A cloud-based and local early warning customized service system

Abstract

The application provides an early warning customized service system based on cloud and local end cooperation, the system comprises: a basic resource layer for integrating cloud computing resources and local hardware resources; a data exchange layer for accessing multi-source heterogeneous meteorological data of satellite broadcast, international meteorological exchange system and local observation system; a digital base layer for cleaning, fusing, storing and managing the gathered data; an algorithm processing layer for identifying meteorological risks and generating early warning decision results; an integrated platform layer for encapsulating computing, data and algorithm capabilities as unified services; a business application layer for realizing comprehensive monitoring, forecast analysis and early warning information release based on the services. The application has the beneficial effects that through the cooperation and deep customization design of the cloud and the edge, the data transmission minimization, the on-demand utilization of computing resources and the efficient, flexible and accurate improvement of early warning capability are realized while ensuring the business continuity in extreme scenarios.

Description

Technical Field

[0001] This invention belongs to the field of meteorological early warning, and in particular relates to an early warning customized service system based on cloud and local terminal collaboration. Background Technology

[0002] As global climate change intensifies, extreme weather and climate events are becoming more frequent and recurring, posing a severe challenge to human society. Effective early warning systems are crucial for disaster mitigation; according to the Global Commission on Adaptation, providing 24-hour advance notice of impending disasters can reduce losses by 30%.

[0003] In 2022, the United Nations launched the "Early Warning for All Initiative," aiming to ensure that everyone on Earth is protected by early warning systems by 2027. The World Health Organization (WMO) has designated this initiative as its flagship project and established risk assessment, monitoring and early warning, information dissemination, and emergency response as the "four pillars" supporting early warning for all. However, early warning capabilities are currently unevenly distributed globally, with only half of the world's countries possessing comprehensive multi-hazard early warning systems. There are significant gaps between developing and developed countries in their monitoring and forecasting, early warning information dissemination, and emergency response capabilities.

[0004] With the widespread adoption of IoT devices and the massive amounts of data generated in real time by various sensors, early warning systems face new technological challenges. Traditional early warning architectures primarily rely on two technical approaches and their inherent drawbacks: Pure cloud-based alert mode: All data must be uploaded to the cloud for processing. While this mode can leverage the powerful computing capabilities of the cloud for complex analysis, limited network bandwidth or excessive cloud load can lead to data transmission delays or even interruptions, making it difficult to meet the demands of alert scenarios with extremely high real-time requirements (such as security monitoring and emergency shutdown of equipment malfunctions). The instability of cloud input / output (IO) capabilities directly affects the timeliness of alerts.

[0005] Pure local alert mode: All calculations and decisions are completed on local devices. While this mode offers excellent real-time response, it has high initial hardware procurement and deployment costs, and generates ongoing maintenance, updates, and manpower costs. It requires a professional IT team for daily maintenance, system upgrades, security protection, and troubleshooting, resulting in a heavy technical burden.

[0006] Furthermore, existing systems typically lack flexibility, making it difficult to provide personalized early warning rules and threshold settings based on the needs of different users or scenarios. For example, in different scenarios such as agricultural frost warnings, urban flooding warnings, and port and shipping warnings, users have varying needs for early warning thresholds, transmission channels, and response methods.

[0007] Therefore, there is an urgent need for a new early warning architecture that can balance real-time performance with intelligent analysis capabilities and dynamically adjust workflows based on network conditions. This architecture should effectively leverage the powerful computing capabilities of the cloud and the real-time response advantages of the edge, achieving optimized allocation of early warning resources and maximizing early warning effectiveness through the deep integration of cloud intelligence and edge computing. This would break through the technical bottlenecks of traditional early warning models and provide more reliable technical support for the global response to climate change challenges. Summary of the Invention

[0008] In view of this, the present invention aims to propose an early warning customized service system based on cloud and local collaboration, in order to solve at least one of the above-mentioned technical problems.

[0009] To achieve the above objectives, the technical solution of the present invention is implemented as follows: The first aspect of this invention proposes an early warning customization service system based on cloud and local collaboration. The system adopts a layered architecture and includes: The basic resource layer is used to integrate cloud computing resources and local hardware resources to build a dynamically allocated computing resource pool. The data exchange layer is used to access multi-source heterogeneous meteorological data from satellite broadcasting, international meteorological exchange systems, and local observation systems, and to standardize and aggregate the data. The digital foundation layer is used to clean, integrate, store, and manage the collected data to form a meteorological data resource system. The algorithm processing layer is used to analyze and calculate the data based on intelligent meteorological models and artificial intelligence algorithms, identify meteorological risks, and generate early warning decision results; The integrated platform layer is used to encapsulate computing, data, and algorithm capabilities into unified services and provide them to the upper layers for invocation. The business application layer is used to implement comprehensive monitoring, forecast analysis, and early warning information dissemination based on the aforementioned services; The cloud provides global meteorological data and algorithm capabilities, while the local terminal integrates cloud data and local data to achieve localized early warning services.

