CN-121984585-A - Multi-agent automatic operation method, equipment and medium of optical transmission system

CN121984585ACN 121984585 ACN121984585 ACN 121984585ACN-121984585-A

Abstract

The invention provides a multi-agent automatic operation method, equipment and medium of an optical transmission system, the method realizes the full-chain autonomous operation of the optical transmission system by constructing a framework for the cooperative work of four intelligent agents, namely execution, monitoring, fault detection and repair. The method comprises the steps of executing an agent to analyze a user natural language instruction and calling a tool function to complete experimental configuration, monitoring the running state of the agent real-time analysis system, enabling a fault detection agent to adopt a multi-mode large model finely tuned for an optical transmission scene, fusing analysis system images and data to achieve fault accurate positioning and severity assessment, and enabling a repair agent to automatically call the tool to execute parameter adjustment or hardware tuning according to a diagnosis result. The invention solves the problems of high cost, response lag, diagnosis dependence on experience and incapability of closed-loop control of the traditional artificial operation and maintenance learning, and realizes highly-automatic and intelligent operation and maintenance from system configuration, state monitoring, intelligent diagnosis to automatic repair.

Inventors

YI LILIN
Jiang Nianheng
Niu Zekun
SHI MINGHUI

Assignees

上海交通大学

Dates

Publication Date: 20260505
Application Date: 20260309

Claims (8)

1. The multi-agent automatic operation method of the optical transmission system is characterized by comprising the following steps of: four agents which work cooperatively are an execution agent, a monitoring agent, a fault detection agent and a repair agent respectively; The execution agent is used for carrying out natural language interaction with a user, analyzing experimental intention, and calling a pre-constructed tool function to complete configuration and experimental start of an optical transmission link; the monitoring intelligent agent is used for monitoring the running state of the system and the experiment log after the experiment is started, judging whether the experiment is normal or not according to the running condition, or triggering a fault processing flow; the fault detection intelligent body adopts a finely-tuned multi-mode large language model as a core, receives images and data generated during system operation, and performs multi-mode analysis to locate faults and predict severity of the faults; The repairing intelligent body is used for calling corresponding tool functions to adjust hardware equipment or experimental parameters according to the output result of the fault detection intelligent body so as to repair system faults and restore performance; The execution agent, the monitoring agent, the fault detection agent and the repairing agent interact and task flow according to preset cooperative logic, so that full-chain automation of the optical transmission system from configuration, operation, monitoring and fault detection to repairing is realized.
2. The automated operation method of multiple agents for an optical transmission system according to claim 1, wherein the executing agent uses a plain text large language model as a core, and incorporates ReAct prompt word frames to implement automated link configuration and experimental operation for thinking and execution.
3. The automated operation method of multiple agents for an optical transmission system according to claim 1, wherein the monitoring agent uses a plain text large language model as a core, and performs operation state judgment based on experimental logs and historical experimental results through guidance of prompt words.
4. The automated operation method of multiple intelligent agents for an optical transmission system according to claim 1, wherein the trimmed multi-modal large language model adopted by the fault detection intelligent agent is obtained by manually adding faults in the optical transmission system in normal operation and recording corresponding images and data and training, so that the model can predict fault labels according to multi-modal information in fault operation.
5. The automated operation method of multiple agents for an optical transmission system according to claim 1, wherein the repair agent uses a plain text large language model as a core, and automatically invokes a tool function to perform hardware tuning or parameter tuning according to fault information.
6. The method of multi-agent automated operation of an optical transmission system according to any one of claims 1 to 5, wherein the collaboration logic comprises: the execution agent wakes up the monitoring agent after the experiment is started; The monitoring agent wakes up the fault detection agent when the system is judged to be abnormal; the fault detection intelligent agent wakes up the repair intelligent agent after fault positioning and evaluation are completed; After the repair agent completes repair, control right is returned to the execution agent or the monitoring agent to continue or re-execute the experiment.
7. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor, when executing the computer program, implements the multi-agent automated method of operation of the optical transmission system of any one of claims 1 to 6.
8. A computer-readable storage medium, on which a computer program is stored, characterized in that the computer program, when executed by a processor, implements a multi-agent automated operation method of an optical transmission system according to any one of claims 1 to 6.

