Search

CN-122022189-A - Multi-agent cooperative control method and system based on brain-like fusion architecture

CN122022189ACN 122022189 ACN122022189 ACN 122022189ACN-122022189-A

Abstract

The invention discloses a multi-agent collaborative management and control method and a system based on a brain-like fusion architecture, which relate to the technical field of industrial safety production and comprise the steps of acquiring multi-source data related to a target industrial system and performing time sequence alignment to obtain time pair Ji Duo source data; the method comprises the steps of obtaining evidence representation of each data source based on Ji Duo source data, obtaining a security situation cognition vector based on evidence conflict assessment and a dynamic credibility adjustment mechanism, constructing a risk field state model based on the security situation cognition vector, combining a dynamic risk threshold value to obtain a risk evolution result, constructing a multi-agent collaborative decision model based on the risk evolution result, generating an optimal collaborative strategy under a security constraint condition, converting and executing based on the optimal collaborative strategy to obtain an execution result, evaluating an execution effect of the strategy based on the execution result, and carrying out feedback and self-learning optimization to keep the optimal collaborative strategy. Long-term continuous scientific management and control under complex safety production and emergency management scenes is realized.

Inventors

  • BAO KAIYANG

Assignees

  • 塔盾信息技术(上海)有限公司

Dates

Publication Date
20260512
Application Date
20260408

Claims (10)

