A meteorological business application system based on a general large model

Abstract

The application provides a meteorological business application system based on a general large model, wherein a data base module is configured to acquire multi-source heterogeneous meteorological data related to meteorological business, perform abnormality detection, format standardization, interpolation correction processing on the multi-source heterogeneous meteorological data, and obtain structured data; real-time collection of various computing power indexes of the meteorological business application system, and output of the structured data and the various computing power indexes to a wisdom brain module; the wisdom brain module is configured to associate and match the structured data with meteorological business standard processes stored in a knowledge base and algorithms preset in an algorithm library based on a dynamic routing mechanism, determine an optimal model combination from a model library in combination with the computing power indexes; and perform processing on the structured data based on the optimal model combination to obtain a business processing result of the meteorological business. The above scheme can effectively improve the accuracy of the meteorological business processing result and optimize the overall business capability of the system.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and more specifically, to a meteorological operational application system based on a general large model. Background Technology

[0002] Meteorological services are a core support for ensuring the orderly operation of key social sectors such as agricultural production, transportation, and disaster prevention. The core requirement is to achieve efficient processing of meteorological data and accurate output of operational results through meteorological operational application systems. Existing typical meteorological operational application systems usually include a data acquisition module and a fixed model processing module: the data acquisition module pulls multi-source meteorological data through a preset interface, performs only simple deduplication and format conversion, and then directly passes it to the fixed model processing module; the latter has a built-in single model for specific operations (such as short-term precipitation forecasting and temperature trend prediction), with fixed parameters and calling logic. After receiving data, it directly runs and outputs results. The system's computing power is monitored and inspected periodically by humans, and the model's running priority is manually adjusted based on the inspection results.

[0003] The aforementioned technologies suffer from significant deficiencies in data processing accuracy: the data acquisition module simply deduplicates and converts the format, resulting in poor consistency, residual outliers, and missing key data in the data input to the model; the fixed model is based on low-quality data processing, which is prone to input bias, and the final output business results (such as forecast bias and misjudgment of disaster level) are not accurate enough, making it difficult to support the meteorological business's demand for data accuracy, and may even lead to false alarms or delays in early warning due to data errors, affecting the scientific nature of decisions such as disaster prevention and mitigation and industry services. Summary of the Invention

[0004] In view of this, the purpose of this invention is to provide a meteorological business application system based on a general large model, which can effectively improve the accuracy of meteorological business processing results and optimize the overall business capabilities of the system.

[0005] This application provides a meteorological operational application system based on a general large model, the meteorological operational application system including a data base module and an intelligent brain module;

[0006] The data base module is used to: acquire multi-source heterogeneous meteorological data related to meteorological operations; perform anomaly detection, format standardization, and interpolation correction on the multi-source heterogeneous meteorological data to obtain structured data; collect various computing power indicators of the meteorological operations application system in real time; and output the structured data and various computing power indicators to the intelligent brain module respectively.

[0007] The intelligent brain module is used to: based on a dynamic routing mechanism, associate and match the structured data with the meteorological business standard process stored in the knowledge base and the preset algorithms in the algorithm library, and determine the optimal model combination from the model library in combination with the computing power index; and process the structured data based on the optimal model combination to obtain the business processing result of the meteorological business.

[0008] Optionally, the meteorological business application system also includes a support middleware module;

[0009] The intelligent brain module is also used to: combine the computing power index and business priority to generate a business scheduling instruction, and output the business processing result and the business scheduling instruction to the support platform module;

[0010] The support platform module is used to: schedule corresponding algorithm resources in the algorithm library of the intelligent brain module through the algorithm platform according to the business scheduling instructions; and integrate the components required for meteorological business applications through the business platform to form an initial resource package; based on the requirements of the meteorological business scenario, screen and adapt the initial resource package to obtain an adapted resource package; use the adapted resource package to support the implementation of meteorological business functions, and collect status data in real time during the execution of meteorological business; combine the status data with the business processing results to generate resource adjustment suggestions and feed them back to the intelligent brain module.

[0011] The intelligent brain module is also used to: correct the model parameters in the optimal model combination according to the resource adjustment suggestions.

[0012] Optionally, the meteorological business application system further includes a business application module;

[0013] The business application module is used to: receive the adapted resource package output by the support platform module; start the workflow engine, and according to the business scenario information carried in the adapted resource package, call the business processing logic of the corresponding module and perform business analysis according to the standardized process; during the analysis, call the business processing results output by the intelligent brain module, and combine the business processing results to complete the deep processing of the corresponding business and generate multimodal business products.

[0014] Optionally, the meteorological business application system further includes a smart service module;

[0015] The intelligent service module is used to: obtain the multimodal business product from the business application module; call preset user profile data, match the multimodal business product with the user profile, and determine the appropriate push channel; and accurately push the multimodal business product to the corresponding user or industry scenario through the determined push channel.

[0016] Optionally, the intelligent service module is further configured to: initiate a streaming computing framework to collect user feedback information on the product in real time under various scenarios; perform structured processing on the feedback information to obtain structured feedback data; transmit the structured feedback data to the support platform module, and the support platform module triggers the system to adjust and optimize the data processing flow, model inference parameters, and business execution flow to form a closed loop.

[0017] Optionally, the data base module includes a station network observation information operation platform, a meteorological big data cloud platform, a meteorological comprehensive business monitoring system, and a meteorological data quality control and processing system;

[0018] The station network observation information operation platform is used to: collect multi-source heterogeneous meteorological data required for target meteorological operations, including satellite remote sensing data, radar echo data, and ground observation station data, and transmit the collected multi-source heterogeneous meteorological data to the meteorological data quality control and processing system;

[0019] The meteorological data quality control and processing system is used to: receive multi-source heterogeneous meteorological data transmitted from the station network observation information operation platform, perform anomaly detection, format standardization, and interpolation correction processing on the multi-source heterogeneous meteorological data in sequence, remove outliers in the data, unify the data format, and fill in data gaps, finally obtain structured data, and transmit the structured data to the meteorological big data cloud platform;

[0020] The meteorological big data cloud platform is used to: receive structured data output from the meteorological data quality control and processing system, classify, store and manage the structured data using a distributed storage architecture, and provide structured data calling interfaces for other modules of the system.

[0021] The meteorological integrated business monitoring system is used to: collect various computing power indicators during the operation of the meteorological business application system in real time, including GPU / CPU resource utilization, data transmission rate, and remaining storage resources; monitor and identify anomalies in the collected computing power indicators in real time; and output the computing power indicators to the intelligent brain module synchronously.

[0022] Optionally, the intelligent brain module includes a knowledge base, a general large model library, an algorithm library, a component library, and a meteorological large model library;

[0023] The knowledge base is used to store standard procedures, historical business data, emergency response plans, and business scenario adaptation rules in the meteorological business field, providing knowledge support for task recognition and model selection of the intelligent brain module.

[0024] The general large model library is used to store general large models with generalized reasoning capabilities, which can realize cross-scenario business processing based on knowledge transfer in the meteorological field.

[0025] The meteorological big model library is used to store specialized models optimized for meteorological business scenarios, including numerical forecasting models, disaster early warning models, and meteorological product generation models. The meteorological big model has the ability to deeply analyze and process meteorological professional data.

[0026] The algorithm library is used to store various algorithms required for meteorological business processing, including data mining algorithms, machine learning algorithms, reinforcement learning algorithms, and spatiotemporal interpolation algorithms, providing algorithmic support for model processing of structured data;

[0027] The component library is used to store the functional components required in the meteorological business processing process, including data preprocessing components, model inference components, and result visualization components, and to support the intelligent brain module in quickly building the business processing chain.

[0028] The intelligent brain module is used to: invoke information and resources from the knowledge base and algorithm base through a dynamic routing mechanism, and select and combine the optimal model combination from the general large model library and the meteorological large model library in combination with computing power indicators, and perform processing on structured data to obtain business processing results.

[0029] Optionally, the supporting middle platform module includes an AI middle platform, a knowledge middle platform, an algorithm middle platform, a data middle platform, a business middle platform, and a component middle platform;

[0030] The data platform is used to: receive structured data output from the data base module, and achieve semantic alignment of multi-source data through semantic mapping, format unification, and data cleaning to form a standardized data resource pool and provide data support for other platforms;

[0031] The algorithm platform is used to: schedule corresponding algorithm resources from the algorithm library of the intelligent brain module according to the business scheduling instructions output by the intelligent brain module, encapsulate and manage the algorithm resources, and provide algorithm call interfaces for business processing.

[0032] The business platform is used to: integrate various business logic components and process templates required for meteorological business applications, build a standardized business processing framework, and support the rapid configuration and expansion of business scenarios;

[0033] The AI ​​platform is used to manage the model resources of the intelligent brain module, including model deployment, model monitoring, and model iteration, while providing AI model calling services to other modules of the system.

[0034] The knowledge platform is used to: connect to the knowledge base of the intelligent brain module, extract, integrate and update meteorological business knowledge, construct a dynamic knowledge graph, and provide knowledge support for business processing and model optimization;

[0035] The component platform is used to: integrate the functional components required by each module of the system, including interface adaptation components, resource scheduling components, and log management components, to achieve component reuse and unified management;

[0036] The supporting middle platform module is used to: integrate data, algorithms, models, and components into an initial resource package through the collaborative work of various middle platforms, and output it to the business application module after scenario adaptation and adjustment, while receiving feedback information to generate resource adjustment suggestions.

[0037] Optionally, the business application modules include an integrated weather and climate module, a decision-making meteorology module, a professional meteorological service module, an agricultural meteorological service module, a watershed meteorological service module, and a weather modification operation platform module;

[0038] The integrated weather and climate module is used to: receive the adapted resource package output by the support platform module, call the business processing results of the intelligent brain module, perform various business analyses such as weather monitoring, climate trend analysis, and short-term climate prediction, and generate an integrated weather and climate analysis report.

[0039] The decision-making meteorological module is used to: combine the disaster early warning results and impact assessment results output by the intelligent brain module with various business processing data, integrate the relevant data of terrain, population and economy, and generate meteorological service short reports and emergency response suggestions for government decision-making.

[0040] The professional meteorological service module is used to generate customized meteorological service products for different industries, including power, transportation, and culture and tourism, by calling the business processing results of industry-specific components and intelligent brain modules in the adapted resource package. These products include power load forecasts, traffic weather warnings, and scenic spot weather guides.

