Satellite-ground integrated intelligent operation and maintenance method and system based on Dify platform and AI Agent

Abstract

The invention discloses a satellite-ground integrated intelligent operation and maintenance method and system based on a Dify platform and an AI Agent. The method comprises the following steps that user questions, recognition entities, operation intentions and key parameters are analyzed through an intention recognition optimization and multi-round RAG recall module; calling a data mining tool to obtain related statistics and detail information, and generating a data processing overall situation summary table; calling a data integrity checking tool, automatically checking task nodes generated from the original data to the data product, and generating a checking report; according to the report, calling a visual chart module to draw a business flow chart marked with error nodes; calling an error node analysis and statistics module, analyzing error reasons in combination with historical fault information, and generating a statistical table; and integrating various information to generate a data processing state monitoring report and sending the report to a user. The method aims at solving the problems that traditional operation and maintenance of a remote sensing satellite ground system depend on manpower, fault response is slow, data analysis is weak, and expandability is poor.

Description

Technical Field

[0001] This invention relates to the technical field of intelligent operation and maintenance of remote sensing satellite ground systems, and in particular to an intelligent operation and maintenance method and system built on the open-source AI application development platform Dify and the AI ​​Agent technology framework. Background Technology

[0002] A remote sensing satellite system consists of a satellite platform system and a payload system. The satellite platform system is responsible for basic functions such as orbit and attitude control, power supply, temperature control, and data downlink. The payload system contains active and passive detection instruments such as optical, microwave, and space weather sensors, responsible for remote sensing observation of the observation area or target. The ground system consists of ground stations, an operations control center, and a data processing center. The ground stations are responsible for receiving data and transmitting telemetry and control commands, the operations control center coordinates mission scheduling and operation management, and the data processing center processes, generates, and distributes remote sensing products.

[0003] Operational remote sensing satellites and their ground systems require 24 / 7 uninterrupted operation, characterized by long operational chains, numerous data transfer links, and high timeliness requirements. Satellite operation in orbit is susceptible to anomalies caused by space weather and environmental factors. Ground station data reception may suffer quality degradation due to severe weather, tracking anomalies, channel link failures, etc. Operation control centers and data processing centers must handle various emergencies such as satellite operational anomalies, surges in traffic, and IT resource environment disruptions. The operational capabilities and efficiency of the space-ground system directly determine its overall system effectiveness.

[0004] Traditional operation and maintenance models rely on manual interpretation, experience-based judgment, and manual operation, which have the following prominent problems:

[0005] 1. High dependence on manual labor: Operation and maintenance personnel need to manually check the system status, analyze satellite telemetry data and ground system logs, which is costly and prone to judgment errors due to human mistakes;

[0006] 2. Slow fault response: Fault location relies on personal experience and the troubleshooting process is cumbersome, resulting in an excessively long average repair time for satellite-ground system anomalies, which affects business continuity.

[0007] 3. Weak data analysis capabilities: It is difficult to conduct real-time and in-depth correlation analysis of multi-dimensional indicators such as satellite operation status, data reception progress, processing quality, and timeliness achievement rate, and it is impossible to accurately identify potential risks;

[0008] 4. Poor scalability: Faced with continuously expanding business scale and exponentially increasing data volume, traditional models are difficult to adapt quickly, becoming the core bottleneck for improving system efficiency.

[0009] The information disclosed in this background section is intended only to enhance the understanding of the overall background of the invention and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0010] The purpose of this invention is to provide a space-ground integrated intelligent operation and maintenance method and system based on the Dify platform and AI Agent. Through natural language interaction, it provides full-process intelligent support for daily operation and maintenance scenarios, significantly improving operation and maintenance efficiency and decision-making accuracy.

[0011] To achieve the above objectives, this invention provides a space-ground integrated intelligent operation and maintenance method based on the Dify platform and AI Agent, comprising the following steps:

[0012] S1: By analyzing user questions through intent recognition optimization and multi-round RAG recall module, the entity, operation intent and key parameters in the question are identified;

[0013] S2: Call the data mining tools in the tool function set to obtain statistical and detailed information related to the key parameters of the question from the API data interface of the business scheduling control platform of the operation control center, and generate a summary table of the overall data processing situation;

[0014] S3: Call the data integrity check tool in the tool function set to automatically check the task nodes between the raw data and the data product generation, determine whether there are any errors, and generate a data integrity check report;

[0015] S4: Based on the data integrity verification report, call the visualization chart module to draw a business process diagram containing all business nodes, and mark the nodes with errors;

[0016] S5: For the marked error nodes, call the error node analysis and statistics module to extract their return codes, input and output files and running log information, analyze the cause of the error in combination with historical fault information, and generate an error node statistics table according to the frequency of error occurrence.