  1. 1. A multi-agent cooperative control method based on a brain-like fusion architecture is characterized by comprising the following steps: acquiring multi-source data related to emergency management of safe production of a target industrial system, and performing time sequence alignment to obtain Ji Duo source data of a time pair; Acquiring evidence representation of each data source based on Ji Duo source data of the time, and acquiring a security situation cognitive vector based on evidence conflict assessment and a dynamic credibility adjustment mechanism; Constructing a risk field state model based on the security situation cognitive vector, and combining a dynamic risk threshold to obtain a risk evolution result; constructing a multi-agent collaborative decision model based on the risk evolution result, and generating an optimal collaborative strategy of each agent under a safety constraint condition; Converting and executing based on the optimal cooperative strategy to obtain an executing result; and evaluating the strategy execution effect based on the execution result, and carrying out feedback and self-learning optimization to keep the optimal cooperative strategy.
  2. 2. The multi-agent collaborative management and control method based on a brain-like fusion architecture as set forth in claim 1, wherein the time pair Ji Duo source data acquisition method is as follows: Cleaning, outlier rejection and format normalization processing are carried out based on the multi-source data, and data identification and timestamp information are generated uniformly to obtain structured multi-source data; extracting event anchor points with business significance based on the structured multi-source data; acquiring an optimal time mapping function of each data source relative to a reference time axis based on the event anchor point; And mapping the time stamp of each data source to a unified reference time axis based on the optimal time mapping function, so as to realize time axis correction of multi-source data and obtain the time pair Ji Duo source data.
  3. 3. The multi-agent collaborative management and control method based on a brain-like fusion architecture as set forth in claim 2, wherein the multi-source data comprises industrial control system operation data, production management and safety management system data, environmental and risk monitoring data, and personnel, equipment and emergency resource status data; The event anchor points with business significance comprise equipment start-stop events, threshold value out-of-limit events, alarm trigger events, operation state change events or a combination thereof.
  4. 4. The multi-agent collaborative management and control method based on a brain-like fusion architecture as set forth in claim 2, wherein the security situation cognitive vector acquisition method is as follows: performing feature modeling on Ji Duo source data by adopting a brain-like evidence fusion mechanism based on the time, and mapping the data into original evidence representations corresponding to all data sources; Acquiring the dynamic reliability of each data source based on the data quality, the evidence conflict degree and the time residual error; Performing weighted correction on the original evidence representation based on the dynamic reliability to obtain corrected evidence support; Obtaining multi-mode evidence conflict degrees based on the corrected evidence support degrees; obtaining a fused evidence quality function through evidence fusion operation based on the multi-modal evidence conflict degree; mapping the evidence quality function into the security situation cognitive vector of the target industrial system at the current moment.
  5. 5. The multi-agent collaborative management and control method based on a brain-like fusion architecture as set forth in claim 4, wherein the risk evolution result obtaining method is as follows: constructing a risk field state model of the target industrial system based on the security situation cognitive vector; Based on the risk field state model, carrying out continuous time long-term reasoning on the risk states of all nodes in the target industrial system to obtain risk field potential energy; Determining a dynamic risk threshold which can be born by the target industrial system through probability safety constraint based on the risk field potential energy; and the risk field potential energy and the dynamic risk threshold jointly form the risk evolution result.
  6. 6. The multi-agent cooperative control method based on the brain-like fusion architecture as claimed in claim 5, wherein the optimal cooperative strategy obtaining method is as follows: Constructing a multi-agent collaborative decision model consisting of a plurality of agents, a global coordination unit and a safety constraint projection mechanism based on the risk evolution result; based on the multi-agent collaborative decision model, a double-layer collaborative optimization and safety constraint projection mechanism is adopted, and an upper layer coordinator generates a global coordination strategy vector; Under the condition of safety constraint, each agent at the lower layer obtains a local optimal strategy corresponding to each agent based on the global coordination strategy vector; And carrying out consistency alignment on the basis of the local optimal strategy and the global coordination strategy vector to obtain the optimal cooperative strategy.
  7. 7. The method for collaborative management and control of multiple agents based on brain-like fusion architecture as set forth in claim 6, wherein said multiple agent collaborative decision-making model comprises a perception agent, an inference agent, a decision-making agent, an execution agent, and a learning agent; The perception agent is used for receiving current security situation information and environment state information of the system; the reasoning intelligent body is used for carrying out risk trend reasoning based on the risk evolution result and the security situation information; the decision agent is used for generating a candidate safety control strategy according to the reasoning result; the execution agent is used for converting the candidate safety control strategy into a specific control instruction and executing the specific control instruction; the learning agent is used for updating and optimizing the strategy parameters according to the execution result and the system feedback; the intelligent agents make collaborative decisions under the constraint of a safety constraint projection mechanism so as to ensure that the generated control strategy meets the safety constraint condition of the system.
  8. 8. The multi-agent cooperative control method based on the brain-like fusion architecture as claimed in claim 7, wherein the execution result obtaining method is as follows: converting the optimal cooperative strategy into an executable control instruction and sending the executable control instruction to the target industrial system for execution; And acquiring the real-time information based on the equipment state change and the system operation information in the execution process to obtain the execution result of actual execution.
  9. 9. The multi-agent cooperative control method based on a brain-like fusion architecture as claimed in claim 8, wherein the method for maintaining an optimal cooperative strategy specifically comprises: Acquiring a causal attributive execution deviation index based on the execution result; Constructing a learning optimization target based on the causal attributive performing deviation index and the dynamic risk threshold; Converting the learning optimization target into a Lagrange form to obtain a Lagrange objective function; And optimizing based on the Lagrangian objective function, and adaptively updating the probability safety constraint penalty coefficient, so that the target industrial system improves the penalty strength of the safety constraint when approaching a risk boundary, and automatically converges to the updated optimal cooperative strategy meeting the safety constraint.
  10. 10. A multi-agent cooperative control system based on a brain-like fusion architecture, which is used for executing the multi-agent cooperative control method based on the brain-like fusion architecture as claimed in any one of claims 1 to 9, and is characterized by comprising a multi-source data processing module, a brain-like fusion cognitive module, a task risk reasoning module, a multi-agent cooperative decision module, a cooperative execution module and a feedback optimization module; The multi-source data processing module is used for acquiring multi-source data related to emergency management of safe production of a target industrial system and performing time sequence alignment to obtain Ji Duo source data of a time pair; The brain-like fusion cognitive module is used for acquiring evidence representation of each data source based on Ji Duo source data of the time, and acquiring a security situation cognitive vector based on evidence conflict assessment and a dynamic credibility adjustment mechanism; the task risk reasoning module is used for constructing a risk field state model based on the security situation cognitive vector and combining a dynamic risk threshold value to obtain a risk evolution result; The multi-agent collaborative decision module is used for constructing a multi-agent collaborative decision model based on the risk evolution result and generating an optimal collaborative strategy of each agent under the safety constraint condition; The cooperative execution module is used for converting and executing based on the optimal cooperative policy to obtain an execution result; And the feedback optimization module is used for evaluating the strategy execution effect based on the execution result, carrying out feedback and self-learning optimization and keeping the optimal cooperative strategy.

