Search

CN-121980384-A - AI-based power fault diagnosis and processing method and system

CN121980384ACN 121980384 ACN121980384 ACN 121980384ACN-121980384-A

Abstract

The invention relates to the technical field of automation and artificial intelligence intersection of a power system, and particularly discloses an AI-based power fault diagnosis and processing method and system. The system comprises a local rapid diagnosis module, a deep analysis diagnosis module, a simulation verification and decision module, a closed loop optimization module and a closed loop optimization module, wherein the local rapid diagnosis module is used for matching real-time fault characteristics with a historical case library to realize common fault second-level response, the deep analysis diagnosis module is used for carrying out deep reasoning on the physical characteristics obtained by fused power grid power flow calculation and an AI model when rapid matching fails to generate a candidate fault list, the simulation verification and decision module is used for sequentially verifying the candidate list through digital twin simulation to determine final diagnosis and generate a personalized treatment scheme, and the closed loop optimization module is used for synchronously updating and self-learning according to field treatment feedback. The invention realizes the whole-flow closed loop from fault perception, intelligent analysis, physical verification, decision generation to knowledge evolution, improves the accuracy of power fault diagnosis and the safety and efficiency of treatment, and has the capability of continuous self-optimization.

Inventors

  • LI WENLONG
  • JIAO XIANGZHEN
  • SUN RUNJIA
  • Ni Xiangjie

Assignees

  • 李文龙

Dates

Publication Date
20260505
Application Date
20260122

Claims (10)

  1. 1. An AI-based power failure diagnosis and treatment system, comprising: the local rapid diagnosis module is used for rapidly matching the fault recording data and the event sequence record acquired in real time with the historical fault case library, calculating to obtain comprehensive similarity, and outputting a diagnosis result and a treatment plan pre-stored in the corresponding historical case if the highest comprehensive similarity exceeds a first preset threshold; The depth analysis diagnosis module is used for starting when the highest comprehensive similarity does not exceed the first preset threshold value, and is configured to execute power flow calculation before and after a fault to generate physical characteristics representing power transfer and voltage abnormality, and input the physical characteristics, the characteristics extracted from fault recording data and event sequence records and power flow topology data into an AI analysis model together to obtain a candidate fault root cause list ordered according to occurrence probability; The simulation verification and decision module is used for calling a corresponding digital twin model in a simulation fault library to simulate each fault hypothesis in the candidate fault root cause list according to a probability sequence, calculating consistency scores of simulation data and real-time data, determining a fault hypothesis with a first consistency score exceeding a second preset threshold value as a final diagnosis conclusion, and generating a fault treatment operation instruction based on the conclusion and the real-time parameters; The closed loop optimization module is used for receiving field feedback after the fault handling operation instruction is executed, evaluating the accuracy of the diagnosis according to the actual fault reasons in the feedback, and synchronously updating and optimizing the historical fault case library, the simulated fault library and the AI analysis model according to the evaluation results.
  2. 2. The AI-based power failure diagnosis and treatment system of claim 1, wherein the local rapid diagnosis module calculates the comprehensive similarity by calculating a first similarity of the current failure and the historical case on waveform characteristics and a second similarity of the current failure and the historical case on action logic characteristics, respectively, and weighting and summing the first similarity and the second similarity, wherein a weight coefficient of the second similarity is higher than a weight coefficient of the first similarity.
  3. 3. The AI-based power failure diagnosis and treatment system of claim 1, wherein the physical characteristics generated by the depth analysis diagnosis module include, in particular, a set of lines for which the amount of active power change identified by the flow calculation exceeds a set threshold, and a set of bus bars for which the amount of voltage change exceeds a set threshold.
  4. 4. The AI-based power failure diagnosis and treatment system according to claim 1, wherein the simulation verification and decision module calculates the consistency score by performing weighted fusion calculation on a waveform similarity between a simulation waveform and an actual measurement waveform and a time sequence matching between a simulation protection action time and an actual measurement protection action time.
  5. 5. The AI-based power failure diagnosis and treatment system of claim 1, wherein the closed-loop optimization module performs an incremental learning on the AI analysis model, and specifically comprises constructing the current failure full-flow data with correct diagnosis as a new case, merging the new case with the historical correct case as an incremental training set, and fine-tuning parameters of the AI analysis model accordingly.
  6. 6. The AI-based power failure diagnosis and treatment system of claim 1, wherein the closed-loop optimization module is further configured to perform an adaptive adjustment to automatically raise the second preset threshold for a corresponding failure scenario in the simulated failure library when a same misjudgment is detected that occurs a preset number of times consecutively for a particular type of failure.
  7. 7. The AI-based power failure diagnosis and treatment system of claim 1, wherein each historical case in the library of historical failure cases is associated with a dynamic confidence score, wherein the confidence scores are dynamically updated according to the diagnosis correctness fed back after the case is matched and quoted, and wherein the local rapid diagnosis module preferentially recommends historical cases with higher confidence scores when matching.
  8. 8. The AI-based power failure diagnosis and treatment system of claim 1, further comprising a flow scheduling module for controlling the sequential execution and data transfer of the local rapid diagnosis module, the deep analysis diagnosis module, the analog verification and decision module, and the closed loop optimization module, setting a timeout threshold for critical processing links, and starting a backup analysis path upon timeout.
  9. 9. The AI-based power failure diagnosis and treatment system of claim 1, wherein the simulation verification and decision module, when generating the failure handling operation instructions, performs the steps of retrieving successful handling cases in the historical failure case library with the final diagnosis conclusion to obtain an operation template, and instantiating and populating the operation template in combination with real-time device parameters and operating environment data obtained from external system queries.
  10. 10. AI-based power failure diagnosis and treatment method, applied to a system according to any of claims 1-9, comprising the steps of: S1, collecting real-time fault data, extracting features, carrying out quick matching with a historical case library, outputting a result if matching is successful, and otherwise, entering S2; S2, performing power grid power flow analysis to generate physical characteristics, and fusing multi-source characteristics to obtain a candidate fault list according to probability sequence through an AI analysis model; S3, performing sequential digital twin simulation verification on the candidate fault list, determining the first verified fault as final diagnosis, and generating a disposal operation instruction according to the final diagnosis; And S4, evaluating the diagnosis accuracy according to feedback after instruction execution, and driving closed-loop optimization of a historical case library, a simulated fault library and an AI analysis model by using feedback information.