[0017] S6: Integrate the overall data processing summary table, data integrity verification report, business process diagram, error cause analysis results, and error node statistics table to generate a data processing status monitoring report and send it to the user.

[0018] In one embodiment of the present invention, in step S1, the intent recognition optimization and multi-round RAG recall module matches the corresponding inference mode according to the task complexity. The inference mode includes the Function Calling mode for tasks with fixed processes and the ReAct framework for complex tasks that require dynamic planning.

[0019] In one embodiment of the present invention, in step S2, the data mining tool accesses and integrates a remote sensing domain knowledge base that includes satellite parameter tables, API documents, and processing procedures. When processing queries, it first retrieves relevant paragraphs and parameters from the knowledge base and uses them as an enhanced context input model.

[0020] In one embodiment of the present invention, in step S2, the remote sensing satellite knowledge base is constructed by combining a relational database with a time-series database and a NoSQL database; the relational database is used to store various satellite configuration parameters, ground system parameter configurations, and operational fault experience information; the time-series database is used to store satellite operation data throughout its entire life cycle and ground system operation data; and the NoSQL database is used to store technical documents and manuals as well as satellite observation images.

[0021] In one embodiment of the present invention, the automated end-to-end verification in step S3 includes the following sub-steps:

[0022] S301: Compare the scanning frame plan under the theoretical normal observation state of each satellite instrument with the actual downlinked data frames, perform data frame consistency check, and determine whether there are missing frames, incorrect frames, discontinuities, out-of-order frames, or incorrect frame sequence numbers.

[0023] S302: Check the operation status of the raw data of each instrument generated in the business system. If the return code is non-zero, it is necessary to further determine whether the non-zero exception return code is a business-defined exception return code or an operating system-level exception return code. If it is a business-defined exception return code, trace the system log to find the reason for the abnormal exit of the data collection program. If it is an operating system-level exception return code, determine the reason for the exception based on the operating system's interrupt return code for the process, and determine whether an error has occurred in the data collection operation.

[0024] S303: Check the job completion status of each data preprocessing task node in each workflow and record information of error nodes;

[0025] S304: Check the completion status of each product generation task node in the workflow and record information of error nodes;

[0026] S305: Compare the number of data distributions that should have been completed with the number that were actually completed to determine if there are any errors in the data distribution;

[0027] S306: Based on the verification results of sub-steps S301 to S305, generate a remote sensing satellite data integrity verification report for nodes with erroneous annotations.

[0028] In one embodiment of the present invention, in step S4, the visualization chart module is based on the ECharts chart library and dynamically renders visualization charts by configuring titles, legends, X-axis, Y-axis, and data series parameters. It supports interactive components such as prompt boxes and toolboxes and can generate standardized tables from structured data.

[0029] In one embodiment of the present invention, step S5 includes the following specific steps:

[0030] S501: Based on the unique ID information of the faulty node, obtain its return code, input / output files, and runtime logs;

[0031] S502: Identify the macroscopic cause of the error by performing a mapping lookup through the program return code specification table;

[0032] S503: Perform in-depth analysis of input and output files and running logs, extract time features using a large language model by calling the Dify workflow, extract spatial features, call the remote sensing satellite domain knowledge base mentioned in step S2 of RAG to find the same abnormal historical fault information and fault cases, retrieve the matching fault causes, and locate the specific cause of the error.

[0033] In one embodiment of the present invention, the historical fault information is obtained through a database retrieval tool, which is used to connect to a database, execute SQL queries, and obtain historical data stored in the database.

[0034] In one embodiment of the present invention, the overall data processing summary table and the business process diagram are both drawn using a visualization chart module. This visualization chart module is based on the ECharts chart library and dynamically renders the visualization chart by configuring the title, legend, X-axis, Y-axis and data series.

