CN-117591962-B - Defect detection model training, GIS equipment defect detection method and related device

CN117591962BCN 117591962 BCN117591962 BCN 117591962BCN-117591962-B

Abstract

The invention discloses a defect monitoring model training, GIS equipment defect detection method and a related device, which are applied to the field of GIS equipment mechanical defect detection, wherein the method adopts improved adaptive noise complete set empirical mode decomposition to conduct characteristic decomposition on GIS equipment mechanical vibration signals under variable frequency current excitation, and utilizes normalized mutual information calculation to conduct validity screening on eigenmode functions and achieve GIS equipment mechanical vibration signal reconstruction; and further extracting a characteristic matrix of the mechanical vibration signal of the reconstructed GIS equipment, and performing model training to obtain a defect detection model. According to the invention, the feature decomposition and the signal reconstruction of the GIS equipment mechanical vibration signal under the excitation of the variable-frequency current are realized through the modal decomposition and the normalized mutual information calculation, and the defect detection model is trained by extracting the features of the reconstructed GIS equipment mechanical vibration signal.

Inventors

HAO JIAN
LIU QINGSONG
LI YING
LI XU
SHAO ZIQI
Che Haolun
XU JING
WANG JIXIANG
Xia Ruochun
ZENG QIAN

Assignees

重庆大学

Dates

Publication Date: 20260505
Application Date: 20231121

Claims (9)

1. A method for training a defect detection model, comprising: Acquiring a GIS equipment mechanical vibration signal under the excitation of variable frequency current; performing modal decomposition on the GIS equipment mechanical vibration signal through an improved adaptive noise complete set empirical mode decomposition algorithm to obtain a plurality of intrinsic mode functions; Calculating normalized mutual information of the GIS equipment mechanical vibration signals and a plurality of eigenvalue functions; removing false eigenmode functions in the eigenmode functions based on the normalized mutual information to obtain target eigenmode functions; linearly superposing a plurality of target eigen mode functions to obtain a mechanical vibration signal of the reconstructed GIS equipment; Extracting the characteristics of the mechanical vibration signals of the reconstructed GIS equipment and a plurality of eigenvalue functions, and constructing a characteristic matrix based on the characteristics; training based on the feature matrix to obtain a defect detection model; the extracting the characteristics of the mechanical vibration signals of the reconstructed GIS equipment and the plurality of eigenvalue functions comprises: Extracting the amplitude, fundamental frequency amplitude, skewness index, kurtosis index, parity response ratio and vibration entropy of the mechanical vibration signal of the reconstructed GIS equipment; Extracting the modal energy ratio and the modal gravity center ratio of a plurality of the intrinsic modal functions; The mode of extracting the amplitude is as follows: ; In the formula, For said amplitude of said reconstructed GIS device mechanical vibration signal, For the reconstruction of the time series of mechanical vibration signals of the GIS device, Sampling point sequence number; the mode of extracting the fundamental frequency amplitude is as follows: ; In the formula, For said fundamental frequency amplitude of said reconstructed GIS device mechanical vibration signal, A frequency domain sequence obtained by Fourier transformation of the reconstructed vibration signal, The frequency is represented by a frequency value, Is the frequency of the excitation current; The way of extracting the even response ratio is as follows: ; In the formula, For said even response ratio of said reconstructed GIS device mechanical vibration signal, The highest frequency of the frequency spectrum of the mechanical vibration signal of the GIS equipment is obtained; the mode for extracting the vibration entropy is as follows: ; In the formula, The vibration entropy of the mechanical vibration signal of the reconstructed GIS equipment is obtained; the mode of extracting the modal energy ratio is as follows: ; In the formula, For the ratio of the modal energies, Is the first of a plurality of the eigenmode functions Order mode Is a frequency domain sequence of (a); the mode of extracting the modal gravity center ratio is as follows: ; In the formula, Is the modal gravity center ratio.
2. The method for training a defect detection model according to claim 1, wherein the training based on the feature matrix to obtain the defect detection model comprises: training a random forest model based on the data set by taking the feature matrix as the data set; and determining the trained random forest model as the defect detection model.
3. The defect detection model training method of claim 2, wherein the training a random forest model based on the data set comprises: training the random forest model based on the data set, and searching optimized model parameters through a firefly algorithm in the training process.
4. The method for training a defect detection model according to claim 1, wherein the training based on the feature matrix to obtain the defect detection model comprises: Performing data dimension reduction on the feature matrix to obtain a dimension reduction feature matrix; and training based on the dimension reduction feature matrix to obtain the defect detection model.
5. The method of claim 4, wherein the performing data dimension reduction on the feature matrix to obtain a dimension-reduced feature matrix comprises: and performing PCA dimension reduction on the feature matrix to obtain the dimension reduction feature matrix.
6. The GIS equipment defect detection method is characterized by comprising the following steps of: Acquiring a mechanical vibration signal of GIS equipment to be detected; performing defect detection on the GIS equipment based on the defect detection model and the mechanical vibration signal of the GIS equipment to be detected; wherein the defect detection model is a model trained according to the defect detection model training method of any one of claims 1 to 5.
7. A defect detection model training device, comprising: the first module is used for acquiring a GIS equipment mechanical vibration signal under the excitation of variable frequency current; The second module is used for carrying out modal decomposition on the GIS equipment mechanical vibration signal through an improved adaptive noise complete set empirical mode decomposition algorithm to obtain a plurality of eigenmode functions; The third module is used for calculating normalized mutual information of the GIS equipment mechanical vibration signals and the plurality of eigenvalue functions; A fourth module, configured to remove a false eigenmode function in the eigenmode functions based on the normalized mutual information, so as to obtain a plurality of target eigenmode functions; a fifth module, configured to linearly superimpose the multiple target eigen mode functions to obtain a mechanical vibration signal of the reconstructed GIS device; A sixth module, configured to extract characteristics of the mechanical vibration signal of the reconstructed GIS device and a plurality of eigen mode functions, and construct a feature matrix based on the characteristics; a seventh module, configured to train to obtain a defect detection model based on the feature matrix; the extracting the characteristics of the mechanical vibration signals of the reconstructed GIS equipment and the plurality of eigenvalue functions comprises: Extracting the amplitude, fundamental frequency amplitude, skewness index, kurtosis index, parity response ratio and vibration entropy of the mechanical vibration signal of the reconstructed GIS equipment; Extracting the modal energy ratio and the modal gravity center ratio of a plurality of the intrinsic modal functions; The mode of extracting the amplitude is as follows: ; In the formula, For said amplitude of said reconstructed GIS device mechanical vibration signal, For the reconstruction of the time series of mechanical vibration signals of the GIS device, Sampling point sequence number; the mode of extracting the fundamental frequency amplitude is as follows: ; In the formula, For said fundamental frequency amplitude of said reconstructed GIS device mechanical vibration signal, A frequency domain sequence obtained by Fourier transformation of the reconstructed vibration signal, The frequency is represented by a frequency value, Is the frequency of the excitation current; The way of extracting the even response ratio is as follows: ; In the formula, For said even response ratio of said reconstructed GIS device mechanical vibration signal, The highest frequency of the frequency spectrum of the mechanical vibration signal of the GIS equipment is obtained; the mode for extracting the vibration entropy is as follows: ; In the formula, The vibration entropy of the mechanical vibration signal of the reconstructed GIS equipment is obtained; the mode of extracting the modal energy ratio is as follows: ; In the formula, For the ratio of the modal energies, Is the first of a plurality of the eigenmode functions Order mode Is a frequency domain sequence of (a); the mode of extracting the modal gravity center ratio is as follows: ; In the formula, Is the modal gravity center ratio.
8. An electronic device, comprising: a memory for storing a computer program; Processor for implementing the defect detection model training method according to any one of claims 1 to 5 and/or the GIS device defect detection method according to claim 6 when executing the computer program.
9. A computer readable storage medium, wherein computer executable instructions are stored in the computer readable storage medium, and when executed by a processor, the computer executable instructions implement the defect detection model training method according to any one of claims 1 to 5, and/or the GIS device defect detection method according to claim 6.

