A Reservoir Greenhouse Gas Accounting Method and System Based on Multi-Agent Collaboration

Abstract

This invention provides a reservoir greenhouse gas accounting method and system based on multi-agent collaboration, belonging to the field of reservoir greenhouse gas accounting technology. The invention uses a central scheduling agent to parse the user's natural language accounting request for the target reservoir, constructs a task dependency graph, and schedules multi-source data acquisition agents, reference reservoir matching agents, emission accounting agents, uncertainty analysis agents, and sensitivity analysis agents to work collaboratively. Basic reservoir parameters and watershed environmental parameters are automatically acquired. The reference reservoir matching agent performs migration and completion of missing reservoir water feature parameters based on multi-dimensional similarity calculation. The emission accounting agent calls the accounting model to calculate the reservoir's greenhouse gas emissions. The uncertainty analysis agent and the sensitivity analysis agent respectively evaluate the confidence interval and key sensitive parameters of the emission results. This invention achieves automated execution of the entire reservoir greenhouse gas accounting process, improving accounting efficiency and result reliability.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of reservoir greenhouse gas accounting and artificial intelligence, and relates to a reservoir greenhouse gas accounting method and system based on multi-agent collaboration. Background Technology

[0002] Greenhouse gases from reservoirs refer to gases emitted into the atmosphere through water-air interface exchange after the reservoir is built, due to the decomposition of vegetation and soil organic matter in the submerged area and changes in the physicochemical processes of the water. These gases primarily include carbon dioxide. methane and nitrous oxide .

[0003] Currently, reservoir greenhouse gas accounting technologies mainly include three categories: field monitoring methods, empirical estimation methods, and parametric modeling methods. Field monitoring methods collect gas samples by deploying floating boxes on the water surface to directly obtain the greenhouse gas exchange flux at the water-air interface; empirical estimation methods are represented by the IPCC emission factor method, which estimates based on parameters such as the climate zone where the reservoir is located, the reservoir age, and the water surface area; parametric modeling methods are represented by the G-Res tool, which introduces the physical characteristics of the reservoir, watershed conditions, and environmental parameters to perform refined modeling and accounting of greenhouse gas emissions. However, the above methods still face the following common problems in practical applications: (1) The accounting process has a low degree of automation and relies heavily on the manual intervention of domain experts. From the determination of accounting targets, the item-by-item query and verification of input parameters, the selection and configuration of accounting models, to the compilation of final results and the writing of reports, there is a lack of effective automatic connection mechanism between each link, which requires manual step-by-step driving. As a result, the single accounting cycle is long and the labor cost is high, making it difficult to support the large-scale, batch greenhouse gas assessment needs of reservoirs. (2) The input data required for accounting are scattered from multiple heterogeneous data sources such as meteorology, hydrology, remote sensing, soil, and geographic information, and the data formats, spatial resolutions and temporal granularities are different. Existing tools lack the ability to uniformly access and automatically integrate multi-source heterogeneous data.