[0010] Furthermore, the basic resource layer adopts a collaborative deployment approach of cloud computing resources and local integrated computing devices. The cloud is used to carry out global meteorological data processing and intelligent model calculation tasks, while the local device is used to perform early warning analysis tasks with high requirements for timeliness and data security.

[0011] Furthermore, the data exchange layer includes a localized data bus, used for unified access, format conversion, data governance, and exchange scheduling of global meteorological data and local observation data distributed from the cloud; The localized data bus supports mixed processing of real-time data streams and offline batch data, and schedules and controls the data transmission and processing order based on data type and business priority.

[0012] Furthermore, the digital base layer includes a multi-level data storage structure for storing raw meteorological data, fused intermediate data, and thematic data for early warning applications, respectively.

[0013] Furthermore, the algorithm processing layer includes an artificial intelligence-based meteorological analysis model, which is used to identify anomalies in multi-source meteorological elements and output corresponding risk levels and response suggestions; The algorithm processing layer generates early warning analysis results with uncertainty descriptions by performing set analysis or weighted fusion of calculation results from different forecast models.

[0014] Furthermore, the integrated platform layer adopts a microservice architecture to decouple data processing services, algorithm computing services, and early warning services, and enables cross-module calls through standardized interfaces; The microservice architecture is deployed based on containerization technology and uses a unified scheduling mechanism to achieve automatic service deployment, elastic scaling, and fault isolation.

[0015] Furthermore, the system also includes a configurable toolbox module for supporting parameterized configuration and componentized combination of early warning rules, algorithm model parameters, and early warning display methods.

[0016] The second aspect of this invention proposes a method for customized early warning services based on cloud-based and local-based collaboration, applied to the customized early warning service system based on cloud-based and local-based collaboration described in the first aspect, comprising the following steps: S1. Acquire global meteorological data in the cloud and perform intelligent model calculations; S2. Access local meteorological observation data on the local end and aggregate it with data sent from the cloud; S3. Clean, merge, and store the collected data to build meteorological data resources; S4. Perform intelligent analysis based on the meteorological data resources to identify meteorological risks and generate early warning decision results; S5. Provide the early warning decision results to business applications through platform services to realize the release of early warning information.

[0017] Furthermore, in steps S1 and S2, the cloud is used to acquire global meteorological data and perform calculations on meteorological models and artificial intelligence models that require high computing resources, while the local end is used to access local meteorological observation data and perform unified access, format conversion, data governance and scheduling processing of cloud data and local data through a localized data bus. The data bus supports both real-time data stream processing and offline batch data processing, and controls the data processing order according to data type or business priority.

[0018] In steps S3 to S5, the collected meteorological data are stored as raw data, intermediate data after fusion processing, and thematic data for early warning applications, respectively. Anomalies of multi-source meteorological elements are identified based on an artificial intelligence meteorological analysis model. By performing aggregate analysis or weighted fusion on the calculation results from different forecast models, an early warning analysis result containing meteorological risk level, response suggestions, and uncertainty description is generated. The early warning analysis result is customized and published according to the configured early warning rules, algorithm parameters, or display method.

[0019] Compared with existing technologies, the early warning customization service system based on cloud and local terminal collaboration described in this invention has the following beneficial effects: This architecture significantly optimizes resource utilization through a collaborative mechanism. Edge nodes do not upload all raw data to the cloud; instead, they execute data filtering and intelligent upload strategies, synchronizing only fragments of abnormal events, key metadata, or aggregation results to the cloud. This model drastically reduces the amount of data transmitted over the network, saving valuable bandwidth resources and cloud storage costs. Furthermore, the powerful computing resources of the cloud can be dynamically allocated to each edge node on demand, avoiding idle waste and achieving efficient utilization of computing resources.

[0020] The system supports deep customization to meet the needs of different users and scenarios. Users can customize early warning rules, thresholds, and algorithm models for specific regions or disaster types (such as the glacial lake outburst floods in Pakistan) through the cloud platform. This flexibility is thanks to the modular "intelligent early warning toolkit" design and configurable parameter system. Business function components are also designed to be decoupled from the underlying technology, facilitating rapid combination and expansion to meet diverse application scenarios.

