories: 1) Environmental and physical layer data: such as weather warnings (lightning, strong winds, heavy rain), geological disaster warnings, and real-time weather data (temperature, humidity, precipitation) collected from interfaces of meteorological, seismological, and other departments, used to correlate the stability risks of outdoor lines and wireless transmission links; 2) Network and interconnection layer data: such as public network egress quality, backbone network congestion status, health of third-party cloud services or content delivery network services, and planned network cutover, upgrade, and maintenance time windows collected from internet monitoring platforms, used to assess internet channel quality and predict planned outages; 3) Business and operation layer data: such as electronic program guides, major event / sports live broadcast previews, real-time viewership ratings or popular content lists, and important guarantee times collected from program production and operation systems, used to predict peak business traffic and identify business periods that require key protection; 4) Policy and control layer data: such as temporary control policies, security policy adjustments, and resource reservation instructions synchronized from network management systems or operation and maintenance management platforms, used to ensure that intelligent analysis and decision-making are consistent with the intentions of upper-level management.

[0043] Through the aforementioned network data acquisition unit, business data acquisition unit, user experience acquisition unit, and external data acquisition unit, it is possible to achieve comprehensive and multi-dimensional unified acquisition of multi-source heterogeneous data from the television transmission network. This can break down data silos and lay the foundation for building a complete and interconnected network data view.

[0044] Data processing module 2 is used to perform spatiotemporal alignment, data governance, and feature engineering on the collected multi-source heterogeneous data to form a network state data stream.

[0045] Specifically, the data processing module 2 is responsible for spatiotemporal alignment, data governance, and feature engineering of the collected multi-source heterogeneous data, transforming the collected raw multi-source heterogeneous data into a high-quality, correlated network state data stream.

[0046] In one embodiment, such as Figure 1 As shown, data processing module 2 includes: a spatiotemporal alignment unit 21, a spatiotemporal data lake 22, a data governance unit 23, and a feature engineering unit 24. These units constitute the data lake and feature engineering layer, serving as the system's "memory and understanding hub," responsible for spatiotemporal alignment, storage, governance, association, and feature extraction of the collected raw multi-source heterogeneous data.

[0047] The spatiotemporal alignment unit 21 is used to add a unified timestamp, device / service identifier and topology location information to the multi-source heterogeneous data collected by the data acquisition module 1 to form a raw data stream with spatiotemporal labels.

[0048] Specifically, each data record is assigned a unified baseline timestamp. Data with its own timestamp is corrected and aligned, while data without a timestamp is based on the system reception time and labeled accordingly. Standardized device / service identifiers are added to each data record (e.g., using a four-level coding system of "region-room-rack-device" to identify devices, and "service type-program ID-session ID" to identify services). The location information of each data record in the network topology diagram is added. Data records with spatiotemporal tags are merged into the same data stream in their timestamp order, effectively solving data fusion obstacles caused by dispersed data sources, inconsistent formats, and misaligned timing, thus ensuring data consistency and relevance at the data entry point.

[0049] Spatiotemporal Data Lake 22 adopts a hybrid storage structure of time-series database and graph database to receive and store raw data streams with spatiotemporal labels. The time-series database is used to store time-series sequence data of performance indicators and status data, and the graph database is used to store network topology relationship data representing the connection relationships of devices, ports and links.

[0050] Specifically, the spatiotemporal data lake 22 adopts a hybrid storage architecture combining time-series databases and graph databases. A data lake is a technical architecture capable of centrally storing raw, multi-source, heterogeneous data. The spatiotemporal data lake 22 specifically refers to a dedicated storage system for storing spatiotemporally aligned raw data streams. The spatiotemporal data lake 22 receives, persistently stores, and manages the raw data streams with spatiotemporal tags output by the spatiotemporal alignment unit 21. For example, the spatiotemporal data lake 22 stores the raw data streams with spatiotemporal tags according to their characteristics: all performance metrics, state telemetry, and event data with time-series characteristics are written to a high-performance time-series database. Each data record is organized using its timestamp as the primary key or index to support efficient writing, compressed storage, and fast aggregation queries by time range for massive amounts of historical and real-time time-series data. Simultaneously, all static and dynamic topology information representing network entities and their inter-entity connections (such as physical connections, logical affiliations, and routing paths between devices, boards, and ports) is modeled as attribute graphs and stored in a graph database. Network entities serve as nodes, and connections as edges. Rich attributes can be attached to nodes and edges, supporting real-time updates of complex network topologies, path queries, impact surface analysis, and other graph traversal operations. Through this hybrid storage design, the Spatiotemporal Data Lake 16 constructs a unified, efficient, and structured data storage and management system, providing a solid and reliable data foundation for upper-level data governance, feature engineering, and collaborative analysis.

[0051] The data governance unit 23 is used to perform data cleaning, standardization, normalization and data alignment operations on the data stored in the spatiotemporal data lake.

[0052] Specifically, the data governance unit 23 performs the following operations on the data stored in the spatiotemporal data lake 22: First, it performs data cleaning, identifying and processing missing values, obvious errors, and noisy data through a rule engine and anomaly detection algorithm. For missing values ​​caused by device offline, it uses forward and backward interpolation or specific value filling, and smooths or removes outliers that exceed a reasonable range. Second, it performs data standardization, converting data items from different manufacturers and protocols (such as power units and status enumeration values) into a unified unit of measurement and enumeration dictionary, and performing consistency verification and mapping on all device / service identifiers. Next, it performs data normalization, standardizing numerical performance indicators (such as latency and traffic) using minimum-maximum scaling or Z-score to bring them into a unified dimension, facilitating model training and comparison. Finally, it performs data alignment, based on a unified spatiotemporal label, performing precise time window matching and correlation splicing on cross-source and cross-layer data required for correlation analysis (such as device CPU utilization, link traffic, and service packet loss rate on the same link at the same time). The data processed in the above way forms a standardized, high-quality dataset, which is then written back into the post-governance data area of ​​the spatiotemporal data lake. This effectively improves data quality and consistency, and eliminates the interference of noise and anomalies on the artificial intelligence analysis model.

[0053] Feature engineering unit 24 is used to extract high-level temporal features from data-governed time-series data, extract graph structure features from data-governed network topology data, and fuse high-level temporal features and graph structure features to form a network state data stream.