[0035] This invention also provides a space-ground integrated intelligent operation and maintenance system based on the Dify platform and AI Agent, comprising:

[0036] The intent recognition optimization and multi-round RAG recall module is used to parse user questions and identify entities, operation intentions and key parameters in the questions;

[0037] The tool function set module is used to obtain statistical and detailed information related to the key parameters of the query from the API data interface of the business scheduling control platform of the operation control center, generate a summary table of the overall data processing situation, and automatically check the task nodes between the raw data and the data product generation to determine whether there are errors and generate a data integrity verification report.

[0038] The visualization chart module is used to draw a business process diagram containing all business nodes based on the data integrity verification report, and to mark the nodes with errors.

[0039] The error node analysis and statistics module is used to extract the return code, input and output files and running log information of the marked error nodes, analyze the cause of the error by combining historical fault information, and generate an error node statistics table according to the frequency of error occurrence.

[0040] The report generation module integrates the overall data processing summary table, data integrity verification report, business process diagram, error cause analysis results, and error node statistics table to generate a data processing status monitoring report and send it to the user.

[0041] Compared with existing technologies, the satellite-ground integrated intelligent operation and maintenance method and system based on the Dify platform and AI Agent according to the present invention, based on the low-code capabilities of the Dify platform and the layered inference architecture of the AI ​​Agent, is highly efficient and quick to develop and deploy, can be rapidly iterated, and has good scalability. Through natural language interaction and scenario-based toolsets, fault location is shortened from hours to minutes, significantly improving operation and maintenance efficiency and reducing manpower input. It integrates remote sensing knowledge base and RAG technology, combined with multi-dimensional visualization analysis, providing accurate and verifiable decision support. The modular encapsulation of tool functions allows for seamless integration with existing systems and flexible expansion to adapt to the needs of business expansion. It has a low barrier to entry, allowing non-professionals to quickly query business operation status, and comprehensively covers the entire satellite operation and maintenance process. Attached Figure Description

[0042] Figure 1 This is a flowchart of a space-ground integrated intelligent operation and maintenance method based on the Dify platform and AI Agent according to an embodiment of the present invention.

[0043] Figure 2 This is a schematic diagram illustrating the normal result obtained from the MERSI data processing flow of Fengyun-3H satellite according to the present invention.

[0044] Figure 3 This is a schematic diagram illustrating an abnormal result in the MERSI data processing flow of Fengyun-3H satellite according to the present invention. Detailed Implementation

[0045] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings, but it should be understood that the scope of protection of the present invention is not limited to the specific embodiments.

[0046] Unless otherwise expressly stated, throughout the specification and claims, the term "comprising" or its variations such as "including" or "comprises" shall be understood to include the stated elements or components without excluding other elements or other components.

[0047] like Figure 1 As shown, a preferred embodiment of the present invention provides a space-ground integrated intelligent operation and maintenance method based on the Dify platform and AI Agent, which is mainly applied to the Dify platform and has been deeply customized based on the Agent capabilities of the Dify platform.

[0048] This invention discloses an integrated space-ground intelligent operation and maintenance method based on the Dify platform and AI Agent, which achieves intelligent question-and-answer support in the operation control center through the following steps:

[0049] S1: Through intent recognition optimization and multi-round RAG recall modules, user questions are analyzed and key parameters of the question are identified.

[0050] The process of parsing user queries includes analyzing data processing status and diagnosing anomalies. Key parameters identified in the query include satellite name, parameter name, and time. This step is an open-ended task requiring complex analysis. It automatically matches a pre-defined data processing status query and analysis workflow template in the Dify platform and triggers its instance.

[0051] To address the varying complexity of remote sensing satellite ground system missions, the intent recognition optimization and multi-round RAG recall modules in this step employ a layered reasoning strategy tailored to different mission objectives, as detailed below:

[0052] For tasks with clear objectives and fixed processes (e.g., querying the current status of Fengyun-3G satellite), the Function Calling mode is preferred. In this mode, the core responsibility is to accurately identify the user's intent and parameters and call a pre-defined single utility function. This mode omits complex step-by-step reasoning processes and acts directly, resulting in extremely high execution efficiency.

[0053] For complex tasks that are highly open-ended and require dynamic planning (e.g., analyzing and diagnosing the causes of data processing anomalies in the Medium Resolution Spectroscopic Imager (MERSI), the ReAct (Reasoning + Acting) framework is activated. In this mode, a cyclical process of thinking-action-observation is carried out, dynamically generating thought chains and autonomously deciding on the next action until the problem is solved, thereby ensuring the success rate of complex tasks.