[0041] The agricultural meteorological service module is used to: combine the agricultural production cycle, crop growth model and meteorological element forecast results output by the intelligent brain module to perform various business analyses such as agricultural meteorological disaster risk assessment, crop growth monitoring and agricultural activity suggestions, and generate agricultural meteorological service products;

[0042] The basin meteorological service module is used to: generate basin meteorological service products such as basin flood warnings and water resource scheduling suggestions by calling basin observation data and the results of precipitation forecasts and flood simulations from the intelligent brain module, in response to the hydrological and meteorological needs of a specific basin.

[0043] The artificial weather modification platform module is used to: receive the cloud water resource analysis and operation condition judgment results output by the intelligent brain module, and combine the operation site distribution and equipment status information to generate artificial weather modification operation plans and operation effect evaluation reports to support the operation of artificial weather modification.

[0044] Optionally, the smart service module includes a transportation service platform, a cultural and tourism service platform, a city and county service platform, a public service platform, a power service platform, an energy service platform, an agricultural service platform, and a government service platform;

[0045] The traffic service platform is used to: receive traffic and meteorological service products output by the business application module, including road icing warnings and visibility forecasts; combine them with user profiles in the traffic industry; and push the products through various channels such as the traffic department's business system, roadside displays, and navigation apps to support traffic safety.

[0046] The cultural and tourism service platform is used to: receive scenic area meteorological service products output by the business application module, including scenic area precipitation forecasts and ultraviolet intensity warnings; combine them with the user profiles of the cultural and tourism industry; and push the products through various channels such as the scenic area's official website, tourism APP, and travel agency service system to assist in the planning of cultural and tourism activities.

[0047] The city and county service platform is used to: receive local meteorological service products output by the business application module, including city and county rainstorm warnings and township meteorological reports; and push products through various channels such as the city and county meteorological department business platform and government affairs APP, based on the needs and characteristics of city and county users, to support grassroots meteorological services.

[0048] The public service platform is used to: receive public meteorological service products output by the business application module, including daily weather forecasts and life weather indices; combine them with public user profiles, including age, occupation, and living habits; and push the products through various channels such as SMS, WeChat official accounts, and meteorological apps to meet the public's daily meteorological information needs.

[0049] The power service platform is used to: receive power meteorological service products output by the business application module, including cooling load forecasts and line icing warnings; and push products through various channels such as the power dispatch system and enterprise-specific service platforms in combination with the needs of power enterprise users to assist in power production dispatch.

[0050] The energy service platform is used to: receive energy meteorological service products output by the business application module, including wind power forecast and photovoltaic irradiance forecast, and push the products through various channels of the energy management system and service platform in combination with the user profile of energy enterprises to support the efficient utilization of new energy.

[0051] The agricultural service platform is used to: receive agricultural meteorological service products output by the business application module, including crop growth period forecasts and meteorological risk warnings for pests and diseases; combine these with farmer user profiles, including crops planted and planting scale; and push the products through various channels such as agricultural APPs and agricultural technology extension platforms to assist agricultural production.

[0052] The government service platform is used to: receive decision-making meteorological service products output by business application modules, including major disaster warnings and emergency response suggestions; and push these products through various channels such as the government collaboration platform and emergency command system, in accordance with the needs of government departments, to support government decision-making and emergency management.

[0053] The technical solution provided in this application includes, but is not limited to, the following beneficial effects:

[0054] The data foundation module effectively improves the consistency and accuracy of the output structured data by performing anomaly detection, format standardization, and interpolation correction on multi-source heterogeneous meteorological data, providing a high-quality data foundation for meteorological business processing. At the same time, it collects and outputs the system's computing power indicators in real time, breaking the lag of traditional manual monitoring of computing power and providing accurate resource status data support for subsequent business processes, thus consolidating the foundation for system operation from both data quality and resource awareness perspectives.

[0055] The intelligent brain module relies on a dynamic routing mechanism to achieve precise association and matching between structured data and meteorological business standard processes and preset algorithms, avoiding the problem of insufficient business adaptability caused by fixed logic. At the same time, it determines the optimal model combination from the model library by combining computing power indicators. This not only ensures the adaptability of model selection to system resources and reduces the risk of computing power waste or insufficient resources, but also improves the accuracy and reliability of meteorological business processing results by processing data through the optimal model combination, thus comprehensively optimizing the system's business processing capabilities.

[0056] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0057] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1 This invention provides a schematic diagram of the structure of a meteorological operational application system according to an embodiment of the invention.

[0059] Figure 2 A schematic diagram of the structure of the second meteorological business application system provided in an embodiment of the present invention is shown;

[0060] Figure 3 This invention provides a schematic diagram of the structure of a third meteorological operational application system.

[0061] Figure 4This invention provides a flowchart of a rainstorm intelligent prediction method based on model scheduling and workflow construction, as illustrated in an embodiment of the present invention.

[0062] Figure 5 This invention provides a schematic diagram of the structure of a fourth meteorological operational application system.

[0063] Figure 6 This invention provides a flowchart of an intelligent early warning system for rainstorm risks based on user profiles, as illustrated in an embodiment of the present invention.

[0064] Figure 7 The flowchart illustrates a method for generating forecast materials based on a meteorological post-assisted architecture, as provided in an embodiment of the present invention. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0066] Example

[0067] To facilitate understanding of this application, the following is combined with... Figure 1 The schematic diagram of the structure of a meteorological operational application system based on a general large model provided by the embodiments of the present invention is shown below, and the contents described herein will be described in detail for the embodiments of this application.

[0068] See Figure 1 As shown, Figure 1 The diagram shows a structural schematic of a meteorological business application system provided by an embodiment of the present invention, wherein the meteorological business application system includes a data base module 101 and a smart brain module 102.

[0069] The data base module is used to: acquire multi-source heterogeneous meteorological data related to meteorological operations; perform anomaly detection, format standardization, and interpolation correction on the multi-source heterogeneous meteorological data to obtain structured data; collect various computing power indicators of the meteorological operations application system in real time; and output the structured data and various computing power indicators to the intelligent brain module respectively.

[0070] Specifically, the core functions of the data base module are the acquisition and processing of multi-source heterogeneous meteorological data for meteorological operations, and the real-time collection of system computing power indicators. The specific technical implementation is as follows:

[0071] It can access multi-dimensional meteorological data from space, air, and ground through standardized interfaces, including satellite remote sensing, radar echoes, ground station observations, and numerical model outputs.

[0072] The system sequentially performs anomaly detection (using statistical tests or machine learning algorithms to remove invalid and extreme values), format standardization (unifying unstructured / semi-structured data into NetCDF format through schema mapping), and interpolation correction (using spatial interpolation to handle uneven spatial distribution of data), outputting structured data.

[0073] The system collects computing power metrics such as GPU / CPU utilization, memory usage, and task response latency in real time, and transmits the structured data and computing power metrics synchronously to the intelligent brain module to provide it with data and resource status support.

[0074] The intelligent brain module is used to: based on a dynamic routing mechanism, associate and match the structured data with the meteorological business standard process stored in the knowledge base and the preset algorithms in the algorithm library, and determine the optimal model combination from the model library in combination with the computing power index; and process the structured data based on the optimal model combination to obtain the business processing result of the meteorological business.

[0075] Specifically, the intelligent brain module uses a dynamic routing mechanism as its core to achieve collaborative scheduling of data, knowledge, algorithms, and models. The specific technical process is as follows:

[0076] After receiving the structured data and computing power indicators from the data base module, the structured data is matched with the meteorological business standard process (including cross-departmental business specification steps) in the knowledge base and the professional algorithms (such as precipitation short-term extrapolation algorithm and disaster risk assessment algorithm) in the algorithm library to determine the business logic and technical framework for data processing.

[0077] By combining real-time computing power indicators (such as GPU resource utilization) to analyze the system's carrying capacity, the model selection is dynamically adapted (high-precision, high-computing-power models are prioritized when GPUs are sufficient, and lightweight models are matched when computing power is tight). The optimal model combination that adapts to business needs and resource status is determined from the model library consisting of general large models and meteorological special large models.

[0078] Based on the optimal model combination, the entire process of structured data is processed (such as calling the meteorological big model to correct numerical forecast results and the general big model to generate risk assessment text in the rainstorm forecast scenario), and outputting business processing results such as forecast and early warning products, risk assessment reports, and service suggestions to support the implementation of subsequent business applications and service scenarios.

[0079] In an optional implementation, see Figure 2 As shown, Figure 2 A schematic diagram of the structure of a second meteorological business application system provided by an embodiment of the present invention is shown, wherein the meteorological business application system further includes a support middleware module 201.

[0080] The intelligent brain module is also used to: generate business scheduling instructions by combining the computing power indicators and business priorities, and output the business processing results and business scheduling instructions to the support platform module.

[0081] Specifically, after completing structured data processing and generating meteorological business processing results, the intelligent brain module will generate business scheduling instructions based on the real-time computing power indicators transmitted by the data base module and the preset business priorities. The specific technical logic is as follows:

[0082] Computing power metrics include system resource status parameters such as GPU / CPU utilization, memory usage, and task response latency; business priorities are divided according to the urgency and importance of meteorological services, such as disaster-related services like rainstorm warnings and typhoon monitoring having higher priority than regular weather broadcasting services.

[0083] The algorithm analyzes the matching relationship between computing power indicators and business priorities. When computing power is sufficient, resources are allocated to high-priority businesses first. When computing power is tight, a resource quota algorithm is used to ensure that core businesses (high priority) run without interruption, while low-priority businesses are subject to resource throttling or queuing.

[0084] The generated business scheduling instructions and meteorological business processing results are synchronously transmitted to the support middleware module. The business scheduling instructions specify the resource allocation rules and business execution priorities, providing a decision-making basis for the subsequent algorithm calls and component integration of the support middleware module.

[0085] The support platform module is used to: schedule corresponding algorithm resources in the algorithm library of the intelligent brain module through the algorithm platform according to the business scheduling instructions; and integrate the components required for meteorological business applications through the business platform to form an initial resource package; based on the requirements of the meteorological business scenario, screen and adapt the initial resource package to obtain an adapted resource package; use the adapted resource package to support the implementation of meteorological business functions, and collect status data in real time during the execution of meteorological business; combine the status data with the business processing results to generate resource adjustment suggestions and feed them back to the intelligent brain module.