[0054] Specifically, the feature engineering unit 24 performs deep feature extraction and fusion on the high-quality time-series data and topological relationship data processed by the data governance unit 23 to construct a high-order feature set that can comprehensively characterize the network's dynamic behavior and structural attributes, forming the final network state data stream. For example, it extracts advanced time-series features such as recent slope, periodic fluctuation amplitude, and correlation with adjacent links from the CRC error time-series data, and extracts graph features such as degree centrality and betweenness centrality of nodes from the topological graph. Through time alignment and feature concatenation, the advanced time-series features and related graph structure features under the same time slice are fused to generate a unified, high-dimensional, and structured network state feature vector, which is output in chronological order as a continuous network state data stream that can simultaneously reflect the network's dynamic behavior and topological dependencies. This provides deep, structured, and high-information-density feature input for the subsequent accurate prediction, diagnosis, and optimization of the intelligent analysis module 3.

[0055] The intelligent analysis module 3 is used to perform collaborative analysis of network status data streams through multiple interconnected artificial intelligence models, thereby generating at least one of the following analysis results: predictive alarms, root cause location results, resource optimization strategies, and quantitative correlation between network performance and user experience.

[0056] Specifically, the intelligent analysis module 3 uses multiple interconnected artificial intelligence models to conduct collaborative analysis of network status data streams, enabling accurate early warning of network faults, automated location of fault root causes, dynamic optimization and scheduling of network resources, and quantitative correlation between network performance and user experience. This transforms the traditional passive response and experience-driven operation and maintenance model into a proactive prevention and intelligent decision-making operation and maintenance model.

[0057] In one embodiment, such as Figure 1 As shown, the intelligent analysis module 3 includes: a prediction unit 31, a diagnosis unit 32, a scheduling and optimization unit 33, and a correlation analysis unit 34. These units constitute the core engine layer of artificial intelligence, serving as the "brain" of the system and responsible for executing core intelligent tasks such as prediction, diagnosis, optimization, and correlation analysis.

[0058] The correlation analysis unit 34 is used to establish and update the quantitative correlation between network performance and user experience based on the network key performance index matrix, user experience quality index matrix and full-link data in the network state data stream, using causal discovery algorithms and attribution analysis models.

[0059] Specifically, the correlation analysis unit 34 employs a causal discovery algorithm and an attribution analysis model as the correlation analysis model. This model uses the network key performance indicator matrix (KPI matrix) and user experience quality indicator matrix (QoE matrix) extracted from the network state data stream, combined with end-to-end data corresponding to historical operation and maintenance events, as input. The KPI matrix covers device performance, link quality, and service flow indicators, while the QoE matrix includes user-perceived events such as lag, black screen, and decoding failure. The end-to-end data includes the network state that triggers intelligent analysis, the generated intelligent analysis results, the issued network control commands, and the network and service quality feedback after their execution. Internally, the causal discovery algorithm first identifies statistically significant causal pairs between network KPIs and user QoE from time-series data. It then uses the actual effects of analytical decisions recorded in the end-to-end data to dynamically evaluate and correct the confidence and timeliness of the identified causal relationships, thereby constructing and continuously updating a dynamic "network-experience" causal graph. Based on this causal graph, the attribution analysis model further quantifies the impact weight and contribution ratio of various user QoE issues when a specific network KPI deteriorates. This quantification process also references the execution effect of resource optimization strategies in the end-to-end data to calibrate the attribution weights. The model ultimately outputs a structured quantitative correlation report, which clearly reveals specific quantitative relationships such as "a 10% increase in core link latency will lead to a 25% increase in live streaming service stuttering rate," and includes a confidence index calculated based on historical end-to-end data. This allows for accurate and verifiable mapping and dynamic attribution of network performance problems and user experience impairments. The resulting real-time knowledge graph provides a reliable basis for continuous optimization of end-to-end quality tracing from the network signal level to the user perception level.

[0060] Prediction unit 31 is used to adopt a model that combines spatiotemporal convolutional networks with attention mechanisms, and analyze the historical alarm time series characteristics, historical performance time series characteristics, and graph structure characteristics in the network state data stream based on the current quantized correlation relationship, so as to generate predictive alarms for network device or link performance degradation.

[0061] Specifically, prediction unit 31 employs a deep learning model integrating a spatiotemporal convolutional network and a self-attention mechanism as its prediction model. This model takes historical alarm time-series features, historical performance time-series features (such as latency, packet loss rate, and traffic), and graph structure features (representing the topological connections between devices and links) from the network state data stream as input, combined with the current quantized correlation relationship. The current quantized correlation relationship is dynamically established and updated by the correlation analysis unit 34, which provides causal weights between network performance indicators and user experience, enabling the prediction model to prioritize performance indicators and network nodes that have a greater impact on user experience. Internally, the spatiotemporal convolutional network captures the temporal dependencies of the device's own performance and the spatial dependencies between neighboring devices propagating along the network topology; the self-attention mechanism dynamically weights the multi-dimensional spatiotemporal features extracted by the spatiotemporal convolutional network, focusing on the key time steps and key network nodes most relevant to the current prediction. When calculating prediction probabilities, the model uses current quantitative correlations to refine its assessment of the severity of degradation for different performance metrics. For example, for metrics highly correlated with core user experience (such as live stream latency), the model assigns a higher risk weight even if the degradation levels are the same. After training, the model outputs predictions of performance degradation of network devices or links within a specified future time window. When the predicted degradation probability exceeds a preset health threshold, a predictive alarm is generated, containing the prediction target, degradation type, occurrence time window, and a severity level dynamically calculated based on degradation speed and impact range. The severity level assessment references the proportion of the metric's impact on user experience in the current quantitative correlation, thus achieving a shift from passive response to proactive early warning.

[0062] The diagnostic unit 32 is used to perform fault propagation analysis on the real-time alarm timing characteristics, real-time performance timing characteristics, and graph structure characteristics in the network status data stream using a heterogeneous graph neural network and based on predictive alarms and current quantization correlation, so as to generate diagnostic results including fault root cause location results, location reliability, and inference path.