[0021] The system successfully addresses the pain points of traditional architectures. The edge computing layer ensures business continuity under extreme conditions, maintaining core early warning functions uninterrupted even during network instability or cloud outages. The cloud layer provides global optimization and decision support, ensuring the accuracy and foresight of early warnings. Attached Figure Description

[0022] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a schematic diagram of the cloud + edge collaborative early warning customized service system architecture described in Embodiment 3 of the present invention. Detailed Implementation

[0023] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0024] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0025] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0026] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0027] The primary objective of this invention is to overcome the shortcomings of existing pure cloud-based early warning modes, such as response delays and insufficient intelligent analysis capabilities of pure local early warning modes, and to provide a cloud + edge collaborative early warning system. This system aims to respond to the UN's initiative on early warning for all and serve the urgent needs of Belt and Road Initiative countries in addressing climate change. By constructing an architecture that dynamically coordinates edge real-time processing with cloud-based intelligent decision-making, it achieves multiple technological benefits: First, it utilizes inference engines deployed on overseas cloud nodes for preliminary data processing and immediate early warning, reducing critical warning latency to milliseconds and significantly improving real-time response. Second, through network status awareness and local disaster recovery mechanisms, it ensures that the local edge system can independently operate core early warning functions when networks are unstable or cloud services are interrupted in some Belt and Road countries and regions, guaranteeing business continuity and system reliability. Third, it leverages the powerful computing power and data aggregation capabilities of the cloud to train complex models and conduct global analysis, and distributes customized early warning rules and algorithm models to users in different countries, regions, and industries, achieving intelligent customization, such as a special development for early warning needs related to the glacial lake outburst flood in Pakistan. Finally, through a collaborative mechanism, it optimizes data upload strategies, enabling overseas cloud nodes to upload only fragments of abnormal events or key metadata, significantly reducing unnecessary raw data stream transmission, effectively saving network bandwidth and cloud storage costs, improving resource utilization efficiency, and helping developing countries with varying infrastructures obtain advanced early warning capabilities at a lower cost.

[0028] Example 1: A customized early warning service system based on cloud and local collaboration, the system adopting a layered architecture, including: The basic resource layer is used to integrate cloud computing resources and local hardware resources to build a dynamically allocated computing resource pool. The data exchange layer is used to access multi-source heterogeneous meteorological data from satellite broadcasting, international meteorological exchange systems, and local observation systems, and to standardize and aggregate the data. The digital foundation layer is used to clean, integrate, store, and manage the collected data to form a meteorological data resource system. The algorithm processing layer is used to analyze and calculate the data based on intelligent meteorological models and artificial intelligence algorithms, identify meteorological risks, and generate early warning decision results; The integrated platform layer is used to encapsulate computing, data, and algorithm capabilities into unified services and provide them to the upper layers for invocation. The business application layer is used to implement comprehensive monitoring, forecast analysis, and early warning information dissemination based on the aforementioned services; The cloud provides global meteorological data and algorithm capabilities, while the local terminal integrates cloud data and local data to achieve localized early warning services.

[0029] The basic resource layer adopts a collaborative deployment approach of cloud computing resources and local integrated computing devices. The cloud is used to carry out global meteorological data processing and intelligent model calculation tasks, while the local device is used to perform early warning analysis tasks with high requirements for timeliness and data security.

[0030] The data exchange layer includes a localized data bus, which is used for unified access, format conversion, data governance, and exchange scheduling of global meteorological data and local observation data sent from the cloud. The localized data bus supports mixed processing of real-time data streams and offline batch data, and schedules and controls the data transmission and processing order based on data type and business priority.

[0031] The digital base layer includes a multi-level data storage structure for storing raw meteorological data, fused intermediate data, and thematic data for early warning applications.

[0032] The algorithm processing layer includes an artificial intelligence-based meteorological analysis model, which is used to identify anomalies in multi-source meteorological elements and output corresponding risk levels and response suggestions. The algorithm processing layer generates early warning analysis results with uncertainty descriptions by performing set analysis or weighted fusion of calculation results from different forecast models.

[0033] The integrated platform layer adopts a microservice architecture, decoupling data processing services, algorithm computing services, and early warning services, and enabling cross-module calls through standardized interfaces; The microservice architecture is deployed based on containerization technology and uses a unified scheduling mechanism to achieve automatic service deployment, elastic scaling, and fault isolation.

[0034] The system also includes a configurable toolbox module, which supports parameterized configuration and componentized combination of early warning rules, algorithm model parameters, and early warning display methods.

[0035] Example 2: A method for customized early warning services based on cloud-based and local-based collaboration, applied to the customized early warning service system based on cloud-based and local-based collaboration described in Example 1, includes the following steps: S1. Acquire global meteorological data in the cloud and perform intelligent model calculations; S2. Access local meteorological observation data on the local end and aggregate it with data sent from the cloud; S3. Clean, merge, and store the collected data to build meteorological data resources; S4. Perform intelligent analysis based on the meteorological data resources to identify meteorological risks and generate early warning decision results; S5. Provide the early warning decision results to business applications through platform services to realize the release of early warning information.

