Water conservancy data processing method and system based on multi-agent cooperation, electronic equipment and storage medium

Abstract

The invention provides a water conservancy data processing method and system based on multi-agent cooperation, electronic equipment and a storage medium, and relates to the technical field of water conservancy industry data processing. According to the method, through cooperation of multiple agents and a water conservancy industry large model, fusion of professional knowledge and processing capacity is achieved, multi-dimensional data processing and multi-task calculation can be completed at the same time, the task completion precision is improved by 30%-40% compared with a single model, and the professionality and consistency of output results are remarkably enhanced; moreover, multi-agent parallel processing replaces a serial process, and a dynamic task allocation mechanism is combined, so that the real-time task response time is shortened by more than 50%; the water conservancy industry large-model-driven adaptive processing strategy reduces the manual rule customization time, and the data preprocessing efficiency is improved by 60%.

Description

Technical Field

[0001] This application relates to the field of data processing technology in the water conservancy industry, and in particular to a water conservancy data processing method, system, electronic device and storage medium based on multi-agent collaboration. Background Technology

[0002] Currently, data processing technology in the water conservancy industry is gradually developing towards intelligence, with the core technology architecture mainly built around three stages: "data acquisition - centralized processing - model application". At the data acquisition level, a multi-source data acquisition network has been formed, centered on sensors, drone inspections, satellite remote sensing, and water conservancy project monitoring systems, enabling real-time acquisition and preliminary transmission of multi-dimensional data on watershed hydrology, water resources, water environment, and water projects.

[0003] In the data processing stage, existing technologies mostly adopt a centralized processing architecture, which involves aggregating, cleaning, and format-converting scattered multi-source data through a data platform before storing it in a distributed database or data warehouse. During processing, traditional data processing algorithms are often relied upon to complete basic data preprocessing, and then combined with water conservancy professional models or general machine learning models to achieve in-depth data analysis, such as water level prediction, flood risk assessment, and water pollution source tracing.

[0004] In recent years, some research has begun to introduce large-scale modeling technology, combining water conservancy industry knowledge bases, historical data, and general large-scale models to construct large-scale water conservancy industry models for scenarios such as water conservancy professional question answering and data report generation. However, the application of existing large-scale water conservancy industry models is mostly in a "single model, single task" mode, that is, a model is optimized for a specific water conservancy task, there is a lack of effective collaboration mechanisms between models, and the data processing flow and model application are relatively disconnected.

[0005] Therefore, the aforementioned technical problems urgently need to be solved. Summary of the Invention

[0006] In view of the above problems, this application is made to provide a water conservancy data processing method, system, electronic device, and storage medium based on multi-agent cooperation to overcome or at least partially solve the above problems. The technical solution is as follows: Firstly, a method for processing water conservancy data based on multi-agent collaboration is provided, the method comprising: A large-scale water conservancy industry model with professional knowledge Q&A, task parsing, and model calling capabilities is pre-built, and the large-scale water conservancy industry model is deployed on core computing nodes; Based on the requirements of the entire water conservancy data processing process, define and deploy a data acquisition intelligent agent, a data preprocessing intelligent agent, a task analysis intelligent agent, at least one professional computing intelligent agent, and a result fusion and output intelligent agent. A real-time communication link between each intelligent agent and the water conservancy industry big model is constructed using a preset communication protocol. A scheduling center is established and integrated into the water conservancy industry big model to store the status information and data processing progress of each intelligent agent. The data acquisition agent starts data acquisition according to a preset cycle or triggering conditions, performs preliminary processing on the acquired data, synchronizes the pre-processed data and acquisition status to the scheduling center, and submits the pre-processed data to the data preprocessing agent. The task analysis agent receives task requirements from users and submits the corresponding requirement text to the water conservancy industry big model. The water conservancy industry big model combines professional knowledge to analyze the task attributes and obtain the analysis results, including task type, priority, required data range, type of agent to be called, and professional model. Based on the analysis results of the water conservancy industry big data model and combined with the real-time status of each intelligent agent, the dispatch center automatically assigns tasks to the corresponding intelligent agents. The data preprocessing agent processes the pre-processed data according to the data processing rules issued by the water conservancy industry big model, synchronizes the processed standardized data to the dispatch center, and sends a data ready signal. The scheduling center pushes processed standardized data to at least one specialized computing agent. Each specialized computing agent performs calculations based on the processed standardized data, generates intermediate results, and feeds them back to the scheduling center. The result fusion and output agent obtains intermediate results from various professional computing agents from the scheduling center, transmits them to the water conservancy industry big model for verification, compares the consistency of different intermediate results, judges the rationality of intermediate results in combination with industry knowledge, removes contradictory data and adds explanations to obtain the processing result; then, according to user needs, the processing result is converted into a visual form for output, and at the same time, a complete report including data source, processing flow, core conclusions and recommended measures is generated.

[0007] In one possible implementation, a large-scale water conservancy industry model with capabilities for professional knowledge question answering, task parsing, and model invocation is pre-built, including: Based on a pre-set large model, incremental pre-training is performed using water conservancy industry corpus, incorporating professional knowledge of the water conservancy industry, and at the same time, the algorithm logic and parameter system of various types of water conservancy professional models are connected to pre-build a large water conservancy industry model with professional knowledge question answering, task parsing, and model calling capabilities.

[0008] In one possible implementation, data acquisition agents are grouped by region or data type, with each group corresponding to a specific data acquisition device, responsible for real-time data acquisition, preliminary filtering, and format standardization.