[0054] S2: Call the data mining tools in the tool function set to obtain statistical information related to the key parameters of the question from the operation control center, obtain detailed information of the statistical information, and generate a summary table of the overall data processing situation.

[0055] This data mining tool is used to obtain key information such as satellite telemetry, ground system reception, processing, faults, latency, and global mosaic from the API data interface provided by the microservice-based operation control center business scheduling and control platform, providing structured raw data support for the upper-layer question and answer system.

[0056] The statistical information includes the amount of data processed, success rate, and failure rate; the detailed information includes task ID, frame count, process ID, logical relationships between jobs, and process completion status.

[0057] In this step, a high-quality remote sensing satellite domain knowledge base was built using the data mining tools in the tool function set. This knowledge base integrates professional texts and technical materials such as satellite parameter tables, API documents, and processing procedures. By using the data mining tools, relevant paragraphs and parameters can be retrieved from the remote sensing domain knowledge base when processing queries. These are then used as an enhanced context input model to improve the accuracy of identifying professional terms and domain concepts such as satellite constellations, payload names, and business processes, thus ensuring the accuracy and timeliness of the responses.

[0058] The remote sensing satellite knowledge base is constructed by combining relational databases with time-series databases and NoSQL databases.

[0059] Relational databases are used to store various satellite configuration parameters, ground system parameter configurations, and operational fault experience information. Satellite configuration parameters include satellite designation, satellite payload observation resolution, spectral range, and observation coverage. Ground system configuration parameters include the configuration of a satellite's ground receiving station network and the data processing job configuration of a certain type of instrument (including job name, job ID, job scheduling and control parameters, and IT resource application information). Operational fault experience information includes anomalies, cause analysis, and handling results from all business processes, including satellite safety, mission planning, data processing, product distribution, and IT equipment support.

[0060] Time-series databases are used to store satellite operation data and ground system operation data throughout the entire life cycle of a satellite (such as a meteorological satellite, whose design life is generally 5-8 years). Satellite operation data includes time-series data such as telemetry information of satellite platforms (such as the working status of satellite power system, thermal control system, telemetry and control system, etc.) and payloads (such as payload scanning status, payload calibration source status, payload temperature control status, etc.). Ground system operation data includes antenna motion status, signal-to-noise ratio, raw received data quality, and data processing flow operation status when ground stations track satellites.

[0061] NoSQL databases are used to store unstructured information such as technical documents and manuals (such as summaries of experience and procedures for handling a certain type of fault or API interface call specifications for the system) and satellite observation images (such as a set of typical normal observation images and a set of abnormal observation images of a certain type of remote sensing satellite payload that has been annotated).

[0062] S3: Call the data integrity check tool in the tool function set to automatically check the task nodes between the raw data and the data product generation, determine whether there are any errors, and generate a data integrity check report.

[0063] The data integrity check tool is used to automatically check the integrity of raw data from Level 0 to Level 2 products in the remote sensing satellite ground system. This data integrity check tool is a process-oriented anomaly detection engine built on the business logic of remote sensing satellite data processing. It can scan the data processing workflow status track by track, automatically capture abnormal processes and record their unique IDs.

[0064] Specifically, this data integrity checking tool can traverse each node of the abnormal workflow according to the preset business logic order and perform full-chain verification, including: data frame consistency verification calculation (comparing theoretical frames with actual frames to check for missing frames, incorrect frames, discontinuities, out-of-order frames, and incorrect frame sequence numbers), L0 job running status check (return code, runtime, logs, distinguishing between business / OS anomalies, such as memory overflow), data preprocessing task completion verification, product generation task completion verification, data distribution verification, and finally generate a remote sensing satellite data integrity verification report.

[0065] Specifically, the automated data integrity check process, which involves calling the data integrity check tools in the tool function set, includes the following steps:

[0066] S301: By fully recording the scanning data frame information of each instrument transmitted from the operation control center, and combining it with the scanning frame plan of each instrument under the theoretical normal observation state of the satellite calculated by the instrument parameters (including working mode, scanning frequency, etc.), the consistency calculation of the data frames of each instrument is performed. The consistency between the frames to be generated and the frames actually transmitted and obtained by the operation control center is compared to determine whether there are any missing or incorrect frames. For example, if some of the data frames actually transmitted and collected by a certain instrument are discontinuous, out of order, or have incorrect frame numbers.