[0036] In steps S1 and S2, the cloud is used to acquire global meteorological data and perform calculations on meteorological and artificial intelligence models that require high computing resources, while the local end is used to access local meteorological observation data and perform unified access, format conversion, data governance, and scheduling processing of cloud data and local data through a localized data bus. The data bus supports both real-time data stream processing and offline batch data processing, and controls the data processing order according to data type or business priority.

[0037] In steps S3 to S5, the collected meteorological data are stored as raw data, intermediate data after fusion processing, and thematic data for early warning applications, respectively. Anomalies of multi-source meteorological elements are identified based on an artificial intelligence meteorological analysis model. By performing aggregate analysis or weighted fusion on the calculation results from different forecast models, an early warning analysis result containing meteorological risk level, response suggestions, and uncertainty description is generated. The early warning analysis result is customized and published according to the configured early warning rules, algorithm parameters, or display method.

[0038] Example 3: like Figure 1 As shown, based on the systems and methods described in Embodiments 1 and 2, a cloud-edge collaborative early warning customized service system is established. Following the information-based and intensive design concept of "large system, large platform, multiple applications," the system adopts a layered construction logic, including a six-layer architecture: basic resource layer, data exchange layer, digital foundation layer, algorithm processing layer, integrated platform layer, business application layer, and user layer, as detailed below: Basic resource layer: By flexibly integrating overseas cloud resources (such as elastic computing, storage, and networking) with local all-in-one hardware, a computing resource pool that can be dynamically allocated on demand is built. It supports all upper-layer services, and its "cloud + edge" dual-mode design can not only take advantage of the elasticity and low cost of the cloud, but also meet the needs of data localization, independent control, and high-performance computing in specific scenarios.

[0039] Data Exchange Layer: This layer is the system's "information artery," responsible for building efficient and reliable data input channels. By integrating multiple sources such as CMACast satellite broadcasting, WIS 2.0 international exchange data, and local data, it achieves standardized access and aggregation of multi-source, heterogeneous meteorological data, providing high-quality data supply to the upper layers and serving as a key guarantee for promoting data fusion and sharing.

[0040] Digital Foundation Layer: This layer is the system's "data hub," dedicated to achieving intelligent management of the entire data lifecycle. It follows a pipeline of "data transmission → data processing → data storage," cleaning, integrating, and standardizing massive amounts of data, and constructing a data resource system including a basic meteorological database and a specialized early warning database, providing a solid data foundation for accurate early warning analysis.

[0041] Algorithm Processing Layer: This is the system's "intelligent brain," with a built-in intelligent meteorological model and various professional algorithms forming the intelligent computing engine. It can quickly calculate standardized meteorological elements, identify anomalies, and optimize the accuracy of early warning products through AI models (such as the landing area correction model), ultimately outputting an intelligent decision result of "risk level + response suggestions."

[0042] Integrated Platform Layer: This layer acts as a "capability middleware," encapsulating the underlying computing resources, data services, and algorithm tools into unified, reusable services (such as map services and early warning release services). It adopts a "capability layering + open collaboration" design, effectively aggregating common technical capabilities, enabling upper-layer applications to be quickly built and flexibly invoked, thereby significantly improving development efficiency and system integration.

[0043] Business Application Layer: As the direct interface for end users, this layer leverages the capabilities provided by the integrated platform layer to develop core business functions such as comprehensive monitoring, model forecasting, and early warning. It ensures that early warning information can reach meteorological personnel in various countries intuitively and efficiently, meeting their practical application needs in disaster prevention and mitigation.

[0044] The data architecture of the aforementioned cloud + edge collaborative early warning customized service system is as follows: The top layer of the architecture comprises diverse data sources, including CMA global observation and forecast data, observation data from other countries, international reanalysis data, and AI large-scale model data. This data is first aggregated into an overseas cloud data resource pool, forming a global, standardized data foundation. The foundational data from the overseas cloud is then transferred to a localized data bus. Simultaneously, local observation and forecast data generated by partner countries are also incorporated into this data bus. The data bus standardizes, integrates, and manages global data from the cloud with local data generated at the endpoint, generating high-quality, usable data products that directly support business applications. These data products are ultimately delivered to local product application services, providing partner countries with accurate and reliable early warning services.

[0045] The deployment design of the aforementioned cloud + edge collaborative early warning customized service system is as follows: The deployment of a layered, decoupled, and data-driven "cloud + edge" deployment model aims to efficiently leverage the global data and intelligent algorithm capabilities of the China Meteorological Administration to empower localized meteorological early warning applications in partner countries. Data is integrated from both the cloud and edge into a "localized data bus" for processing and then used to serve edge applications, achieving a complete collaborative closed loop.