[0009] In one possible implementation, the data preprocessing agent is subdivided into a structured data processing unit and an unstructured data processing unit. The structured data processing unit is responsible for missing value imputation and outlier identification, while the unstructured data processing unit is responsible for remote sensing image segmentation and document keyword extraction.

[0010] In one possible implementation, specialized computational agents perform calculations based on processed, standardized data to generate intermediate results, including: Various professional computational intelligent agents receive model parameters from the large model of the water conservancy industry; Each specialized computing agent performs specialized calculations based on the processed standardized data and model parameters, generating intermediate results. If additional data is needed during the calculation process, each specialized computing agent initiates a data request to the corresponding agent through the scheduling center and receives the supplementary data returned by the corresponding agent.

[0011] In one possible implementation, the result fusion and output agent, for real-time early warning tasks, simultaneously pushes early warning information to the water conservancy monitoring platform and the smart terminals of relevant responsible persons.

[0012] In one possible implementation, the result fusion and output agent collects user feedback and actual monitoring data, and transmits them to the water conservancy industry big model. The water conservancy industry big model dynamically optimizes the processing rules and model parameters of subsequent tasks by comparing and analyzing the causes of errors, and updates the processing strategy of the agent to achieve self-iterative optimization.

[0013] Secondly, a water conservancy data processing system based on multi-agent collaboration is provided, the system comprising: The first deployment unit is used to pre-build a large water conservancy industry model with professional knowledge question answering, task parsing, and model calling capabilities, and deploy the large water conservancy industry model on the core computing node; The second deployment unit is used to define and deploy data acquisition intelligent agents, data preprocessing intelligent agents, task analysis intelligent agents, at least one professional computing intelligent agent, and result fusion and output intelligent agents according to the needs of the entire water conservancy data processing process. Establishment unit, used to build real-time communication links between each intelligent agent and the water conservancy industry big model using a preset communication protocol, establish scheduling center, integrate the scheduling center into the water conservancy industry big model, and be responsible for storing the status information and data processing progress of each intelligent agent; The collaborative processing unit is used by the data acquisition agent to initiate data acquisition according to a preset cycle or triggering conditions, perform preliminary processing on the acquired data, synchronize the pre-processed data and acquisition status to the scheduling center, and submit the pre-processed data to the data preprocessing agent; and The task analysis agent receives task requests from users and submits the corresponding request text to a large-scale water conservancy industry model. The model then analyzes the task attributes using its professional knowledge, obtaining information including task type, priority, required data range, the type of agent to be invoked, and the analysis results from the professional model. Based on the analysis results of the large-scale water conservancy industry model and combined with the real-time status of each agent, the dispatch center automatically assigns tasks to the corresponding agents; and The data preprocessing agent processes the initially processed data according to the data processing rules issued by the large-scale water conservancy model, synchronizes the standardized processed data to the dispatch center, and sends a data ready signal; and The scheduling center pushes processed, standardized data to at least one specialized computing agent. Each specialized computing agent performs calculations based on the processed, standardized data, generates intermediate results, and feeds them back to the scheduling center. The result fusion and output agent obtains intermediate results from various professional computing agents from the scheduling center, transmits them to the water conservancy industry big model for verification, compares the consistency of different intermediate results, judges the rationality of intermediate results in combination with industry knowledge, removes contradictory data and adds explanations to obtain the processing result; then, according to user needs, the processing result is converted into a visual form for output, and at the same time, a complete report including data source, processing flow, core conclusions and recommended measures is generated.

[0014] Thirdly, an electronic device is provided, comprising a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the water conservancy data processing method based on multi-agent cooperation as described above.

[0015] Fourthly, a storage medium is provided, the storage medium storing a computer program, wherein the computer program is configured to execute the water conservancy data processing method based on multi-agent cooperation as described above when running.

[0016] By utilizing the aforementioned technical solutions, this application provides a water conservancy data processing method, system, electronic device, and storage medium based on multi-agent collaboration. This method achieves the integration of professional knowledge and processing capabilities through the collaboration of multiple agents and a large-scale water conservancy industry model. It can simultaneously complete multi-dimensional data processing and multi-task computation, improving task completion accuracy by 30%-40% compared to a single model, and significantly enhancing the professionalism and consistency of the output results. Furthermore, parallel processing by multiple agents replaces the serial process, and combined with a dynamic task allocation mechanism, it reduces the real-time task response time by more than 50%. The adaptive processing strategy driven by the large-scale water conservancy industry model reduces the time for manual rule customization, improving data preprocessing efficiency by 60%. In addition, distributed deployment and fault tolerance mechanisms reduce the system failure rate by 80%, and process interruption time is controlled within 10 seconds. When adding new tasks, the agents are plug-and-play, shortening the system expansion cycle from the traditional 1-3 months to 1-2 weeks, significantly reducing development costs. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below.

[0018] Figure 1 A flowchart of a water conservancy data processing method based on multi-agent cooperation provided in an embodiment of this application is shown; Figure 2 A structural diagram of a water conservancy data processing system based on multi-agent collaboration provided in an embodiment of this application is shown. Detailed Implementation

[0019] Exemplary embodiments of the present application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this application will be thorough and complete, and will fully convey the scope of the present application to those skilled in the art.

[0020] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such use can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the term "comprising" and its variations should be interpreted as open-ended terms meaning "including but not limited to."

