5G + rural human settlement environment long-acting management and control platform
By combining 5G IoT terminal clusters, dense heterogeneous networks, and edge computing layers, the problems of limited data collection range and poor timeliness in traditional environmental monitoring have been solved. Real-time, full-area, high-precision collection and reliable transmission of rural environmental data have been achieved, improving the comprehensiveness and long-term effectiveness of environmental management.
Patent Information
- Application Number
- CN202511151773.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-18
- Publication Date
- 2025-11-11
AI Technical Summary
Traditional monitoring methods suffer from limited data collection scope and poor timeliness, transmission networks are prone to coverage blind spots, data processing lacks preprocessing capabilities, and intelligent analysis models have a single dimension, resulting in insufficient comprehensiveness and long-term effectiveness of environmental management strategies.
By employing 5G IoT terminal clusters, a dense heterogeneous network architecture, an edge computing layer, and an intelligent analysis hub, combined with multi-source sensing, preprocessing, deep learning, and blockchain technologies, we can achieve real-time collection, reliable transmission, intelligent analysis, and decision optimization of environmental data.
It has enabled real-time, comprehensive, and high-precision collection and reliable transmission of environmental data in rural areas, improved the comprehensiveness and timeliness of environmental situation awareness, enhanced the dimensional richness of environmental assessment models and the scientific nature of decision optimization, and formed a highly efficient management and control mechanism across the entire chain.
Smart Images
Figure CN120935525A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of management and control platform technology, specifically a 5G+ long-term management and control platform for rural living environment. Background Technology
[0002] The integration of information technology and ecological environment governance is deepening, especially in the field of rural living environment management. How to utilize next-generation communication technologies, the Internet of Things, and artificial intelligence to achieve accurate data collection, efficient transmission, and intelligent analysis of environmental data has become a key direction for improving rural environmental governance. With the maturity of 5G technology and the development of edge computing and artificial intelligence algorithms, building an intelligent environmental management platform covering the entire chain of data perception, transmission, processing, and decision-making has become an important research topic in the industry.
[0003] Traditional monitoring methods often rely on single-type sensors or manual inspections, resulting in limited data collection scope and poor timeliness, making it difficult to achieve comprehensive, real-time environmental information acquisition. Regarding transmission networks, ordinary communication architectures are prone to coverage blind spots in complex rural terrains and lack service isolation and quality-of-service (QoS) assurance mechanisms, leading to low reliability of critical environmental data transmission. In data processing, the edge side lacks effective preprocessing capabilities, with raw data being directly transmitted to the central end, increasing computational pressure and potentially affecting analytical accuracy due to outliers. At the intelligent analysis level, traditional models are one-dimensional, making it difficult to integrate multi-source heterogeneous data to construct a comprehensive environmental assessment system. Decision optimization relies on empirical methods and lacks scientific strategy generation mechanisms based on advanced algorithms such as deep reinforcement learning. Furthermore, it lacks a composite knowledge system encompassing knowledge graphs, historical cases, and policies and regulations, resulting in insufficient comprehensiveness and long-term effectiveness of environmental management strategies. Summary of the Invention
[0004] To address the shortcomings of existing technologies, this invention provides a 5G+ rural living environment long-term management and control platform to solve the problems mentioned in the background technology, such as increased computational pressure, potential impact of outliers on analysis accuracy, and insufficient comprehensiveness and long-term effectiveness of environmental management and control strategies.
[0005] To achieve the above objectives, the present invention provides the following technical solution: a 5G+ rural living environment long-term management and control platform, comprising:
[0006] Multi-source sensing layer: Composed of a group of 5G IoT terminals deployed in rural areas, including an environmental monitoring sensor array, an intelligent image acquisition device and a drone inspection system. The sensor array integrates a PM2.5 / PM10 dual-mode detection module, a COD water quality monitoring unit and a soil moisture monitoring component.
