CN-121612888-B - Intelligent recognition system for abnormal state of transformer

CN121612888BCN 121612888 BCN121612888 BCN 121612888BCN-121612888-B

Abstract

The invention relates to the technical field of intelligent inspection and state diagnosis of power equipment, in particular to an intelligent recognition system for abnormal states of a transformer; the system comprises a data acquisition unit, an environment parameter estimation unit, an optical decoupling unit, a state diagnosis unit, an interference quantification unit and a coupling diagnosis unit, wherein the system captures a high-dynamic pixel stream and estimates the environment parameter, the core is that inverse optical transmission decoupling is adopted to physically decompose the pixel stream into an equipment inherent reflectivity map and an environment illumination and interference map, the system performs defect diagnosis on the reflectivity map, simultaneously quantifies the interference map to evaluate the reliability, and finally couples to generate an interpretable diagnosis report.

Inventors

PENG BO
PENG YIHAO
Gao Angang
LI LULU
ZHANG YUHUA
HU ZHENGUO
LI DANGHUI

Assignees

陕西瑞博圣泰电器有限公司

Dates

Publication Date: 20260512
Application Date: 20260202

Claims (4)

1. An intelligent recognition system for abnormal states of a transformer is characterized by comprising: The scene perception and data acquisition unit is used for capturing a high dynamic range visual image stream of a transformer target area to obtain an original high dynamic pixel stream; The physical environment parameter estimation unit is used for receiving the original high-dynamic pixel stream, extracting environment parameters representing physical interference and generating an environment state vector; The inverse optical transmission decoupling unit is used for receiving the original high-dynamic pixel stream and the environment state vector and decomposing the original high-dynamic pixel stream into an equipment inherent reflectivity map and an environment illumination and interference map; The equipment material state diagnosis unit is used for carrying out defect detection on the equipment inherent reflectivity graph, identifying a defect area and generating a defect characteristic signal; The interference quantification and credibility generating unit is used for receiving the environment illumination and interference graph, analyzing interference signals and generating an environment interference degree index; The coupling diagnosis result generating unit is used for receiving the defect characteristic signals and the environment interference degree indexes, performing physical interpretable coupling on the defect characteristic signals and the environment interference degree indexes, and generating an interpretable diagnosis report; The physical environment parameter estimation unit generating the environment state vector includes: comprehensively determining an atmospheric scattering coefficient for representing fog or smoke concentration by analyzing the global contrast reduction degree and color cast of the original high-dynamic pixel stream; determining a dynamic interference pattern for characterizing rain or strong wind by detecting high frequency, transient high light or low frequency jitter of the flow in the original high dynamic pixel stream; generating the environment state vector based on the atmospheric scattering coefficient and the dynamic interference mode; the interference quantization and credibility generating unit is used for: the energy of the environment illumination and interference graph is analyzed, the statistical characteristics of the intensity, the range and the frequency of an interference signal are included, and a comprehensive evaluation value serving as a credibility adjustment factor is generated to serve as the environment interference degree index which is used for quantitatively reflecting the pollution degree of the original image by meteorological noise; the coupling diagnosis result generating unit is used for: taking the environmental interference index as a diagnosis confidence regulator, and actively regulating and controlling the output logic of the defect characteristic signal; When the defect characteristic signal indicates that a defect exists, judging the environment interference degree index and a trusted interference threshold; the process of the coupling diagnosis result generating unit for making the determination includes: if the environmental interference index is lower than the credible interference threshold, outputting a high credibility defect; if the environmental interference index is higher than the credible interference threshold, actively reducing the confidence coefficient of the diagnosis result, or marking the diagnosis result as a suspected defect under high interference, or triggering a request for manual review; Environmental interference index = Wherein Quantized eigenvalues of interference intensity, range and frequency calculated according to the above statistical eigenvalue analysis The method is characterized in that preset weights which are optimized on the verification data set according to a calibration method and used for representing the importance of different interference features are obtained, and the value is defined as an environmental interference degree index.
2. The intelligent transformer abnormal state identification system according to claim 1, wherein the reverse optical transmission decoupling unit is configured to: dynamically selecting and configuring a physical optical model by taking the environmental state vector as a control input; When the environmental state vector indicates a high scattering coefficient, a reverse decoupling algorithm based on an atmospheric scattering physical model is initiated to perform the decomposition.
3. The intelligent transformer abnormal state identification system according to claim 1, wherein the equipment material state diagnosis unit is configured to: and analyzing the inherent reflectivity map of the equipment by adopting a high-sensitivity defect detection algorithm to accurately identify the area which accords with the early oil stain or hairline crack characteristics in the inherent reflectivity map of the equipment.
4. A transformer abnormal state intelligent recognition system according to claim 3, wherein said trusted interference threshold is a benchmark set by comparing decoupling quality at different interference intensities with final diagnostic accuracy on a verification data set for statistical analysis.