[0021] To address the aforementioned technical problems, embodiments of this application provide a water conservancy data processing method based on multi-agent collaboration, such as... Figure 1As shown, the water conservancy data processing method based on multi-agent collaboration may include the following steps S101 to S109: Step S101: Pre-build a large water conservancy industry model with professional knowledge Q&A, task parsing, and model calling capabilities, and deploy the large water conservancy industry model on the core computing node; Step S102: Based on the requirements of the entire water conservancy data processing process, define and deploy a data acquisition intelligent agent, a data preprocessing intelligent agent, a task analysis intelligent agent, at least one professional computing intelligent agent, and a result fusion and output intelligent agent. Step S103: A real-time communication link between each intelligent agent and the large water conservancy industry model is constructed using a preset communication protocol. A scheduling center is established and integrated into the large water conservancy industry model. The scheduling center is responsible for storing the status information and data processing progress of each intelligent agent. Here, the preset communication protocol can be the MQTT (Message Queuing Telemetry Transport) communication protocol, which is a lightweight publish-subscribe message transmission protocol designed specifically for the Internet of Things environment. Step S104: The data acquisition agent starts data acquisition according to a preset cycle or trigger condition, performs preliminary processing on the acquired data, synchronizes the pre-processed data and acquisition status to the scheduling center, and submits the pre-processed data to the data preprocessing agent. Step S105: The task analysis agent receives the task requirements from the user and submits the corresponding requirement text to the water conservancy industry big model. The water conservancy industry big model combines professional knowledge to analyze the task attributes and obtain the analysis results including task type, priority, required data range, type of agent to be called and professional model. Step S106: The dispatch center automatically assigns tasks to the corresponding intelligent agents based on the analysis results of the water conservancy industry big model and the real-time status of each intelligent agent. Step S107: The data preprocessing agent processes the pre-processed data according to the data processing rules issued by the water conservancy industry big model, synchronizes the processed standardized data to the dispatch center, and sends a data ready signal. In step S108, the scheduling center pushes processed standardized data to at least one specialized computing agent. Each specialized computing agent performs calculations based on the processed standardized data, generates intermediate results, and feeds them back to the scheduling center. Step S109: The result fusion and output agent obtains the intermediate results of various professional computing agents from the scheduling center, transmits them to the water conservancy industry big model for verification, compares the consistency of different intermediate results, judges the rationality of intermediate results in combination with industry knowledge, removes contradictory data and adds explanations to obtain the processing result; then, according to user needs, the processing result is converted into a visual form for output, and at the same time, a complete report including data source, processing flow, core conclusions and recommended measures is generated.

[0022] This embodiment achieves the integration of professional knowledge and processing capabilities through the collaboration of multiple agents and a large-scale water conservancy industry model. It can simultaneously complete multi-dimensional data processing and multi-task computation, improving task completion accuracy by 30%-40% compared to a single model, and significantly enhancing the professionalism and consistency of the output results. Furthermore, parallel processing by multiple agents replaces the serial process, and combined with a dynamic task allocation mechanism, the response time for real-time tasks is reduced by more than 50%. The adaptive processing strategy driven by the large-scale water conservancy industry model reduces the time for manual rule customization, and improves data preprocessing efficiency by 60%. In addition, distributed deployment and fault tolerance mechanisms reduce the system failure rate by 80%, and process interruption time is controlled within 10 seconds. When adding new tasks, the agents are plug-and-play, and the system expansion cycle is shortened from the traditional 1-3 months to 1-2 weeks, significantly reducing development costs.

[0023] This application embodiment provides a possible implementation method. The above step S101 pre-constructs a large water conservancy industry model with professional knowledge question answering, task parsing, and model calling capabilities, which may specifically include the following step A1: Step A1: Based on the pre-set large model, incremental pre-training is performed using water conservancy industry corpus, incorporating professional knowledge of the water conservancy industry, and at the same time, the algorithm logic and parameter system of various types of water conservancy professional models are connected to pre-build a large water conservancy industry model with the ability to answer professional knowledge questions, analyze tasks, and call models.

[0024] The pre-defined large model here can be an open-source large model, also known as a large language model (LLM). First, as the name suggests, it is large in scale, with network parameters reaching tens or hundreds of billions or even more. Second, it is generalizable, meaning it is not limited to specific problems or domains. Third, it is emergent, meaning it generates unexpected new capabilities. The parameter size of the large model can be flexibly configured according to actual needs, for example, 671 billion, 14 billion, 32 billion, 7 billion, or 1.5 billion, etc. The choice of parameter size mainly depends on computing resource conditions and model performance requirements; this embodiment does not impose any restrictions on this.

[0025] This embodiment upgrades the general, perception-level large model into a domain expert brain with water conservancy professional knowledge and computing capabilities, thereby laying a core foundation for building a smart water conservancy system.

[0026] This application provides a possible implementation method in which data acquisition agents are grouped according to region or data type. Each group of agents corresponds to a specific data acquisition device and is responsible for real-time data acquisition, preliminary filtering, and format standardization. The data acquisition device can be a sensor, drone, satellite, monitoring system, etc., and this embodiment does not limit it.

[0027] This application provides a possible implementation method in which the data preprocessing agent is subdivided into a structured data processing unit and an unstructured data processing unit. The structured data processing unit is responsible for missing value imputation and outlier identification; the unstructured data processing unit is responsible for remote sensing image segmentation and document keyword extraction. This embodiment's technical solution of subdividing the data preprocessing agent into structured and unstructured data processing units achieves integrated, efficient, and intelligent preprocessing of multi-source heterogeneous data through specialized division of labor and collaborative processing, laying a solid foundation for subsequent data analysis and applications.