[0007] 5G transmission network: It adopts a dense heterogeneous network architecture, including macro base stations, micro base stations and visible light communication nodes, and is equipped with a network slicing management module to achieve service isolation and set up a QoS guarantee mechanism;
[0008] Edge computing layer: Deploys edge nodes with AI acceleration chips, with a built-in preprocessing engine to perform data cleaning, outlier detection and spatiotemporal alignment operations, and is configured with lightweight deep learning models;
[0009] Intelligent Analysis Hub: Composed of a distributed computing cluster, it includes: an environmental situation awareness subsystem, which constructs a multi-dimensional environmental assessment model based on a spatiotemporal convolutional neural network; a decision optimization engine, which uses deep reinforcement learning algorithms to generate control strategies and integrates multi-objective optimization algorithms to balance environmental benefits; and a long-term management database, which constructs a composite knowledge system including an environmental governance knowledge graph, a historical case library, and a policy and regulation library.
[0010] Terminal Interaction Layer: Develop cross-platform application systems, integrate GIS visualization modules, public participation interfaces and government regulatory portals, and configure blockchain evidence storage components.
[0011] Preferably, the multi-source sensing layer is configured with an adaptive sampling mechanism, which dynamically adjusts the sensor's working mode by calculating the environmental entropy value, and automatically triggers the sampling mode when the pollution concentration exceeds the threshold.
[0012] This mechanism combines environmental spatiotemporal characteristics with multi-source data fusion to optimize sampling frequency and regional coverage strategies, achieving a balance between the dual objectives of accurate pollution source tracking and low-power operation.
[0013] Preferably, the 5G transmission network is equipped with a mobility management unit, which employs a predictive handover algorithm and combines UAV positioning data to enable the monitoring terminal to switch between base stations;
[0014] The predictive handover algorithm analyzes the overlap between the real-time trajectory of the drone and the coverage area of the base station, dynamically predicts the handover timing based on the base station load, and introduces a multi-path transmission protocol to improve the reliability of data transmission, ensuring the continuity and low latency of monitoring data in high-speed mobile scenarios.
[0015] Preferably, the preprocessing engine of the edge computing layer integrates an unsupervised anomaly detection model, uses an improved isolated forest algorithm to identify sensor fault data, and configures a dynamic threshold adjustment mechanism to adapt to seasonal environmental changes;
[0016] The improved isolated forest algorithm enhances anomaly detection accuracy by introducing spatiotemporal feature encoding, while the dynamic threshold adjustment mechanism constructs a threshold surface based on historical seasonal data, automatically adapting to environmental changes such as rainy / dry seasons and avoiding misjudgments caused by seasonal fluctuations.
[0017] Preferably, the environmental situation awareness subsystem of the intelligent analysis center introduces a transfer learning mechanism, uses urban environmental data to pre-train a model, and achieves adaptation to rural scenarios through feature alignment;
[0018] The transfer learning mechanism adopts a domain-adaptive approach to reduce the differences in the distribution of urban and rural environmental data, and combines the unique environmental elements of rural areas to reconstruct the feature space, thereby improving the model's generalization ability and fine-grained environmental assessment accuracy in complex rural scenarios.
[0019] Preferably, the decision optimization engine configuration strategy interpretation module uses the SHAP value analysis method to generate a feasibility assessment report of control measures to help decision-makers understand AI suggestions;
[0020] SHAP value analysis incorporates multiple dimensions such as environmental benefits, economic costs, and social acceptance into its explanatory framework. It uses visual charts to show the contribution of each factor to the strategy and generates accessible explanatory texts based on real-world cases of rural governance, thereby enhancing grassroots decision-makers' trust in AI recommendations and their willingness to implement them.
[0021] Preferably, the long-term management and control database establishes an environmental governance effect prediction model, trains an LSTM network based on historical intervention data, and realizes the ex-ante effect evaluation of management and control strategies;
[0022] The LSTM network input integrates multi-timescale features with environmental element data, and by combining it with causal relationship chains in the knowledge graph, it enhances the predictive model's responsiveness and interpretability to complex environmental interventions.
[0023] Preferably, the blockchain evidence storage component adopts a consortium blockchain architecture, sets up smart contracts to automatically execute data uploading operations, and configures national cryptographic algorithms to ensure data transmission security;
[0024] The consortium blockchain nodes are composed of multiple authoritative entities, including environmental protection departments, township governments, and third-party monitoring agencies. The smart contract triggering conditions cover the entire process of data collection, transmission, and analysis. The on-chain data is encrypted and stored using the national cryptographic SM4 algorithm to ensure the authority, immutability, and compliance of the stored data.