[0086] Specifically, the support platform module, as the core connecting the intelligent brain module and the front-end business applications, conducts resource coordination and optimization feedback based on business scheduling instructions and business processing results. The specific technical process is as follows:

[0087] After receiving business scheduling instructions and business processing results, the algorithm platform accurately schedules and matches the algorithm resources for the current business from the algorithm library associated with the intelligent brain module (such as the precipitation forecast business scheduling precipitation short-term extrapolation algorithm and numerical model correction algorithm); at the same time, the business platform integrates the components required for meteorological business (including data interaction components, visualization rendering components, and message push components), and combines the scheduled algorithm resources with the integrated components to form an initial resource package.

[0088] Optimize the initial resource package based on the specific needs of meteorological business scenarios—such as agricultural meteorological service scenarios—by screening resources related to crop growth models and farmland soil moisture analysis, eliminating irrelevant algorithms and components, and adapting and adjusting the parameters of the retained resources to generate an adapted resource package.

[0089] After adaptation, the resource package directly supports the implementation of meteorological business functions (such as driving the agricultural meteorological service system to generate crop disaster risk assessment reports and supporting the watershed meteorological service platform to complete precipitation trend forecasts). During the process, status data such as business processing progress, actual resource utilization rate, and accuracy feedback of business results are collected in real time. These data are then analyzed together with the business processing results output by the intelligent brain module to identify problems such as resource waste (e.g., idle algorithm resources) or insufficient resources (e.g., component response delays). Targeted resource adjustment suggestions such as adjusting computing power allocation and replacing inefficient components are generated and fed back to the intelligent brain module.

[0090] The intelligent brain module is also used to: correct the model parameters in the optimal model combination according to the resource adjustment suggestions.

[0091] Specifically, the intelligent brain module uses the resource adjustment suggestions (including feedback on issues such as model accuracy and computing power consumption) output by the middle platform module as the core input for parameter correction, clarifying the direction of model optimization (such as improving prediction accuracy and reducing computing power consumption).

[0092] If the suggestion points out that the model's prediction accuracy for extreme weather is insufficient and the system has additional computing power redundancy, the model's ability to fit and predict extreme weather characteristics can be improved by adjusting parameters such as increasing the number of model iterations and expanding the feature input dimensions. If the suggestion mentions that the model's computing power consumption is too high and affects the execution of other business operations, model parameters can be optimized by using techniques such as model quantization (e.g., FP16 / INT8 hybrid quantization) and knowledge distillation (transferring knowledge from complex models to lightweight models) to reduce the computing power consumption during model inference.

[0093] By correcting the parameters mentioned above, the optimal model combination can be dynamically adjusted, ensuring that subsequent meteorological operations not only utilize resources efficiently but also improve the accuracy of the results.

[0094] In an optional implementation, see Figure 3 As shown, Figure 3A schematic diagram of the structure of a third meteorological business application system provided in an embodiment of the present invention is shown, wherein the meteorological business application system further includes a business application module 301.

[0095] The business application module is used to: receive the adapted resource package output by the support platform module; start the workflow engine, and according to the business scenario information carried in the adapted resource package, call the business processing logic of the corresponding module and perform business analysis according to the standardized process; during the analysis, call the business processing results output by the intelligent brain module, and combine the business processing results to complete the deep processing of the corresponding business and generate multimodal business products.

[0096] Specifically, the business application module takes the adapted resource package output by the support platform module as input, drives scenario-based business processing through the workflow engine, and achieves in-depth business processing by combining the analysis results of the intelligent brain module, ultimately generating multimodal business products. The specific process is as follows:

[0097] The system receives the adapted resource package output by the support platform module. This resource package has been screened and adapted according to the needs of meteorological business scenarios. It includes scenario-matching algorithm resources, component resources and configuration information, providing accurate resource support for subsequent processing.

[0098] The workflow engine is activated, first parsing the business scenario information carried in the resource package (such as agricultural meteorological crop disaster risk assessment, watershed meteorological precipitation trend forecast, public daily weather broadcast, etc.); then, based on the scenario information, the corresponding module's business processing logic is automatically invoked (the agricultural scenario invokes the crop growth model analysis and disaster impact assessment logic of the Huinong module, and the watershed scenario invokes the watershed hydrological coupling calculation and flood risk early warning logic of the Zhihe module), ensuring that the processing logic matches the scenario requirements; at the same time, business analysis is advanced according to the preset standardized process (including data import, model calculation, and preliminary result generation), ensuring the standardization of processing and the consistency of results.

[0099] During the business analysis process, the system calls the business processing results (including core information such as numerical forecasts, disaster risk level assessments, and impact range predictions) generated by the intelligent brain module based on the optimal model combination in real time, and integrates them with the related data in the local business processing logic (such as population distribution and infrastructure location data) to complete in-depth business processing (such as analyzing the scope of traffic interruption and the risk of building damage in the rainstorm disaster impact assessment by combining the rainstorm location and intensity prediction results).

[0100] Based on the deep processing results, multimodal products are generated according to scenarios and user needs. For example, in the scenario of rainstorm disaster early warning for decision-making departments, comprehensive decision-making products including text reports, risk level maps, and impact analysis charts are output; in the scenario of daily weather services for the public, text forecasts, temperature change curves, and short videos of life indexes are output; and in the scenario of meteorological services for agricultural crops, growth suggestions and soil moisture monitoring heat maps with combined text and graphics are output to meet diverse meteorological service needs.

[0101] See Figure 4 As shown, Figure 4 This document illustrates a flowchart of a rainstorm intelligent forecasting system based on model scheduling and workflow, as provided in an embodiment of the present invention. The system uses a timed (every 6 hours) task or a real-time event (radar echo > 45 dBZ) to intelligently trigger a trigger. The workflow engine generates an instance number, writes metadata (reporting time, region, responsible person), and initiates the process. The workflow engine calls the data acquisition module to pull multi-source observation data, transmits it to the quality control assimilation module for anomaly detection, spatial interpolation, and other operations, and outputs it as NetCDF to object storage. The workflow engine listens for file completion events and triggers the next stage. The workflow engine calls the model inference interface of the model accumulator, passes the NetCDF from the previous stage as the initial model value, submits a high-performance computing job, and returns the model output path upon completion. Simultaneously, 10 ensemble members are started, managed by the workflow engine using a Map / Reduce model. The workflow engine feeds the output results of each ensemble forecast member to the model inference interface of the intelligent corrector to obtain a high-resolution precipitation field and uncertainty interval. Real-time radar data streams trigger the model inference interface of the short-term extrapolator, generating 0–2 hour extrapolated data, which is then fused with the numerical model results and updated to the Redis cache. The workflow engine transmits the fused results, along with terrain, population, and pipeline data, to the decision module's interface, outputting tiered warning signals, affected areas, and recommended measures. The workflow engine calls the information publisher interface to generate images, GeoJSON, and text based on the decision module's results, and calls SMS, WeChat, and large-screen interfaces to push information to recipients. After the real-time data is stored in the database, the workflow engine calls the verification and evaluation unit to calculate TS and ETS scores, writes them back to the metadata database, triggers automatic fine-tuning tasks, and forms a business closed loop. The entire process is driven by the workflow engine, with each module (trigger, collector, etc.) linked sequentially, covering the complete meteorological business chain from data acquisition and model calculation to product release and verification.

[0102] In an optional implementation, see Figure 5 As shown, Figure 5 A schematic diagram of the structure of the fourth meteorological business application system provided by the embodiment of the present invention is shown, wherein the meteorological business application system further includes a smart service module 501.

[0103] The intelligent service module is used to: obtain the multimodal business product from the business application module; call preset user profile data, match the multimodal business product with the user profile, and determine the appropriate push channel; and accurately push the multimodal business product to the corresponding user or industry scenario through the determined push channel.

[0104] Specifically, the smart service module is based on the scenario-based multimodal business products output by the business application module. Through user profile matching and precise channel push, it realizes scenario-based and personalized delivery of meteorological services. The specific process is as follows:

[0105] The system receives multimodal business products that have undergone in-depth scenario-based processing from the business application module. These products include text reports, charts, maps, and short videos. Specifically, they include crop weather advice packages for agricultural scenarios, road traffic weather warning videos for transportation scenarios, and lifestyle index reminders for public scenarios, providing product support for subsequent services.

[0106] The system calls upon pre-set user profile data, which is constructed based on historical service records, user attributes, and industry needs. The individual user dimension includes geographical location, meteorological factors of interest (temperature, precipitation, air quality, etc.), and service preferences (push time period, product form). The industry user dimension (agriculture, transportation, power, etc.) includes business scenario characteristics (e.g., agricultural users are concerned about the meteorological impact on crop growth cycle, and transportation users are concerned about road icing warnings) and decision-making needs (e.g., power users need line anti-icing planning). By matching the content attributes (business type, coverage area, information focus) of multimodal business products with the user profile needs characteristics, the target user group of each product is determined.

[0107] Based on the characteristics of the target user group, the appropriate push channels are determined: daily weather short videos for the general public are matched with APP push and WeChat official accounts; crop disaster risk graphic packages for agricultural users are matched with SMS and industry-specific service platforms; and comprehensive reports on rainstorm disasters for government decision-making departments are matched with intranet emails and large screen push. The products are then accurately pushed to users or industry scenarios through the corresponding channels. For example, rainstorm defense short videos are pushed to the public in rainy areas in the south, crop frost warning graphic packages are pushed to agricultural users in the main grain-producing areas of Northeast China, and road surface temperature and icing risk warnings are pushed to highway management departments to meet differentiated needs.

