Satellite measurement and control ground station intelligent task automatic operation method based on large model

Abstract

The invention provides a satellite measurement and control ground station intelligent task automatic operation method based on a large model, relates to the technical field of satellite measurement and control and large language model LLM application, and solves the limitation problem of preset logic automatic measurement and control. According to the method, knowledge bases oriented to different subsystems are formed according to demand analysis information of measurement and control tasks; lLM is trained based on satellite measurement and control ground station domain knowledge, and a multi-role station control agent which takes the trained LLM as a basic framework and is used for automatic operation of measurement and control tasks is constructed; the intelligent agent is deployed in a satellite measurement and control ground station and calls multiple types of service equipment. When a measurement and control task is executed, the multi-role station control agent issues a measurement and control task requirement, a measurement and control task plan is generated, actions are executed according to knowledge bases corresponding to different subsystems, capturing, tracking, measurement and control and data receiving and recording of transit satellites are organized and completed, the execution condition of the measurement and control task is automatically analyzed, and a prompt and solution scheme is generated when the measurement and control task is abnormal.

Description

Technical Field

[0001] The present invention relates to the field of satellite measurement and control and large language model (LLM) application technology, and in particular to a large model-based satellite measurement and control ground station intelligent task automatic operation method. Background Art

[0002] In the field of aerospace measurement and control, with the continuous advancement of technology and the increase in task complexity, the requirements for the degree of automation of ground stations are also increasing. At present, the development trend of intelligent equipment is to achieve functions such as comprehensive perception of the environment, automatic control, trend prediction and optimized decision-making. These systems provide strong auxiliary decision support for equipment operation and maintenance by uniformly collecting, managing and deeply analyzing various types of data. However, in the actual operation of aerospace measurement and control ground stations, although the research on artificial intelligence AI and deep learning algorithms is quite mature and has a large amount of literature as a theoretical basis, the direct application of advanced AI technologies such as large language models LLM to actual scenarios, especially for specific application solutions for aerospace measurement and control ground stations, is still in the exploratory stage.

[0003] Traditional aerospace ground station control software design relies on pre-set fixed processes to perform satellite control tasks. This automation method based on preset logic can meet the needs of routine tasks, but it often seems powerless when faced with unexpected situations such as abnormal satellite status or equipment failure. Once an abnormal situation occurs, it is difficult for the existing automation system to recover to normal working state on its own, and usually requires experienced on-duty personnel to intervene and deal with it. This situation not only increases the cost of human intervention, but may also affect the success rate of the mission due to slow response speed.

[0004] In addition, differences in geographical location, functional requirements, and equipment configurations at different measurement and control sites pose challenges to the adaptability of automation processes. This means that in order to ensure that each site can operate efficiently, a customized automation solution must be tailored for it, which undoubtedly increases the complexity and cost of development and maintenance work. At the same time, such a customized process also increases the possibility of errors, further increasing the difficulty of management.

[0005] In summary, how to use the latest artificial intelligence technology and large language models to improve the intelligence level of aerospace measurement and control ground stations, especially to enhance their adaptive capabilities in abnormal situations, has become an urgent problem to be solved, but this problem can be solved by designing a task automatic operation mode combined with a large language model from scratch. Therefore, by introducing more flexible and intelligent technical means, the need for manual intervention can be effectively reduced, the reliability and efficiency of the system can be improved, and the challenges brought by future space missions can be better met. Summary of the invention

[0006] Based on the current status of the background technology, the purpose of the present invention is to provide artificial intelligence technology and large language model support for the automatic operation of satellite measurement and control tasks, and therefore proposes a method for automatic operation of intelligent tasks of satellite measurement and control ground stations based on large models. The present invention abandons the traditional software mechanized process design method, and instead designs intelligent bodies of different positions based on the site management method, and the intelligent bodies collaborate to complete the automatic operation control of various measurement and control tasks; the present invention therefore solves the limitation problem in the automatic measurement and control of traditional preset logic, is applicable to various business functions in satellite measurement and control ground stations, and successfully implements artificial intelligence technology and large language models in the field of aerospace measurement and control.