[0004] Therefore, there is an urgent need for a technical solution that can automatically complete data acquisition, parameter integration, and accounting execution to reduce reliance on manual labor and improve accounting efficiency. In recent years, intelligent agent technology based on large language models has developed rapidly. It has the capabilities of autonomous task decomposition, multi-agent collaboration, and external tool invocation, providing a new technical path for realizing the automation and intelligence of reservoir greenhouse gas accounting processes. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a reservoir greenhouse gas accounting method and system based on multi-agent collaboration. This method constructs multiple functionally specialized agents and uses a collaborative orchestration mechanism to achieve task allocation and data transfer among the agents, thereby automating the entire reservoir greenhouse gas accounting process. The agents involved in this method include: a central scheduling agent, a multi-source data acquisition agent, a reference reservoir matching agent, an emissions accounting agent, an uncertainty analysis agent, and a sensitivity analysis agent. Each agent collaborates to complete the accounting task through a collaborative orchestration mechanism. This method and system can solve the technical problems of low automation in existing reservoir greenhouse gas accounting processes, difficulty in automatically acquiring and fusing multi-source heterogeneous data, and interruptions in the accounting process due to the lack of some key parameters.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A reservoir greenhouse gas accounting method based on multi-agent cooperation, the method specifically includes the following steps: S1. Steps for constructing a knowledge base in the reservoir domain: Collect unstructured document data related to the target reservoir, perform structured processing on the original documents through a document parsing engine to generate standardized corpus, and import the standardized corpus into a knowledge base constructed based on retrieval enhancement generation technology to complete document segmentation, vectorization and index construction, forming a knowledge base in the reservoir domain that can be queried by various intelligent agents through natural language. S2. Calculation Request Parsing Steps: The central scheduling agent receives the natural language calculation request Q submitted by the user and parses it into a structured task description. ,in For the accounting target, For the set of information known to the user, The set of constraints is defined; an execution plan is generated based on the structured task description, and a task dependency graph is constructed. ,in For a set of task nodes, Based on the dependencies between task nodes, tasks are assigned to each agent in the order of dependency. S3. Basic Parameter Acquisition Steps: Based on the target reservoir identifier and the constraints in the structured task description, the multi-source data acquisition agent obtains the set of basic reservoir parameters from external data sources. Each parameter is encapsulated as a structured field. ,in For parameter values, For data source identification, For confidence level, For the range of uncertainty, Mark the data source type; S4. Missing Parameter Completion Steps: The parameters required for calculation are classified into three categories according to their properties: reservoir inherent attribute parameters, watershed environmental parameters, and reservoir water body characteristic parameters; the reference reservoir matching agent is based on the feature vector of the target reservoir. Reference reservoir database Multidimensional similarity calculation of various reference reservoirs in the data, and retrieval of the most similar reference reservoir. When the similarity meets the preset threshold, The corresponding parameter value is used as the fill-in value for the missing field, and its source type is marked. Set the migration to complete the data; when the similarity is lower than the threshold, do not perform completion and provide feedback to the user; S5. Watershed Environmental Parameter Acquisition Steps: Based on the geographical coordinates of the target reservoir, the multi-source data acquisition agent acquires meteorological sequence data, soil attribute data, and spatial geographic data through meteorological data interfaces, soil data interfaces, and geographic information data interfaces, respectively, to form a set of watershed environmental parameters. ; S6. Emissions Accounting Steps: The emissions accounting agent summarizes the basic parameter set and the watershed environmental parameter set into the accounting input parameter set. According to the constraints Determine the accounting model from the pre-set accounting model library. Perform greenhouse gas emission calculations; S7. Uncertainty Analysis Steps: The uncertainty analysis agent will calculate the input parameters. In this context, each parameter is determined according to its uncertainty range. The data is mapped to random variables, and the Monte Carlo method is used for multiple sampling and calculations to obtain the confidence interval of the emission results. S8. Sensitivity Analysis Steps: The sensitivity analysis agent calculates the sensitivity index of each input parameter to the emission results, sorts them from largest to smallest sensitivity, and outputs the ranking results of key parameters. S9. Report generation steps: The central dispatching agent summarizes the emission accounting results, uncertainty analysis results, and sensitivity analysis results to generate a structured accounting report.

[0007] Furthermore, in step S1, the unstructured document data includes accounting guidelines, operation and maintenance reports, and scientific and technological literature; the structuring process includes performing layout analysis on the documents, automatically identifying the structured elements in the documents, and converting them into standardized corpus.

[0008] Furthermore, in step S2, the task dependency graph is a directed acyclic graph, and the central scheduling agent performs parallel scheduling of task nodes with no dependency relationship and monitors the execution status of each agent in real time.

[0009] Furthermore, in step S4, when the inherent attribute parameters of the reservoir are missing, feedback is sent to the user requesting their provision; the watershed environmental parameters are automatically acquired by the multi-source data acquisition agent through a spatial data interface; and when the reservoir water body characteristic parameters cannot be directly obtained, they are completed through migration by matching with a reference reservoir; the feature vector This includes the target reservoir's geographical location, storage capacity, climate zone, age, and intended use.

[0010] Furthermore, in step S5, the multi-source data acquisition agent acquires watershed environmental parameters in the following priority order: spatial data interface first, geographic information processing second, and text retrieval as a fallback.