[0025] Preferably, the cross-platform application system is configured with a self-evolutionary learning module to continuously optimize the analysis model through online incremental learning, and a model version rollback mechanism is set up;
[0026] Incremental learning is triggered by scenarios such as significant changes in environmental data distribution and policy and regulatory updates. The model version rollback mechanism retains multiple historical versions and associates them with performance evaluation indicators. When the performance of the new model deteriorates, it automatically switches to the optimal historical version to ensure the stability and adaptability of the system.
[0027] Compared with existing technologies, this invention provides a 5G+ rural living environment long-term management and control platform, which has the following beneficial effects:
[0028] This 5G+ rural living environment long-term management and control platform, through the deployment of environmental monitoring sensor arrays, intelligent image acquisition devices and drone inspection systems in a multi-source perception layer, combined with the ultra-dense heterogeneous architecture of the 5G transmission network and network slicing management module, realizes real-time, full-domain, high-precision collection and reliable transmission of environmental data in rural areas, ensuring the comprehensiveness and timeliness of environmental situation awareness;
[0029] Meanwhile, the edge computing layer cleans, detects anomalies, and aligns the raw data in time and space through a built-in preprocessing engine and a lightweight deep learning model. Combined with the spatiotemporal convolutional neural network and deep reinforcement learning algorithm of the intelligent analysis center, it effectively improves the dimensionality of the environmental assessment model and the scientific nature of the decision optimization engine. In turn, it balances environmental benefits through the knowledge graph of the long-term management database and multi-objective optimization algorithms, forming a highly efficient management mechanism covering the entire chain of data collection, transmission, analysis, and decision-making. Attached Figure Description
[0030] Figure 1 This is a schematic diagram of the structure of the present invention. Detailed Implementation
[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0032] Please see Figure 1 This invention provides a technical solution: a 5G+ rural living environment long-term management and control platform, comprising:
[0033] Multi-source sensing layer: Composed of a group of 5G IoT terminals deployed in rural areas, including an environmental monitoring sensor array, intelligent image acquisition device and drone inspection system. The sensor array integrates a PM2.5 / PM10 dual-mode detection module, a COD water quality monitoring unit and a soil moisture monitoring component.
[0034] A layered deployment strategy is adopted, with high-precision sensors and low-power devices working together. The sampling frequency is dynamically adjusted through the 5G network, and data pre-screening is achieved by combining edge computing nodes to reduce the amount of invalid data transmission. The image acquisition device is equipped with an adaptive illumination adjustment algorithm to improve the recognition accuracy under complex weather conditions. The drone inspection system integrates a path planning algorithm to support automatic patrol in key areas and rapid response to sudden pollution incidents.
[0035] 5G transmission network: It adopts a dense heterogeneous network architecture, including macro base stations, micro base stations and visible light communication nodes, and is equipped with a network slicing management module to achieve service isolation and set up a QoS guarantee mechanism;
[0036] The ultra-dense heterogeneous network forms a three-dimensional network by using macro base stations to cover a wide area, micro base stations to enhance local capacity, and visible light communication nodes to supplement indoor / obstructed area coverage; the network slicing management module dynamically allocates resources based on AI algorithms, allocating dedicated virtual networks for different service types such as environmental monitoring, image transmission, and control commands; the QoS guarantee mechanism ensures low-latency transmission of critical data through priority scheduling and bandwidth reservation.
[0037] Edge computing layer: Deploys edge nodes with AI acceleration chips, with a built-in preprocessing engine to perform data cleaning, outlier detection and spatiotemporal alignment operations, and is configured with lightweight deep learning models;
[0038] The edge nodes adopt a heterogeneous computing architecture, integrating AI acceleration chips and general-purpose processors to support parallel model inference; the preprocessing engine combines a rule engine with unsupervised learning algorithms to achieve dynamic data cleaning threshold adjustment; the lightweight deep learning model adopts model pruning and quantization techniques to adapt to the computing power limitations of edge devices, while supporting online incremental learning to adapt to changes in environmental data distribution.