[0007] The present invention adopts the following technical solutions to achieve the purpose:

[0008] A method for automatically running intelligent tasks of a satellite measurement and control ground station based on a large model comprises: forming a plurality of knowledge bases for different subsystems according to measurement and control task demand analysis information; training a large language model LLM based on the domain knowledge of the satellite measurement and control ground station, and constructing a multi-role station control intelligent agent which takes the trained large language model LLM as the basic framework and is used for automatically running measurement and control tasks; deploying the multi-role station control intelligent agent in the satellite measurement and control ground station, and calling and controlling multiple types of business equipment in the satellite measurement and control ground station by the multi-role station control intelligent agent; when the satellite measurement and control ground station performs measurement and control tasks, the multi-role station control intelligent agent publishes measurement and control task requirements, generates measurement and control task plans, executes actions according to the knowledge bases corresponding to different subsystems, organizes and completes the capture, tracking, measurement and control, data reception and recording of transit satellites, automatically analyzes the measurement and control task execution status, and generates prompts and solutions in the event of abnormalities.

[0009] Furthermore, the demand analysis information of the measurement and control task is determined by a technician with professional knowledge of measurement and control, and the demand analysis information includes a process file recorded in natural language; LLM is trained based on the process file, and after the training is completed, when executing the measurement and control task, it uses the logical description in the process file as a basis to obtain the status of multiple types of business equipment under the subsystem and complete the corresponding call control action.

[0010] Preferably, the contents of the process file include an overall process and a subsystem process. The overall process is used to maintain the global control content of the measurement and control task unchanged, and the subsystem process is used to respond to different satellite measurement and control ground stations, or the independent control content of different business equipment in a satellite measurement and control ground station; both the overall process and the subsystem process include control commands for business equipment; the lower limit of the level of the control command is any specific control action or operation of any specific business equipment.

[0011] Furthermore, the multi-role station control agent includes a stationmaster agent, a task scheduling agent, a satellite management agent and a subsystem agent; the stationmaster agent receives the measurement and control task requirements proposed by the user and then publishes them, continuously responding to new tasks or maintenance requirements; the task scheduling agent responds to the newly published measurement and control task requirements, and schedules the measurement and control tasks by establishing measurement and control task plans, calculating orbits and allocating resources, while observing whether there are actions that meet the time requirements in the existing measurement and control task plans and executing them; the satellite management agent receives the scheduled measurement and control task requirements, and provides the satellite information and task information corresponding to the measurement and control task; the subsystem agent, based on the received measurement and control task requirements, satellite information and task information, performs business control on the business equipment of the satellite measurement and control ground station and completes specific business actions.

[0012] Specifically, before executing the measurement and control task, the corresponding task objectives and prompt word templates are pre-configured for each intelligent agent in the multi-role station control intelligent agent based on the knowledge base of different subsystems, and the role attributes of each intelligent agent are correspondingly associated; each intelligent agent solves the specific problems corresponding to its own role when executing the measurement and control task based on the configured task objectives and prompt word templates.

[0013] Preferably, each intelligent agent in the multi-role station control intelligent agent is configured with an independent running process, and when executing the measurement and control task, each intelligent agent listens separately to see whether there is any input corresponding to its own role. If such input appears, it is processed on the basis of LLM according to the configured task goal to obtain the corresponding result of the input, and the result is returned to form a new input, which is listened and processed by other intelligent agents.

[0014] In summary, due to the adoption of this technical solution, the beneficial effects of the present invention are as follows:

[0015] The method of the present invention proposes an operation framework based on a large model and an intelligent agent. Through training and learning, the intelligent agent can extract patterns from the data and make decisions and predictions autonomously. It only needs to provide it with tools with ultimate control capabilities, such as various types of equipment in the site, to automatically run the measurement and control tasks. The organization and use of equipment tools and the completion of business scenario requirements are all determined by the collaboration of the intelligent agent. Developers only need to conduct business demand analysis and configure the corresponding professional knowledge base for the intelligent agent. Compared with the traditional automated operation mode, although the application scenarios are the same, the application logic of the large model LLM is different from the fixed preset programming code, and the subsequent expansion capabilities brought about will also be significantly different. With the continuous development of artificial intelligence technology and computing power hardware, the aerospace measurement, operation and control solutions brought about by artificial intelligence interaction and control under the guidance of the method of the present invention will have more and more obvious advantages.

[0016] The method of the present invention allows end users or system designers to directly participate in the implementation and modification of system task processes without relying too much on coding skills and programmers; and operators can control the system by directly inputting natural language. In the research and development stage, developers no longer need to encode a specific task scenario or process, but only need to let the corresponding intelligent agent understand the knowledge or steps of the process and provide equipment monitoring tools, and the intelligent agent can automatically run and complete the work.