[0063] Specifically, the diagnostic unit 32 employs a heterogeneous graph neural network as its diagnostic model. This model takes as input real-time alarm timing characteristics, real-time performance timing characteristics, and graph structure characteristics (representing the topological connections between devices and links) from the network status data stream. Simultaneously, the diagnostic model relies on predictive alarms generated by the prediction unit 31 and the current quantized correlation relationships dynamically maintained by the correlation analysis unit 34. Predictive alarms focus on high-risk network areas and time segments, while the current quantized correlation relationships provide weights for the potential impact of different performance degradation modes on user experience in fault propagation analysis. This guides the diagnostic model to prioritize searching for and verifying fault propagation paths that may lead to severe user experience impairment. The heterogeneous graph neural network accurately simulates the spread of faults in the network by modeling differentiated propagation patterns between different types of nodes (such as encoders, optical transceivers, and multiplexers) and different types of edges (such as physical connections and logical bearers). During the propagation simulation, the diagnostic model combines the current quantized correlation relationships to assess the worst-case user experience impact that faults may cause along different paths, using this assessment as a crucial consideration in path search and root cause probability calculation. The model calculates and outputs a fault causal probability graph, which, based on the input topology, marks the probability of each node being the root cause of the fault. The node with the highest root cause probability, highlighted in the fault causal probability graph, is identified as the root cause of the fault. The model outputs the location result, including the location confidence and an interpretable inference path presented as a node sequence. The inference path clearly marks the predicted impact strength on various user experience indicators (such as lag and black screen) when the fault propagates along the path. This strength value comes from the attribution analysis results in the current quantitative correlation, thereby achieving automated, interpretable, and business-oriented accurate diagnosis of complex faults, greatly improving troubleshooting efficiency, accuracy, and the credibility of decision-making.

[0064] The scheduling optimization unit 33 is used to employ a multi-agent deep deterministic strategy gradient algorithm, and based on the diagnostic results and the current quantization correlation, analyze the temporal characteristics of the entire network state in the network state data stream based on the current network state, so as to generate at least one resource optimization strategy among dynamic routing strategy, adaptive coding strategy and protection switching pre-decision.

[0065] Specifically, the scheduling optimization unit 33 employs a multi-agent deep deterministic policy gradient algorithm as its scheduling optimization model. This model models logical domains with different functions in the television transmission network (such as source reception, core scheduling, and signal distribution domains) as collaborative agents. The model takes the real-time network-wide status (including traffic, load, fault status, and priority of different services for each link and device) from the network state data stream as input. Simultaneously, the scheduling optimization model uses the diagnostic results generated by the diagnostic unit 32 (including fault root cause location, confidence level, and inference path) and the current quantitative correlation relationships dynamically maintained by the correlation analysis unit 34. The diagnostic results determine the fault root causes to be avoided, the high-risk paths to be isolated, and the potential impact range of the faults, providing key constraints on the action space of each agent. The current quantitative correlation relationships provide the quantitative impact weights of different network performance index degradations on the user experience of various services, providing precise business value basis for the decision-making priorities of each agent during resource competition. Based on shared environmental state information and the aforementioned constraints and value basis, each agent conducts collaborative decision-making learning through a centralized training and distributed execution framework, aiming to jointly optimize the overall network resource efficiency and service quality. During the decision-making process, when evaluating alternative paths, each intelligent agent (such as the routing domain agent) will combine the fault propagation path information in the diagnostic results to avoid scheduling services to high-risk paths affected by the root cause; while when adjusting the bitrate, the coding domain agent will make precise adjustments based on the sensitivity of the service type to specific performance indicators (such as latency and packet loss) in the current quantization correlation, with the goal of minimizing the user-perceived quality degradation.

[0066] After online or offline training, the scheduling optimization model can output a joint optimization strategy, which may include at least one of the following: a dynamic routing strategy, used to calculate and select the end-to-end transmission path with the optimal latency and highest reliability for specific high-priority service flows in real time. Its path selection logic will explicitly avoid fault nodes and affected areas indicated by diagnostic results, and prioritize the core services with the greatest impact on user experience identified in the current quantitative correlation; an adaptive coding strategy, used to dynamically adjust the source coding bitrate to match the available bandwidth based on the real-time network congestion level without significantly affecting the user's subjective viewing quality. The magnitude and rhythm of the bitrate adjustment will refer to the specific impact model of the service bitrate change on user experience score in the current quantitative correlation; and protection switching pre-decision, used to pre-calculate the optimal backup path and switching parameters before a potential fault occurs, so that lossless or smooth switching can be performed immediately when a fault actually occurs. The selection of the backup path and the setting of the switching parameters will comprehensively consider the fault impact range predicted by the diagnostic results and the service importance revealed by the current quantitative correlation, thereby achieving dynamic, intelligent, interpretable, and service experience-oriented collaborative optimization of network resources.

[0067] The closed-loop mechanism of "dynamic cognition guides precise perception, perception results drive optimized decision-making, and decision effect feedback strengthens cognition" in the intelligent analysis module 3 enables the system to form a synergistic enhancement effect in prediction accuracy, diagnostic precision, scheduling applicability, and quality correlation, ultimately achieving a fundamental leap from passive response to proactive and adaptive intelligent operation and maintenance.

[0068] After the intelligent analysis module 3 completes the collaborative analysis of the network status data stream and generates various analysis results, the system process enters the instruction execution module 4, which is responsible for converting the intelligent analysis results into safe, reliable, and executable control actions.

[0069] The instruction execution module 4 is used to generate network control instructions based on the analysis results and send the network control instructions to the target devices in the television transmission network to drive the target devices to perform corresponding operations, thereby realizing at least one of the predictive maintenance, fault repair and dynamic resource scheduling of the television transmission network; wherein at least some of the network control instructions are security verified before being sent.

[0070] Specifically, the instruction execution module 4 generates network control instructions based on the analysis results, and after performing security verification on the instructions that need to be verified (such as critical or high-risk instructions), it issues them together with the instructions that do not need to be verified. Under the premise of ensuring operational safety, it realizes at least one of predictive maintenance, rapid fault repair and dynamic resource scheduling of the television transmission network, and finally transforms intelligent decision-making into precise and safe network control operations.

[0071] In one embodiment, such as Figure 1 As shown, the instruction execution module 4 includes a strategy management unit 41, a digital twin sandbox 42, and an intelligent control unit 43. These units constitute the intelligent decision-making and automated execution layer, serving as the system's "decision-making and execution organs," responsible for transforming the intelligent analysis results output by the artificial intelligence analysis model into safe and reliable operations.

[0072] The strategy management unit 41 is used to verify and merge the analysis results and generate network control commands.

[0073] Specifically, the policy management unit 41 performs comprehensive decision verification and policy fusion on multiple analysis results to generate executable and verifiable network control commands. For example, the policy management unit 41 receives multi-dimensional analysis results including predictive alarms, root cause analysis, resource optimization strategies, and network-experience correlation reports. It then performs multiple verifications on these results based on pre-set business assurance policies, network security rules, device operating procedures, and network operation constraints to ensure the accuracy and rationality of the network control commands. Verification content may include policy conflict detection (such as resource competition between different optimization objectives), security compliance review (such as whether the command complies with the principle of least privilege), operational risk assessment (such as whether the command may cause business interruption), and cost-benefit assessment. After successful verification, the executable and high-priority suggestions from each intelligent analysis result are logically fused and prioritized to generate a unified and standardized sequence of network control commands. Necessary execution context, rollback contingency plans, and permission tags are then attached to provide structured input for subsequent command verification and execution.