[0028] This application embodiment provides a possible implementation method in which each specialized computing agent performs calculations based on the processed standardized data to generate intermediate results in step S108, which may specifically include the following steps B1 and B2: Step B1: Each specialized computational agent receives the model parameters issued by the large water conservancy industry model. In step B2, each specialized computing agent performs specialized calculations based on the processed standardized data and model parameters to generate intermediate results. If additional data is needed during the calculation process, each specialized computing agent initiates a data request to the corresponding agent through the scheduling center and receives the supplementary data returned by the corresponding agent.

[0029] In this embodiment, the model parameters are uniformly distributed by the large model of the water conservancy industry. This means that all professional computing agents follow the same set of verified and up-to-date computing standards. This avoids conflicts or deviations in computing results caused by inconsistent parameters, and ensures the authority and comparability of the final decision-making basis. If supplementary data is needed, a request is initiated, making the computing process change from static and preset to dynamic and on-demand. When an agent finds insufficient data in the computing process, it can request supplementation from other agents in real time, forming a closed-loop "perception-computation-re-computation" process, which greatly enhances the system's adaptability to complex and ever-changing scenarios.

[0030] This application provides a possible implementation method in which the result fusion and output intelligent agent, for real-time early warning tasks, simultaneously pushes early warning information to the water conservancy monitoring platform and the intelligent terminals of relevant responsible persons. In this way, an efficient, reliable, and actionable closed loop for early warning information transmission is constructed, thereby greatly improving the speed and effectiveness of emergency response.

[0031] This application provides a possible implementation method in which the result fusion and output agent collects user feedback and actual monitoring data, and transmits them to the water conservancy industry big model; the water conservancy industry big model dynamically optimizes the processing rules and model parameters of subsequent tasks by comparing and analyzing the causes of errors, and updates the processing strategy of the agent to achieve self-iterative optimization.

[0032] The above introduces Figure 1 The embodiments shown have various implementation methods for each step. The following will further explain the water conservancy data processing method based on multi-agent cooperation in this application through specific embodiments.

[0033] The existing technology has the following technical problems: (1) Low data processing efficiency and difficulty in adapting to the characteristics of multi-source heterogeneous data: Water conservancy data has the characteristics of wide sources, many types (structured data such as monitoring indicators, unstructured data such as remote sensing images, semi-structured data such as document reports, etc.) and large differences in real-time requirements. Under the centralized processing architecture, the serial processing mode of data aggregation, cleaning and analysis is prone to data congestion, resulting in delays in the response of tasks with high real-time requirements. At the same time, traditional processing algorithms have poor adaptability to multiple types of data and require a large number of processing rules to be manually customized.

[0034] (2) The application of large models in the water conservancy industry is limited and the task adaptability is insufficient: existing large models in the water conservancy industry are mostly optimized for single tasks. When faced with complex water conservancy scenarios, single models are difficult to take into account the professional needs of each task, and the model accuracy and processing effect decrease. Moreover, the model is disconnected from the data processing link and cannot dynamically adjust the processing strategy according to the data characteristics, resulting in unstable quality of model input data and affecting the reliability of output results.

[0035] (3) Lack of flexible collaboration mechanism and low utilization of resources and knowledge: Each water conservancy professional model and the large model store professional knowledge and processing capabilities separately and operate independently, making it impossible to achieve knowledge sharing and capability complementarity.

[0036] (4) Poor fault tolerance and scalability: The centralized architecture relies on the core processing node. Once the node fails, the entire data processing process is interrupted. At the same time, when new water conservancy tasks are added or new types of data are connected, the existing processing architecture and model need to be reconstructed on a large scale, which results in a long development cycle and high cost.

[0037] In response to the problems existing in the prior art, this specific embodiment has made the following innovations: (1) Construct a collaborative architecture of a large water conservancy industry model and multiple intelligent agents: The pre-trained large water conservancy industry model is used as the core knowledge base. Data collection, preprocessing, task analysis, professional calculation, and result output are encapsulated as independent intelligent agents. Each intelligent agent can achieve real-time interaction with the large model through a unified communication protocol.

[0038] (2) Dynamic task allocation and collaboration mechanism of intelligent agents: Based on the task parsing capability of the water conservancy industry big model, combined with the functional attributes of each intelligent agent (such as data acquisition intelligent agent, image processing intelligent agent, hydrological calculation intelligent agent) and load status, the dynamic splitting and optimal allocation of tasks are realized, while supporting temporary collaboration between intelligent agents (such as data preprocessing intelligent agent providing customized data to professional computing intelligent agent).

[0039] (3) Adaptive data processing strategy driven by large model: The water conservancy industry large model automatically generates suitable data processing rules (such as outlier judgment criteria and data fusion methods) based on the input data type and task requirements (such as real-time early warning and trend analysis), and sends them to the corresponding intelligent agents for execution without human intervention.

[0040] (4) Distributed fault tolerance and expansion mechanism: Each agent adopts distributed deployment. When a single agent fails, the large model can quickly call redundant agents or redistribute tasks to ensure process continuity. When adding new tasks or data types, only agents with corresponding functions need to be added, and the interface adaptation is completed through the large model without reconstructing the overall architecture.

[0041] This specific embodiment includes the following steps: Step 1: System initialization and agent configuration.

[0042] (1) Construct a knowledge base for a large-scale water conservancy model: Based on an open-source and universal large-scale model, incremental pre-training is carried out using water conservancy industry corpus (including water conservancy professional literature, standards, historical data reports, engineering cases, etc.), integrating professional knowledge in fields such as hydrology and water resources, water conservancy engineering, and water environment; at the same time, the algorithm logic and parameter system of various types of water conservancy professional models are connected to form a large-scale water conservancy model with professional knowledge question answering, task parsing, and model calling capabilities, which is deployed on core computing nodes.