[0046] The overseas cloud platform serves as the global data and intelligence hub for the entire system. It is responsible for collecting and aggregating data from multiple authoritative sources, including global observation and model forecast data from CMA, observational data from other countries, mainstream international numerical model products, and AI models and algorithms developed by CMA. The platform's core function is to standardize, quality-control, and integrate this massive and diverse global data, forming a unified and reliable data resource pool. This provides downstream users with standardized data products and intelligent algorithm support, demonstrating the system's global perspective and powerful computing capabilities on the cloud side.

[0047] The localized data bus serves as a core hub and intelligent bridge connecting the cloud and the edge. It receives standardized global data products from overseas cloud platforms and integrates them with local data resources in partner countries. This bus is not merely a data channel; it undertakes critical tasks of data fusion, governance, exchange, and management, ensuring that data from different sources can be processed and utilized in a standardized manner. It delivers organized and processed observational, forecasting, modeling, and application monitoring data downstream, making it a crucial link in achieving precise alignment and collaborative computing between global capabilities and local needs.

[0048] The local data and application service system of the cooperating countries is the ultimate carrier for realizing the system's value at the "end" side. It first integrates the local data resources of the cooperating countries, such as local observation network data, local forecast products, model algorithms customized for their own geographical and climatic conditions, and basic geographic information. Building on this foundation, the system fully utilizes high-quality data products provided by the localized data bus, which are deeply integrated with global data, to ultimately construct and operate specialized meteorological product application services (such as weather warnings) serving various industries within the country (e.g., aviation, agriculture, disaster prevention and mitigation), thus realizing the localization and operational application of advanced global meteorological capabilities.

[0049] Example 4: The cloud + edge collaborative early warning customized service system described in Example 3 includes at least the following key technologies: Key Technology 1: Microservices Built upon a hybrid microservice architecture based on Kubernetes and Spring Cloud, its core technical logic lies in fully leveraging the synergistic advantages of container orchestration platforms and microservice governance frameworks. Specifically, Docker is used to standardize the encapsulation of applications and their dependencies, ensuring environmental consistency; the powerful container orchestration capabilities of the Kubernetes cluster enable automatic deployment, elastic scaling, load balancing, fault self-healing, and high availability of microservices; and the Spring Cloud framework handles complex logic such as service governance, configuration management, circuit breaking, and rate limiting in distributed systems. This architecture combines the underlying infrastructure management capabilities of Kubernetes with the business-oriented microservice integration capabilities of Spring Cloud, ultimately supporting high-performance, high-concurrency, and easily maintainable cloud-native agile development and deployment through continuous integration and continuous delivery (CI / CD) pipelines.

[0050] Key Technology 2: 2D Map Engine The platform's underlying 2D map engine is deeply customized based on OpenLayers. Through source code-level optimization of core modules such as the rendering engine, data parsing, and memory management, it significantly improves the rendering performance and interactive efficiency of massive amounts of meteorological data. To achieve flexibility in the technology stack and broad business support, the platform adopts a high-level abstraction and adapter design pattern, encapsulating common map functions into a unified interface. This decouples business components from the OpenLayers underlying layer and allows seamless support for Leaflet engine application requirements through a dedicated adapter. This technical approach, combining "deep kernel optimization" and "high-level abstraction adaptation," ensures high performance while maximizing the reusability of business components, reducing development and maintenance costs, and supporting the sustainable evolution of the platform's technology stack.

[0051] Key Technology 3: Componentization The design employs a layered approach (platform general components, meteorological basic components, and business application components) combined with classification to ensure high cohesion and low coupling of component functions. A unified meteorological software application repository enables component registration, release, version control, and shared access. Technically, it supports multi-language development (such as C, C++, Java, and Python) and leverages the pilot-scale simulation environment of the meteorological big data cloud platform to complete automated testing, functional performance verification, and access assessment of components, thereby ensuring component quality and business applicability. Its innovation is particularly evident in its metadata-driven component discovery mechanism and standard interface specifications, enabling flexible assembly and dynamic invocation of components at different granularities to quickly respond to diverse business logic needs. Ultimately, through strict version management and exit mechanisms, a sustainable evolution loop for components is formed.

[0052] Key Technology 4: Distributed Task Scheduling Employing a centralized scheduling model, the engine decouples the scheduling center from the task executors, achieving separation of scheduling and execution. The core engine ensures high availability of the scheduler itself through master-slave election and distributed locking mechanisms, avoiding single points of failure. Simultaneously, relying on automatic task node registration and heartbeat detection mechanisms, it dynamically manages the executor cluster, monitors node status in real time, and removes abnormal instances, ensuring high availability and elastic scalability at both the scheduling and execution levels. To meet diverse business scenarios, the engine incorporates rich scheduling strategies (such as polling, random, failover, consistent hashing, and sharded broadcasting) and can compensate for missed schedules. Finally, by supporting multiple trigger-driven methods including time, event, API, and manual intervention, the engine achieves unified, reliable, and intelligent orchestration and scheduling of complex task flows.