[0011] Furthermore, in step S6, the accounting model library includes the IPCC Tier 1 emission factor method and the G-Res parameterized model. The model library also supports registration and access to other models. The IPCC Tier 1 emission factor method refers to the Tier 1 accounting method specified in the Intergovernmental Panel on Climate Change (IPCC) guidelines, which estimates greenhouse gas emissions from reservoirs based on default emission factors and activity data such as reservoir area, climate zone, and reservoir age. The G-Res parameterized model is a parameterized model for estimating greenhouse gas emissions from reservoirs, which estimates greenhouse gas emissions based on parameters such as reservoir morphology, hydrological operation, climate conditions, reservoir age, trophic status, and pre-inundation land use. and Emissions are estimated empirically or semi-empirically; the emission calculation includes calculating separately... Emissions and Emissions and convert to equivalent ,in , for The global warming potential coefficient.

[0012] Furthermore, in step S8, the sensitivity index is a first-order sensitivity index based on variance decomposition: ,in To provide input parameters Conditional expectations of emission results under given conditions.

[0013] Furthermore, the method also includes a feedback correction step: the user submits correction feedback F to the accounting report, and the central scheduling agent parses the correction feedback to generate a correction set. And based on the task dependency graph Identify the set of downstream task nodes affected: Only for The corresponding task steps are re-executed, keeping the existing results of the unaffected steps unchanged.

[0014] This invention also provides a reservoir greenhouse gas accounting system based on multi-agent collaboration, the system comprising: The knowledge base construction module is used to collect unstructured document data related to the target reservoir, perform structured processing on it, and then import it into a knowledge base built based on retrieval enhancement generation technology to form a domain knowledge base that can be retrieved through natural language. The central scheduling agent is used to receive users' natural language accounting requests and parse them into structured task descriptions, construct task dependency graphs based on the structured task descriptions and assign tasks to each agent, monitor the execution status of each agent, and summarize the output results of each agent to generate a structured accounting report. A multi-source data acquisition intelligent agent is used to obtain a set of basic reservoir parameters and a set of watershed environmental parameters from external data sources, and encapsulate each parameter into a structured field containing parameter value, data source identifier, data confidence level, uncertainty range and data source type label; The reference reservoir matching agent is used to search for the most similar reference reservoir when the water body feature parameters of the reservoir are missing and cannot be directly obtained from public data sources. Based on the multidimensional similarity calculation between the feature vector of the target reservoir and each reference reservoir in the reference reservoir database, the agent fills in the missing field with the corresponding parameter value when the similarity meets the preset threshold and marks it as migration and completion data; otherwise, the agent will provide feedback to the user to request the data. An emissions accounting agent is used to determine the accounting model from a pre-set accounting model library based on constraints, and to perform greenhouse gas emissions calculations based on the set of accounting input parameters. An uncertainty analysis agent is used to map the input parameters of the calculation to random variables and perform multiple sampling and calculations using the Monte Carlo method to obtain the confidence interval of the emission results; A sensitivity analysis agent is used to calculate the sensitivity index of each input parameter to the emission results and output the ranking results of key parameters.

[0015] The beneficial effects of this invention are as follows: First, by using a centrally dispatching intelligent agent to collaboratively orchestrate knowledge base construction, reference reservoir matching, multi-source data collection, and accounting analysis, the traditional manually operated reservoir greenhouse gas accounting process is transformed into an automated, intelligent process. This reduces the human costs of data preparation and accounting execution, and improves accounting efficiency. Second, by using a multi-source data collection intelligent agent to automatically connect to multiple external data interfaces and integrate various types of environmental data, the problem of scattered input data from different data sources and difficulty in obtaining it uniformly is solved. For situations where reservoir water characteristic parameters are difficult to obtain directly from public data sources, a reference reservoir matching intelligent agent performs migration and completion based on multi-dimensional similarity calculations, and sets similarity thresholds to control the completion quality, ensuring the executability of the accounting process and the reliability of the results. Third, by using uncertainty analysis and sensitivity analysis intelligent agents to perform probability assessments and key parameter identification on the accounting results, the output results not only include estimated emission values ​​but also corresponding confidence intervals and key sensitive parameter information, providing quantitative basis for subsequent data supplementation and decision-making. Fourth, through the feedback and correction mechanism, users can submit correction feedback on the accounting report. The central scheduling agent only re-executes the affected downstream task nodes, avoiding repeated calculations throughout the process and improving interaction efficiency.