[0043] (2) Configure multi-agent module: Based on the requirements of the entire process of water conservancy data processing, define and deploy the following functional agents. Each agent has independent data processing capabilities and standardized communication interfaces, and can interact with the water conservancy industry large model and other agents: ① Data acquisition agents: Grouped by region or data type (such as watershed monitoring group, engineering monitoring group, remote sensing data group, etc.). Each group of agents corresponds to specific data acquisition equipment (such as sensors, drones, satellites, monitoring systems, etc.), and is responsible for real-time data acquisition, preliminary filtering and format standardization, and synchronizing the data status to the large model.

[0044] ② Data preprocessing intelligent agent: It is subdivided into structured data processing unit and unstructured data processing unit. The structured data processing unit is responsible for missing value filling and outlier identification; the unstructured data processing unit is responsible for remote sensing image segmentation and document keyword extraction. The processing rules are dynamically issued by the water conservancy industry big model.

[0045] ③ Task Analysis Intelligent Agent: Receives task requirements input by users (such as water level forecast for a certain basin in the next 24 hours, source tracing of water pollution in reservoirs), analyzes the task type, core requirements (such as forecast accuracy, response time), required data types and professional models through the water conservancy industry big data model, and generates a task processing list (such as water level forecasting requires calling hydrological models, required data: rainfall, water level and topography data of the basin in the past 7 days).

[0046] ④ Specialized computational intelligent agents: Corresponding to various water conservancy professional models, such as hydrological computational intelligent agents, water quality computational intelligent agents, and engineering safety computational intelligent agents, they receive pre-processed target data and model parameters issued by the large water conservancy industry model (such as the confluence coefficient of the hydrological model, which is optimized by the large model based on historical similar flood events), perform specialized calculations, and output intermediate results (such as water level change curves and pollutant diffusion paths).

[0047] ⑤ Results Fusion and Output Agent: Collect intermediate results from various professional computational agents, verify the results through a large water conservancy industry model (such as determining whether the water quality assessment results match the hydrological simulation), remove redundant information and visualize the data (such as generating water level forecast line graphs, pollution source heat maps, etc.), and finally output reports or API (Application Programming Interface) data according to user needs.

[0048] (3) Establish a communication and scheduling center: The real-time communication link between the intelligent agent and the water conservancy industry big model is constructed using the MQTT communication protocol. The scheduling center is integrated into the water conservancy industry big model and is responsible for storing the status information of each intelligent agent (such as load rate, current task, fault status, etc.) and data processing progress.

[0049] Step 2: Multi-source data collection and initial uploading.

[0050] (1) The data acquisition agent starts data acquisition according to the preset cycle (e.g., real-time monitoring data is collected once every 5 seconds, and remote sensing data is collected once every 2 hours) or the trigger condition (e.g., encrypted acquisition during rainstorm warning): sensor data is directly transmitted to the corresponding acquisition agent through the 5G / BeiDou communication module; UAV inspection data is uploaded after being compressed by the edge computing node; satellite remote sensing data is called through the official interface and synchronized to the remote sensing data group agent.

[0051] (2) The data acquisition agent performs preliminary processing on the data: removes null values ​​and data with format errors generated by the offline sensor, and converts the output format of different devices into a unified JSON structure of data identifier-acquisition time-region-value-data type.

[0052] (3) The data acquisition agent will synchronize the processed data and acquisition status to the dispatch center. If data acquisition fails (such as sensor failure), it will immediately send a fault alarm to the dispatch center, which will be recorded and marked by the water conservancy industry big model.

[0053] Step 3: Task analysis and agent scheduling.

[0054] (1) Task input and parsing: Users input task requirements through terminals such as water conservancy monitoring platforms and mobile APPs. After receiving the task analysis intelligent agent, the task text is transmitted to the water conservancy industry big model. The big model combines professional knowledge to parse the task attributes: clarify the task type, priority, required data range, type of intelligent agent to be called and professional model.

[0055] (2) Dynamic task allocation: The scheduling center allocates tasks based on the analysis results of the large model and the real-time status of each agent: priority is given to assigning agents with a load rate of less than 30% to high-priority tasks; if the target data is not collected, the emergency collection process of the corresponding data collection agent is immediately triggered; if an agent fails, the task is automatically assigned to a redundant agent.

[0056] (3) Processing rule distribution: The water conservancy industry big model generates personalized data processing rules according to task requirements and distributes them to the corresponding intelligent agents. For example, in the flood risk assessment task, the water conservancy industry big model distributes "abnormal rainfall judgment standard: rainfall exceeding 50 mm in 1 hour is abnormal" to the data preprocessing intelligent agent and "confluence coefficient value of 1.2" to the hydrological calculation intelligent agent.

[0057] Step 4: Multi-agent collaborative data processing and computation.

[0058] (1) Data preprocessing: The data preprocessing agent receives the raw data uploaded by the acquisition agent and performs processing according to the rules issued by the large model: The structured data processing unit uses time-series interpolation to complete the missing water level data, identifies and marks abnormal rainfall values ​​by comparing historical data of the same period with water conservancy standards; The unstructured data processing unit segments the watershed remote sensing image, extracts water area change data, synchronizes the processed standardized data to the dispatch center, and sends a data ready signal.