[0053] Key Technology 5: Rapid Access and Processing of Multi-Source Heterogeneous Data The platform accesses a wide range of data sources, diverse types, and varied organizational structures, encompassing various categories of data from the Tianqing Big Data Cloud Platform, CMACast, and local sources, including site monitoring, satellite, numerical models, and service products. This data is presented in multiple formats such as GRIB, HDF, NC, and structured text. To meet the diverse service-level needs and timeliness requirements of different data sources, the system, based on the characteristics of the data sources and a data processing and distribution framework, has designed a rapid access and processing mechanism for diverse and heterogeneous data. This mechanism comprises three core modules: raw data acquisition, data parsing and processing, and data storage and transmission.

[0054] The raw data acquisition module relies on a data processing and distribution framework, employing differentiated acquisition methods tailored to the characteristics and business requirements of different data types. For highly structured data such as site observations, the platform acquires the latest content through data interface services. For gridded data such as numerical patterns, the platform performs streaming processing based on dynamically acquired latest data to ensure data real-time performance and integrity. Furthermore, for similar data obtainable from multiple channels, the platform prioritizes timeliness, automatically selecting the data source with the highest timeliness and shortest processing time.

[0055] The data parsing and processing module aims to unify the parsing of raw data from diverse sources and formats into standard formats such as COGTIFF and JSON, which are universally applicable to the platform. It also incorporates various quality control and feature extraction methods to meet diverse business needs. The platform acquires raw data using dynamic streaming files, ensuring that parsing begins immediately upon data arrival. For different data types, the platform is equipped with corresponding parsing and processing components, such as a universal template configuration numerical model parsing component for standard GRIB and NC format data, and a satellite disk map parsing component with geographic information registration and reference system conversion capabilities. Based on this, the platform establishes a multi-level processing workflow based on processes and threads: a process model is used between different tasks to achieve transaction isolation and disaster recovery rollback, while a thread pool is used within the same task to improve efficiency and reduce error handling overhead. Simultaneously, the module also reserves several data parsing and processing frameworks to accommodate various localized data access requirements.

[0056] The data management and transmission module is responsible for storing, managing, and distributing the processed data in real time. Based on key characteristics such as data structure and storage method, the platform adopts categorized storage and transmission strategies: unstructured data is uniformly stored using a combination of database indexes and file storage, and transmitted via file methods tailored to different storage types such as NAS and object storage; structured data employs a two-tiered storage approach—long-term storage using a business database combined with short-term storage using a cache database—and data transmission based on structured text. Regarding data management, the platform implements hierarchical management and periodic cleanup based on data's business purpose and access frequency. This categorized management approach improves data transmission speed and management efficiency to a certain extent, thereby ensuring real-time data transmission and low-latency access.

[0057] Example 4: The cloud + edge collaborative early warning customized service system described in Example 3 consists of the following components: This system constructs an integrated, data-driven, and intelligent collaborative technical architecture covering the entire chain of "monitoring—forecasting—early warning—service." Relying on a comprehensive observation network (including Fengyun meteorological satellites, next-generation weather radars, and national automatic weather stations, among other space-based and ground-based facilities), it collects multi-dimensional data in real time. Through multi-source intelligent forecasting engines (such as AI models like "Fenglei" and "Fengqing"), it fuses and analyzes massive amounts of observational data and numerical forecast products to generate gridded forecast products ranging from hourly to seasonal scales. At the localized application layer, the disaster identification and dissemination module, based on edge computing capabilities and combined with regional characteristics, dynamically determines disaster risk thresholds and generates early warning products. The system also provides a modular toolbox, supporting functions such as early warning rule configuration, product editing, and GIS visualization, allowing users to flexibly customize service processes according to industry needs. Finally, AI intelligent components serve as the core driver, using a knowledge base (integrating historical cases, defense guidelines, etc.) and machine learning models (such as disaster identification and location correction algorithms) to automatically generate and optimize early warning products, forming a closed-loop chain from monitoring, analysis, decision-making to service.

[0058] Integrated observation module: Achieving high-precision, multi-source, all-weather meteorological monitoring globally, the early weather warning system, through an "observation-as-a-service" engineering design, will overcome the geographical and technological barriers of traditional observation resources. It will provide low-cost, high-efficiency observation support for countries along the Belt and Road Initiative, helping to build an integrated early warning network of "global monitoring - regional collaboration - local response." By monitoring convective cloud clusters, tropical cyclones, and other hazardous weather systems in real time, the platform can provide users with intuitive, high-definition images of weather evolution. Simultaneously, the platform will integrate station observations and grid analysis data to achieve comprehensive global monitoring of multiple elements (temperature, wind speed, humidity, air pressure, precipitation, etc.). Furthermore, the monitoring targets will be expanded to include climate and ecological indicators such as ENSO, IOD, MJO, drought index, and NDVI. This will provide long-term references for agriculture, ecological protection, water resource management, and other fields, making it an infrastructure for multi-domain integrated applications.