[0017] Therefore, the method of the present invention improves the flexibility and variability of satellite measurement and control ground stations, lowers technical barriers, reduces dependence on operator proficiency, and reduces the risk of operational errors, providing new possibilities for optimizing and improving the mission processes of aerospace measurement, operation and control ground stations.

[0018] The method of the present invention can also unify the automatic operation implementation methods of various measurement and control ground stations. Regardless of whether their mission requirements or specific configurations are the same, the intelligent body itself has professional aerospace measurement and control knowledge and can handle various mission requirements. At the same time, it can obtain the design characteristics, equipment configuration and operating experience of each measurement and control ground station to achieve differentiated execution processes, thereby solving the limitations of different manufacturers and inconsistent operating processes, and helping to accumulate operating experience, summarize lessons, and optimize the task management and control capabilities of the measurement and control ground station.

[0019] In summary, the method of the present invention has the advantages of strong adaptability, good scalability, simple deployment, short development cycle and low-code development. Its application will increase customer experience, reduce operation and maintenance costs, and change the current software development model in the field of aerospace measurement, operation and control, so that the corresponding subsystem or overall designer can directly modify the operation logic, process and external interface of the site software by modifying the document, without the need for them to have professional programming language and program deployment knowledge. This method helps to speed up the development process and reduce costs. System designers or customers can directly participate in the design and implementation of the software without affecting the flexibility and stability of the system. AI technology can thus be smoothly implemented in the field of aerospace measurement and control, and provide technical reserves and core competitiveness for subsequent related technical fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A schematic diagram of the architecture of a multi-role station control intelligent agent constructed by the method of the present invention;

[0021] Figure 2 A schematic diagram of the dynamic process editing process in the method of the present invention;

[0022] Figure 3 It is a schematic diagram of the process of dynamically expanding the equipment in the method of the present invention;

[0023] Figure 4 is a schematic diagram of the health management process in the method of the present invention;

[0024] Figure 5 It is a flow chart of the service equipment monitoring function in the method of the present invention;

[0025] Figure 6 A schematic diagram of a planned receiving automatic operation process in the method of the present invention;

[0026] Figure 7 A schematic diagram of an automatic operation process for preparing a task in the method of the present invention;

[0027] Figure 8 A schematic diagram of the automatic running process of starting a task in the method of the present invention;

[0028] Fig. 9 A schematic diagram of the automatic running process of task completion in the method of the present invention;

[0029] Fig.10 A schematic diagram of the process of tracking and starting automatic operation in the method of the present invention;

[0030] Fig.11 It is a schematic diagram of tracking and ending the automatic operation process in the method of the present invention. DETAILED DESCRIPTION

[0031] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0032] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] Example

[0034] A method for automatically running intelligent tasks of a satellite measurement and control ground station based on a large model, the method comprising: forming a plurality of knowledge bases for different subsystems according to measurement and control task demand analysis information; training a large language model LLM based on satellite measurement and control ground station domain knowledge, and constructing a multi-role station control intelligent agent with the trained large language model LLM as a basic architecture and used for automatically running measurement and control tasks; deploying the multi-role station control intelligent agent in the satellite measurement and control ground station, and calling and controlling multiple types of business equipment in the satellite measurement and control ground station by the multi-role station control intelligent agent; when the satellite measurement and control ground station performs measurement and control tasks, the multi-role station control intelligent agent publishes measurement and control task requirements, generates measurement and control task plans, executes actions according to knowledge bases corresponding to different subsystems, organizes and completes the capture, tracking, measurement and control, data reception and recording of transit satellites, automatically analyzes the measurement and control task execution status, and generates prompts and solutions in the event of an abnormality.

[0035] Traditional measurement and control task process control requires process coding, configuration or graphical dragging for specific task scenarios. Due to the variety of requirements and different equipment, the task processes of various satellite measurement and control ground stations cannot be unified, resulting in repeated development work; in addition, users cannot modify the process or customize new processes for control. Therefore, this embodiment uses the large language model LLM for relevant design processing, and focuses on preventing the illusion of LLM caused by unclear processes during design, thereby avoiding the problem of unstable process control.