[0059] (2) Professional calculation: The dispatch center pushes the preprocessed data and model parameters to the professional calculation agent. The agent calls the corresponding SWAT hydrological model to simulate the runoff generation and confluence process of the watershed, calculates the reservoir inflow and water level change trend in the next 12 hours, generates intermediate results (such as the water level will rise to 15.2m in the next 6 hours, exceeding the flood limit by 0.2m), and feeds them back to the dispatch center.

[0060] (3) Cross-agent collaboration: If a specialized computing agent needs to supplement data (e.g., a water quality computing agent needs water flow velocity data for hydrological calculation), it can send a data request to the corresponding agent through the scheduling center without manual intervention. For example, in a pollution source tracing task, the water quality computing agent requests water flow velocity data along the pollutant diffusion path from the hydrological computing agent, and the hydrological computing agent provides real-time feedback, realizing data linkage processing.

[0061] Step 5: Result fusion, verification and output.

[0062] (1) Results fusion: The results fusion agent obtains the intermediate results of each professional computing agent from the scheduling center and transmits them to the water conservancy industry big model for verification: compare the consistency of the results of different models, judge the rationality of the results in combination with industry knowledge (such as whether the pollution source tracing path conforms to the water flow direction), eliminate contradictory data and supplement explanations.

[0063] (2) Data visualization and output: The result fusion agent converts the processing results into a visualization form (such as line chart, heat map, 3D model, etc.) according to user needs, and generates a complete report including data source, processing flow, core conclusions and recommended measures; for real-time early warning tasks, the early warning information (such as flood risk level, early warning area, etc.) is pushed to the water conservancy monitoring platform and the mobile terminals of relevant responsible persons to ensure that the information is quickly reached.

[0064] Step Six: Feedback and Model Optimization.

[0065] After the results are output, the results fusion agent collects user feedback (such as a large error between the forecast result and the actual water level) and actual monitoring data (such as the actual water level of the reservoir in the next 12 hours), and transmits it to the water conservancy industry big model. The big model compares and analyzes the reasons for the error (such as unreasonable model parameters or imperfect data preprocessing rules), dynamically optimizes the processing rules and model parameters of subsequent tasks (such as adjusting the confluence coefficient of the hydrological model), and updates the processing strategy of the agent, so as to realize the system's self-iterative optimization.

[0066] This specific embodiment can achieve the following technical effects: (1) Improve data processing efficiency and real-time performance: Multi-agent parallel processing replaces serial processes, and combined with dynamic task allocation mechanism, the response time of real-time tasks (such as flood warning data processing) is shortened by more than 50%; the adaptive processing strategy driven by the large model in the water conservancy industry reduces the time for manual rule customization and improves data preprocessing efficiency by 60%.

[0067] (2) Enhanced ability to handle complex scenarios: The collaboration between multiple agents and the large model of the water conservancy industry has achieved the integration of professional knowledge and processing capabilities. For example, in the scenario of integrated watershed management, multi-dimensional data processing and multi-task calculation can be completed simultaneously. The accuracy of task completion is 30%-40% higher than that of a single model, and the professionalism and consistency of the output results are significantly enhanced.

[0068] (3) Improve system fault tolerance and scalability: Distributed deployment and fault tolerance mechanism reduce system failure rate by 80% and process interruption time is controlled within 10 seconds; when adding new tasks, the intelligent agent is plug-and-play, and the system expansion cycle is shortened from the traditional 1-3 months to 1-2 weeks, greatly reducing development costs.

[0069] (4) Reduce reliance on manual labor and operating costs: Adaptive data processing strategies and intelligent collaboration mechanisms reduce manual intervention by more than 90%, especially in data preprocessing and task scheduling, where no professional personnel are required to be on duty, which can reduce manual operation costs by more than 40% per year.

[0070] It should be noted that the sequence numbers of the steps in the above embodiments do not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. In practical applications, all the above possible implementation methods can be arbitrarily combined in a combined manner to form possible embodiments of this application, which will not be described in detail here.

[0071] Based on the water conservancy data processing method based on multi-agent collaboration provided in the above embodiments, and based on the same inventive concept, this application also provides a water conservancy data processing system based on multi-agent collaboration.

[0072] Figure 2This is a structural diagram of a water conservancy data processing system based on multi-agent collaboration provided in an embodiment of this application. Figure 2 As shown, the water conservancy data processing system based on multi-agent collaboration may specifically include a first deployment unit 210, a second deployment unit 220, an establishment unit 230, and a collaborative processing unit 240.

[0073] The first deployment unit 210 is used to pre-build a large water conservancy industry model with professional knowledge question answering, task parsing and model calling capabilities, and deploy the large water conservancy industry model on the core computing node; The second deployment unit 220 is used to define and deploy data acquisition intelligent agents, data preprocessing intelligent agents, task analysis intelligent agents, at least one professional computing intelligent agent, and result fusion and output intelligent agents according to the needs of the entire process of water conservancy data processing. Establishment unit 230 is used to build a real-time communication link between each intelligent agent and the water conservancy industry big model using a preset communication protocol, establish a scheduling center, integrate the scheduling center into the water conservancy industry big model, and be responsible for storing the status information and data processing progress of each intelligent agent. The collaborative processing unit 240 is used to initiate data acquisition by the data acquisition agent according to a preset period or triggering conditions, perform preliminary processing on the acquired data, synchronize the pre-processed data and acquisition status to the scheduling center, and submit the pre-processed data to the data preprocessing agent; and The task analysis agent receives task requests from users and submits the corresponding request text to a large-scale water conservancy industry model. The model then analyzes the task attributes using its professional knowledge, obtaining information including task type, priority, required data range, the type of agent to be invoked, and the analysis results from the professional model. Based on the analysis results of the large-scale water conservancy industry model and combined with the real-time status of each agent, the dispatch center automatically assigns tasks to the corresponding agents; and The data preprocessing agent processes the initially processed data according to the data processing rules issued by the large-scale water conservancy model, synchronizes the standardized processed data to the dispatch center, and sends a data ready signal; and The scheduling center pushes processed, standardized data to at least one specialized computing agent. Each specialized computing agent performs calculations based on the processed, standardized data, generates intermediate results, and feeds them back to the scheduling center. The result fusion and output agent obtains intermediate results from various professional computing agents from the scheduling center, transmits them to the water conservancy industry big model for verification, compares the consistency of different intermediate results, judges the rationality of intermediate results in combination with industry knowledge, removes contradictory data and adds explanations to obtain the processing result; then, according to user needs, the processing result is converted into a visual form for output, and at the same time, a complete report including data source, processing flow, core conclusions and recommended measures is generated.