[0074] The digital twin sandbox 42 is used to simulate and deduce network control commands in a digital twin model when the network control commands are preset commands, so as to verify the security of the network control commands.

[0075] Specifically, assuming the preset instructions are critical or high-risk network control instructions, the digital twin sandbox 42 performs a full-process simulation in a high-fidelity digital twin model of the television transmission network before the actual execution of the critical or high-risk network control instructions, in order to verify the effectiveness and harmlessness of the network control instructions. Based on the stored data in the spatiotemporal data lake 22, the digital twin sandbox 42 constructs and dynamically updates a virtual image highly consistent with the physical network, whose model covers network elements, transmission links, service flows, and protocol interactions. When a network control instruction (such as route switching, power adjustment, or protection switching) generated by the policy management unit 41 is determined to be critical or high-risk, the network control instruction will be imported into the sandbox environment. The sandbox uses the current network state as the initial condition, executes the network control instruction in the digital twin model, and simulates the state changes of network devices, service flow path migration, performance index fluctuations, and potential cascading effects after the execution of the network control instruction through a simulation engine. During the simulation, the sandbox continuously detects whether abnormal situations such as service interruption, service quality degradation, policy conflicts, or security violations occur. After the simulation, the sandbox generates a verification report, which clarifies whether the network control command has passed security verification. The report includes the key state sequence during the simulation, the identified potential risks, and optimization suggestions for command parameters or execution timing. This ensures that the network control command has undergone sufficient security simulation verification before being executed in a real network, effectively reducing network operation risks caused by command errors or conflicts.

[0076] The intelligent control unit 43 is used to issue network control commands to the target device (such as a network element device or a matrix controller) through a standardized northbound interface or script; wherein the issued network control commands include network control commands that do not require verification and / or verified network control commands.

[0077] Specifically, the intelligent control unit 43 issues network control commands in a unified and automated manner through a standardized northbound interface or script, thereby enabling the reliable issuance of network control commands that do not require verification and / or verified network control commands.

[0078] The instruction execution module 4 can significantly reduce the risk of misoperation and secondary failures while improving the speed of operation and maintenance response and the level of automation, thereby ensuring the operational safety, service continuity and execution determinism of the television transmission network in the process of implementing predictive maintenance, fault repair and resource scheduling.

[0079] In one embodiment, such as Figure 1 As shown, the aforementioned television transmission network management system also includes: a model optimization module ( Figure 1 (Not shown in the image) is used to collect the full-link data corresponding to each operation and maintenance event, from collaborative analysis to the execution of network control commands, after each operation and maintenance event is completed, and to iteratively optimize multiple artificial intelligence models based on the full-link data; where an operation and maintenance event refers to a complete network control operation process triggered by the intelligent analysis module and completed by the command execution module.

[0080] Specifically, the model optimization module collects end-to-end data from intelligent analysis to network control command execution and feedback, including network status triggering intelligent analysis, generated intelligent analysis results, issued network control commands, and subsequent network and service quality feedback. This complete event sequence is marked as training samples and stored in the case library. The module periodically extracts new samples from the case library and, based on the real results and feedback contained in the end-to-end data recorded in the samples, performs incremental training and iterative optimization on the prediction model, diagnostic model, scheduling optimization model, and correlation analysis model. The optimization process specifically includes: adjusting the weights and confidence levels of causal relationships in the correlation analysis model using real user experience feedback and network status results from the end-to-end data; correcting the thresholds and feature weights of the prediction model based on the actual occurrence of predictive alarms; optimizing the propagation and inference parameters of the diagnostic model based on the actual accuracy of fault root cause localization; and updating the reward function and policy network of the scheduling optimization model by referencing the actual execution effect of the scheduling optimization strategy. Through this continuous learning based on real feedback, the system's prediction accuracy, diagnostic precision, scheduling applicability, and correlation reliability can all continuously and autonomously improve with actual operation, thereby driving the continuous evolution of the entire system's intelligent analysis and decision-making capabilities.

[0081] In one embodiment, such as Figure 1 As shown, the above-mentioned television transmission network management system also includes: a human-computer interaction module 5, which provides a human-computer interaction interface and visualizes and alarms the intermediate process data and output results of the intelligent analysis module 3 and the instruction execution module 4.

[0082] Specifically, through the visual interface provided by the human-computer interaction module 5, maintenance personnel can intuitively and in real time grasp the network status, intelligent analysis process, and execution results, greatly enhancing system transparency and user-friendliness. At the same time, this module ensures that key information can be delivered in a timely manner through multi-dimensional alarm notifications, providing maintenance personnel with an effective entry point for auxiliary decision-making and manual intervention. While fully leveraging the system's automation and intelligence capabilities, it retains necessary human-computer collaboration and final decision-making control, thereby achieving an efficient combination of intelligent maintenance and expert experience, improving the reliability of the overall maintenance process and the efficiency of human-computer collaboration.

[0083] In one embodiment, the human-computer interaction module 5 includes at least one of the following functional units: a global situation map unit, a fault causal chain unit, a predictive maintenance dashboard unit, and a model operation and maintenance center unit.

[0084] To provide a comprehensive, intuitive, and user-friendly intelligent operation and maintenance interface, such as Figure 1 As shown, the human-computer interaction module 5 can be configured simultaneously with a global situation map unit 51, a fault cause-effect chain unit 52, a predictive maintenance dashboard unit 53, and a model operation and maintenance center unit 54. These units constitute the interaction and presentation layer, serving as the system's "human-computer interaction interface," responsible for providing global visualization, intelligent insight, and control entry points.

[0085] It is understood that in practical applications, the above functional units can be flexibly combined and selected according to specific operation and maintenance roles, focus scenarios and display requirements, which also falls within the scope of feasible embodiments of the present invention.

[0086] The four functional units of the human-computer interaction module 5 will be described in detail below.

[0087] The global situation map unit 51 is used to map the real-time health, load and alarm level of network devices or links in a three-dimensional network topology view by the color, size and dynamic effects, and to scroll and display predictive alarms and root cause location results.