[0036] In this embodiment, the demand analysis information of the measurement and control task is determined by a technician with professional knowledge of measurement and control, and the demand analysis information includes a process file recorded in natural language; these process files should be written as logically clear as possible so that LLM can process them using its natural language processing capabilities. LLM is trained based on the process file, and after the training is completed, when executing the measurement and control task, it obtains the status of multiple types of business equipment under the subsystem based on the logical description in the process file and completes the corresponding call control action.

[0037] Here, as the preferred method of this embodiment, the content of the process file includes the overall process and the subsystem process. The overall process is used to maintain the global control content of the measurement and control task unchanged, and the subsystem process is used to deal with the independent control content of each different satellite measurement and control ground station, or each different business equipment in a satellite measurement and control ground station. In both the overall process and the subsystem process, control commands for business equipment are included; the lower limit of the level of the control command is any specific control action or operation of any specific business equipment, that is, clearly explain what kind of control action or operation a business equipment should perform at a certain moment in the process, to ensure that LLM will not have hallucinations because it cannot find the tool that accurately corresponds to the equipment.

[0038] The following is a detailed introduction to the multi-role station control agent constructed and used in this embodiment. In terms of business capabilities, the multi-role station control agent is responsible for monitoring and controlling the entire satellite measurement and control ground station, which is equivalent to using the large language model LLM to complete the functions of the station control software of the traditional measurement and control ground station system.

[0039] like Figure 1 As shown in the figure, the multi-role station control agent includes the stationmaster agent, task scheduling agent, satellite management agent and subsystem agent. The stationmaster agent receives the measurement and control task requirements proposed by the user and then publishes them, continuously responding to new tasks or maintenance requirements; the task scheduling agent responds to the newly released measurement and control task requirements, and schedules the measurement and control tasks by establishing measurement and control task plans, calculating orbits and allocating resources, and at the same time observes whether there are actions that meet the time requirements in the existing measurement and control task plans and executes them; the satellite management agent receives the measurement and control task requirements after scheduling, and provides the satellite information and task information corresponding to the measurement and control task; the subsystem agent performs business control on the business equipment of the satellite measurement and control ground station on the basis of receiving the measurement and control task requirements, satellite information and task information, and completes specific business actions.

[0040] Figure 1 In the text, LLM stands for Large Language Model and Agent stands for Agent. Large Language Model is an artificial intelligence system that can process and generate natural language. It understands human language by learning a large amount of text data and can generate coherent and meaningful answers or content based on the context. Agent refers to a software entity with a certain degree of autonomy. It can perceive its state, perform tasks, make decisions and take actions to achieve goals in a specific environment. Agents usually have the ability to learn and can optimize their behavior strategies over time. In addition to the specific agent roles mentioned above in this embodiment, Figure 1 The virtual company software agent and health management agent also shown will be described in detail later.

[0041] From the perspective of software architecture, the multi-role station control agent has multiple specific agents, so it contains corresponding roles, actions, knowledge bases and other elements. Its basic architecture is the common LLM and is unified and integrated through LLM.

[0042] Each role in the multi-role station control agent is an agent, which is configured with its own knowledge base and actions before construction or when used, and publishes its own task requirements, which are accepted and completed by other corresponding agents. Therefore, the multi-role station control agent is equivalent to a control team of a measurement and control station, each specific agent role is equivalent to the personnel configuration of each position, the knowledge base is equivalent to the professional knowledge and job description of the personnel in each position, and the actions are equivalent to the skills and executable operations of the personnel in each position.

[0043] In this embodiment, before executing the measurement and control task, the corresponding task objectives and prompt word templates are configured for each agent in the multi-role station control agent in advance according to the knowledge base of different subsystems, and the role attributes of each agent are correspondingly associated; each agent solves the specific problems corresponding to its own role when executing the measurement and control task according to the configured task objectives and prompt word templates. The following Table 1 lists and explains the specific roles in the multi-role station control agent, which can be referred to simultaneously, and its content can be referred to and the agent can be configured accordingly in actual application.

[0044] Table 1 Schematic diagram of role allocation of multi-role station control agents

[0045]

[0046]

[0047] Through the above method, the initial task requirements are gradually realized by defining multiple roles. When users use multi-role station control agents, the general entry point is the requirements of the stationmaster agent, and the subsequent roles will perform relevant processing according to the decomposed process steps based on this original task requirement.