[0067] S302: Check the job running status of the raw instrument data L0 generated in the business system (the check includes L0 generation job return code, job running time, logs, etc.). If the return code is non-zero (in the Linux standard, return code 0 represents the process ending normally), it is necessary to further determine whether the non-zero exception return code is a business-defined exception return code or an operating system-level exception return code. If it is a business-defined exception return code, trace the system log to find the reason for the abnormal exit of the aggregation program. If it is an operating system-level exception return code, determine the reason for the exception based on the operating system's interrupt return code for the process (e.g., in the Linux operating system, return code 139 indicates abnormal exit of the process due to memory overflow, and 127 indicates forced exit by the operating system). Based on this, determine whether an error has occurred in the data aggregation job and complete the L0 data check.

[0068] S303: Data preprocessing (the process of locating and calibrating remote sensing data based on L0 data to generate L1 level data) task completion verification, checking the job completion status of each data preprocessing task node in each workflow, and recording information of error nodes;

[0069] S304: Verification of the completion status of the product generation (process of generating L2 level data based on L1 data inversion of various geophysical parameters) task. Check the completion status of each product generation task node in each workflow and record the information of the error nodes.

[0070] S305: Data distribution status verification, compare the number of data to be distributed with the number of data actually distributed, and determine whether there is an error in the data distribution.

[0071] S306: Based on the full-chain verification results of the above steps, generate a data integrity inspection report for remote sensing satellites, pointing out the nodes where errors (missing, abnormal or interrupted) occurred, providing a basis for data quality assessment and problem tracking.

[0072] S4: Based on the data integrity verification report, draw a business process diagram that includes all business nodes, and mark the nodes with errors in the process diagram to visually show the abnormal locations.

[0073] Figure 2 The diagram shows the normal data processing workflow of the Medium Resolution Spectroscopic Imager (MERSI) on the Fengyun-3H satellite (FY-3H). It includes nodes for multi-station raw data aggregation and stitching, quality inspection, positioning, calibration, and calibration result image generation. In the normal workflow, each node is marked in blue after completion, and the workflow ends normally.

[0074] Figure 3As shown, when an anomaly occurs at a certain job node, the flowchart will mark the faulty node in red and draw feedback information about the job anomaly, including the job name, ID, anomaly status and return code, and CPU and memory usage, so as to quickly locate the cause of the anomaly.

[0075] S5: For the marked error nodes, extract the return code, input / output files and log information of the error nodes, and analyze the cause of the node error; perform statistics and summary of errors in each stage, and generate an error node statistics table according to the frequency of error occurrence.

[0076] Specifically, step S5 includes the following steps:

[0077] S501: Based on the unique ID information of the faulty node, obtain its return code, input / output files, runtime logs, and other key information;

[0078] S502: Identify the macroscopic cause of the error by performing a mapping lookup through the program return code specification table;

[0079] S503: Performs in-depth analysis of input / output files and runtime logs. By calling the Dify workflow, it uses a large language model to extract time features (determine the time window of the fault) and extract spatial features (locate the possible fault links). It calls RAG to search the remote sensing satellite domain knowledge base mentioned in S2 to find the same or similar abnormal historical fault information and fault cases, and retrieves the matching fault causes. By leveraging the reasoning capabilities of the large language model, it can locate the specific cause of the error, which helps maintenance personnel quickly locate faults and identify high-frequency fault points.

[0080] Historical fault information is obtained through a database retrieval tool, which connects to the database, executes SQL queries, and retrieves historical data stored in the database, ensuring the security and efficiency of data access.

[0081] In this step, the large language model undergoes specific optimization (especially the fusion of RAG and domain knowledge, utilizing a remote sensing domain knowledge base) and in-depth analysis to diagnose the specific causes of errors. For example, if a return code indicates a memory overflow, the system will further analyze the logs to pinpoint the specific module or data volume causing the problem. Simultaneously, the system statistically summarizes all errors and generates an error node statistics table based on their frequency of occurrence.

[0082] S6: Integrate all information, generate a data processing status monitoring report, and send it to the user.

[0083] The data processing status monitoring report includes a summary table of the overall data processing status, a data integrity verification report, a business process diagram, and a table showing the causes of errors at analysis nodes and statistics of error nodes.