Description

AI-based power fault diagnosis and processing method and system Technical Field The invention belongs to the technical field of power system automation and artificial intelligence intersection, and particularly relates to a method and a device for intelligent diagnosis and treatment scheme generation and system self-optimization of power grid faults. Background Safe and stable operation of modern power systems is highly dependent on rapid, accurate diagnosis and effective handling of faults. The traditional fault diagnosis mode is mainly based on alarm information of a dispatching automation System (SCADA), action reports (SOE) of a protection device and fault recording data, and operators perform manual analysis and decision. This approach has significant limitations: First, the diagnostic process is highly dependent on the personal experience and skill level of the operator. In the face of massive alarm information, it is difficult for people to quickly correlate multi-source data and accurately position fault points, and particularly for complex faults or new faults, analysis is long in time consumption and easy to misjudge, so that fault processing delay is caused. Second, the formulation of fault handling schemes lacks standardized, systematic support. The existing method depends on operation rules and personal memory, is easy to miss operation steps, have wrong sequence or improper safety measures due to negligence or insufficient experience, has potential safety hazards, and is difficult to effectively accumulate and share excellent treatment experience. Again, the prior art approach makes more isolated use of fault information. Most diagnosis methods are based on simple logic rules, only analyze electric quantity waveforms, or only pay attention to protection action logic, and fail to perform deep fusion and collaborative analysis on electric transient characteristics, protection control system behaviors, power grid trend changes and network topology, so that the depth and accuracy of diagnosis are limited. Finally, existing systems lack self-perfecting capabilities. The traditional expert system or auxiliary decision tool knowledge base is fixed, cannot automatically learn from actually occurring fault cases and treatment feedback, is difficult to adapt to dynamic changes of power grid structures, equipment and operation modes, and diagnosis and treatment strategies cannot be optimized continuously. Therefore, a power failure diagnosis and processing technology capable of automatically fusing multi-source information, intelligently reasoning and deciding and having continuous learning and evolution capability is needed. Disclosure of Invention In order to overcome the above-mentioned drawbacks of the prior art, embodiments of the present invention provide an AI-based power failure diagnosis and processing method and system, so as to solve the above-mentioned problems set forth in the background art. In order to achieve the above purpose, the invention provides a power failure diagnosis and processing system based on AI, comprising: the local rapid diagnosis module is used for rapidly matching the fault recording data and the event sequence record acquired in real time with the historical fault case library, calculating to obtain comprehensive similarity, and outputting a diagnosis result and a treatment plan pre-stored in the corresponding historical case if the highest comprehensive similarity exceeds a first preset threshold; The depth analysis diagnosis module is used for starting when the highest comprehensive similarity does not exceed the first preset threshold value, and is configured to execute power flow calculation before and after a fault to generate physical characteristics representing power transfer and voltage abnormality, and input the physical characteristics, the characteristics extracted from fault recording data and event sequence records and power flow topology data into an AI analysis model together to obtain a candidate fault root cause list ordered according to occurrence probability; The simulation verification and decision module is used for calling a corresponding digital twin model in a simulation fault library to simulate each fault hypothesis in the candidate fault root cause list according to a probability sequence, calculating consistency scores of simulation data and real-time data, determining a fault hypothesis with a first consistency score exceeding a second preset threshold value as a final diagnosis conclusion, and generating a fault treatment operation instruction based on the conclusion and the real-time parameters; The closed loop optimization module is used for receiving field feedback after the fault handling operation instruction is executed, evaluating the accuracy of the diagnosis according to the actual fault reasons in the feedback, and synchronously updating and optimizing the historical fault case library, the simulated fault library and the AI ana