[0088] Specifically, the global situational awareness unit 51 obtains network topology and status data with spatiotemporal labels from the data processing module 2 and renders it according to preset visual encoding rules: the colors of devices and links dynamically change according to their health scores (e.g., green for healthy, yellow for warning, and red for fault); the size of the graphic can map real-time traffic load or device importance; simultaneously, the topology map highlights currently active predictive alarms and located root cause nodes of faults with dynamic effects such as pulses and flashes. The view supports multi-level drill-down, and operations personnel can click on any node to view its detailed performance indicator trends, related alarm lists, and impact range analysis. The scrolling information bar continuously updates and categorizes the latest outputs from the prediction unit 31 and the diagnostic unit 32. Through this view, operations personnel can intuitively grasp the real-time operating status and potential risks of the entire network in a three-dimensional topology form, achieving "a complete picture from one map," significantly improving the efficiency of perception and decision-making speed regarding the macro-situation of complex networks.

[0089] Fault causal chain unit 52 is used to display the propagation chain, root cause node and reasoning path of fault events in a visualization method that combines timeline and topology graph.

[0090] Specifically, the fault causal chain unit 52 receives the diagnostic results (including fault root cause location, confidence level, and inference path) output from the diagnostic unit 32, and constructs a fault propagation diagram integrating temporal dimension and topology structure in a two-dimensional view based on the node sequence and timestamps recorded in the inference path: the left side of the view is a vertical time axis, marking key time points from fault occurrence and propagation to detection; the center of the view is a horizontally expanded network topology sub-graph, highlighting the devices and links affected sequentially along the inference path starting from the root cause node; the connection between nodes is indicated by arrows to represent the fault propagation direction, and the fault type, performance index degradation value, and propagation delay can be marked on the connection. Maintenance personnel can interactively expand any node to view the performance index comparison of that node before and after fault propagation, as well as the confidence level of that node as an intermediate link in the propagation calculated by the diagnostic model. Through this view, the complex fault propagation process and root cause location results can be presented in a visual and traceable causal chain, greatly improving the intuitiveness, interpretability, and trust in the diagnostic results of fault troubleshooting.

[0091] Predictive maintenance dashboard unit 53 is used to display a list of risky devices or links, a predicted performance degradation trend curve, and a recommended maintenance plan for a specified future period.

[0092] Specifically, the predictive maintenance dashboard unit 53 receives predictive alarms generated by the prediction unit 31, aggregates and prioritizes them, and generates a risk list presented in tabular form. Each record in the list includes the risk device / link identifier, the predicted degradation type, the expected occurrence time window, the predicted probability, the severity level, and the potential business impact. Simultaneously, for high-risk items in the list, the predictive maintenance dashboard unit 53 extracts historical and predicted time-series data of their key performance indicators from the network status data stream, and plots a comparison curve of historical trends, current status, and future predictions to visually demonstrate the degradation development trend. Based on the above information, the predictive maintenance dashboard unit 53 calls the built-in rule engine or simple optimization algorithm, combined with preset maintenance resources, time windows, and business tolerance strategies, to generate one or more recommended maintenance plans for each high-risk prediction. The plan includes suggested maintenance operations (such as equipment inspection, parameter adjustment, and component replacement), suggested execution time, estimated required resources, and expected post-maintenance effects. Maintenance personnel can confirm, modify, or directly dispatch the plan as a work order. This dashboard enables the transformation of predictive analytics results into specific, actionable maintenance plans, driving a shift in operations and maintenance from reactive to proactive planning, thereby achieving closed-loop management of predictive maintenance.

[0093] The Model Operation and Maintenance Center Unit 54 is used to provide performance monitoring views, version management interfaces, sample annotation tools, and retraining trigger entry points for multiple artificial intelligence models.

[0094] Specifically, the Model Operation and Maintenance Center Unit 54 centrally displays real-time operational metrics for predictive models, diagnostic models, scheduling and optimization models, and correlation models in a dashboard format. These metrics include inference time, accuracy / recall curves, data feature distribution offset alarms, and resource consumption. The version management interface records the training configuration, dataset version, iteration history, and deployment status of each model, supporting model version rollback, comparison, and one-click release. The sample annotation tool allows operations experts to extract corresponding end-to-end raw data from the case library for review of cases with model prediction errors or low confidence, and to add or correct result labels to form high-quality correction samples. When model performance metrics consistently fall below thresholds, or when a sufficient number of new labeled samples are received, operations personnel can manually or automatically initiate the incremental training process for the corresponding model through the retraining trigger entry, scheduling the training task and resources to the backend computing platform. Through the Model Operation and Maintenance Center Unit 54, a unified visualization tool platform is provided for performance monitoring, iterative optimization, and lifecycle management of multiple artificial intelligence models, effectively lowering the threshold for model operation and maintenance and ensuring the continuity and reliability of the system's intelligent analysis capabilities.

[0095] To better understand the present invention, the working principle of the intelligent analysis module 3 of the television transmission network management system of the present invention will be specifically explained below with reference to specific embodiments.

[0096] like Figure 2 As shown, the intelligent analysis module, through the coordinated linkage of the prediction unit, diagnosis unit, and scheduling optimization unit, combined with the digital twin sandbox and correlation analysis unit, forms a closed-loop intelligent analysis process of "prediction-diagnosis-scheduling-verification-feedback".

[0097] 1) Data Input Layer: Fusion of Historical and Real-time Data First, historical data (such as traffic, alarms, and device status over the past 7 / 30 days) and real-time data (current business traffic, link load, and device performance indicators) are collected from various monitoring devices, business systems, and log platforms of the television transmission network to form a multi-source heterogeneous data source, which serves as the input basis for subsequent analysis.

[0098] 2) Prediction Unit: ST-GCN / Attention Mechanism Model The prediction unit employs a model combining a spatiotemporal graph convolutional network (ST-GCN) and an attention mechanism to process the input historical and real-time data. The ST-GCN model excels at capturing spatial relationships in network topology (such as physical connections between links and service flow dependencies) and temporal trends (such as hourly / daily periodic fluctuations in traffic), learning the evolutionary patterns of network states. The attention mechanism focuses on key nodes / links (such as core switches and high-bandwidth links), enhancing sensitivity to their state changes. Through joint modeling of spatiotemporal features, the prediction unit outputs predictive alerts (such as "a link will exceed 80% load in 10 minutes" or "a device's CPU utilization will exceed the threshold in 5 minutes"), proactively detecting potential faults or resource bottlenecks.