[0016] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0017] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 This is an overall flowchart of a reservoir greenhouse gas accounting method based on multi-agent collaboration according to the present invention; Figure 2 This is a flowchart illustrating the task parsing and scheduling process of the central scheduling intelligent agent in this invention. Figure 3 This is a flowchart illustrating the implementation of greenhouse gas accounting in the Three Gorges Reservoir in an embodiment of the present invention. Detailed Implementation

[0018] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0019] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0020] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0021] The reservoir greenhouse gas accounting method based on multi-agent cooperation proposed in this invention, such as... Figure 1 As shown, the method specifically includes the following steps: User input calculation request: The user submits a natural language calculation request Q for the target reservoir by inputting the calculation request. The calculation request includes at least the name or geographical coordinates of the target reservoir, and may also include known basic parameters, calculation time range and result display requirements.

[0022] The central scheduling agent generates execution plans and task scheduling: such as Figure 2 As shown, the central scheduling agent receives the natural language processing request Q submitted by the user and calls the large language model to parse it into a structured task description:

[0023] in, For accounting targets, such as annual emissions or monthly emissions; A set of known information provided to the user; This is a set of constraints, such as model selection constraints and time range. The structured task description serves as the shared context input for subsequent agents.

[0024] Based on this, the central scheduling agent generates an execution plan P according to the structured task description and constructs a task dependency graph, which is a directed acyclic graph, represented as:

[0025] in This represents a set of task nodes, where each node corresponds to a subtask executed by an agent. This indicates the dependencies between task nodes, meaning that a subsequent task can only be executed after the preceding task is completed. The central scheduling agent, based on... Determine the execution order of each task, schedule tasks with no dependencies in parallel, and monitor the execution status of each task.

[0026] The multi-source data acquisition agent obtains basic reservoir parameters from a search engine interface, a domain knowledge base, and an external data source set S, based on the target reservoir identifier and the constraints in the structured task description. The basic parameters include at least the reservoir name, geographical coordinates, water surface area, reservoir age, and purpose. To ensure data traceability, the multi-source data acquisition agent encapsulates each parameter as a structured field:

[0027] in For parameter values, For data source identification, Confidence level, characterizing the degree of credibility of this parameter. For the range of uncertainty, This is used to characterize whether the parameter is obtained by completing a similar reservoir. The confidence level... Based on the data source type of the parameters The parameters are assigned values ​​based on their source hierarchy, following these principles: parameters from official authoritative data sources have the highest confidence level; parameters automatically obtained through spatial data interfaces have the next highest confidence level; parameters extracted through web searches or literature reviews have the next lowest confidence level; and parameters completed through reference reservoir migration have the lowest confidence level, with their confidence level positively correlated with matching similarity—higher similarity results in higher confidence level for the completed parameters. This confidence level is included in the accounting report as an indicator of parameter quality, allowing users to assess the data reliability of each parameter. The aforementioned uncertainty range... Prioritize using the error range, standard deviation, or coefficient of variation provided by the data source itself. When the data source does not provide uncertainty information, estimate the uncertainty based on the reliability of the data source and the parameter's own variation characteristics: assign a smaller uncertainty range to parameters with clear numerical records; determine the uncertainty range based on typical coefficients of variation reported in the literature for parameters with spatial interpolation or temporal extrapolation; for parameters supplemented by migration of a reference reservoir, increase the uncertainty range accordingly based on the uncertainty of the corresponding parameter in the reference reservoir to reflect the additional uncertainty introduced by the migration supplementation. The output base parameter set is as follows: , where n represents the number of parameters.