[0108] See Figure 6 As shown, Figure 6This document illustrates a flowchart of an intelligent rainstorm risk warning system based on user profiles, provided by an embodiment of the present invention. The system triggers a rainstorm risk warning request via user / system input (using natural language, shortcut buttons, etc.), entering a natural language parsing stage. The AI ​​semantic understanding module transforms the input into a structured intent of "spatiotemporal range + key elements," corresponding to the "out-of-control range & key elements" determination in the diagram, thus clarifying the direction of the warning demand. After intent parsing, automatic task orchestration is triggered. The planning layer scheduler generates a task DAG, splits and allocates resources; simultaneously, it links air-space-ground data ETL (intelligent ETL pipeline accesses multi-source data, processes and writes it into a spatiotemporal data lake) and historical scenario retrieval (spatiotemporal knowledge graph recalls similar historical event features), providing data and prior references for subsequent analysis. The task orchestration output and data input chain risk model, combined with real-time observation streams (adjustments are triggered if differences exceed thresholds), sequentially runs multiple model deductions to obtain a risk level matrix; then, based on the matrix, exposure and vulnerability calculations are performed to assess the risk impact. After calculation, sensitive data is checked. If sensitive information is found, tiered desensitization is performed. Then, multimodal product generation (outputting visualizations, text, etc., according to templates) is initiated. Finally, the warning product is accurately delivered to users through multiple channels, including system pop-ups and SMS. The process relies on a streaming computing framework for real-time monitoring, tracking discrepancies between observations and forecasts. Exceeding thresholds triggers incremental recalculation and second-level update pushes, forming a minute-level closed loop of "observation → model → product → feedback," continuously iterating and optimizing the warning system. The entire process, based on the modules shown in the diagram (user / system triggering, natural language parsing, etc.), connects the entire chain from data collection and model inference to product delivery, supporting the efficient operation of intelligent rainstorm risk warnings.

[0109] In an optional implementation, the intelligent service module is further configured to: initiate a streaming computing framework to collect user feedback information on the product in real time under various scenarios; perform structured processing on the feedback information to obtain structured feedback data; transmit the structured feedback data to the support platform module, which then triggers the system to adjust and optimize the data processing flow, model inference parameters, and business execution flow to form a closed loop.

[0110] Specifically, the smart service module activates a streaming computing framework to capture user feedback in real time across various scenarios—from individual users, including clicks, favorites, and text comments on weather products within the app; and from industry users, including service request adjustment suggestions submitted through a dedicated platform (such as agricultural users suggesting the need to supplement soil moisture data for crop meteorology) and product accuracy feedback (such as traffic users pointing out deviations in road icing warning times), achieving real-time capture of feedback information across all scenarios.

[0111] For unstructured feedback data such as text comments and voice suggestions, natural language processing technologies (such as keyword extraction and intent recognition) are used to extract key information (such as "need to increase soil moisture data" and "early warning period deviation"), and transform it into structured feedback data containing dimensions such as feedback type (demand adjustment / accuracy feedback), demand points, and product names involved, to ensure that the information conforms to the data interaction specifications of subsequent stages of the system and can be effectively identified and called.

[0112] The structured feedback data is transmitted to the support platform module. The support platform module then combines historical business execution status data (such as business processing progress and result accuracy) and resource usage data (such as computing power consumption and component response efficiency) to conduct multi-dimensional analysis and trigger system optimization.

[0113] At the data processing level: if feedback indicates that specific meteorological data (such as soil moisture) is missing, optimize the data collection and processing steps of the data base module and add corresponding data source access;

[0114] At the model inference parameter level: if the feedback shows that the accuracy of the disaster warning is insufficient, adjust the parameters such as the number of iterations and feature weights of the model in the intelligent brain module;

[0115] At the business execution process level: If feedback mentions that the product generation cycle is too long, optimize the task scheduling logic of the workflow engine of the business application module (such as adjusting task priority and merging redundant links) to improve process efficiency, and finally form a complete closed loop of user feedback-data flow-system optimization to continuously iterate the system service capabilities.

[0116] In one optional implementation, the data base module includes a station network observation information operation platform, a meteorological big data cloud platform, a meteorological integrated business monitoring system, and a meteorological data quality control and processing system.

[0117] The information operation platform for station network observation is used to: collect multi-source heterogeneous meteorological data required for target meteorological operations, including satellite remote sensing data, radar echo data, and ground observation station data, and transmit the collected multi-source heterogeneous meteorological data to the meteorological data quality control and processing system.

[0118] Specifically, the station network observation information operation platform accesses multi-source meteorological data from "space, air, and ground" through standardized interfaces. The space-based data includes remote sensing data such as satellite cloud images and temperature profiles, the space-based data includes radar echo data, and the ground-based data includes measured data such as air temperature and precipitation from ground stations. After collection, the raw data from different sources and formats are transmitted in real time to the meteorological data quality control and processing system according to protocols such as FTP and MQTT to support subsequent data governance.

[0119] The meteorological data quality control and processing system is used to: receive multi-source heterogeneous meteorological data transmitted from the station network observation information operation platform, perform anomaly detection, format standardization, and interpolation correction processing on the multi-source heterogeneous meteorological data in sequence, remove outliers in the data, unify the data format, and fill in data gaps, and finally obtain structured data, and transmit the structured data to the meteorological big data cloud platform.

[0120] Specifically, the meteorological data quality control and processing system uses statistical testing methods and machine learning models to screen data, eliminating invalid and extreme outliers to ensure data accuracy. It converts data from different formats, such as satellite remote sensing, radar echoes, and ground observations, into NetCDF format using standardized tools, clarifying variable naming, precision, and metadata specifications to resolve compatibility issues. Spatial interpolation algorithms, such as Kriging and inverse distance weighting, are used to fill data gaps caused by uneven distribution or malfunctions at different stations, meeting operational accuracy and coverage requirements. After processing, the structured data is transmitted to the meteorological big data cloud platform via standardized interfaces to support subsequent storage and retrieval.

[0121] The meteorological big data cloud platform is used to: receive structured data output from the meteorological data quality control and processing system, classify, store and manage the structured data using a distributed storage architecture, and provide structured data call interfaces for other modules of the system.

[0122] Specifically, after receiving structured data from the meteorological data quality control and processing system, the meteorological big data cloud platform stores the data according to data type (observation / forecast data), time dimension (real-time / historical data), and spatial dimension (regional / station data) based on a distributed storage architecture. It also establishes a data indexing and management mechanism to ensure rapid data retrieval and access. The platform provides standardized structured data access interfaces for the intelligent brain module, support platform module, and business application module, enabling each module to obtain data on demand, breaking down data flow barriers, and ensuring efficient data flow and sharing within the system.

[0123] The meteorological integrated business monitoring system is used to: collect various computing power indicators during the operation of the meteorological business application system in real time, including GPU / CPU resource utilization, data transmission rate, and remaining storage resources; monitor and identify anomalies in the collected computing power indicators in real time; and output the computing power indicators to the intelligent brain module synchronously.

[0124] Specifically, the meteorological integrated business monitoring system continuously acquires key computing power indicators of the meteorological business application system through real-time acquisition technology. These indicators include GPU / CPU resource utilization (reflecting computing resource usage), data transmission rate (reflecting data flow efficiency), and remaining storage resources (reflecting available storage space). The system monitors the acquired computing power indicators in real time and identifies anomalies, issuing timely warnings when indicators exceed normal thresholds (such as excessively high GPU utilization or insufficient storage). Simultaneously, the system outputs complete computing power indicators to the intelligent brain module, providing a basis for analyzing resource status, formulating scheduling strategies, and determining the optimal model combination, ensuring the system operates stably under reasonable resource allocation.

[0125] In one optional implementation, the intelligent brain module includes a knowledge base, a general large model library, an algorithm library, a component library, and a meteorological large model library.

[0126] The knowledge base is used to store standard procedures, historical business data, emergency response plans, and business scenario adaptation rules in the meteorological business field, providing knowledge support for task recognition and model selection of the intelligent brain module.

[0127] Specifically, the knowledge base stores key information from the entire meteorological operational process in a structured format, including standard meteorological operational procedures (daily weather forecast production, disaster warning issuance specifications, etc.), historical operational data (past observation data for the same period, historical disaster handling data, etc.), emergency response plans (response plans and procedures for disasters such as rainstorms, typhoons, and cold waves), and operational scenario adaptation rules (model selection and algorithm matching rules for different scenarios). When the intelligent brain module performs task identification (such as distinguishing between rainstorm warnings and regular weather broadcasts) and model selection, it can quickly retrieve the corresponding knowledge from the knowledge base, providing industry standards and historical experience support for decision-making and ensuring the compliance and rationality of operational processes.

[0128] The general-purpose large model library is used to store general-purpose large models with generalized reasoning capabilities. These general-purpose large models can achieve cross-scenario business processing based on knowledge transfer from the meteorological field.

[0129] Specifically, the general-purpose large model library stores general-purpose large models with strong generalization and reasoning capabilities. These models possess cross-domain understanding and reasoning abilities and can be adapted to meteorological operational scenarios through knowledge transfer learning from the meteorological domain. For example, the text understanding capabilities of general-purpose large models can be transferred to weather forecast text generation, and the multimodal processing capabilities can be transferred to the integration of meteorological imagery products. This enables cross-scenario operational processing, covering the flexible adaptation needs of different scenarios in meteorological operations and compensating for the limitations of dedicated models in scenario adaptability.

[0130] The meteorological big model library is used to store specialized models optimized for meteorological business scenarios, including numerical forecasting models, disaster early warning models, and meteorological product generation models. The meteorological big model has the ability to deeply analyze and process meteorological professional data.

[0131] Specifically, the meteorological big model library stores three types of key specialized models based on the core needs of meteorological operations: numerical forecasting models for numerical simulation and trend prediction of meteorological elements, disaster warning models for the identification and early warning of disasters such as rainstorms, severe convection, and typhoons, and meteorological product generation models that can automatically generate weather forecast reports and disaster risk maps. Its professional advantages and supporting role are as follows: Compared to general-purpose big models, meteorological big models have stronger in-depth analytical capabilities in meteorological professional data processing. They can accurately identify professional characteristics such as the correlation between radar echo intensity and precipitation, and the correspondence between satellite cloud images and weather systems, providing core model support for the professional processing of meteorological operations.

[0132] The algorithm library is used to store various algorithms required for meteorological business processing, including data mining algorithms, machine learning algorithms, reinforcement learning algorithms, and spatiotemporal interpolation algorithms, providing algorithmic support for model processing of structured data.

[0133] Specifically, the algorithm library covers two core categories of algorithms: data processing algorithms (such as data mining algorithms for uncovering patterns in meteorological data and machine learning algorithms for data classification and prediction) and business optimization algorithms (such as reinforcement learning algorithms for dynamically adjusting business processes and spatiotemporal interpolation algorithms for filling spatial gaps in meteorological data). All algorithms are encapsulated with standardized interfaces. When the intelligent brain module calls the model to process structured data, it can invoke the corresponding algorithm to provide technical support for model operation based on needs such as data feature extraction, model parameter optimization, and data gap filling, ensuring that the model processes meteorological data efficiently and accurately.