[0099] 3) Diagnostic Unit: Heterogeneous Graph Neural Network (HetGNN) The diagnostic unit receives "predictive alarms" from the prediction unit and, in conjunction with the heterogeneous characteristics of the network topology (such as diverse device types, different link protocols, and differences in service layers), uses a heterogeneous graph neural network (HetGNN) for analysis. HetGNN can handle heterogeneous network structures of "node types (such as routers, servers, and terminals) and edge types (such as physical links, logical tunnels, and service flows)," learn the propagation path and root cause characteristics of faults in heterogeneous topologies, and output root cause location-probability (such as "the probability that the hardware failure of link L-01 is the root cause is 85%" and "the probability that the configuration error of device R-03 is the root cause is 70%)," thus clarifying the triggering source and propagation chain of the fault.

[0100] 4) Scheduling Optimization Unit: Joint Optimization Strategy (MADDPG) The scheduling optimization unit employs the Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm, which, combined with the current network state (real-time topology, resource load, service priority, etc.), formulates dynamic optimization strategies. MADDPG is a reinforcement learning algorithm suitable for multi-agent collaboration. It treats "devices, links, and service flows" in the network as multiple agents, which learn optimal resource scheduling, service switching, and fault avoidance strategies through interaction with the environment (network state). It outputs execution plans (such as "switching service B-01 from link L-01 to backup link L-02" or "adjusting the routing weight of device R-03 to reduce load") to guide subsequent actions.

[0101] 5) Digital Twin Sandbox: Pre-verification and Execution The contingency plan is not issued directly, but is first pre-verified in a digital twin sandbox: the digital twin sandbox builds a digital mirror of the network to simulate changes in network status after the contingency plan is executed (such as traffic redistribution, load migration, service interruption duration, etc.); after the verification is passed, the contingency plan is then issued to the actual network devices or control systems for execution to ensure operational safety (avoiding secondary failures caused by misscheduling).

[0102] 6) Association Analysis Unit: Knowledge Graph, Causal Discovery, and Attribution Analysis After the contingency plan is executed, the system will feed back the actual execution results (such as load changes after resource scheduling and alarm status after fault repair) to the correlation analysis unit. The correlation analysis unit analyzes the accuracy of the entire process of "prediction-diagnosis-scheduling-execution" based on knowledge graph (storing the correlation between network devices, services, and alarms), causal discovery algorithm (mining the causal chain between faults and indicators and operations), and attribution analysis model (locating the reasons for the deviation between the actual effect and the contingency plan).

[0103] Finally, the correlation analysis unit feeds back the analysis to the prediction unit, diagnosis unit, and scheduling optimization unit, forming a closed-loop optimization (such as correcting the feature weights of the prediction model, adjusting the topological relationships of the diagnosis model, and optimizing the reward function of the scheduling strategy), continuously improving the system's intelligent analysis capabilities.

[0104] Through the above process, the following can be achieved: 1) Predictive maintenance: anticipating potential risks and preventing failures; 2) Rapid fault repair: accurately locating root causes and shortening service interruption time; 3) Dynamic resource scheduling: allocating resources on demand and improving network utilization; 4) Iterative optimization: continuously optimizing artificial intelligence models and analysis strategies through feedback loops. Ultimately, the reliability (reduced failures, faster recovery), resource efficiency (load balancing, low idle rate), and user experience (smooth service, fewer buffering issues) of the television transmission network are significantly improved.

[0105] This invention provides a television transmission network management system, comprising: a data acquisition module, a data processing module, an intelligent analysis module, and an instruction execution module. The data acquisition module collects multi-source heterogeneous data from the television transmission network in real time. The data processing module performs spatiotemporal alignment, data governance, and feature engineering on the collected multi-source heterogeneous data to form a high-quality, correlated network status data stream, achieving deep integration and unified governance of cross-domain data. The intelligent analysis module uses multiple interconnected artificial intelligence models to collaboratively analyze the network status data stream, enabling accurate early warning of network faults, automated location of fault root causes, dynamic optimization and scheduling of network resources, and quantitative correlation between network performance and user experience. This transforms the traditional passive response and experience-driven operation and maintenance model into a proactive prevention and intelligent decision-making operation and maintenance model. The instruction execution module generates network control instructions based on the analysis results. After verifying the instructions that require verification (e.g., critical or high-risk instructions), it issues them along with the instructions that do not require verification. Under the premise of ensuring operational safety, this achieves at least one of predictive maintenance, rapid fault repair, and dynamic resource scheduling for the television transmission network. Through the closed-loop collaboration of the aforementioned modules, the system can proactively identify and address potential risks, significantly reducing unplanned service interruptions. Its intelligent scheduling strategy enables on-demand, real-time allocation of network resources, avoiding resource idleness or congestion under static configuration. Simultaneously, user-experience-oriented quantitative correlation and root cause analysis ensure that any optimization and repair actions directly and accurately target user-perceived quality shortcomings. Ultimately, the synergistic effect of these technologies significantly improves the operational reliability, resource utilization efficiency, and user experience quality of the television transmission network.

[0106] In some embodiments, the television transmission network management system of the present invention can be implemented in a combination of hardware and software. As an example, the television transmission network management system of the present invention can be a processor in the form of a hardware decoding processor, which is programmed to execute the operation and maintenance method of the television transmission network management system of the present invention. For example, the processor in the form of a hardware decoding processor can be one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0107] The modules described in the embodiments of this invention can be implemented in software or hardware. The names of the modules are not, in some cases, limiting the scope of the module itself.

[0108] Please refer to Figure 3 , Figure 3 This is a flowchart illustrating the operation and maintenance method of the television transmission network management system provided by the present invention. Figure 3 As shown, the method may include the following steps S301-S304.

[0109] S301. Real-time acquisition of multi-source heterogeneous data in the television transmission network.

[0110] S302. Perform spatiotemporal alignment, data governance, and feature engineering on the collected multi-source heterogeneous data to form a network state data stream.

[0111] S303. By collaboratively analyzing network status data streams through multiple interconnected artificial intelligence models, at least one of the following analytical results is generated: predictive alarms, root cause location results, resource optimization strategies, and quantitative correlation between network performance and user experience.

[0112] S304. Generate network control commands based on the analysis results, and send the network control commands to the target devices in the television transmission network to drive the target devices to perform corresponding operations, thereby realizing at least one of the following: predictive maintenance, fault repair, and dynamic resource scheduling of the television transmission network; wherein at least some of the network control commands have been security verified before being sent.

[0113] In one embodiment, the method further includes: After each operation and maintenance event is completed, the full-link data corresponding to that operation and maintenance event, from collaborative analysis to the execution of network control commands, is collected, and multiple artificial intelligence models are iteratively optimized based on the full-link data.

[0114] The operation and maintenance method of the television transmission network management system of the present invention will be specifically described below through specific embodiments.