[0028] Missing parameter completion: The central scheduling agent completes the collected parameters. Completeness checks are performed, and parameters are categorized into three types based on their nature: Inherent reservoir properties (such as water surface area and total capacity) are unique to the target reservoir; if missing, the user is prompted to provide them. Watershed environmental parameters (such as precipitation and soil organic carbon content) are automatically acquired by the multi-source data acquisition agent through a spatial data interface. Reservoir water body characteristic parameters (such as total phosphorus concentration and trophic status) are mainly controlled by the reservoir's physical morphology, climate conditions, and watershed input characteristics; they exhibit statistical similarity among reservoirs of the same climate zone and size. When these parameters cannot be directly obtained, a reference reservoir matching agent is triggered to perform migration and completion. The reference reservoir matching is based on the target reservoir's feature vector. (Including geographical location, reservoir capacity, climate zone, reservoir age, and purpose, etc.) in reference reservoir databases. Search for the reference reservoirs most similar to the target reservoir. :

[0029] in Based on the comprehensive similarity calculated using the aforementioned feature dimensions, if the highest similarity score is below a preset threshold, completion will not be performed; the parameter will be marked as uncompleted and feedback will be sent to the user. If the threshold requirement is met, completion will be performed. The corresponding parameter value is used as the completion value for the missing field. and in its source identification The data marked in the middle is the migration completion data, to distinguish it from the actual measured data.

[0030] The multi-source data acquisition intelligent agent obtains watershed environmental parameters: the multi-source data acquisition intelligent agent further obtains a set of watershed environmental parameters based on the coordinates or spatial location of the target reservoir. The data should include at least the catchment area, precipitation, runoff, soil properties, and land use type. The preferred approach is to obtain the data in the order of spatial data interface first, followed by geographic information processing, and finally text retrieval as a fallback, to improve parameter coverage and stability. The output also uses structured fields. Form. Before the accounting is executed, the emission accounting agent determines the trophic status level of the target reservoir based on the collected water quality indicators and environmental parameters. T As one of the input parameters for emissions accounting.

[0031] The emissions accounting agent performs greenhouse gas emissions calculations: The central dispatch agent summarizes the outputs of the above-mentioned preliminary tasks and transmits them to the emissions accounting agent to form a set of accounting input parameters. The emissions accounting agent is based on constraints. From the preset model library Determine the accounting model Calculate separately and Emissions, and converted to equivalent.

[0032]

[0033]

[0034]

[0035] in, For data parameters, for The global warming potential coefficient.

[0036] Uncertainty analysis agents and sensitivity analysis agents execute in parallel: After obtaining the emission accounting results, the central scheduling agent schedules the uncertainty analysis agents and sensitivity analysis agents in parallel to perform analysis tasks. The uncertainty analysis agent first denotes the set of input parameters involved in the accounting as... Each parameter All are represented in the form of structured field objects. Based on the above fields, the uncertainty analysis agent maps the input parameters into a vector of random variables. According to each parameter Determine its probability distribution After selecting the accounting model Under these conditions, the emissions accounting process is expressed as follows: For uncertainty analysis agents, the Monte Carlo propagation method is preferred for uncertainty propagation: [The text abruptly ends here, likely due to an incomplete sentence or a formatting error.] Second sampling, satisfying and calculate .

[0037] The sensitivity analysis agent employs a global sensitivity analysis method based on variance decomposition to identify the sensitivity of emissions results. Key input parameters with significant impact. The first-order sensitivity exponent for each parameter is defined as:

[0038] in To provide input parameters Conditional expectation of emission results under the given conditions This involves variance calculation. The sensitivity analysis agent calculates the sensitivity index for each parameter. Then, press Sort the data from largest to smallest and output the sorting results of key parameters to characterize the degree of influence of each input parameter on the calculation results.

[0039] Accounting Report Generation and Feedback Correction: The central dispatching agent summarizes the emission accounting results, uncertainty analysis results, and sensitivity analysis results to generate a structured accounting report. At least including: emissions of each gas Confidence intervals of emission results The report includes key parameter descriptions and sensitivity rankings, and preferably also includes explanations of parameter sources and confidence levels, model descriptions, and result interpretations. Users can submit feedback for revisions. The central scheduling agent analyzes and corrects the feedback, identifies the parameters or constraints that need to be corrected, and generates a correction set. Recalculation range identifier. Central scheduling agent relies on dependency graph. Only downstream task nodes affected by the correction will be re-executed:

[0040] That is, only rescheduling For the corresponding task steps, the existing results of the unaffected steps should remain unchanged to reduce redundant calculations.