[0059] Multi-source intelligent forecasting module; This system enables intelligent forecasting across multiple models and sources, integrating major global numerical models such as CMA_GFS, ECMWF_IFS, NCEP_GFS, JMA-GSM, and ICON. It also incorporates AI meteorological models, leveraging artificial intelligence to create intelligent forecast products. This allows for comprehensive diagnosis of multiple timeframes and factors (such as precipitation, circulation patterns, dynamic conditions, water vapor conditions, dust storm forecasts, and wave forecasts), and the generation of diagnostic reports with a single click. Furthermore, it incorporates technologies such as multi-model comparison, model bias analysis, and ensemble forecast stability assessment, providing forecasters with interactive and visualized analytical tools.

[0060] Local disaster identification and reporting module: The local disaster identification and dissemination module has constructed a seamless early warning technology system covering the interannual climate scale to the short-term weather scale. At the climate prediction level, the platform, based on reanalysis data and multi-source models, enables the prediction of key factors such as ENSO, monsoon, and sea surface temperature anomalies, supporting medium- and long-term climate anomaly early warning. At the weather forecast level, relying on the intelligent digital weather forecasting system (GOWFS), it provides high-precision hourly / 3-hourly element forecasts for the next 10 days, with a focus on accurately capturing extreme temperatures, strong winds, and other hazardous weather events. For major disasters such as tropical cyclones, it further integrates multi-model track forecasts, AI intelligent algorithms, and ensemble probability analysis to generate comprehensive forecast products that include track, intensity, and probability of impact, significantly improving forecast accuracy.

[0061] At the information dissemination level, the platform is committed to transforming early warning information into a direct driving force for emergency action. It personalizes early warning rules and thresholds based on different users or scenarios, intelligently identifies risk categories and regions, automatically generates structured early warning products, and achieves customized information delivery across borders and regions through a cloud-edge-device collaborative dissemination architecture. The system supports targeted dissemination through multiple channels such as large screens and mobile applications, ensuring rapid sharing of early warning information through local nodes even in environments with limited communication, ultimately forming a closed loop from intelligent disaster perception to precise information delivery.

[0062] Toolbox: The toolkit is designed with openness, customizability, and scalability as its core principles, aiming to provide users with flexible and autonomous configuration capabilities for different business scenarios. Through open parameter configuration, the platform allows users to adjust warning rules, threshold parameters, and display styles according to specific needs, achieving personalized customization of business logic without modifying the underlying code. At the algorithm level, the system supports flexible customization of models and algorithms. Users can combine, adjust, or import custom models based on pre-built algorithm libraries, and ensure algorithm reliability through continuous integration and pilot-scale testing processes. Meanwhile, the display interface is highly customizable. Users can personalize interactive elements such as menu items, sidebars, map layers, and pop-up dialog boxes by dragging and dropping components, customizing layouts, and configuring themes. Relying on notification and push message mechanisms, it achieves multi-platform adaptation and hierarchical delivery of warning information. Furthermore, the toolkit adopts a modular architecture, achieving loosely coupled integration of functional components through standardized interfaces, significantly improving the system's adaptability in cross-platform deployment, business iteration, and ecosystem expansion.

[0063] AI intelligent components: AI-powered intelligent functions have built a comprehensive capability platform based on meteorological knowledge base and Large Language Model (LLM), forming a dual-core application system covering intelligent analysis of forecasts and early warnings and decision support for business processes. It integrates multi-dimensional knowledge such as meteorological disaster case databases, regional climate backgrounds, and personalized early warning rules, and combines the basic capabilities of generative LLM, such as document generation, thought chain reasoning, and RAG retrieval enhancement, to provide intelligent drive for upper-level business.

[0064] The core functions are manifested in two main directions: (1) The intelligent forecast and early warning analysis module focuses on professional judgment and provides functions such as intelligent weather map recognition, extreme weather circulation pattern analysis, matching of similar cases of disastrous weather, and extreme value record reminders, which significantly improves the foresight and accuracy of disaster identification; (2) The forecaster assistant module deeply integrates with the business process, supports the production and release of meteorological special reports, the auxiliary generation of weather consultation materials, and the workflow arrangement and duty assistance based on AI-Agent, which effectively optimizes business efficiency.

[0065] Furthermore, this module boasts robust capabilities for localized service integration and component-based encapsulation. It supports the encapsulation of local observational data and core early warning models (such as GLOF warnings and monsoon forecasts) from partner countries into standardized AI application components. Based on template libraries and intelligent agent technology, it automatically generates multilingual service products tailored to regional needs. Forecasters can complete the entire process of subjective revision, review, and release of products through an interactive interface, achieving precise adaptation and efficient empowerment of globally advanced AI capabilities with local business scenarios.