[0084] In this invention, both the overall data processing summary table and the business process diagram are drawn using a visualization chart module. This visualization chart module is based on the ECharts chart library and dynamically renders visual charts such as line charts, bar charts, and pie charts by configuring core parameters such as title, legend, X-axis, Y-axis, and data series. For example, in the timeliness assessment chart of each stage from the reception and transmission of raw data from the ground station to the generation of L0, L1, and L2 level data by satellite observation instruments, different colors are used to distinguish the timeliness of each stage for legend parameter configuration. The X-axis represents the satellite observation time series parameters, and the Y-axis represents the timeliness parameters of each stage. It also supports the generation of standardized tables from the acquired structured data. Charts can be embedded in Q&A responses and interactively displayed through interactive components such as tooltips and toolboxes to enhance users' intuitive understanding of the data.

[0085] According to another preferred embodiment of the present invention, a space-ground integrated intelligent operation and maintenance system based on the Dify platform and AI Agent is disclosed, the system comprising:

[0086] The intent recognition optimization and multi-round RAG recall module is used to parse user questions and identify entities, operation intentions and key parameters in the questions;

[0087] The tool function set module is used to obtain statistical and detailed information related to the key parameters of the query from the API data interface of the business scheduling control platform of the operation control center, generate a summary table of the overall data processing situation, and automatically check the task nodes between the raw data and the data product generation to determine whether there are errors and generate a data integrity verification report.

[0088] The visualization chart module is used to draw a business process diagram containing all business nodes based on the data integrity verification report, and to mark the nodes with errors.

[0089] The error node analysis and statistics module is used to extract the return code, input and output files and running log information of the marked error nodes, analyze the cause of the error by combining historical fault information, and generate an error node statistics table according to the frequency of error occurrence.

[0090] The report generation module integrates the overall data processing summary table, data integrity verification report, business process diagram, error cause analysis results, and error node statistics table to generate a data processing status monitoring report and send it to the user.

[0091] The foregoing description of specific exemplary embodiments of the invention is for illustrative and explanatory purposes. These descriptions are not intended to limit the invention to the precise forms disclosed, and it will be apparent that many changes and variations can be made in accordance with the foregoing teachings. The exemplary embodiments were chosen and described in order to explain the specific principles of the invention and its practical application, thereby enabling those skilled in the art to implement and utilize various different exemplary embodiments of the invention, as well as various different choices and variations. The scope of the invention is intended to be defined by the claims and their equivalents.

Claims

1. A space-ground integrated intelligent operation and maintenance method based on the Dify platform and AI Agent, characterized in that, Includes the following steps: S1: By analyzing user questions through intent recognition optimization and multi-round RAG recall module, the entity, operation intent and key parameters in the question are identified; S2: Call the data mining tools in the tool function set to obtain statistical and detailed information related to the key parameters of the question from the API data interface of the business scheduling control platform of the operation control center, and generate a summary table of the overall data processing situation; S3: Call the data integrity check tool in the tool function set to automatically check the task nodes between the raw data and the data product generation, determine whether there are any errors, and generate a data integrity check report; S4: Based on the data integrity verification report, call the visualization chart module to draw a business process diagram containing all business nodes, and mark the nodes with errors; S5: For the marked error nodes, call the error node analysis and statistics module to extract their return codes, input and output files and running log information, analyze the cause of the error in combination with historical fault information, and generate an error node statistics table according to the frequency of error occurrence. S6: Integrate the overall data processing summary table, data integrity verification report, business process diagram, error cause analysis results, and error node statistics table to generate a data processing status monitoring report and send it to the user.

2. The space-ground integrated intelligent operation and maintenance method based on the Dify platform and AI Agent as described in claim 1, characterized in that, In step S1, the intent recognition optimization and multi-round RAG recall module matches the corresponding inference mode according to the task complexity. The inference mode includes the Function Calling mode for tasks with fixed processes and the ReAct framework for complex tasks that require dynamic planning.

3. The integrated space-ground intelligent operation and maintenance method based on the Dify platform and AI Agent as described in claim 1, characterized in that, In step S2, the data mining tool accesses and integrates a remote sensing domain knowledge base that includes satellite parameter tables, API documents, and processing procedures. When processing queries, it first retrieves relevant paragraphs and parameters from the knowledge base and uses them as an enhanced context input model.