[0115] Please refer to Figure 4 , Figure 4 This is a flowchart illustrating the operation and maintenance method for handling "insufficient optical power alarms and potential service interruption risks" in the television transmission network management system provided by this invention. Figure 4 As shown, the method may include the following steps S401-S406.

[0116] S401, Acquisition and synchronous sensing of optical power degradation data.

[0117] Continuous collection of multi-source heterogeneous data across the entire television transmission network. For example, by deploying monitoring probes in the television transmission network, real-time collection of various performance indicators, including the "optical transmission power" of a certain video optical transceiver device A, is performed and synchronously written into the spatiotemporal data lake to complete the real-time monitoring and perception of the attenuation of key performance parameters of optical modules.

[0118] Step S402: Intelligent prediction and risk warning of optical power degradation trend.

[0119] Specifically, the prediction engine analyzes the time-series data of the collected "optical emission power" metric: it identifies that although the metric is currently within the normal range, it exhibits a continuous and stable linear downward trend. Based on this trend and historical patterns, the prediction engine's built-in Spatiotemporal Graph Convolutional Network (ST-GCN) model predicts that the optical power will fall below a critical value in approximately 3 hours, posing a risk of service interruption. A predictive alert work order with a "critical" level is then automatically generated and pushed to the operations and maintenance monitoring dashboard and the mobile terminals of relevant personnel.

[0120] Step S403: Locating and analyzing the root causes of optical power degradation (optical module aging).

[0121] Specifically, the diagnostic engine automatically starts after the maintenance personnel confirm the predictive alarm work order. The diagnostic process focuses on the alarming device, combining its network topology connections, the status of upstream and downstream devices, and environmental information in the same rack (such as temperature and power supply) for comprehensive analysis. The heterogeneous graph neural network (HetGNN) model built into the diagnostic engine outputs a structured diagnostic result: "Root cause probability: Aging of the TX optical module in optical transceiver A (85%); slight decrease in upstream input optical power (10%)", and lists supporting evidence, including that the alarm in optical transceiver A occurred independently, and that its historical decay curve of transmitted optical power has a high degree of matching with the typical aging curve of optical modules.

[0122] Step S404: Prepare a scheduling plan to deal with the aging of optical modules.

[0123] Specifically, the scheduling optimization engine is triggered synchronously based on prediction and diagnostic conclusions. The Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm model employed within the scheduling optimization engine rapidly simulates and derives the "optimal handover plan" in a digital twin sandbox environment, based on the current network-wide service load and redundant paths: during off-peak hours, video service traffic carried on the at-risk optical path is seamlessly switched over via a pre-configured backup route. After this plan is generated, it is submitted to the policy manager for final approval.

[0124] Step S405: Optical module replacement decision and service protection switchover execution.

[0125] Specifically, operations and maintenance personnel review the predicted alarms, root cause analysis reports, and switchover plans. After confirming that everything is correct, they approve and execute the switchover with a single click through the human-machine interface. The automated controller then executes the verified switchover operation within a preset time window, and monitors business performance indicators in real time throughout the process to ensure a smooth and seamless business switchover.

[0126] Step S406: Closed-loop feedback and model iteration optimization for optical module aging early warning.

[0127] Specifically, the entire operational event, from optical power degradation early warning to protective switching and maintenance decisions, was automatically marked as a high-quality positive sample and stored in a dedicated case library. In subsequent periodic training tasks, these newly added samples will be used to incrementally train and optimize multiple core AI models, including prediction, diagnosis, and scheduling. This will make the prediction model more sensitive to the "optical power degradation caused by optical module aging" pattern, the diagnostic model more accurate in root cause localization, and the generation and execution of scheduling strategies more efficient, thereby continuously improving the ability to warn and handle similar hidden hardware risks.

[0128] This embodiment can proactively address optical module aging failures before they actually impact services, transforming traditional reactive repairs into intelligent prevention, significantly improving the reliability, resource utilization efficiency, and user experience of television transmission networks.

[0129] An electronic device according to an embodiment of the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any of the above-mentioned operation and maintenance methods for a television transmission network management system. That is, an electronic device according to an embodiment of the present invention may include, but is not limited to: a processor and a memory; the memory is used to store the computer program; the processor is used to execute the operation and maintenance method for a television transmission network management system shown in any embodiment of the present invention by calling the computer program.

[0130] In one alternative embodiment, an electronic device is provided, such as Figure 5 As shown, Figure 5 The illustrated electronic device 4000 includes a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, for example, via a bus 4002. Optionally, the electronic device 4000 may further include a transceiver 4004, which can be used for data interaction between the electronic device and other electronic devices, such as sending and / or receiving data. It should be noted that in practical applications, the transceiver 4004 is not limited to one type, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present invention.

[0131] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this invention. Processor 4001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0132] Bus 4002 may include a path for transmitting information between the aforementioned components. Bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 4002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The bus 4002 is represented by only one thick line, but this does not mean that there is only one bus or one type of bus.

[0133] The memory 4003 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0134] The memory 4003 stores application code (computer program) for executing the present invention, and its execution is controlled by the processor 4001. The processor 4001 executes the application code stored in the memory 4003 to implement the content shown in the foregoing method embodiments.

[0135] Among them, electronic devices can also be terminal devices, which can be any device that can install applications, including at least one of smartphones, tablets, laptops, desktop computers, smart speakers, smartwatches, smart TVs, and smart in-vehicle devices.

[0136] It should be noted that, Figure 5 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0137] An embodiment of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements any of the above-described operation and maintenance methods for a television transmission network management system.

[0138] Alternatively, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, a floppy disk, and an optical data storage device, etc.

[0139] In an exemplary embodiment, a computer program product or computer program is also provided, which includes computer instructions stored in a computer-readable storage medium. The processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the aforementioned operation and maintenance method for the television transmission network management system.

[0140] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0141] It should be understood that the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0142] The computer-readable storage medium provided in this invention can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EEPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0143] The aforementioned computer-readable storage medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the method shown in the above embodiments.

[0144] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this invention.

[0145] It should be noted that the terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and represent a limitation on a specific order or sequence. Where appropriate, the order of use for similar objects can be interchanged so that the embodiments of this application described herein can be implemented in an order other than that shown or described.

[0146] Those skilled in the art will recognize that this invention can be implemented as a system, method, or computer program product. Therefore, this invention can be specifically implemented in the following forms: it can be entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, this invention can also be implemented as a computer program product contained in one or more computer-readable media, which includes computer-readable program code.