[0041] Example: In this embodiment, greenhouse gas emissions from the Three Gorges Reservoir are taken as the research object. As the world's largest hydroelectric project, the Three Gorges Reservoir's greenhouse gas emissions research spans a long period, involves diverse data sources, and has complex data formats, making it suitable as a typical case for verifying the method of this invention. The overall implementation process can be found in [reference needed]. Figure 3 .

[0042] The implementation process of this invention is as follows: (1) User submits calculation request: The user submits a natural language calculation request Q through the input system: "Please calculate the CO2 and CH4 emissions of the Three Gorges Reservoir, using the G-Res model. Please use relevant tools to automatically complete the key parameters and output the uncertainty interval and sensitivity ranking." The user's known information includes at least the reservoir name. The system can automatically obtain the other parameters. If the parameters cannot be obtained, the agent will further question the user in the feedback process and ask the user to provide the parameters. (2) Central scheduling agent parses the request and generates an execution plan: The central scheduling agent parses the calculation request Q and generates a structured task description. ,in To "calculate CO2 and CH4 emissions", The reservoir name is "Three Gorges Reservoir". C contains "model priority = G-Res, parameters are automatically acquired, and uncertainty interval and sensitivity ranking are output". Then the central scheduling agent constructs a task dependency graph G=(V,E), where V contains the following task nodes: knowledge base retrieval, basic parameter collection, missing parameter completion, watershed environmental parameter collection, trophic status determination, emission accounting, uncertainty analysis, sensitivity analysis, and report summary; E represents the execution dependency relationship between each node. The central scheduling agent determines the execution order based on this and performs parallel scheduling for tasks without dependency relationships. (3) Basic data collection: The multi-source data collection agent starts the knowledge base retrieval and web page search tasks, using "Three Gorges Reservoir" as the keyword, to retrieve relevant accounting guidelines, technical reports, literature and web pages, etc., to obtain existing research data and accounting method reference information of Three Gorges Reservoir and obtain basic parameters of the reservoir. This includes geographic coordinates, dam site elevation, normal water level, total reservoir capacity, water surface area, year of construction, and intended use. All parameters are encapsulated as structured fields. The central dispatching agent performed integrity checks based on three parameter categories. Since the Three Gorges Reservoir is a world-class water conservancy project with ample relevant research literature, all three parameter categories could be obtained from publicly available data sources. Therefore, the reference reservoir matching agent was not triggered, and all parameters... , indicating that all data were obtained directly from the query. (4) Collection of watershed environmental parameters: The multi-source data collection agent further obtains meteorological sequence data such as temperature, precipitation, wind speed and surface radiation through the meteorological data interface based on the geographical coordinates of the Three Gorges Reservoir; obtains soil organic carbon content in the reservoir area and upstream watershed through the soil data interface; and obtains watershed boundaries, topographic slope and land use type distribution through the geographic information data interface. The above data are summarized to form a set of watershed environmental parameters. Each parameter also uses a structured field. Formal encapsulation. (5) The emission accounting agent determines the trophic status level of the Three Gorges Reservoir based on the collected water quality indicators and environmental parameters. TAs one of the reference parameters. (6) Emission accounting: The central dispatching agent will use the basic parameters as one of the reference parameters. With watershed environmental parameters The data is summarized and transmitted to the emissions accounting agent. The emissions accounting agent then processes the data according to the constraints. Select the G-Res model and perform emission calculations. (7) Uncertainty and Sensitivity Analysis: The central scheduling agent schedules the uncertainty analysis agent and the sensitivity analysis agent in parallel. The uncertainty analysis agent will calculate the input parameters. Each parameter is based on its The data is mapped to random variables, and 10,000 samples are taken using the Monte Carlo method to obtain the confidence intervals for the emission results. The sensitivity analysis agent uses a variance decomposition-based method to calculate the first-order sensitivity index of each parameter, according to... Sort the data from largest to smallest to identify the key parameters that have the most significant impact on the Three Gorges Reservoir emission accounting results (such as water surface area, water temperature, sediment organic carbon content, etc.), so that users can focus on them in subsequent data supplementation. (8) The central dispatching agent summarizes the output results of the above steps and generates a structured accounting report R, including: Three Gorges Reservoir CO2 emission, CH4 emission, CO2 equivalent, emission result confidence interval, ranking of key sensitive parameters, and explanation of the data source and confidence level of each parameter. After reviewing the report, users can submit correction feedback F, and the central dispatching agent will parse the feedback content and re-execute some tasks.