[0074] This application embodiment provides a possible implementation, wherein the first deployment unit 210 is further configured to: Based on a pre-set large model, incremental pre-training is performed using water conservancy industry corpus, incorporating professional knowledge of the water conservancy industry, and at the same time, the algorithm logic and parameter system of various types of water conservancy professional models are connected to pre-build a large water conservancy industry model with professional knowledge question answering, task parsing, and model calling capabilities.

[0075] This application provides a possible implementation method. The second deployment unit 220 is also used to: group data acquisition agents according to region or data type, with each group of agents corresponding to a specific data acquisition device, responsible for real-time data acquisition, preliminary filtering and format standardization.

[0076] This application embodiment provides a possible implementation, wherein the second deployment unit 220 is further configured to: The data preprocessing agent is subdivided into structured data processing units and unstructured data processing units. The structured data processing unit is responsible for missing value imputation and outlier identification; the unstructured data processing unit is responsible for remote sensing image segmentation and document keyword extraction.

[0077] This application embodiment provides a possible implementation, wherein the collaborative processing unit 240 is further configured to: Various professional computational intelligent agents receive model parameters from the large model of the water conservancy industry; Each specialized computing agent performs specialized calculations based on the processed standardized data and model parameters, generating intermediate results. If additional data is needed during the calculation process, each specialized computing agent initiates a data request to the corresponding agent through the scheduling center and receives the supplementary data returned by the corresponding agent.

[0078] This application embodiment provides a possible implementation, wherein the collaborative processing unit 240 is further configured to: The result fusion and output of the intelligent agent for real-time early warning tasks simultaneously pushes early warning information to the water conservancy monitoring platform and the intelligent terminals of relevant responsible persons.

[0079] This application embodiment provides a possible implementation, wherein the collaborative processing unit 240 is further configured to: The results fusion and output process involves the intelligent agent collecting user feedback and actual monitoring data, which is then transmitted to the water conservancy industry big data model. The big data model analyzes the causes of errors, dynamically optimizes the processing rules and model parameters for subsequent tasks, and updates the processing strategy of the intelligent agent to achieve self-iterative optimization.

[0080] Based on the same inventive concept, this application also provides an electronic device, including a processor and a memory, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the water conservancy data processing method based on multi-agent cooperation of any of the above embodiments.

[0081] Based on the same inventive concept, this application also provides a storage medium storing a computer program, wherein the computer program is configured to execute the water conservancy data processing method based on multi-agent cooperation of any of the above embodiments when running.

[0082] Based on the same inventive concept, this application also provides a computer program product, including a computer program configured to execute the work order quality inspection method based on a large language model of any of the above embodiments at runtime.

[0083] Those skilled in the art will clearly understand that the specific working process of the systems, devices, and modules described above can be referred to the corresponding process in the foregoing method embodiments. For the sake of brevity, it will not be repeated here.

[0084] Those skilled in the art will understand that the technical solution of this application, or all or part of it, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several program instructions to cause an electronic device (e.g., a personal computer, server, or network device) to execute all or part of the steps of the methods described in the embodiments of this application when running the program instructions. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0085] Alternatively, all or part of the steps of the foregoing method embodiments can be implemented by hardware (such as electronic devices like personal computers, servers, or network devices) associated with program instructions. The program instructions can be stored in a computer-readable storage medium. When the program instructions are executed by the processor of the electronic device, the electronic device executes all or part of the steps of the methods described in the embodiments of this application.

[0086] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that within the spirit and principles of this application, modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the corresponding technical solutions to leave the protection scope of this application.