Description

Intelligent recognition system for abnormal state of transformer Technical Field The invention relates to the technical field of intelligent inspection and state diagnosis of power equipment, in particular to an intelligent identification system for abnormal states of a transformer. Background In the visual inspection operation and maintenance of power equipment such as transformers, early defects such as greasy dirt and cracks are required to be identified by a system through analysis of images, the existing identification models based on data driving are usually black boxes, the degree of interference suffered by the models in diagnosis cannot be quantified, and a physical interpretable diagnosis basis cannot be provided, so that the physical confusion of equipment state signals and environment interference signals is solved, the sensitive and reliable identification of the tiny defects in the early stage under the high-interference environment is realized. Disclosure of Invention In order to solve the technical problems, the invention provides an intelligent recognition system for abnormal states of a transformer, and specifically, the technical scheme of the invention comprises the following steps: The scene perception and data acquisition unit is used for capturing a high dynamic range visual image stream of a transformer target area to obtain an original high dynamic pixel stream; The physical environment parameter estimation unit is used for receiving the original high-dynamic pixel stream, extracting environment parameters representing physical interference and generating an environment state vector; The inverse optical transmission decoupling unit is used for receiving the original high-dynamic pixel stream and the environment state vector and decomposing the original high-dynamic pixel stream into an equipment inherent reflectivity map and an environment illumination and interference map; The equipment material state diagnosis unit is used for carrying out defect detection on the equipment inherent reflectivity graph, identifying a defect area and generating a defect characteristic signal; The interference quantification and credibility generating unit is used for receiving the environment illumination and interference graph, analyzing interference signals and generating an environment interference degree index; and the coupling diagnosis result generating unit is used for receiving the defect characteristic signals and the environment interference degree indexes, performing physical interpretable coupling on the defect characteristic signals and the environment interference degree indexes, and generating an interpretable diagnosis report. Preferably, the physical environment parameter estimation unit generating the environment state vector includes: comprehensively determining an atmospheric scattering coefficient for representing fog or smoke concentration by analyzing the global contrast reduction degree and color cast of the original high-dynamic pixel stream; determining a dynamic interference pattern for characterizing rain or strong wind by detecting high frequency, transient high light or low frequency jitter of the flow in the original high dynamic pixel stream; and generating the environmental state vector based on the atmospheric scattering coefficient and the dynamic interference pattern. Preferably, the reverse optical transmission decoupling unit is configured to: dynamically selecting and configuring a physical optical model by taking the environmental state vector as a control input; When the environmental state vector indicates a high scattering coefficient, a reverse decoupling algorithm based on an atmospheric scattering physical model is initiated to perform the decomposition. Preferably, the device material state diagnosis unit is configured to: and analyzing the inherent reflectivity map of the equipment by adopting a high-sensitivity defect detection algorithm to accurately identify the area which accords with the early oil stain or hairline crack characteristics in the inherent reflectivity map of the equipment. Preferably, the interference quantization and reliability generating unit is configured to: And generating a comprehensive evaluation value serving as a credibility adjustment factor by analyzing the energy of the environment illumination and interference graph, including the statistical characteristics of the intensity, the range and the frequency of the interference signal, so as to serve as the environment interference degree index, wherein the environment interference degree index is used for quantitatively reflecting the pollution degree of the original image by the meteorological noise. Preferably, the coupling diagnosis result generating unit is configured to: taking the environmental interference index as a diagnosis confidence regulator, and actively regulating and controlling the output logic of the defect characteristic signal; And when the defect charact