Search

CN-121979150-A - Offline virtual laboratory and intelligent digital application iteration method and system based on intelligent agent

CN121979150ACN 121979150 ACN121979150 ACN 121979150ACN-121979150-A

Abstract

The invention discloses an offline virtual laboratory and intelligent digital application iteration method and system based on an agent. The method comprises the steps of constructing an off-line virtual laboratory consistent with a site in a non-production environment, keeping consistency of data and configuration through a synchronization mechanism, automatically scheduling working condition arrangement, simulation deduction and rule check by using an intelligent agent, carrying out process deduction in a mode of combining data driving and similar working condition retrieval, completing index calculation and safety verification, carrying out full life cycle management on a log-intelligence application, and realizing cross-site quick replication through templatized encapsulation. The system can realize low-risk and high-efficiency iterative verification and popularization of intelligent digital application on the premise of not interfering with on-site production, and remarkably improves the safety and engineering level of industrial intelligent landing.

Inventors

  • LIU YONGCAI
  • Liu Langtian
  • WU ZHONGHUA
  • LI NAN
  • WANG FEI

Assignees

  • 深圳市佳运通电子有限公司

Dates

Publication Date
20260505
Application Date
20260204

Claims (10)

  1. 1. An agent-based offline virtual laboratory and intelligent application iteration system, comprising: the synchronous and isolation interface layer is used for acquiring field real-time data, historical data, system configuration, rule logic and service ledger from the field data source layer, and performing desensitization processing and authority labeling on the acquired data to form data and configuration resources which can be safely used in an offline environment; The data base and the consistency management layer are connected with the synchronous and isolation interface layer and are used for receiving and storing the data and configuration resources, and version management, difference detection and data management are carried out on the data and the configuration resources based on a unified object model so as to generate and maintain a consistency data base containing version data, configuration, rules, working condition portraits and dependency graphs; The deduction verification and experiment execution layer is connected with the data base and the consistency treatment layer and is used for arranging experimental cases based on working condition portraits in the consistency data base, calling a simulation deduction executor, utilizing a data-driven deduction model, similar working condition retrieval and rule constraint check, executing simulation deduction and verification of digital intelligence application and strategies including security strategies in an offline environment, and generating an index calculation result and a verification report; The application iteration and delivery layer is connected with the data base, the consistency management layer and the deduction verification and experiment execution layer and is used for carrying out full life cycle standing book management on the intelligent logarithmic application, carrying out application dependency analysis and point location mapping based on a dependency graph in the consistency data base, configuring and integrating joint debugging on the application in a joint debugging sandbox through a standard interface, and generating an online evaluation and optimization suggestion of the application based on an index calculation result and a verification report generated by the deduction verification and experiment execution layer; The intelligent agent arrangement and flow control layer is connected with the synchronization and isolation interface layer, the data base and consistency control layer, the deduction verification and experiment execution layer and the application iteration and delivery layer, and is used for carrying out automatic arrangement and scheduling on the whole flow of data synchronization, consistency control, experiment arrangement, deduction verification, application joint debugging and evaluation delivery through the intelligent agent, and carrying out flow control based on the unified object model and the dependency graph; The unified portal and the role workbench are connected with the agent arrangement and flow control layer and are used for providing a man-machine interaction interface so as to drive and monitor the automatic flow of the agent arrangement and flow control layer and safely transmit the application or strategy approval after off-line verification to the site through the approval issuing interface of the synchronous and isolation interface layer.
  2. 2. The agent-based offline virtual laboratory and digital intelligence application iteration system of claim 1, wherein said synchronization and isolation interface layer is configured to: establishing a synchronization strategy combining periodic synchronization and on-demand synchronization, wherein the synchronization object comprises historical time sequence data, system configuration parameters, a point table structure, an alarm rule, interlocking logic and an emergency flow; Performing unified coding, time alignment and difference detection on the synchronized data, and generating a version number and a verification abstract for each synchronization; and carrying out desensitization treatment and authority labeling on the sensitive data.
  3. 3. The agent-based offline virtual laboratory and digital intelligence application iteration system of claim 1, wherein said data base and consistency management layer comprises: The unified object model module is used for distributing offline object identifiers for the measuring points, the equipment, the rule entries and the application examples; the version management and difference detection module is used for carrying out version storage and difference comparison on the configuration snapshot, the point table and the rules; And the dependency graph module is used for maintaining the dependency relationship among the objects and generating a regression verification plan when the change occurs.
  4. 4. The agent-based offline virtual laboratory and digital intelligence application iteration system of claim 1, wherein said deduction verification and experiment execution layer comprises: The data-driven deduction model module adopts a combination mode of data model, statistical rule and similar working condition retrieval and replay to carry out process simulation; the rule and constraint checking module is used for carrying out feasibility checking and risk marking on the deduction result; And the security policy verification module is used for carrying out scripted verification on the alarm rule and the interlocking logic.