4. The integrated space-ground intelligent operation and maintenance method based on the Dify platform and AI Agent as described in claim 1, characterized in that, In step S2, the remote sensing satellite knowledge base is constructed by combining a relational database with a time-series database and a NoSQL database. The relational database is used to store various satellite configuration parameters, ground system parameter configurations, and operational fault experience information. The time-series database is used to store satellite operation data throughout its entire life cycle and ground system operation data. The NoSQL database is used to store technical documents and manuals as well as satellite observation images.

5. The integrated space-ground intelligent operation and maintenance method based on the Dify platform and AI Agent as described in claim 1, characterized in that, The automated end-to-end verification in step S3 includes the following sub-steps: S301: Compare the scanning frame plan under the theoretical normal observation state of each satellite instrument with the actual downlinked data frames, perform data frame consistency check, and determine whether there are missing frames, incorrect frames, discontinuities, out-of-order frames, or incorrect frame sequence numbers. S302: Check the operation status of the raw data of each instrument generated in the business system. If the return code is non-zero, it is necessary to further determine whether the non-zero exception return code is a business-defined exception return code or an operating system-level exception return code. If it is a business-defined exception return code, trace the system log to find the reason for the abnormal exit of the data collection program. If it is an operating system-level exception return code, determine the reason for the exception based on the operating system's interrupt return code for the process, and determine whether an error has occurred in the data collection operation. S303: Check the job completion status of each data preprocessing task node in each workflow and record information of error nodes; S304: Check the completion status of each product generation task node in the workflow and record information of error nodes; S305: Compare the number of data distributions that should have been completed with the number that were actually completed to determine if there are any errors in the data distribution; S306: Based on the verification results of sub-steps S301 to S305, generate a remote sensing satellite data integrity verification report for nodes with erroneous annotations.

6. The space-ground integrated intelligent operation and maintenance method based on the Dify platform and AI Agent as described in claim 1, characterized in that, In step S4, the visualization chart module is based on the ECharts chart library and dynamically renders visualization charts by configuring parameters such as title, legend, X-axis, Y-axis, and data series. It supports interactive components such as prompt boxes and toolboxes and can generate standardized tables from structured data.

7. The space-ground integrated intelligent operation and maintenance method based on the Dify platform and AI Agent as described in claim 1, characterized in that, Step S5 includes the following specific steps: S501: Based on the unique ID information of the faulty node, obtain its return code, input / output files, and runtime logs; S502: Identify the macroscopic cause of the error by performing a mapping lookup through the program return code specification table; S503: Perform in-depth analysis of input and output files and running logs, extract time features using a large language model by calling the Dify workflow, extract spatial features, call the remote sensing satellite domain knowledge base mentioned in step S2 of RAG to find the same abnormal historical fault information and fault cases, retrieve the matching fault causes, and locate the specific cause of the error.

8. The integrated space-ground intelligent operation and maintenance method based on the Dify platform and AI Agent as described in claim 7, characterized in that, The historical fault information is obtained through a database retrieval tool, which is used to connect to the database, execute SQL queries, and retrieve the historical data stored in the database.

9. The integrated space-ground intelligent operation and maintenance method based on the Dify platform and AI Agent as described in claim 1, characterized in that, The overall data processing summary table and business process diagram are both drawn using a visualization chart module. This visualization chart module is based on the ECharts chart library and dynamically renders visualization charts by configuring titles, legends, X-axis, Y-axis, and data series.

10. A space-ground integrated intelligent operation and maintenance system based on the Dify platform and AI Agent, characterized in that, include: The intent recognition optimization and multi-round RAG recall module is used to parse user questions and identify entities, operation intentions and key parameters in the questions; The tool function set module is used to obtain statistical and detailed information related to the key parameters of the query from the API data interface of the business scheduling control platform of the operation control center, generate a summary table of the overall data processing situation, and automatically check the task nodes between the raw data and the data product generation to determine whether there are errors and generate a data integrity verification report. The visualization chart module is used to draw a business process diagram containing all business nodes based on the data integrity verification report, and to mark the nodes with errors. The error node analysis and statistics module is used to extract the return code, input and output files and running log information of the marked error nodes, analyze the cause of the error by combining historical fault information, and generate an error node statistics table according to the frequency of error occurrence. The report generation module integrates the overall data processing summary table, data integrity verification report, business process diagram, error cause analysis results, and error node statistics table to generate a data processing status monitoring report and send it to the user.

Images