Description

Multi-agent cooperative control method and system based on brain-like fusion architecture Technical Field The invention relates to the technical field of industrial safety production, in particular to a multi-agent cooperative control method and system based on a brain-like fusion architecture. Background Along with the continuous expansion of industrial production scale and continuous improvement of production system complexity, the problems of safety risk management and emergency disposal in the production operation process of high-risk industrial fields such as energy, chemical industry, metallurgy, mine, manufacturing industry, industrial park and the like are increasingly complex, and higher requirements are put forward on the safety management and risk control technology. In the industrial production operation process, various factors such as equipment state, process parameters, personnel behaviors, environmental conditions, external disturbance and the like act on the system operation state together, and the factors are mutually coupled in the time dimension and the space topological structure, so that the accident risk presents the characteristics of multisource, multimode and dynamic evolution. However, the existing security management technology generally relies on single monitoring index or short-time window analysis, lacks fusion modeling capability of multi-source risk information in space-time dimension, and also is difficult to continuously infer the whole process of accident inoculation, risk evolution and emergency treatment. The prior art mainly surrounds monitoring early warning, information integration and auxiliary decision development, but has certain technical limitations in complex industrial safety and emergency management scenes, and is characterized in that (1) multisource safety production and emergency management information are dispersed in different systems and lack of a unified fusion modeling mechanism, so that overall safety situation cognition is difficult to form, (2) accident risks and emergency events have long-term evolution characteristics, the prior art is difficult to support continuous reasoning and management of cross time scales, (3) multiple roles and lack of effective collaborative decision and linkage mechanisms among multiple systems, complex emergency tasks depend on manual coordination, and (4) management strategy adaptability is insufficient, and continuous optimization and evolution are difficult to realize in long-term operation and multiple event handling processes. Therefore, how to realize long-term continuous perception, dynamic reasoning, collaborative decision and adaptive optimization of safe production running state, risk evolution process and emergency treatment behavior, and further realize long-term continuous scientific management and control in complex safe production and emergency management scenes is a problem to be solved by those skilled in the art. Disclosure of Invention In order to overcome the problems or at least partially solve the problems, the invention provides a multi-agent collaborative management and control method and a system based on a brain-like fusion architecture, which are used for realizing long-term continuous perception, dynamic reasoning, collaborative decision and self-adaptive optimization of safe production running states, risk evolution processes and emergency disposal behaviors by constructing a unified intelligent collaborative architecture of safe production and emergency management and introducing a brain-like fusion cognitive mechanism, a long-task continuous reasoning mechanism, a multi-agent collaborative decision mechanism and an execution-feedback-self-learning closed-loop mechanism under the architecture, thereby realizing long-term continuous collaborative management and control and self-adaptive optimization under complex safe production and emergency management scenes. In order to achieve the above purpose, the present invention adopts the following technical scheme: In a first aspect, an embodiment of the present invention provides a multi-agent collaborative management and control method based on a brain-like fusion architecture, including: acquiring multi-source data related to emergency management of safe production of a target industrial system, and performing time sequence alignment to obtain Ji Duo source data of a time pair; Acquiring evidence representation of each data source based on Ji Duo source data of the time, and acquiring a security situation cognitive vector based on evidence conflict assessment and a dynamic credibility adjustment mechanism; Constructing a risk field state model based on the security situation cognitive vector, and combining a dynamic risk threshold to obtain a risk evolution result; constructing a multi-agent collaborative decision model based on the risk evolution result, and generating an optimal collaborative strategy of each agent under a safety constraint