  5. 5. The agent-based offline virtual laboratory and digital intelligence application iteration system of claim 1, wherein said application iteration and delivery layer comprises: the application ledger and life cycle management module is used for recording application versions, dependency points and running states; the dependence analysis and point location mapping module is used for analyzing application dependence and completing point location mapping verification; And the joint debugging sandbox module is used for carrying out interface joint debugging and compatibility verification on the application in an offline environment.
  6. 6. The agent-based offline virtual laboratory and digital intelligence application iteration system of claim 1, wherein agents in said agent orchestration and process management layer are for: Calling a data query, working condition arrangement and deduction execution tool through a tool interface layer; Generating a mission plan comprising a data version, a configuration version and a model version; binding evidence chain elements such as data windows, key curves, rule trigger records and the like when outputting the conclusion.
  7. 7. The agent-based offline virtual laboratory and digital intelligence application iteration system of claim 1, further comprising: The security authority and audit module is used for providing identity authentication, role authority control and whole-process audit trail; and the operation monitoring and operation maintenance guaranteeing module is used for performing task queue monitoring, resource quota management and service health check.
  8. 8. The agent-based offline virtual laboratory and digital intelligence application iteration system of claim 1, wherein said deduction verification and experiment execution layer is further configured to: establishing a unified index system at least comprising indexes related to yield, energy consumption, volatility and alarm triggering; and comparing and analyzing the multi-scheme parallel deduction result to generate an evaluation report containing difference comparison and a risk list.
  9. 9. The agent-based offline virtual laboratory and digital intelligence application iteration system of claim 1, wherein said application iteration and delivery layer is further configured to: packaging the mature application into a template package containing parameterized configuration and dependency declarations; performing point location mapping check and working condition adaptation check during cross-site migration; Regression verification is completed before the new site is online.
  10. 10. An intelligent agent-based offline virtual laboratory and intelligent application iteration method is characterized by comprising the following steps of: Step one, demand triggering and target definition, namely receiving triggering requests from scientific research experiments, technical improvement verification, strategy parameter adjustment, safety check, application access or field change regression, and determining target sites or device ranges, target problems, expected output types and constraint conditions, wherein the expected output types comprise reports, suggestions, configuration draft or regression conclusions, and the constraint conditions comprise safety boundaries, data ranges and time windows; Selecting a configuration version, a point table version, a rule version, a model version and an index caliber version based on the target determined in the first step, and generating a synchronization list and a synchronization strategy, wherein the synchronization strategy comprises period synchronization or on-demand synchronization, full-quantity synchronization or increment synchronization and synchronization range cutting, and establishing a version binding relation of the process; Step three, data and configuration synchronization and consistency verification, namely synchronizing time sequence data, event data, configuration snapshot, point table metadata, rule logic and emergency flow assets from a field domain according to the synchronization list and the synchronization strategy generated in the step two; Step four, data management, object modeling and working condition portrait updating, which is to perform time alignment, unit dimension unification processing, quality identification and source tracing on the data synchronized in the step three and passing through verification; step five, working condition arrangement and test case generation, namely receiving key parameters and strategies input by a user or an intelligent agent, generating a working condition script and converting the working condition script into a standardized test case, defining a comparison baseline, an evaluation index set and a passing criterion, storing the test case into a case library and binding version information; Step six, simulation deduction execution and rule or constraint checking, namely selecting at least one deduction route of a data model, a statistical rule, similar working condition replay or a combined route according to the test case generated in the step five, and executing time propulsion and disturbance injection; Calculating the yield, the energy consumption, the fluctuation degree, the stability, the alarm or interlocking triggering times and the risk exposure index according to a unified index system based on the deduction result generated in the step six, outputting a transverse comparison result and a recommendation conclusion if multiple schemes are deduced in parallel, and definitely recommending a premise, an applicable boundary and a risk list to form a standardized evaluation report and an evidence chain for archiving; Step eight, safety strategy script verification and regression verification, namely importing and loading alarm rules, interlocking logic and emergency flows when safety check is needed, constructing a limit working condition or fault scene script and executing verification, outputting protection action sequence, response time, coverage and completeness conclusion, generating candidate improvement schemes and triggering regression use case set to re-verify if defects are found, and until passing criteria or forming risk acceptance suggestions are met; step nine, applying joint debugging delivery and online closed loop, namely finishing the generation of a dependence analysis and point location mapping check sum configuration package when application iteration is involved, generating an online material after the offline joint debugging sandbox passes verification and entering an approval process, continuously evaluating the application effect according to the monitoring index and outputting an optimization suggestion after the online, and packaging the online material into a template package and starting a cross-site copying process when the application effect is stable and meets preset conditions.