[0048] In addition, as a preferred method of this embodiment, each agent in the multi-role station control agent is configured with an independent running process, and when executing the measurement and control task, each agent listens to whether there is an input corresponding to its own role. If the input appears, it processes the corresponding result of the input based on the configured task goal and LLM, and returns the result to form a new input, which is listened to and processed by other agents. In this way, once other roles in the system listen to the input related to their own tasks, they can obtain the tasks and execute them; this method is similar to the personnel management and organizational structure of the current measurement and control station, and each position responds to its corresponding new task requirements in real time.

[0049] In this embodiment, before and during the execution of the measurement and control task, additional role help components are pre-configured for the multi-role station control intelligent agent to assist the multi-role station control intelligent agent in executing the measurement and control task; these role help components may include typical search role components and prompt decomposition roles, etc. Prompt decomposition refers to breaking down a complex prompt into smaller, easier to manage and process components, so as to better understand and optimize the role of each part. This method can help improve the quality and relevance of the content generated by the large language model.

[0050] In this embodiment, before and during the execution of the measurement and control task, additional role help components are pre-configured for the multi-role station control intelligent agent to assist the multi-role station control intelligent agent in executing the measurement and control task; for the multiple agent roles of the multi-role station control intelligent agent and the multiple specific component roles in the role help component, each has a py file corresponding to its own skills; the py file has pre-defined templates for the corresponding skills, and each role processes the input and output of the corresponding measurement and control task part through the skills it possesses, and completes the measurement and control task of its own part.

[0051] Example 2

[0052] Based on Example 1, this example provides a relatively detailed introduction to some specific agent roles in the multi-role station control agent, as well as the corresponding processes during the implementation of the method.

[0053] First, in the multi-role station control agent, the dynamic process editing process of the subsystem agent is divided into the development stage and the operation stage, which can be viewed simultaneously. Figure 2 Schematic diagram of this process. This process is used to configure and edit the action execution sequence of multiple types of business equipment in the satellite measurement and control ground station; first, the knowledge base of the new process is configured in the development stage. The customer runs the RAG terminal to input the process documents related to the demand analysis information. After the process documents are vectorized using the embedded model, they are stored as the knowledge base of the corresponding subsystem; then the sequence design of the new process is completed in the operation stage. The stationmaster agent combines with the task scheduling agent to input the measurement and control task requirements, read the corresponding knowledge base from the embedded model, and return the text related to the measurement and control task requirements. The task scheduling agent integrates the process requirements, forms the corresponding prompt words and hands them over to LLM for processing; LLM generates the process and forms a stack of actions to be executed, determines the corresponding tool for the next business equipment to be called, and hands it over to the task scheduling agent and the subsystem agent to execute the relevant actions until the end of the operation stage.

[0054] In this embodiment, the multi-role station control agent also includes a virtual software company agent, which can be simultaneously referenced Figure 3 The process diagram shows that the virtual software company agent is used to dynamically expand the new business equipment in the satellite measurement and control ground station and develop the corresponding business functions, so as to realize the equipment monitoring interface editing function. The virtual software company agent generates prompts corresponding to the writing code and the test code according to the measurement and control task requirements corresponding to the new business equipment and functions, and hands them over to the LLM to generate the monitoring interface code and the test code; after the virtual software company agent tests and confirms, the corresponding code is added to the tool list that can be called for control, and the tool list corresponding to the business equipment of the subsystem agent is updated for use when executing the measurement and control task.

[0055] In this embodiment, the multi-role station control agent also includes a health management agent, which can be viewed simultaneously. Figure 4The process diagram shows that the health management agent is used to perform health management detection on each agent in the multi-role station control agent and multiple types of business equipment in the satellite measurement and control ground station. Based on the health detection instructions from the station master agent, the health management agent performs fault detection operations according to the preset cycle, obtains health management knowledge from the preset health management knowledge base, generates prompt words corresponding to health management, and hands them over to LLM to generate a health management process; the task scheduling agent and subsystem agent obtain the corresponding task plan information and detection point status based on the health management process, automatically perform fault detection and obtain the detection results, and the health management agent then generates a health management report to end the health management process.

[0056] The following is an introduction to the various satellite measurement and control task functions that can be achieved after the specific agent roles in the multi-role station control agent are combined in the method of this embodiment.

[0057] The monitoring function flow of various types of business equipment in the satellite measurement and control ground station can be found in Figure 5 As shown in the figure, the webmaster agent serves as the recipient of the user's input device control requirements. On this basis, it adds context and generates corresponding prompt words. LLM decomposes tasks, extracts business equipment names and related commands, calls the tools of the corresponding subsystem agents, implements the call of business equipment interfaces, and controls them to complete corresponding measurement and control operations.