[0134] The component library is used to store the functional components required in the meteorological business processing process, including data preprocessing components, model inference components, and result visualization components, and to support the intelligent brain module in quickly building the business processing chain.

[0135] Specifically, the component library is divided into three core component categories based on function: data preprocessing components for further cleaning and feature filtering of structured data, model inference components for loading models and performing data inference calculations, and result visualization components for converting business processing results into charts / maps and other forms. These components are highly reusable and composable. When the intelligent brain module builds a business processing chain, there is no need to repeatedly develop basic functions; components can be directly called to quickly build the process, significantly improving the efficiency of chain construction and operational stability.

[0136] The intelligent brain module is used to: invoke information and resources from the knowledge base and algorithm base through a dynamic routing mechanism, and select and combine the optimal model combination from the general large model library and the meteorological large model library in combination with computing power indicators, and perform processing on structured data to obtain business processing results.

[0137] Specifically, in actual business processing, the AI ​​module achieves collaborative invocation and integration of resources from various sub-libraries through a dynamic routing mechanism. First, the AI ​​module receives structured data and computing power metrics output from the data base module. Then, it initiates the dynamic routing mechanism: on one hand, it calls upon standard processes and scenario adaptation rules from the knowledge base, combining them with corresponding algorithm resources from the algorithm library to clarify the technical path and methodological framework for the current business processing; on the other hand, based on the computing power metrics, it analyzes the current computing resource status of the system (such as GPU / CPU utilization, available memory, etc.) and selects models from the general large model library and the meteorological large model library that are suitable for the current resource status and business needs—for example, when computing power is sufficient and cross-scenario business needs to be processed, the general large model and the meteorological large model are prioritized; when computing power is limited and specialized business needs to be processed, a lightweight meteorological-specific model is selected. Through this selection and combination, after forming the optimal model combination, the AI ​​module calls functional components from the component library to build a complete business processing link, inputs structured data into the optimal model combination for processing, and finally obtains business processing results that meet the meteorological business needs, providing core intelligent output for the subsequent operation of the middle platform module and business application module.

[0138] In one optional implementation, the supporting middleware module includes an AI middleware, a knowledge middleware, an algorithm middleware, a data middleware, a business middleware, and a component middleware.

[0139] The data platform is used to: receive structured data output from the data base module, and achieve semantic alignment of multi-source data through semantic mapping, format unification, and data cleaning to form a standardized data resource pool, providing data support for other platforms.

[0140] Specifically, the data platform receives pre-processed structured data from the data base module. This data may have semantic differences from multiple sources (e.g., the same meteorological element is described differently in different data sources). A unified data semantic system is established through semantic mapping technology to match the fields and meanings of data from different sources. At the same time, format unification and data cleaning are carried out to eliminate subtle format differences and filter residual invalid information, ultimately forming a standardized data resource pool with unified structure and consistent semantics. The standardized data resource pool provides on-demand data support for the AI ​​platform, algorithm platform, and business platform. For example, it provides standardized observation data for AI platform model training and unified format business data for algorithm platform algorithm operation, ensuring the consistency and accuracy of data use across all platforms.

[0141] The algorithm platform is used to: schedule corresponding algorithm resources from the algorithm library of the intelligent brain module according to the business scheduling instructions output by the intelligent brain module, encapsulate and manage the algorithm resources, and provide algorithm call interfaces for business processing.

[0142] Specifically, the algorithm platform uses the business scheduling instructions output by the intelligent brain module as its core basis (the instructions specify the type of algorithm required for the current business, such as spatiotemporal interpolation algorithm for precipitation prediction and machine learning classification algorithm for disaster assessment). It accurately schedules matching algorithm resources from the algorithm library of the intelligent brain module; it standardizes and encapsulates the scheduled algorithms, transforming them into service interfaces that can be directly called; at the same time, it establishes an algorithm lifecycle management mechanism to monitor the algorithm's running status and version iteration in real time; and it provides convenient algorithm calling interfaces for the business platform, business application modules, etc., supporting each module to quickly call the required algorithms without having to worry about the underlying implementation details, effectively improving the efficiency of algorithm application.

[0143] The business platform is used to: integrate various business logic components and process templates required for meteorological business applications, build a standardized business processing framework, and support the rapid configuration and expansion of business scenarios.

[0144] Specifically, the business platform integrates the core resources required for meteorological business applications, including business logic components such as forecast result verification and disaster impact analysis, as well as process templates such as rainstorm warnings and agricultural meteorological services, to build a standardized business processing framework. The standardized framework supports rapid configuration according to different business scenarios. For example, in the cultural tourism meteorological service scenario, components for analyzing scenic area meteorological elements and modules for associating tourist flow can be added. At the same time, it supports horizontal expansion of business scenarios. When adding new meteorological services, there is no need to reconstruct the overall framework. Only the corresponding business components and templates need to be added for access, ensuring that the system can quickly respond to diverse business needs.

[0145] The AI ​​platform is used to manage the model resources of the intelligent brain module, including model deployment, model monitoring, and model iteration, while providing AI model calling services to other modules of the system.

[0146] Specifically, the AI ​​platform primarily connects to the general-purpose large model library and the meteorological large model library of the intelligent brain module. During deployment, it provides a containerized deployment environment, encapsulating models as elastically scalable microservices to ensure stable model operation. In model monitoring, it collects real-time metrics such as model inference latency, accuracy, and resource consumption. If performance degradation or operational anomalies are detected, it promptly issues warnings and activates backup models. During model iteration, it fine-tunes and optimizes models by combining updated business knowledge from the knowledge platform with user feedback data, improving the accuracy of models in handling meteorological business. It also provides standardized AI model invocation services for other system modules such as the business application module and the intelligent brain module, supporting each module in calling corresponding model resources according to its business needs.

[0147] The knowledge platform is used to: connect to the knowledge base of the intelligent brain module, extract, integrate and update meteorological business knowledge, construct a dynamic knowledge graph, and provide knowledge support for business processing and model optimization.

[0148] Specifically, the knowledge platform connects to the knowledge base of the intelligent brain module, using natural language processing and knowledge extraction algorithms to extract key knowledge (such as procedures for handling new meteorological disasters and updated meteorological observation standards) from newly added meteorological business data, industry standards, and emergency plans. The extracted new knowledge is integrated with existing knowledge base information to eliminate knowledge conflicts and redundancy, ensuring the consistency and accuracy of the knowledge system. A dynamic knowledge graph is constructed, visually presenting meteorological business knowledge in an "entity-relationship" format (such as a "rainstorm disaster-affected area-emergency measures" association graph), facilitating rapid retrieval and application by various modules. Dynamically updated knowledge resources support business processing (such as optimization of business platform process templates) and model optimization (such as knowledge injection for AI platform model training), ensuring that system business processing always adheres to the latest industry standards and experience.

[0149] The component platform is used to integrate the functional components required by each module of the system, including interface adaptation components, resource scheduling components, and log management components, so as to realize the reuse and unified management of components.

[0150] Specifically, the component platform integrates common functional components required by various system modules, such as the data base, intelligent brain module, and business application module. This includes interface adaptation components for converting different module interface protocols, resource scheduling components for dynamically allocating computing resources such as CPU / GPU, and log management components for recording system operation logs to facilitate troubleshooting. A unified management system is established for all components, including component registration, version control, and access control, ensuring component traceability and security. Through a component reuse mechanism, modules do not need to repeatedly develop the same functions and can directly call components from the component platform, reducing system development costs while ensuring the consistency and stability of component functions. For example, the business application module and the support platform module can share a single log management component to ensure a unified system log format.

[0151] The supporting middle platform module is used to: integrate data, algorithms, models, and components into an initial resource package through the collaborative work of various middle platforms, and output it to the business application module after scenario adaptation and adjustment, while receiving feedback information to generate resource adjustment suggestions.

[0152] Specifically, in actual operation, the support platform module completes resource integration and scenario adaptation through the collaborative work of six sub-platforms: the data platform provides standardized data, the algorithm platform schedules and adapts algorithms, the AI ​​platform provides usable models, the knowledge platform supplements business knowledge, the business platform builds standardized processes, and the component platform provides general functional components. The resources of each platform are collaboratively integrated to form an initial resource package covering "data-algorithm-model-knowledge-process-function". Subsequently, the support platform module, based on the specific meteorological business scenario requirements (such as agricultural meteorological services needing to focus on crop growth models, and watershed meteorological services needing to focus on hydrological coupling algorithms), filters (removes algorithms and models irrelevant to the scenario) and adapts and adjusts (optimizes model parameters and adjusts business process nodes) the initial resource package to obtain an adapted resource package, which is then output to the business application module to provide direct support for the implementation of business functions. Meanwhile, the support platform module receives structured feedback data transmitted by the intelligent service module, combines it with status information during business execution, comprehensively analyzes issues such as resource utilization efficiency and business processing accuracy, generates targeted resource adjustment suggestions (such as optimizing the semantic mapping rules of the data platform and updating the algorithm version of the algorithm platform), and feeds the suggestions back to the intelligent brain module and each sub-platform, driving continuous optimization of each link of the system and forming a closed loop of "resource integration - business support - feedback optimization".

[0153] In one optional implementation, the business application modules include an integrated weather and climate module, a decision-making meteorology module, a professional meteorological service module, an agricultural meteorological service module, a watershed meteorological service module, and a weather modification operation platform module.

[0154] The integrated weather and climate module is used to: receive the adapted resource package output by the support platform module, call the business processing results of the intelligent brain module, perform various business analyses such as weather monitoring, climate trend analysis, and short-term climate prediction, and generate an integrated weather and climate analysis report.