[0043] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A reservoir greenhouse gas accounting method based on multi-agent collaboration, characterized in that, The method specifically includes the following steps: S1. Steps for constructing a knowledge base in the reservoir domain: Collect unstructured document data related to the target reservoir, perform structured processing on the original documents through a document parsing engine to generate standardized corpus, and import the standardized corpus into a knowledge base constructed based on retrieval enhancement generation technology to complete document segmentation, vectorization and index construction, forming a knowledge base in the reservoir domain that can be queried by various intelligent agents through natural language. S2. Calculation Request Parsing Steps: The central scheduling agent receives the natural language calculation request Q submitted by the user and parses it into a structured task description. ,in For the accounting target, For the set of information known to the user, The set of constraints is defined; an execution plan is generated based on the structured task description, and a task dependency graph is constructed. ,in For a set of task nodes, Based on the dependencies between task nodes, tasks are assigned to each agent in the order of dependency. S3. Basic Parameter Acquisition Steps: Based on the target reservoir identifier and the constraints in the structured task description, the multi-source data acquisition agent obtains the set of basic reservoir parameters from external data sources. Each parameter is encapsulated as a structured field. ,in For parameter values, For data source identification, For confidence level, For the range of uncertainty, Mark the data source type; S4. Missing Parameter Completion Steps: The parameters required for calculation are classified into three categories according to their properties: reservoir inherent attribute parameters, watershed environmental parameters, and reservoir water body characteristic parameters; the reference reservoir matching agent is based on the feature vector of the target reservoir. Reference reservoir database Multidimensional similarity calculation of various reference reservoirs in the data, and retrieval of the most similar reference reservoir. When the similarity meets a preset threshold, The corresponding parameter value is used as the fill-in value for the missing field, and its source type is marked. Set the migration to complete the data; when the similarity is lower than the threshold, do not perform completion and provide feedback to the user; S5. Watershed Environmental Parameter Acquisition Steps: The multi-source data acquisition agent, based on the geographical coordinates of the target reservoir, acquires meteorological sequence data, soil attribute data, and spatial geographic data through meteorological data interfaces, soil data interfaces, and geographic information data interfaces respectively, forming a set of watershed environmental parameters. ; S6. Emissions Accounting Steps: The emissions accounting agent summarizes the basic parameter set and the watershed environmental parameter set into the accounting input parameter set. According to the constraints Determine the accounting model from the pre-set accounting model library. Perform greenhouse gas emission calculations; S7. Uncertainty Analysis Steps: The uncertainty analysis agent will calculate the input parameters. In this context, each parameter is determined according to its uncertainty range. The data is mapped to random variables, and the Monte Carlo method is used for multiple sampling and calculations to obtain the confidence interval of the emission results. S8. Sensitivity Analysis Steps: The sensitivity analysis agent calculates the sensitivity index of each input parameter to the emission results, sorts them from largest to smallest sensitivity, and outputs the ranking results of key parameters. S9. Report generation steps: The central dispatching agent summarizes the emission accounting results, uncertainty analysis results, and sensitivity analysis results to generate a structured accounting report.

2. The reservoir greenhouse gas accounting method based on multi-agent collaboration according to claim 1, characterized in that, In step S1, the unstructured document data includes accounting guidelines, operation and maintenance reports, and scientific and technological literature; the structuring process includes performing layout analysis on the documents, automatically identifying the structured elements in the documents, and converting them into standardized corpus.

3. The reservoir greenhouse gas accounting method based on multi-agent collaboration according to claim 2, characterized in that, In step S2, the task dependency graph is a directed acyclic graph. The central scheduling agent performs parallel scheduling of task nodes with no dependency relationship and monitors the execution status of each agent in real time.