[0058] For receiving and automatically running the measurement and control plan, see Figure 6 As shown in the figure, after a new plan is created through the stationmaster agent terminal, LLM will organize the plan, and after the satellite management agent queries the satellite information and default parameters, LLM will continue to generate the corresponding plan. The task scheduling agent adds the new plan generated by LLM to the list and checks the time nodes. While executing the plan, it continuously traverses the plan and completes the operation execution at each time node.

[0059] For the preparation process of the measurement and control task, the automatic operation process can be found in Figure 7 As shown in the figure, the task scheduling agent generates the preparation steps based on the LLM and the knowledge base corresponding to the task preparation; after obtaining the measurement and control task information, the satellite management agent queries the measurement and control task information and macro parameters, and the LLM takes over and generates macro control steps, forms relevant commands and matches the corresponding macro control tools of each subsystem agent; the subsystem agents thus complete their respective business functions and prepare to implement relevant controls to execute tasks.

[0060] The process of starting and ending the measurement and control task is similar, and can be viewed simultaneously. Figure 8 and Fig. 9As shown in the figure, both are started by the task scheduling agent. LLM generates relevant steps based on the knowledge base of the corresponding process of starting or ending the measurement and control task. After the task starts, the subsystem agent processes each business device based on the process steps; after the task ends, the task scheduling agent records the work report of this task and clears the relevant plan.

[0061] During the measurement and control mission, taking satellite tracking as an example, Fig.10 and Fig.11 The processes of starting and ending tracking are shown respectively. After the task scheduling agent determines that tracking has started, LLM generates the process steps for starting tracking based on the corresponding knowledge base, and the subsystem agent obtains the relevant characteristic parameters at this time; after LLM organizes these characteristic parameters, the subsystem agent then sets and starts the corresponding business equipment to realize functions related to tracking business such as information scanning and reporting. When the tracking is completed, LLM generates the process steps for ending tracking based on the corresponding knowledge base, and the subsystem agent controls various business equipment to stop the data processing and sending process, and the information is saved. Various business equipment enters standby mode and waits for the next measurement and control task.

Claims

1. A method for automatically running intelligent tasks of a satellite measurement and control ground station based on a large model, characterized in that: include: Based on the demand analysis information of the measurement and control tasks, multiple knowledge bases for different subsystems are formed; Based on the domain knowledge of satellite measurement and control ground stations, the large language model LLM is trained to build a multi-role station control agent that uses the trained large language model LLM as the basic architecture and is used for automatic operation of measurement and control tasks. The multi-role station control intelligent agent is deployed in the satellite tracking and control ground station, and the multi-role station control intelligent agent calls and controls the various subsystem equipment in the satellite tracking and control ground station; when the satellite tracking and control ground station performs tracking and control tasks, the multi-role station control intelligent agent publishes tracking and control task requirements, generates tracking and control task plans, executes actions based on the knowledge base corresponding to different subsystems, organizes and completes the capture, tracking, tracking and control, data reception and recording of transit satellites, automatically analyzes the execution of tracking and control tasks, and generates prompts and solutions in case of abnormalities.

2. The method for automatically running intelligent tasks of a satellite measurement and control ground station according to claim 1, characterized in that: The demand analysis information of the measurement and control task is determined by technical personnel with professional knowledge of measurement and control. The demand analysis information includes process files recorded in natural language. LLM is trained based on the process files, and after the training is completed, when executing the measurement and control task, it uses the logical description in the process file as the basis to obtain the status of each subsystem device under the subsystem and complete the corresponding call control action.

3. The method for automatically running intelligent tasks of a satellite measurement and control ground station according to claim 2, characterized in that: The contents of the process file include the overall process and subsystem process. The overall process is used to maintain the global control content of the measurement and control task unchanged, and the subsystem process is used to respond to different satellite measurement and control ground stations, or the independent control content of different business equipment in a satellite measurement and control ground station; both the overall process and the subsystem process include control commands for each subsystem equipment; the lower limit of the level of the control command is any specific control action or operation of any specific business equipment.