[0155] Specifically, the integrated weather and climate module receives the adapted resource package output by the support platform module. This resource package includes dedicated resources such as weather monitoring algorithms, climate trend analysis components, and short-term climate prediction models. It also calls upon the business processing results previously output by the intelligent brain module (such as multi-source observation data integration and analysis results, and climate model simulation results) to provide resources and data foundation for business analysis. The module then conducts three major business processes in a standardized manner: weather monitoring (real-time tracking of changes in meteorological elements such as temperature, precipitation, and wind speed), climate trend analysis (identifying long-term climate evolution patterns by combining historical climate data), and short-term climate prediction (predicting climate anomaly trends for the next 1-3 months). Through the collaborative calculation of data and models from multiple stages, an integrated weather and climate analysis report is generated, covering actual weather conditions, climate trend assessment, and short-term prediction conclusions. This report provides basic meteorological basis for subsequent analysis of other business modules and also provides basic climate information reference for the public and industry.

[0156] The decision-making meteorological module is used to: combine the disaster early warning results and impact assessment results output by the intelligent brain module with various business processing data, integrate the relevant data of terrain, population and economy, and generate meteorological service short reports and emergency response suggestions for government decision-making.

[0157] Specifically, the decision-making meteorological module, on the one hand, deeply utilizes the business processing data output by the intelligent brain module, covering the warning levels and impact range of meteorological disasters such as rainstorms and typhoons, as well as the assessment results such as the economic losses and the number of affected people that may be caused by the disasters; on the other hand, it integrates non-meteorological data related to meteorological disasters, including topographic data that identifies areas prone to flash floods and landslides, population distribution data that locates high-risk areas for disasters, and economic data that clarifies disaster-sensitive industries and key protection targets.

[0158] By spatially overlaying and quantitatively analyzing meteorological and related data, two core decision support products are generated: short meteorological service reports for routine government decision-making (including analysis of the impact of meteorological element changes on urban operations and agricultural production), and emergency response recommendations for sudden disasters (clarifying emergency activation levels, personnel evacuation plans, and material allocation directions), providing precise meteorological support for government departments to formulate disaster prevention and mitigation policies and carry out emergency management.

[0159] The professional meteorological service module is used to generate customized meteorological service products for different industries, including power, transportation, and culture and tourism, by calling the business processing results of industry-specific components and intelligent brain modules in the adapted resource package. These products include power load forecasts, traffic weather warnings, and scenic area weather guides.

[0160] Specifically, the professional meteorological service module extracts industry-specific components from the adapted resource package supporting the middle platform module according to the differentiated needs of different industries. For example, the load forecasting component for the power industry, the road icing early warning component for the transportation industry, and the scenic area comfort assessment component for the cultural and tourism industry. At the same time, it calls the business processing results generated by the intelligent brain module for each industry, including the temperature-load correlation analysis results for the power industry, the precipitation-road traffic capacity assessment results for the transportation industry, and the scenic area precipitation probability prediction results for the cultural and tourism industry.

[0161] By combining industry-specific components with targeted business processing results, customized meteorological service products are generated for various industries. For the power industry, power load forecasts can help optimize power generation scheduling; for the transportation industry, traffic weather warnings can alert to risks such as road icing and fog; and for the cultural and tourism industry, scenic spot weather guides can provide reference for tourists' travel and scenic spot operations, effectively solving the pain points of meteorological services in various industries.

[0162] The agricultural meteorological service module is used to: combine the agricultural production cycle, crop growth model and meteorological element forecast results output by the intelligent brain module to perform various business analyses such as agricultural meteorological disaster risk assessment, crop growth monitoring and agricultural activity suggestions, and generate agricultural meteorological service products.

[0163] Specifically, the agricultural meteorological service module accurately calls upon the forecast results of key agricultural meteorological elements output by the intelligent brain module, covering temperature, precipitation, sunshine, accumulated temperature, etc. for a period of time in the future; on the other hand, it integrates core agricultural production data, including clarifying the agricultural production cycle of key growth stages such as crop sowing, jointing, and grain filling, as well as crop growth models that simulate the crop growth status and yield formation process under different meteorological conditions, laying a data foundation for subsequent business analysis.

[0164] Based on the integrated data, three types of business analyses are carried out sequentially: agricultural meteorological disaster risk assessment (such as predicting the impact of frost and drought on crops), crop growth monitoring (assessing crop growth status by combining remote sensing data and meteorological conditions), and agricultural activity recommendations (identifying suitable periods for sowing, fertilizing, and harvesting based on meteorological forecasts). Ultimately, agricultural meteorological service products containing disaster warnings, growth analysis, and agricultural guidance are generated to provide precise meteorological services for farmers, agricultural cooperatives, and agricultural management departments, helping to improve the quality and efficiency of agricultural production.

[0165] The basin meteorological service module is used to: generate basin meteorological service products such as basin flood warnings and water resource scheduling suggestions by calling basin observation data and the precipitation forecast and flood simulation results of the intelligent brain module in response to the hydrological and meteorological needs of a specific basin.

[0166] Specifically, the basin meteorological service module first accesses dedicated observation data for a specific basin, covering water level and flow data from hydrological stations within the basin, as well as precipitation and evaporation data from meteorological stations. At the same time, it deeply utilizes the business processing results output by the intelligent brain module, such as future precipitation forecasts that are accurate to the precipitation intensity and duration of each sub-region within the basin, and flood simulation results that predict the evolution of floods and changes in water levels under different precipitation scenarios.

[0167] By combining basin-specific observation data with the intelligent processing results of the intelligent brain module, a comprehensive analysis of the basin's hydrological and meteorological conditions is carried out, ultimately generating two types of core service products: basin flood warnings that clearly define the time, scope, and magnitude of floods (providing early warning information for flood control departments), and water resource scheduling suggestions that propose flood discharge and water replenishment plans based on precipitation forecasts and reservoir water storage conditions, thus helping to rationally utilize water resources and effectively prevent and control flood disasters in the basin.

[0168] The artificial weather modification platform module is used to: receive the cloud water resource analysis and operation condition judgment results output by the intelligent brain module, and combine the operation site distribution and equipment status information to generate artificial weather modification operation plans and operation effect evaluation reports to support the operation of artificial weather modification.

[0169] Specifically, the artificial weather modification platform relies on two types of key data to carry out its work: first, the business processing results output by the intelligent brain module, which includes cloud water resource analysis to identify cloud systems with potential for artificial rain enhancement and hail suppression, as well as the judgment of operational conditions to determine whether the meteorological conditions for artificial rain enhancement and hail suppression are met; second, basic information on artificial weather modification operations, including the distribution of operation sites with clear locations and coverage areas, and the equipment status to understand the availability of equipment such as rocket launchers and anti-aircraft guns. By integrating these two types of data, the platform lays the foundation for subsequent work.

[0170] Based on integrated data, operational feasibility analysis and scheme design are conducted to generate two types of core products: artificial weather modification operation schemes that specify key parameters such as operation time, location, equipment, and dosage; and operation effect evaluation reports that combine post-operation meteorological observation data and cloud system changes to analyze the actual effect of the operation. These products directly support the operation of artificial weather modification and provide technical support for alleviating drought and mitigating hail disasters.

[0171] In one optional implementation, the smart service module encompasses a transportation service platform, a cultural and tourism service platform, a city and county service platform, a public service platform, a power service platform, an energy service platform, an agricultural service platform, and a government service platform.

[0172] The traffic service platform is used to: receive traffic and meteorological service products output by the business application module, including road icing warnings and visibility forecasts; combine these with user profiles in the traffic industry; and push these products through various channels such as the traffic department's business system, roadside displays, and navigation apps to support traffic safety.

[0173] Specifically, the traffic service platform receives traffic-specific meteorological products output by the business application module. The core layer includes road icing warnings and road segment visibility forecasts, while the auxiliary layer covers precipitation-road capacity assessments and wind safety alerts for freight vehicles, forming a structured dataset. Based on pre-defined user profiles in the traffic industry, the platform adapts and pushes data through multiple channels—outputting road network meteorological risk heat maps to the traffic management system to support control decisions, issuing real-time road segment warnings via roadside terminals, and providing navigation apps with route weather risk and congestion avoidance interfaces, thus constructing a full-scenario traffic meteorological safety support system.

[0174] The cultural and tourism service platform is used to: receive scenic area meteorological service products output by the business application module, including scenic area precipitation forecasts and ultraviolet intensity warnings; combine these with user profiles in the cultural and tourism industry; and push these products through various channels such as the scenic area's official website, tourism APP, and travel agency service system to assist in the planning of cultural and tourism activities.

[0175] Specifically, the cultural tourism service platform receives products from the business application module centered on refined meteorological services for scenic spots. These include 24-hour precipitation forecasts for scenic spots with precise time and region information, UV intensity warnings that provide sun protection advice based on altitude and season, a scenic spot comfort index that comprehensively assesses temperature, humidity, and wind speed, and alerts on the impact of special weather conditions such as fog and snow on tour activities. The platform also enables precise product distribution based on user profiles within the cultural tourism industry—pushing products to scenic spot operators through their official websites and internal management systems to assist in adjusting opening hours and suspending high-risk activities; pushing personalized tips to tourists through travel apps and OTA platforms (such as providing tips on temperature differences and rain protection for mountain hikers); and pushing weather warnings for group tour destinations to travel agencies through the service system to assist in adjusting itineraries and providing scientific meteorological basis for cultural tourism activity planning.

[0176] The municipal and county service platform is used to: receive localized meteorological service products output by the business application module, including municipal and county rainstorm warnings and township meteorological reports; and push these products through various channels such as the municipal and county meteorological department business platform and government affairs APP, based on the needs and characteristics of municipal and county users, to support grassroots meteorological services.

[0177] Specifically, the city and county service platforms receive localized meteorological service products output by the business application modules. These include city and county-level rainstorm warnings that clearly differentiate between urban and rural areas, township-level meteorological reports that provide precipitation and strong wind forecasts for agricultural townships and warnings of urban flooding risks, as well as localized disaster prevention guidelines covering geological disaster meteorological risks in mountainous townships and wheat lodging warnings in plain townships. The platforms also match corresponding push channels based on the different user needs of city and county users—data support products are pushed to grassroots meteorological workers through the city and county meteorological department's business platform to assist in service delivery; warning information and prevention suggestions are pushed to grassroots cadres through government apps and township government groups to help fulfill local responsibilities; and meteorological information is pushed through community bulletin boards and village broadcasts to ensure coverage of the grassroots population.

[0178] The public service platform is used to: receive public meteorological service products output by the business application module, including daily weather forecasts and life weather indices; combine these with public user profiles, including age, occupation, and lifestyle habits; and push these products through various channels such as SMS, WeChat official accounts, and meteorological apps to meet the public's daily meteorological information needs.