Claims

1. A water conservancy data processing method based on multi-agent cooperation, characterized in that, The method includes: A large-scale water conservancy industry model with professional knowledge Q&A, task parsing, and model calling capabilities is pre-built, and the large-scale water conservancy industry model is deployed on core computing nodes; Based on the requirements of the entire water conservancy data processing process, define and deploy data acquisition intelligent agents, data preprocessing intelligent agents, task analysis intelligent agents, at least one professional computing intelligent agent, and result fusion and output intelligent agents; A real-time communication link between each intelligent agent and the water conservancy industry big model is constructed using a preset communication protocol. A scheduling center is established and integrated into the water conservancy industry big model to store the status information and data processing progress of each intelligent agent. The data acquisition agent starts data acquisition according to a preset cycle or triggering conditions, performs preliminary processing on the acquired data, synchronizes the pre-processed data and acquisition status to the scheduling center, and submits the pre-processed data to the data preprocessing agent. The task analysis agent receives task requirements from users and submits the corresponding requirement text to the water conservancy industry big model. The water conservancy industry big model combines professional knowledge to analyze the task attributes and obtain the analysis results, including task type, priority, required data range, type of agent to be called, and professional model. Based on the analysis results of the water conservancy industry big data model and combined with the real-time status of each intelligent agent, the dispatch center automatically assigns tasks to the corresponding intelligent agents. The data preprocessing agent processes the pre-processed data according to the data processing rules issued by the water conservancy industry big model, synchronizes the processed standardized data to the dispatch center, and sends a data ready signal. The scheduling center pushes processed standardized data to at least one specialized computing agent. Each specialized computing agent performs calculations based on the processed standardized data, generates intermediate results, and feeds them back to the scheduling center. The result fusion and output agent obtains intermediate results from various professional computing agents from the scheduling center, transmits them to the water conservancy industry big model for verification, compares the consistency of different intermediate results, judges the rationality of intermediate results in combination with industry knowledge, removes contradictory data and adds explanations to obtain the processing result; then, according to user needs, the processing result is converted into a visual form for output, and at the same time, a complete report including data source, processing flow, core conclusions and recommended measures is generated.

2. The method according to claim 1, characterized in that, A large-scale water conservancy industry model with pre-built capabilities including professional knowledge question answering, task parsing, and model invocation is included: Based on a pre-set large model, incremental pre-training is performed using water conservancy industry corpus, incorporating professional knowledge of the water conservancy industry, and at the same time, the algorithm logic and parameter system of various types of water conservancy professional models are connected to pre-build a large water conservancy industry model with professional knowledge question answering, task parsing, and model calling capabilities.

3. The method according to claim 1, characterized in that, Data acquisition agents are grouped by region or data type. Each group of agents corresponds to a specific data acquisition device and is responsible for real-time data acquisition, preliminary filtering, and format standardization.

4. The method according to claim 1, characterized in that, The data preprocessing agent is divided into a structured data processing unit and an unstructured data processing unit. The structured data processing unit is responsible for missing value imputation and outlier identification; the unstructured data processing unit is responsible for remote sensing image segmentation and document keyword extraction.

5. The method according to claim 1, characterized in that, Each specialized computational agent performs calculations based on the processed, standardized data, generating intermediate results, including: Various professional computational intelligent agents receive model parameters from the large model of the water conservancy industry; Each specialized computing agent performs specialized calculations based on the processed standardized data and model parameters, generating intermediate results. If additional data is needed during the calculation process, each specialized computing agent initiates a data request to the corresponding agent through the scheduling center and receives the supplementary data returned by the corresponding agent.

6. The method according to claim 1, characterized in that, The result fusion and output of the intelligent agent for real-time early warning tasks simultaneously pushes early warning information to the water conservancy monitoring platform and the intelligent terminals of relevant responsible persons.

7. The method according to claim 1, characterized in that, The results fusion and output process involves the intelligent agent collecting user feedback and actual monitoring data, which is then transmitted to the water conservancy industry big data model. The big data model analyzes the causes of errors, dynamically optimizes the processing rules and model parameters for subsequent tasks, and updates the processing strategy of the intelligent agent to achieve self-iterative optimization.

8. A water conservancy data processing system based on multi-agent cooperation, characterized in that, The system includes: The first deployment unit is used to pre-build a large water conservancy industry model with professional knowledge question answering, task parsing, and model calling capabilities, and deploy the large water conservancy industry model on the core computing node; The second deployment unit is used to define and deploy data acquisition intelligent agents, data preprocessing intelligent agents, task analysis intelligent agents, at least one professional computing intelligent agent, and result fusion and output intelligent agents according to the needs of the entire water conservancy data processing process. Establishment unit, used to build real-time communication links between each intelligent agent and the water conservancy industry big model using a preset communication protocol, establish scheduling center, integrate the scheduling center into the water conservancy industry big model, and be responsible for storing the status information and data processing progress of each intelligent agent; The collaborative processing unit is used by the data acquisition agent to initiate data acquisition according to a preset cycle or triggering conditions, perform preliminary processing on the acquired data, synchronize the pre-processed data and acquisition status to the scheduling center, and submit the pre-processed data to the data preprocessing agent; and The task analysis agent receives task requests from users and submits the corresponding request text to a large-scale water conservancy industry model. The model then analyzes the task attributes using its professional knowledge, obtaining information including task type, priority, required data range, the type of agent to be invoked, and the analysis results from the professional model. Based on the analysis results of the large-scale water conservancy industry model and combined with the real-time status of each agent, the dispatch center automatically assigns tasks to the corresponding agents; and The data preprocessing agent processes the initially processed data according to the data processing rules issued by the large-scale water conservancy model, synchronizes the standardized processed data to the dispatch center, and sends a data ready signal; and The scheduling center pushes processed, standardized data to at least one specialized computing agent. Each specialized computing agent performs calculations based on the processed, standardized data, generates intermediate results, and feeds them back to the scheduling center. The result fusion and output agent obtains intermediate results from various professional computing agents from the scheduling center, transmits them to the water conservancy industry big model for verification, compares the consistency of different intermediate results, judges the rationality of intermediate results in combination with industry knowledge, removes contradictory data and adds explanations to obtain the processing result; then, according to user needs, the processing result is converted into a visual form for output, and at the same time, a complete report including data source, processing flow, core conclusions and recommended measures is generated.

9. An electronic device, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the water conservancy data processing method based on multi-agent cooperation as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium stores a computer program, wherein the computer program is configured to execute the water conservancy data processing method based on multi-agent cooperation as described in any one of claims 1 to 7 when it runs.

Images