[0039] Intelligent Analysis Hub: Composed of a distributed computing cluster, it includes: an environmental situation awareness subsystem, which constructs a multi-dimensional environmental assessment model based on a spatiotemporal convolutional neural network; a decision optimization engine, which uses deep reinforcement learning algorithms to generate control strategies and integrates multi-objective optimization algorithms to balance environmental benefits; and a long-term management database, which constructs a composite knowledge system including an environmental governance knowledge graph, a historical case library, and a policy and regulation library.
[0040] The environmental situation awareness subsystem integrates geographic information and time-series data through a spatiotemporal convolutional neural network to support cross-regional environmental trend prediction; the decision optimization engine combines deep reinforcement learning and multi-objective optimization algorithms to dynamically weigh environmental benefits, governance costs, social acceptance, and other dimensions; the long-term management database uses a graph database to store knowledge graphs and combines natural language processing technology to achieve automatic policy and regulation parsing and case matching.
[0041] Terminal Interaction Layer: Develop cross-platform application systems, integrate GIS visualization modules, public participation interfaces and government regulatory portals, and configure blockchain evidence storage components;
[0042] The cross-platform application system adopts a responsive design, supporting seamless adaptation to PC, mobile, and large-screen devices; the GIS visualization module integrates 3D modeling and heat map rendering technologies, supporting multi-level environmental data overlay display; the public participation interface enables functions such as pollution reporting and submission of governance suggestions through mini-programs and official accounts, and enhances participation enthusiasm by combining a credit score mechanism; the government supervision portal integrates a process engine, supporting the issuance of control instructions and tracking of execution progress; the blockchain evidence storage component adopts a consortium blockchain architecture, hashing and uploading key information such as monitoring data, decision records, and execution results to the blockchain to ensure full traceability.
[0043] This platform utilizes an environmental monitoring sensor array, intelligent image acquisition device, and drone inspection system deployed in a multi-source perception layer. Combined with the ultra-dense heterogeneous architecture of the 5G transmission network and network slicing management module, it achieves real-time, full-domain, high-precision acquisition and reliable transmission of environmental data in rural areas, ensuring the comprehensiveness and timeliness of environmental situation awareness. Simultaneously, the edge computing layer cleans, detects anomalies, and aligns the raw data in time and space through a built-in preprocessing engine and lightweight deep learning model. This, along with the spatiotemporal convolutional neural network and deep reinforcement learning algorithm in the intelligent analysis center, effectively enhances the dimensional richness of the environmental assessment model and the scientific nature of the decision optimization engine's strategy. Furthermore, by using the knowledge graph of the long-term management database and multi-objective optimization algorithms to balance environmental benefits, it forms a highly efficient management mechanism covering the entire chain of data acquisition, transmission, analysis, and decision-making.
[0044] The multi-source sensing layer is configured with an adaptive sampling mechanism, which dynamically adjusts the sensor's working mode by calculating the environmental entropy value. When the pollution concentration exceeds the threshold, the sampling mode is automatically triggered.
[0045] This mechanism combines environmental spatiotemporal characteristics with multi-source data fusion to optimize sampling frequency and regional coverage strategies, achieving a balance between the dual objectives of accurate pollution source tracking and low-power operation.
[0046] The 5G transmission network is equipped with a mobility management unit, which uses a predictive handover algorithm and combines drone positioning data to enable the monitoring terminal to switch between base stations;
[0047] The predictive handover algorithm analyzes the overlap between the real-time trajectory of the drone and the coverage area of the base station, dynamically predicts the handover timing based on the base station load, and introduces a multi-path transmission protocol to improve the reliability of data transmission, ensuring the continuity and low latency of monitoring data in high-speed mobile scenarios.
[0048] The preprocessing engine of the edge computing layer integrates an unsupervised anomaly detection model, uses an improved isolated forest algorithm to identify sensor fault data, and configures a dynamic threshold adjustment mechanism to adapt to seasonal environmental changes.