[0179] Specifically, the public service receiving business application module outputs public meteorological service products, covering daily weather forecasts for the next 7 days, including temperature, precipitation, and wind; life meteorological indices such as ultraviolet radiation, car washing, clothing recommendations, and cold prevention; and special weather tips such as heatstroke prevention and cold wave protection, as well as holiday travel weather guides. Based on public user profiles, for example, it pushes weather and health tips for morning exercise to the elderly based on age, and rain warnings for going to school to teenagers; it pushes high temperature and rainstorm protection suggestions to outdoor workers based on occupation, and temperature difference and sun protection tips to office workers; the push channels cover SMS, WeChat official accounts, and meteorological apps.

[0180] The power service platform is used to: receive power meteorological service products output by the business application module, including cooling load forecasts and line icing warnings; and, in conjunction with the needs of power enterprise users, push these products through various channels such as the power dispatch system and the enterprise's dedicated service platform to assist in power production dispatch.

[0181] Specifically, the power service platform receives power meteorological service products output by the business application module. These products focus on power production and transmission safety, including cooling load forecasts that predict peak electricity load based on temperature changes to assist power generation planning; line icing warnings that indicate the thickness and risk level of icing on transmission lines and prompt de-icing operations; warnings about the impact of high temperatures on transformers; and safety risk warnings about strong winds on transmission towers. The platform also provides services tailored to the different needs of power companies: pushing load forecasts and grid risk warnings to the power dispatch center through the power dispatch system to assist in formulating power generation plans and cross-regional power allocation schemes; pushing line icing and lightning strike risk warnings to transmission operation and maintenance departments through the company's dedicated service platform to assist in arranging inspections and protection work; and pushing weather impact warnings during peak residential electricity consumption periods to power supply companies to assist in preparing for power supply security, providing crucial meteorological support for power production dispatch.

[0182] The energy service platform is used to: receive energy meteorological service products output by the business application module, including wind power forecasts and photovoltaic irradiance forecasts; combine these with energy enterprise user profiles; and push these products through various channels of the energy management system and service platform to support the efficient utilization of new energy sources.

[0183] Specifically, the energy service platform receives energy meteorological service products focused on the efficiency and safety of new energy production from the business application module. These include wind power forecasts that predict the future power generation of wind farms by combining wind speed and direction, and photovoltaic irradiance forecasts that predict the solar irradiance intensity and power generation potential of the area where photovoltaic power plants are located. It also includes early warnings of the impact of extreme weather events such as strong winds and sandstorms on new energy equipment. Based on user profiles of energy companies, the platform pushes wind speed changes and power forecasts to wind power companies through their energy management systems to assist in optimizing wind turbine start-up and grid connection scheduling. It also pushes irradiance forecasts and shading analysis to photovoltaic companies to assist in adjusting photovoltaic array angles and power generation plans. Finally, it pushes regional new energy power generation forecasts to energy regulatory authorities through the service platform to assist in coordinating the supply of new and traditional energy sources and supporting the efficient utilization of new energy.

[0184] The agricultural service platform is used to: receive agricultural meteorological service products output by the business application module, including crop growth period forecasts and meteorological risk warnings for pests and diseases; combine these with farmer user profiles, including planted crops and planting scale; and push these products through various channels such as agricultural APPs and agricultural technology extension platforms to assist agricultural production.

[0185] Specifically, the agricultural service platform receives agricultural meteorological service products related to crop growth and agricultural activities from the business application module. These include crop growth period forecasts that predict the timing of wheat jointing and corn grain filling to guide fertilization and irrigation; meteorological risk warnings for pests and diseases that combine temperature and humidity to predict the timing and extent of pest and disease occurrence and provide prevention and control suggestions; and agricultural activity suitability indices that assess soil moisture content and determine suitable meteorological conditions for sowing and harvesting. The platform also conducts precise push notifications based on farmer user profiles—by crop type, it sends notifications about rainfall impacts during the heading stage to rice farmers and frost warnings during the flowering stage to fruit tree farmers; and by planting scale, it sends concise agricultural suggestions to small farmers via an agricultural app and detailed meteorological data and decision support reports to large-scale farms via an agricultural technology extension platform, providing comprehensive support for agricultural production.

[0186] The government service platform is used to: receive decision-making meteorological service products output by business application modules, including major disaster warnings and emergency response suggestions; and push these products through various channels such as the government collaboration platform and emergency command system, in accordance with the needs of government departments, to support government decision-making and emergency management.

[0187] Specifically, the government service platform receives decision-making meteorological service products output by the business application module, including major disaster warnings such as red alerts for typhoons and rainstorms that clearly define the scope of impact and risk level, emergency response suggestions such as school closures, work stoppages, and personnel evacuation recommendations corresponding to the warning level, disaster impact assessment reports such as economic loss predictions and affected population statistics, and post-disaster reconstruction meteorological suggestions such as the impact of post-disaster precipitation on reconstruction construction; combined with the needs of government departments, it pushes products—real-time disaster warnings and disposal suggestions to emergency management departments through the emergency command system to assist in initiating emergency response and dispatching rescue forces.

[0188] This application also provides a meteorological job support architecture based on a multi-dimensional knowledge graph and an intelligent agent workflow engine. This architecture can be deeply integrated into all aspects of meteorological operational application systems, providing technical support for efficient operation. In this meteorological job support architecture, a dynamic knowledge graph (covering job responsibilities, system permissions, and emergency procedures) is constructed by extracting entity relationships from meteorological operational procedures using a BERT-BiLSTM-CRF model. This graph can directly connect to the meteorological job management system and the emergency response system. On the one hand, it provides the job management system with accurate "job-permission" matching criteria, enabling automated allocation and adjustment of system permissions. On the other hand, in the emergency response system, it can quickly retrieve corresponding emergency procedure knowledge, assisting the intelligent agent in providing standardized emergency response guidance to forecasters and emergency personnel, reducing human error. The developed GNN workflow engine, based on an improved Petri net, combines reinforcement learning and RPA-OCR technology and can be integrated into a cross-operational meteorological system collaboration platform. In meteorological operations, frequent cross-system operations are required between observation data systems, numerical forecasting systems, and product generation systems. This engine optimizes cross-system operation paths and enables automatic switching between operational systems via RPA-OCR (e.g., automatically retrieving data from the observation data system to the forecasting system), reducing manual switching costs and improving cross-system operational efficiency. The multimodal product generation system can interface with the meteorological product release system. Its spatiotemporal fusion forecasting and mapping capabilities, implemented using a Conv LSTM model, support the system in automatically generating professional meteorological forecast and early warning graphic products (such as rainstorm warning maps and temperature trend maps) and industry service products (such as agricultural meteorological reports and traffic meteorological guidance). This directly meets the meteorological product release system's demand for diverse and high-precision products, reducing forecasters' workload in manual mapping and product editing. The collaborative interaction layer can be connected to the meteorological collaborative work system and the emergency response system respectively. In the collaborative work system, the multi-party review channel based on smart contracts (zero-knowledge proof to ensure privacy and traceability) and the real-time collaborative space integrating the Operational Transformation algorithm can meet the needs of multiple reviewers of meteorological products (such as forecasters' initial compilation and reviewers' verification) and multiple plotting (such as disaster impact range plotting), and resolve collaboration conflicts. Meanwhile, the meteorological post digital twin system (3D visualization + emergency response simulation engine) can be connected to the emergency response system to simulate emergency scenarios under the entire business process (such as typhoon emergency response process), and assist in carrying out emergency simulation and on-the-job training.

[0189] See Figure 7 As shown, Figure 7The flowchart of a method for generating forecast materials based on a meteorological post-assisted architecture provided by an embodiment of the present invention is shown. The process starts with a timed / manual trigger, corresponding to the "workflow orchestration and triggering" stage: In the intranet workflow engine, the "daily weather forecast automation" process can be started by a timed trigger (06:00 every day) or temporarily triggered by user input instructions to start the entire process.

[0190] After being triggered, the system retrieves forecast data. Relying on the data interface in "Requirements Analysis and Interface Protocol", it pulls element forecast data from the local meteorological database in real time through a message queue. Then, it checks whether the data verification passes. If the verification fails, an alarm is triggered and the process terminates. If the verification passes, it proceeds to the next step to ensure data quality.

[0191] After the data verification is successful, the intranet large model interface is called according to the "workflow orchestration and triggering" logic (corresponding to the call to the large model to generate forecast text and plotting configuration): prompt words containing city, original message, etc. are sent to the model deployed on the intranet, and the model returns text forecast and plotting configuration objects; then the image rendering is performed according to the configuration, that is, the rendering node (headless browser container of graphics card pool) generates the image within 3 seconds according to the configuration, and the process corresponds to the "visualization rendering" stage.

[0192] After the image is rendered, it is uploaded to the intranet object storage to obtain the internal link (compress the image and upload the link). Then, according to the "multi-channel publishing" logic, the forecast text and image link are merged into rich text and integrated into publishable content.

[0193] After the rich text is generated, it is pushed to the recipient through channels such as intranet FTP and email. At the same time, the structured data is written to the local time series database for use in the dashboard, covering the needs of "multi-channel publishing" and data retention.

[0194] Finally, the process ends; if any step fails during the entire process, the "monitoring and rollback" mechanism will trigger an alarm through the internal network signaling system and automatically roll back the historical version to ensure business continuity.

[0195] The entire process revolves around the modules shown in the diagram (timed / manual triggering, forecast data capture, etc.), combined with details such as intranet interfaces and large model calls, to achieve automated, closed-loop management of weather forecasts from data acquisition to release.