4. The method for automatically running intelligent tasks of a satellite measurement and control ground station according to claim 1, characterized in that: The multi-role station control agent includes the stationmaster agent, the task scheduling agent, the satellite management agent and the subsystem agent; the stationmaster agent receives the measurement and control task requirements proposed by the user and then publishes them, continuously responding to new tasks or maintenance requirements; the task scheduling agent responds to the newly published measurement and control task requirements, and schedules the measurement and control tasks by establishing measurement and control task plans, calculating orbits and allocating resources, while observing whether there are actions that meet the time requirements in the existing measurement and control task plans and executing them; the satellite management agent receives the measurement and control task requirements after scheduling, and provides the satellite information and task information corresponding to the measurement and control task; the subsystem agent, based on the received measurement and control task requirements, satellite information and task information, performs business control on the business equipment of the satellite measurement and control ground station and completes specific business actions.

5. The method for automatically running intelligent tasks of a satellite measurement and control ground station according to claim 4, characterized in that: Before executing the measurement and control task, the corresponding task objectives and prompt word templates are pre-configured for each intelligent agent in the multi-role station control intelligent agent based on the knowledge base of different subsystems, and the role attributes of each intelligent agent are correspondingly associated; each intelligent agent solves the specific problems corresponding to its own role when executing the measurement and control task based on the configured task objectives and prompt word templates.

6. The method for automatically running intelligent tasks of a satellite measurement and control ground station according to claim 5, characterized in that: Each agent in the multi-role station control agent is configured with an independent running process. When executing the measurement and control task, each agent listens for the input corresponding to its own role. If the input appears, it processes the corresponding result of the input based on the configured task goal and LLM, and returns the result to form a new input, which is listened and processed by other agents.

7. The method for automatically running intelligent tasks of a satellite measurement and control ground station according to claim 4, characterized in that: In the multi-role station control agent, the dynamic process editing process of the subsystem agent is divided into the development stage and the operation stage. This process is used to configure and edit the action execution sequence of various types of business equipment in the satellite measurement and control ground station. First, the knowledge base of the new process is configured in the development stage. The customer runs the RAG terminal to input the process documents related to the demand analysis information. After the process documents are vectorized using the embedded model, they are stored as the knowledge base of the corresponding subsystem. Then, the sequence design of the new process is completed in the operation stage. The stationmaster agent combines with the task scheduling agent to input the measurement and control task requirements, read the corresponding knowledge base from the embedded model, and return the text related to the measurement and control task requirements. The task scheduling agent integrates the process requirements, forms the corresponding prompt words and hands them over to LLM for processing. LLM generates the process and forms a stack of actions to be executed, determines the corresponding tool for the next business equipment to be called, and hands it over to the task scheduling agent and the subsystem agent to execute the relevant actions until the end of the operation stage.

8. The method for automatically running intelligent tasks of a satellite measurement and control ground station according to claim 4, characterized in that: Before and during the execution of the measurement and control task, additional role help components are pre-configured for the multi-role station control agent to assist the multi-role station control agent in executing the measurement and control task; for the multiple agent roles of the multi-role station control agent and the multiple specific component roles in the role help component, each has a py file corresponding to its own skills; the py file has pre-defined templates for the corresponding skills, and each role processes the input and output of the corresponding measurement and control task part through its own skills to complete the measurement and control task of its own part.

9. The method for automatically running intelligent tasks of a satellite measurement and control ground station according to claim 4, characterized in that: The multi-role station control agent also includes a virtual software company agent, which is used to dynamically expand new business equipment in the satellite measurement and control ground station and develop corresponding business functions; the virtual software company agent generates prompts for writing codes and testing codes based on the measurement and control task requirements corresponding to the new business equipment and functions, and hands them over to LLM to generate monitoring interface code and test code; after testing and confirmation by the virtual software company agent, the corresponding code is added to the list of tools that can be called for control, and the subsystem agent's tool list corresponding to the business equipment is updated for use when executing measurement and control tasks.

10. The method for automatically running intelligent tasks of a satellite measurement and control ground station according to claim 4, characterized in that: The multi-role station control agent also includes a health management agent, which is used to perform health management detection of each agent in the multi-role station control agent and multiple types of business equipment in the satellite measurement and control ground station; Based on the health detection instructions from the station manager agent, the health management agent performs fault detection operations according to the preset cycle, obtains health management knowledge from the preset health management knowledge base, generates prompt words corresponding to health management, and hands them over to LLM to generate health management processes; The task scheduling agent and subsystem agent obtain the corresponding task plan information and detection point status based on the health management process, automatically perform fault detection and obtain the detection results. The health management agent then generates a health management report and ends the health management process.

Images