[0049] The improved isolated forest algorithm enhances anomaly detection accuracy by introducing spatiotemporal feature encoding, while the dynamic threshold adjustment mechanism constructs a threshold surface based on historical seasonal data, automatically adapting to environmental changes such as rainy / dry seasons and avoiding misjudgments caused by seasonal fluctuations.
[0050] The environmental situation awareness subsystem of the intelligent analysis center introduces a transfer learning mechanism, uses urban environmental data to pre-train a model, and achieves adaptation to rural scenarios through feature alignment.
[0051] The transfer learning mechanism adopts a domain-adaptive approach to reduce the differences in the distribution of urban and rural environmental data, and combines the unique environmental elements of rural areas to reconstruct the feature space, thereby improving the model's generalization ability and fine-grained environmental assessment accuracy in complex rural scenarios.
[0052] The decision optimization engine configuration strategy interpretation module uses the SHAP value analysis method to generate a feasibility assessment report of control measures to help decision-makers understand AI suggestions;
[0053] SHAP value analysis incorporates multiple dimensions such as environmental benefits, economic costs, and social acceptance into its explanatory framework. It uses visual charts to show the contribution of each factor to the strategy and generates accessible explanatory texts based on real-world cases of rural governance, thereby enhancing grassroots decision-makers' trust in AI recommendations and their willingness to implement them.
[0054] A long-term management and control database is used to establish an environmental governance effectiveness prediction model. An LSTM network is trained based on historical intervention data to achieve ex-ante effect evaluation of management and control strategies.
[0055] The LSTM network input integrates multi-timescale features with environmental element data, and by combining it with causal relationship chains in the knowledge graph, it enhances the predictive model's responsiveness and interpretability to complex environmental interventions.
[0056] The blockchain evidence storage component adopts a consortium blockchain architecture, sets up smart contracts to automatically execute data uploading operations, and configures national cryptographic algorithms to ensure data transmission security.
[0057] The consortium blockchain nodes are composed of multiple authoritative entities, including environmental protection departments, township governments, and third-party monitoring agencies. The smart contract triggering conditions cover the entire process of data collection, transmission, and analysis. The on-chain data is encrypted and stored using the national cryptographic SM4 algorithm to ensure the authority, immutability, and compliance of the stored data.
[0058] Cross-platform application systems are configured with self-evolutionary learning modules to continuously optimize and analyze models through online incremental learning, and a model version rollback mechanism is set up.
[0059] Incremental learning is triggered by scenarios such as significant changes in environmental data distribution and policy and regulatory updates. The model version rollback mechanism retains multiple historical versions and associates them with performance evaluation indicators. When the performance of the new model deteriorates, it automatically switches to the optimal historical version to ensure the stability and adaptability of the system.
[0060] Data acquisition phase: The multi-source sensing layer, through the deployment of environmental monitoring sensor arrays, intelligent image acquisition devices and drone inspection systems in rural areas, combined with an adaptive sampling mechanism, achieves real-time, full-area, and high-precision acquisition of environmental data in rural areas, balancing the goals of accurate pollution source tracking and low-power operation.
[0061] Data transmission phase: The 5G transmission network adopts a dense heterogeneous network architecture, is configured with a network slicing management module and QoS guarantee mechanism, and sets up a mobility management unit to ensure reliable, continuous and low-latency transmission of environmental data.
[0062] Data preprocessing stage: Edge computing layer deploys edge nodes with AI acceleration chips, and the built-in preprocessing engine performs data cleaning, outlier detection and spatiotemporal alignment operations to reduce the amount of invalid data transmission and improve data quality.
[0063] Data analysis and decision-making phase:
[0064] The environmental situation awareness subsystem of the intelligent analysis center integrates geographic information and time series data to support cross-regional environmental trend prediction.
[0065] The decision optimization engine dynamically weighs multiple dimensions such as environmental benefits, governance costs, and social acceptance to help decision-makers understand AI recommendations.
[0066] Knowledge storage and effect prediction stage: The long-term management database is constructed as a composite knowledge system including an environmental governance knowledge graph, a historical case database, and a policy and regulation database, providing knowledge support and effect prediction for decision-making.