[0196] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0197] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. All should be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A meteorological operational application system based on a general large model, characterized in that, The meteorological operational application system includes a data base module and an intelligent brain module; The data base module is used to: acquire multi-source heterogeneous meteorological data related to meteorological operations; perform anomaly detection, format standardization, and interpolation correction on the multi-source heterogeneous meteorological data to obtain structured data; collect various computing power indicators of the meteorological operations application system in real time; and output the structured data and various computing power indicators to the intelligent brain module respectively. The intelligent brain module is used to: based on a dynamic routing mechanism, associate and match the structured data with the meteorological business standard procedures stored in the knowledge base and the preset algorithms in the algorithm library, and determine the optimal model combination from the model library in combination with the computing power index; and process the structured data based on the optimal model combination to obtain the business processing result of the meteorological business. The intelligent brain module includes a knowledge base, a general large model library, an algorithm library, a component library, and a meteorological large model library; The knowledge base is used to store standard procedures, historical business data, emergency response plans, and business scenario adaptation rules in the meteorological business field, providing knowledge support for task recognition and model selection of the intelligent brain module. The general-purpose large model library is used to store general-purpose large models with generalized reasoning capabilities. These general-purpose large models are used to achieve cross-scenario business processing based on knowledge transfer from the meteorological field. The meteorological big model library is used to store specialized models optimized for meteorological business scenarios, including numerical forecasting models, disaster early warning models, and meteorological product generation models. The meteorological big model has the ability to deeply analyze and process meteorological professional data. The algorithm library is used to store various algorithms required for meteorological business processing, including data mining algorithms, machine learning algorithms, reinforcement learning algorithms, and spatiotemporal interpolation algorithms, providing algorithmic support for model processing of structured data; The component library is used to store the functional components required in the meteorological business processing process, including data preprocessing components, model inference components, and result visualization components, and to support the intelligent brain module in quickly building the business processing chain. The intelligent brain module is used to: invoke information and resources from the knowledge base and algorithm base through a dynamic routing mechanism, and select and combine the optimal model combination from the general large model library and the meteorological large model library in combination with computing power indicators, and perform processing on structured data to obtain business processing results.

2. The system according to claim 1, characterized in that, The meteorological operational application system also includes a support middleware module; The intelligent brain module is also used to: combine the computing power index and business priority to generate a business scheduling instruction, and output the business processing result and the business scheduling instruction to the support platform module; The support platform module is used to: schedule the corresponding algorithm resources in the algorithm library of the intelligent brain module through the algorithm platform according to the business scheduling instruction, and integrate the components required for meteorological business applications through the business platform to form an initial resource package; and filter and adapt the initial resource package according to the meteorological business scenario requirements to obtain an adapted resource package. The adapted resource package is used to support the implementation of meteorological business functions. At the same time, the status data during the execution of meteorological business is collected in real time. The status data and the business processing results are combined to generate resource adjustment suggestions and feed them back to the intelligent brain module. The intelligent brain module is also used to: correct the model parameters in the optimal model combination according to the resource adjustment suggestions.

3. The system according to claim 2, characterized in that, The meteorological business application system also includes a business application module; The business application module is used to: receive the adapted resource package output by the support platform module; start the workflow engine, and according to the business scenario information carried in the adapted resource package, call the business processing logic of the corresponding module and perform business analysis according to the standardized process; during the analysis, call the business processing results output by the intelligent brain module, and combine the business processing results to complete the deep processing of the corresponding business and generate multimodal business products.

4. The system according to claim 3, characterized in that, The meteorological business application system also includes a smart service module; The intelligent service module is used to: obtain the multimodal business product from the business application module; call preset user profile data, match the multimodal business product with the user profile, and determine the appropriate push channel; and accurately push the multimodal business product to the corresponding user or industry scenario through the determined push channel.

5. The system according to claim 4, characterized in that, The intelligent service module is also used to: initiate a streaming computing framework to collect user feedback information on the product in real time under various scenarios; perform structured processing on the feedback information to obtain structured feedback data; transmit the structured feedback data to the support platform module, and the support platform module triggers the system to adjust and optimize the data processing flow, model inference parameters, and business execution flow to form a closed loop.

6. The system according to claim 1, characterized in that, The data base module includes a station network observation information operation platform, a meteorological big data cloud platform, a meteorological comprehensive business monitoring system, and a meteorological data quality control and processing system; The station network observation information operation platform is used to: collect multi-source heterogeneous meteorological data required for target meteorological operations, including satellite remote sensing data, radar echo data, and ground observation station data, and transmit the collected multi-source heterogeneous meteorological data to the meteorological data quality control and processing system; The meteorological data quality control and processing system is used to: receive multi-source heterogeneous meteorological data transmitted from the station network observation information operation platform, perform anomaly detection, format standardization, and interpolation correction processing on the multi-source heterogeneous meteorological data in sequence, remove outliers in the data, unify the data format, and fill in data gaps, finally obtain structured data, and transmit the structured data to the meteorological big data cloud platform; The meteorological big data cloud platform is used to: receive structured data output from the meteorological data quality control and processing system, classify, store and manage the structured data using a distributed storage architecture, and provide structured data calling interfaces for other modules of the system. The meteorological integrated business monitoring system is used to: collect various computing power indicators during the operation of the meteorological business application system in real time, including GPU / CPU resource utilization, data transmission rate, and remaining storage resources; monitor and identify anomalies in the collected computing power indicators in real time; and output the computing power indicators to the intelligent brain module synchronously.

7. The system according to claim 2, characterized in that, The supporting middle platform module includes an AI middle platform, a knowledge middle platform, an algorithm middle platform, a data middle platform, a business middle platform, and a component middle platform; The data platform is used to: receive structured data output from the data base module, and achieve semantic alignment of multi-source data through semantic mapping, format unification, and data cleaning to form a standardized data resource pool and provide data support for other platforms; The algorithm platform is used to: schedule corresponding algorithm resources from the algorithm library of the intelligent brain module according to the business scheduling instructions output by the intelligent brain module, encapsulate and manage the algorithm resources, and provide algorithm call interfaces for business processing. The business platform is used to: integrate various business logic components and process templates required for meteorological business applications, build a standardized business processing framework, and support the rapid configuration and expansion of business scenarios; The AI ​​platform is used to manage the model resources of the intelligent brain module, including model deployment, model monitoring, and model iteration, while providing AI model calling services to other modules of the system. The knowledge platform is used to: connect to the knowledge base of the intelligent brain module, extract, integrate and update meteorological business knowledge, construct a dynamic knowledge graph, and provide knowledge support for business processing and model optimization; The component platform is used to: integrate the functional components required by each module of the system, including interface adaptation components, resource scheduling components, and log management components, to achieve component reuse and unified management; The supporting middle platform module is used to: integrate data, algorithms, models, and components into an initial resource package through the collaborative work of various middle platforms, and output it to the business application module after scenario adaptation and adjustment, while receiving feedback information to generate resource adjustment suggestions.

8. The system according to claim 3, characterized in that, The business application modules include an integrated weather and climate module, a decision-making meteorology module, a professional meteorological service module, an agricultural meteorological service module, a watershed meteorological service module, and a weather modification operation platform module. The integrated weather and climate module is used to: receive the adapted resource package output by the support platform module, call the business processing results of the intelligent brain module, perform various business analyses such as weather monitoring, climate trend analysis, and short-term climate prediction, and generate an integrated weather and climate analysis report. The decision-making meteorological module is used to: combine the disaster early warning results and impact assessment results output by the intelligent brain module with various business processing data, integrate the relevant data of terrain, population and economy, and generate meteorological service short reports and emergency response suggestions for government decision-making. The professional meteorological service module is used to generate customized meteorological service products for different industries, including power, transportation, and culture and tourism, by calling the business processing results of industry-specific components and intelligent brain modules in the adapted resource package. These products include power load forecasts, traffic weather warnings, and scenic spot weather guides. The agricultural meteorological service module is used to: combine the agricultural production cycle, crop growth model and meteorological element forecast results output by the intelligent brain module to perform various business analyses such as agricultural meteorological disaster risk assessment, crop growth monitoring and agricultural activity suggestions, and generate agricultural meteorological service products; The basin meteorological service module is used to: generate basin meteorological service products such as basin flood warnings and water resource scheduling suggestions by calling basin observation data and the results of precipitation forecasts and flood simulations from the intelligent brain module, in response to the hydrological and meteorological needs of a specific basin. The artificial weather modification platform module is used to: receive the cloud water resource analysis and operation condition judgment results output by the intelligent brain module, and combine the operation site distribution and equipment status information to generate artificial weather modification operation plans and operation effect evaluation reports to support the operation of artificial weather modification.

9. The system according to claim 4, characterized in that, The smart service module covers transportation service platform, cultural tourism service platform, city and county service platform, public service platform, power service platform, energy service platform, agricultural service platform, and government service platform; The traffic service platform is used to: receive traffic and meteorological service products output by the business application module, including road icing warnings and visibility forecasts; combine them with user profiles in the traffic industry; and push the products through various channels such as the traffic department's business system, roadside displays, and navigation apps to support traffic safety. The cultural and tourism service platform is used to: receive scenic area meteorological service products output by the business application module, including scenic area precipitation forecasts and ultraviolet intensity warnings; combine them with the user profiles of the cultural and tourism industry; and push the products through various channels such as the scenic area's official website, tourism APP, and travel agency service system to assist in the planning of cultural and tourism activities. The city and county service platform is used to: receive local meteorological service products output by the business application module, including city and county rainstorm warnings and township meteorological reports; and push products through various channels such as the city and county meteorological department business platform and government affairs APP, based on the needs and characteristics of city and county users, to support grassroots meteorological services. The public service platform is used to: receive public meteorological service products output by the business application module, including daily weather forecasts and life weather indices; combine them with public user profiles, including age, occupation, and living habits; and push the products through various channels such as SMS, WeChat official accounts, and meteorological apps to meet the public's daily meteorological information needs. The power service platform is used to: receive power meteorological service products output by the business application module, including cooling load forecasts and line icing warnings; and push products through various channels such as the power dispatch system and enterprise-specific service platforms in combination with the needs of power enterprise users to assist in power production dispatch. The energy service platform is used to: receive energy meteorological service products output by the business application module, including wind power forecast and photovoltaic irradiance forecast, and push the products through various channels of the energy management system and service platform in combination with the user profile of energy enterprises to support the efficient utilization of new energy. The agricultural service platform is used to: receive agricultural meteorological service products output by the business application module, including crop growth period forecasts and meteorological risk warnings for pests and diseases; combine these with farmer user profiles, including crops planted and planting scale; and push the products through various channels such as agricultural APPs and agricultural technology extension platforms to assist agricultural production. The government service platform is used to: receive decision-making meteorological service products output by business application modules, including major disaster warnings and emergency response suggestions; and push these products through various channels such as the government collaboration platform and emergency command system, in accordance with the needs of government departments, to support government decision-making and emergency management.

Images