Description

Offline virtual laboratory and intelligent digital application iteration method and system based on intelligent agent Technical Field The invention relates to the technical field of industrial digitization, in particular to an offline virtual laboratory and intelligent digital application iteration method and system based on an intelligent agent. Background The existing industrial field intelligent counting system, control strategy and intelligent application have the contradiction in engineering landing that the field working condition is the truest and the constraint is the most complete, but any trial and error can bring production fluctuation, production stopping loss and safety risk, the offline environment is safer and more controllable, the offline environment is gradually deviated from the field due to the fact that configuration, point table and version dependence are difficult to continuously synchronize with typical working condition data, and the offline test result is lack of representativeness and difficult to migrate to guide field decision. Meanwhile, iteration of intelligent application and control strategies often presents punctiform construction and decentralized operation and maintenance states, and reproducible, auditable and quantifiable iteration methodologies and system support are lacked, so that scheme evaluation caliber is not uniform, verification evidence links are incomplete, popularization and replication costs are high, and effects are difficult to close after online. Specifically, the prior art has disadvantages in that: The prior art lacks an offline test vehicle that can be brought into close proximity to the field for long periods of time and safely isolated. In the prior art, new processes, new parameters and new control strategies often have to be verified on site or semi-site, disturbance is easy to cause to continuous production, uncontrollable safety risks are easy to introduce, while the existing offline simulation or test environments are mostly built at one time or only partially synchronize historical time sequence data, the relation of field software versions, application components, configuration parameters and point tables cannot be systematically re-carved, drifting occurs along with continuous change of the site, and finally, the situation that offline verification conclusion is difficult to use for site decision is formed. The prior art lacks a systematic synchronization and consistency maintenance mechanism for covering the history statistics rule logic of typical working conditions of a configuration point table. To make an offline virtual laboratory representative, not only focused on curve data playback, but also key assets such as system configuration, point table structure, unit range, sampling and quality identification, parameter templates, application version and dependency list, interface definition, alarm rules, interlocking logic, emergency flow and the like must be brought into synchronization and version management, and consistency verification, difference detection, change influence analysis and rollback mechanisms are provided, otherwise, the offline environment inevitably generates structural deviation with the site, so that the test cannot be reproduced, the result cannot be traced back, and the risk cannot be estimated. In the prior art, the deduction capability is difficult to fall to the ground under the condition that a mechanism model is incomplete. Many sites or devices lack a fine mechanism model capable of being maintained for a long time, if a complex mechanism equation is strongly relied on to carry out full-flow simulation, the problems of long modeling period, difficult parameter identification, high maintenance cost, difficult cross-site migration and the like are brought, and the offline capability is difficult to build and reuse in a large scale. The high-risk limit working condition and the fault scene in the prior art are difficult to verify repeatedly at low cost. The industrial safety boundary is often reflected in scenes such as limit liquid level, limit pressure, equipment failure, frequent start and stop, and the like, but the related verification risk of field development is high, the examination and approval is complex, the organization cost and the production stopping cost are high, so that the protection strategy is often remained on the document examination or a small number of spot detection layers, and the system coverage and quantitative conclusion are difficult to form. The scientific research test and scheme comparison in the prior art lack the unified caliber index system and the automatic comparison capability. The new process and the new strategy often need parallel experiments of multiple schemes, but the traditional mode depends on personal experience and scattered tools, the index caliber is not uniform, the organization of the test packet is not standard, the data preparation and reproduction cost is