[0147] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A television transmission network management system, characterized in that, include: The data acquisition module is used to acquire multi-source heterogeneous data in the television transmission network in real time; The data processing module is used to perform spatiotemporal alignment, data governance, and feature engineering on the collected multi-source heterogeneous data to form a network state data stream. The intelligent analysis module is used to perform collaborative analysis on the network status data stream through multiple interconnected artificial intelligence models, thereby generating at least one of the following analysis results: predictive alarms, fault root cause location results, resource optimization strategies, and quantitative correlation between network performance and user experience. The instruction execution module is used to generate network control instructions based on the analysis results and send the network control instructions to the target device in the television transmission network to drive the target device to perform corresponding operations, thereby realizing at least one of predictive maintenance, fault repair and dynamic resource scheduling of the television transmission network; wherein at least some of the network control instructions are security verified before being sent.

2. The television transmission network management system according to claim 1, characterized in that, The system also includes: The model optimization module is used to collect the full-link data corresponding to each operation and maintenance event, from collaborative analysis to network control command execution, after each operation and maintenance event is completed, and to iteratively optimize the multiple artificial intelligence models based on the full-link data. The operation and maintenance event refers to a complete network control operation process triggered by the intelligent analysis module and completed by the instruction execution module.

3. The television transmission network management system according to claim 1, characterized in that, The data acquisition module includes at least two of the following units: The network data acquisition unit is used to collect operating status data, performance index data, and configuration parameter data of various network element devices in the television transmission network. The service data acquisition unit is used to capture the service data stream flowing through the television transmission network by probes deployed at multiple key nodes of the television transmission network, and to collect network transport layer indicators and multi-level error monitoring information conforming to the TR101-290 standard from the service data stream through deep packet inspection or dedicated protocol analysis. The user experience acquisition unit is used to collect at least one type of user experience quality data from the logs of the technical monitoring equipment and in conjunction with the quality evaluation of the bitstream analyzer, including signal loss, cyclic redundancy check error, black screen, still frame, and mute. An external data acquisition unit is used to connect to an external data source through an external data interface to acquire external data related to the operation of the television transmission network.

4. The television transmission network management system according to claim 1, characterized in that, The data processing module includes: The spatiotemporal alignment unit is used to add a unified timestamp, device / service identifier, and topology location information to the multi-source heterogeneous data collected by the data acquisition module, forming a raw data stream with spatiotemporal tags; The spatiotemporal data lake adopts a hybrid storage structure of time-series database and graph database to receive and store the raw data stream with spatiotemporal tags; wherein, the time-series database is used to store time-series sequence data of performance indicators and status data, and the graph database is used to store network topology relationship data representing the relationships between devices, ports and links. The data governance unit is used to perform data cleaning, standardization, normalization and data alignment operations on the data stored in the spatiotemporal data lake; The feature engineering unit is used to extract high-level temporal features from data-governed time-series data, extract graph structure features from data-governed network topology data, and fuse the high-level temporal features and the graph structure features to form the network state data stream.

5. The television transmission network management system according to claim 2, characterized in that, The intelligent analysis module includes: The correlation analysis unit is used to establish and update the quantitative correlation between network performance and user experience based on the network key performance index matrix, user experience quality index matrix and the full-link data in the network state data stream, using causal discovery algorithms and attribution analysis models. The prediction unit is used to employ a model that combines a spatiotemporal convolutional network with an attention mechanism, and to analyze the historical alarm time-series characteristics, historical performance time-series characteristics, and graph structure characteristics in the network state data stream based on the current quantized correlation, so as to generate predictive alarms for network device or link performance degradation. The diagnostic unit is used to employ a heterogeneous graph neural network and, based on the predictive alarm and the current quantized correlation, to perform fault propagation analysis on the real-time alarm timing characteristics, real-time performance timing characteristics, and graph structure characteristics in the network state data stream, so as to generate diagnostic results including the fault root cause location results, location confidence, and inference path. The scheduling optimization unit is used to employ a multi-agent deep deterministic policy gradient algorithm, and based on the diagnostic results and the current quantized correlation, analyze the temporal characteristics of the entire network state in the network state data stream based on the current network state, so as to generate at least one resource optimization strategy among dynamic routing strategy, adaptive coding strategy, and protection switching pre-decision.

6. The television transmission network management system according to claim 1, characterized in that, The instruction execution module includes: The strategy management unit is used to verify and merge the analysis results and generate the network control commands. A digital twin sandbox is used to simulate and deduce the network control command in a digital twin model when the network control command is a preset command, so as to verify the security of the network control command. The intelligent control unit is used to issue network control commands to the target device through a standardized northbound interface or script; wherein the issued network control commands include network control commands that do not require verification and / or verified network control commands.

7. The television transmission network management system according to claim 1, characterized in that, The system also includes: The human-computer interaction module is used to provide a human-computer interaction interface and to visualize and alarm the intermediate process data and output results of the intelligent analysis module and the instruction execution module.

8. The television transmission network management system according to claim 7, characterized in that, The human-computer interaction module includes at least one of the following functional units: The global situation map unit is used to map the real-time health, load and alarm level of network devices or links in a three-dimensional network topology view by using the color, size and dynamic effects of the network devices or links, and to scroll and display the predictive alarms and the root cause location results of the faults. The fault cause-effect chain unit is used to visualize the propagation chain, root cause node, and reasoning path of a fault event in a way that combines timeline and topology graph. The predictive maintenance dashboard unit is used to display a list of risky devices or links, a predicted performance degradation trend curve, and a recommended maintenance plan for a specified future period. The model operation and maintenance center unit is used to provide performance monitoring views, version management interfaces, sample annotation tools, and retraining trigger entry points for the multiple artificial intelligence models.

9. A method for operating and maintaining a television transmission network management system, characterized in that, include: Real-time acquisition of multi-source heterogeneous data in television transmission networks; The collected multi-source heterogeneous data is subjected to spatiotemporal alignment, data governance, and feature engineering to form a network state data stream; The network status data stream is analyzed collaboratively by multiple interconnected artificial intelligence models to generate at least one of the following analysis results: predictive alarms, root cause location results, resource optimization strategies, and quantitative correlation between network performance and user experience. Based on the analysis results, network control commands are generated and sent to target devices in the television transmission network to drive the target devices to perform corresponding operations, thereby achieving at least one of predictive maintenance, fault repair, and dynamic resource scheduling of the television transmission network; wherein at least some of the network control commands are security verified before being sent.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the operation and maintenance method of the television transmission network management system as described in claim 9.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the operation and maintenance method of the television transmission network management system as described in claim 9.

Images