Description

Multi-agent automatic operation method, equipment and medium of optical transmission system Technical Field The present invention relates to the technical field of optical transmission systems, and in particular, to a method, an apparatus, and a medium for automatically operating multiple agents in an optical transmission system. Background Optical transmission systems are the core of modern communication networks, whose operation and maintenance face a number of challenges. The traditional mode is highly dependent on manual operation, has the defects of high learning cost, low response speed, dependence on personal experience in fault diagnosis, incapability of forming an automatic closed loop from problem discovery to repair and the like, and is difficult to meet the high-efficiency and intelligent operation and maintenance requirements. Existing automation techniques focus on a single link, such as parameter configuration using scripts or simple alarm response based on rules. The schemes have insufficient flexibility and adaptability, and particularly when detecting system faults, only structured data such as logs and indexes are usually analyzed, but key unstructured visual information such as spectrum images is ignored, so that deep fusion of multi-mode data cannot be realized, and the accuracy and the comprehensiveness of fault positioning are limited. In recent years, the development of a large language model in task planning and multi-mode large model in graphic understanding provides new possibility for constructing an intelligent operation and maintenance system. However, how to combine these capabilities with the knowledge depth of the optical transmission field, designing a full-chain automation framework capable of cooperatively completing from configuration, monitoring, multi-mode fault diagnosis to autonomous repair is still a technical problem to be solved. At present, a unified multi-agent cooperative framework is lacking so as to realize professional labor division and organic cooperation of each link, and finally, the high automation of system operation and maintenance is achieved. Disclosure of Invention Aiming at the defects in the prior art, the invention aims to provide a multi-agent automatic operation method, equipment and medium of an optical transmission system. The multi-agent automatic operation method of the optical transmission system provided by the invention comprises the following steps: four agents which work cooperatively are an execution agent, a monitoring agent, a fault detection agent and a repair agent respectively; The execution agent is used for carrying out natural language interaction with a user, analyzing experimental intention, and calling a pre-constructed tool function to complete configuration and experimental start of an optical transmission link; the monitoring intelligent agent is used for monitoring the running state of the system and the experiment log after the experiment is started, judging whether the experiment is normal or not according to the running condition, or triggering a fault processing flow; the fault detection intelligent body adopts a finely-tuned multi-mode large language model as a core, receives images and data generated during system operation, and performs multi-mode analysis to locate faults and predict severity of the faults; The repairing intelligent body is used for calling corresponding tool functions to adjust hardware equipment or experimental parameters according to the output result of the fault detection intelligent body so as to repair system faults and restore performance; The execution agent, the monitoring agent, the fault detection agent and the repairing agent interact and task flow according to preset cooperative logic, so that full-chain automation of the optical transmission system from configuration, operation, monitoring and fault detection to repairing is realized. Preferably, the execution agent adopts a plain text large language model as a core, and combines ReAct prompt word frames to realize automatic link configuration and experiment operation of thinking and execution. Preferably, the monitoring agent adopts a plain text large language model as a core, and carries out running state judgment through guidance of prompt words based on experimental logs and historical experimental results. Preferably, the trimmed multi-mode large language model adopted by the fault detection intelligent agent is obtained by manually adding faults in a normally operating optical transmission system, recording corresponding images and data and training, so that the model can predict fault labels according to multi-mode information during fault operation. Preferably, the repair agent adopts a plain text large language model as a core, and automatically invokes a tool function according to fault information to complete hardware tuning or parameter adjustment. Preferably, the collaboration logic specifically includes: the executi