[0067] Data storage and system optimization phase:
[0068] The blockchain evidence storage component in the terminal interaction layer hashes and uploads key information such as monitoring data, decision records, and execution results to the blockchain to ensure traceability throughout the entire process.
[0069] Cross-platform application systems are configured with self-evolutionary learning modules that automatically optimize models or roll back to the best historical version based on scenarios such as changes in environmental data distribution and updates to policies and regulations, ensuring the stability and adaptability of system operation.
[0070] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.
[0071] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A 5G+ long-term management and control platform for rural living environment, characterized in that, include: Multi-source sensing layer: Composed of a group of 5G IoT terminals deployed in rural areas, including an environmental monitoring sensor array, an intelligent image acquisition device and a drone inspection system. The sensor array integrates a PM2.5 / PM10 dual-mode detection module, a COD water quality monitoring unit and a soil moisture monitoring component. 5G transmission network: It adopts a dense heterogeneous network architecture, including macro base stations, micro base stations and visible light communication nodes, and is equipped with a network slicing management module to achieve service isolation and set up a QoS guarantee mechanism; Edge computing layer: Deploys edge nodes with AI acceleration chips, with a built-in preprocessing engine to perform data cleaning, outlier detection and spatiotemporal alignment operations, and is configured with lightweight deep learning models; Intelligent Analysis Hub: Composed of a distributed computing cluster, it includes: an environmental situation awareness subsystem, which constructs a multi-dimensional environmental assessment model based on a spatiotemporal convolutional neural network; a decision optimization engine, which uses deep reinforcement learning algorithms to generate control strategies and integrates multi-objective optimization algorithms to balance environmental benefits; and a long-term management database, which constructs a composite knowledge system including an environmental governance knowledge graph, a historical case library, and a policy and regulation library. Terminal Interaction Layer: Develop cross-platform application systems, integrate GIS visualization modules, public participation interfaces and government regulatory portals, and configure blockchain evidence storage components.
2. The 5G+ rural living environment long-term management and control platform according to claim 1, characterized in that: The multi-source sensing layer is configured with an adaptive sampling mechanism, which dynamically adjusts the sensor's working mode by calculating the environmental entropy value. When the pollution concentration exceeds the threshold, the sampling mode is automatically triggered.
3. The 5G+rural living environment long-term management and control platform according to claim 1, characterized in that: The 5G transmission network is equipped with a mobility management unit, which uses a predictive handover algorithm and combines drone positioning data to enable the monitoring terminal to switch between base stations.
4. The 5G+rural living environment long-term management and control platform according to claim 1, characterized in that: The preprocessing engine of the edge computing layer integrates an unsupervised anomaly detection model, uses an improved isolated forest algorithm to identify sensor fault data, and configures a dynamic threshold adjustment mechanism to adapt to seasonal environmental changes.
5. The 5G+ rural living environment long-term management and control platform according to claim 1, characterized in that: The environmental situation awareness subsystem of the intelligent analysis center introduces a transfer learning mechanism, uses urban environmental data to pre-train a model, and achieves adaptation to rural scenarios through feature alignment.
6. The 5G+ rural living environment long-term management and control platform according to claim 1, characterized in that: The decision optimization engine configuration strategy interpretation module uses the SHAP value analysis method to generate a feasibility assessment report of control measures, assisting decision-makers in understanding AI recommendations.
7. The 5G+ rural living environment long-term management and control platform according to claim 1, characterized in that: The long-term management and control database establishes an environmental governance effect prediction model, trains an LSTM network based on historical intervention data, and realizes the ex-ante effect evaluation of management and control strategies.
8. The 5G+ rural living environment long-term management and control platform according to claim 1, characterized in that: The blockchain evidence storage component adopts a consortium blockchain architecture, sets up smart contracts to automatically execute data uploading operations, and configures national cryptographic algorithms to ensure data transmission security.
9. A 5G+ rural living environment long-term management and control platform according to claim 1, characterized in that: The cross-platform application system is equipped with a self-evolutionary learning module, which continuously optimizes and analyzes the model through online incremental learning, and sets up a model version rollback mechanism.