[0066] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

[0067] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A customized early warning service system based on cloud and local terminal collaboration, characterized in that, The system adopts a layered architecture, including: The basic resource layer is used to integrate cloud computing resources and local hardware resources to build a dynamically allocated computing resource pool. The data exchange layer is used to access multi-source heterogeneous meteorological data from satellite broadcasting, international meteorological exchange systems, and local observation systems, and to standardize and aggregate the data. The digital foundation layer is used to clean, integrate, store, and manage the collected data to form a meteorological data resource system. The algorithm processing layer is used to analyze and calculate the data based on intelligent meteorological models and artificial intelligence algorithms, identify meteorological risks, and generate early warning decision results; The integrated platform layer is used to encapsulate computing, data, and algorithm capabilities into unified services and provide them to the upper layers for invocation. The business application layer is used to implement comprehensive monitoring, forecast analysis, and early warning information dissemination based on the aforementioned services; The cloud provides global meteorological data and algorithm capabilities, while the local terminal integrates cloud data and local data to achieve localized early warning services.

2. The early warning customized service system based on cloud and local terminal collaboration according to claim 1, characterized in that: To run complex artificial The basic resource layer adopts a collaborative deployment approach of cloud computing resources and local integrated computing devices. The cloud is used to carry out global meteorological data processing and intelligent model calculation tasks, while the local device is used to perform early warning analysis tasks with high requirements for timeliness and data security.

3. The early warning customized service system based on cloud and local terminal collaboration according to claim 1, characterized in that: The data exchange layer includes a localized data bus, which is used for unified access, format conversion, data governance, and exchange scheduling of global meteorological data and local observation data sent from the cloud. The localized data bus supports mixed processing of real-time data streams and offline batch data, and schedules and controls the data transmission and processing order based on data type and business priority.

4. The early warning customized service system based on cloud and local terminal collaboration according to claim 1, characterized in that: The digital base layer includes a multi-level data storage structure for storing raw meteorological data, fused intermediate data, and thematic data for early warning applications.

5. The early warning customized service system based on cloud and local terminal collaboration according to claim 1, characterized in that: The algorithm processing layer includes an artificial intelligence-based meteorological analysis model, which is used to identify anomalies in multi-source meteorological elements and output corresponding risk levels and response suggestions. The algorithm processing layer generates early warning analysis results with uncertainty descriptions by performing set analysis or weighted fusion of calculation results from different forecast models.

6. The early warning customized service system based on cloud and local terminal collaboration according to claim 1, characterized in that: The integrated platform layer adopts a microservice architecture, decoupling data processing services, algorithm computing services, and early warning services, and enabling cross-module calls through standardized interfaces; The microservice architecture is deployed based on containerization technology and uses a unified scheduling mechanism to achieve automatic service deployment, elastic scaling, and fault isolation.

7. The early warning customized service system based on cloud and local terminal collaboration according to claim 1, characterized in that: The system also includes a configurable toolbox module, which supports parameterized configuration and componentized combination of early warning rules, algorithm model parameters, and early warning display methods.

8. A method for customized early warning services based on cloud-based and local terminal collaboration, applied to the customized early warning service system based on cloud-based and local terminal collaboration as described in any one of claims 1-7, characterized in that, Includes the following steps: S1. Acquire global meteorological data in the cloud and perform intelligent model calculations; S2. Access local meteorological observation data on the local end and aggregate it with data sent from the cloud; S3. Clean, merge, and store the collected data to build meteorological data resources; S4. Perform intelligent analysis based on the meteorological data resources to identify meteorological risks and generate early warning decision results; S5. Provide the early warning decision results to business applications through platform services to realize the release of early warning information.

9. The method for customized early warning services based on cloud and local terminal collaboration according to claim 8, characterized in that: In steps S1 and S2, the cloud is used to acquire global meteorological data and perform calculations on meteorological and artificial intelligence models that require high computing resources, while the local end is used to access local meteorological observation data and perform unified access, format conversion, data governance, and scheduling processing of cloud data and local data through a localized data bus. The data bus supports both real-time data stream processing and offline batch data processing, and controls the data processing order according to data type or business priority.

10. The method for customized early warning services based on cloud and local terminal collaboration according to claim 8, characterized in that: In steps S3 to S5, the collected meteorological data are stored as raw data, intermediate data after fusion processing, and thematic data for early warning applications, respectively. Anomalies of multi-source meteorological elements are identified based on an artificial intelligence meteorological analysis model. By performing aggregate analysis or weighted fusion on the calculation results from different forecast models, an early warning analysis result containing meteorological risk level, response suggestions, and uncertainty description is generated. The early warning analysis result is customized and published according to the configured early warning rules, algorithm parameters, or display method.

Images