4. The reservoir greenhouse gas accounting method based on multi-agent collaboration according to claim 3, characterized in that, In step S4, when the inherent attribute parameters of the reservoir are missing, feedback is sent to the user requesting their provision. The watershed environmental parameters are automatically acquired by the multi-source data acquisition agent through a spatial data interface. When the reservoir's water body characteristic parameters cannot be directly obtained, they are completed through migration by matching with a reference reservoir. The feature vector... This includes the target reservoir's geographical location, storage capacity, climate zone, age, and intended use.

5. The reservoir greenhouse gas accounting method based on multi-agent collaboration according to claim 4, characterized in that, In step S5, the multi-source data acquisition agent acquires watershed environmental parameters in the following priority order: spatial data interface first, geographic information processing second, and text retrieval as a fallback.

6. The reservoir greenhouse gas accounting method based on multi-agent collaboration according to claim 5, characterized in that, In step S6, the accounting model library includes the IPCC Tier 1 emission factor method and the G-Res parameterized model. The library also supports registration and access to other models. The IPCC Tier 1 emission factor method refers to the Tier 1 accounting method specified in the Intergovernmental Panel on Climate Change (IPCC) guidelines, which estimates greenhouse gas emissions from reservoirs based on default emission factors and reservoir area, climate zone, and reservoir age activity data. The G-Res parameterized model is a parameterized model for estimating greenhouse gas emissions from reservoirs, which estimates greenhouse gas emissions based on parameters such as reservoir morphology, hydrological operation, climate conditions, reservoir age, trophic status, and pre-inundation land use. and Emissions are estimated empirically or semi-empirically; the emission calculation includes calculating separately... Emissions and Emissions and convert to equivalent ,in , for The global warming potential coefficient.

7. The reservoir greenhouse gas accounting method based on multi-agent collaboration according to claim 6, characterized in that, In step S8, the sensitivity index is a first-order sensitivity index based on variance decomposition: ,in To provide input parameters Conditional expectations of emission results under given conditions.

8. The reservoir greenhouse gas accounting method based on multi-agent collaboration according to claim 7, characterized in that, The method also includes a feedback correction step: the user submits correction feedback F to the accounting report, and the central scheduling agent parses the correction feedback to generate a correction set. And based on the task dependency graph Identify the set of downstream task nodes affected: Only for The corresponding task steps are re-executed, keeping the existing results of the unaffected steps unchanged.

9. A reservoir greenhouse gas accounting system based on multi-agent collaboration, characterized in that, The system includes: The knowledge base construction module is used to collect unstructured document data related to the target reservoir, perform structured processing on it, and then import it into a knowledge base built based on retrieval enhancement generation technology to form a domain knowledge base that can be retrieved through natural language. The central scheduling agent is used to receive users' natural language accounting requests and parse them into structured task descriptions, construct task dependency graphs based on the structured task descriptions and assign tasks to each agent, monitor the execution status of each agent, and summarize the output results of each agent to generate a structured accounting report. A multi-source data acquisition intelligent agent is used to obtain a set of basic reservoir parameters and a set of watershed environmental parameters from external data sources, and encapsulate each parameter into a structured field containing parameter value, data source identifier, data confidence level, uncertainty range and data source type label; The reference reservoir matching agent is used to search for the most similar reference reservoir when the water body feature parameters of the reservoir are missing and cannot be directly obtained from public data sources. Based on the multidimensional similarity calculation between the feature vector of the target reservoir and each reference reservoir in the reference reservoir database, the agent fills in the missing field with the corresponding parameter value when the similarity meets the preset threshold and marks it as migration and completion data; otherwise, the agent will provide feedback to the user to request the data. An emissions accounting agent is used to determine the accounting model from a pre-set accounting model library based on constraints, and to perform greenhouse gas emissions calculations based on the set of accounting input parameters. An uncertainty analysis agent is used to map the input parameters of the calculation to random variables and perform multiple sampling and calculations using the Monte Carlo method to obtain the confidence interval of the emission results; A sensitivity analysis agent is used to calculate the sensitivity index of each input parameter to the emission results and output the ranking results of key parameters.

Images