Description

Defect detection model training, GIS equipment defect detection method and related device Technical Field The present invention relates to the field of mechanical defect detection of GIS devices, and in particular, to a defect detection model training method, a GIS device defect detection apparatus, an electronic device, and a computer readable storage medium. Background The gas-insulated metal-enclosed switchgear (Gas Insulated Switchgear, GIS) has the unique advantages of supporting high-voltage and large-capacity transmission of electric energy, compact structure, convenient installation and maintenance and the like, and has become an important component of a power system. The mechanical defect is one of important factors causing the GIS equipment to fail, and the mechanical failure rate reaches 39.3% in the GIS equipment with the voltage level of more than 126 kV. GIS equipment has various mechanical faults, complex conditions and serious harm. The frequency conversion excitation vibration analysis method can excavate the inherent attribute of the mechanical structure and enhance the structural characteristic characterization and mechanical defect detection effect of the GIS equipment, but the existing study on the mechanical defect detection method of the GIS equipment is mostly based on power frequency current, the structural frequency response characteristic of the GIS equipment cannot be represented, and the mechanical defect detection accuracy rate of the GIS equipment under single frequency current is lower. Disclosure of Invention The invention aims to provide a defect detection model training, GIS equipment defect detection method and a related device, which are applied to the field of GIS equipment mechanical defect detection, the method realizes characteristic decomposition and signal reconstruction of GIS equipment mechanical vibration signals under variable-frequency current excitation through modal decomposition and normalized mutual information calculation, the defect detection model is trained by extracting the characteristics of the mechanical vibration signals of the reconstructed GIS equipment, so that compared with the detection of the GIS mechanical defects based on single power frequency current in the prior art, the detection precision is improved. In order to solve the technical problems, the invention provides a defect detection model training method, which comprises the following steps: Acquiring a GIS equipment mechanical vibration signal under the excitation of variable frequency current; performing modal decomposition on the GIS equipment mechanical vibration signal through an improved adaptive noise complete set empirical mode decomposition algorithm to obtain a plurality of intrinsic mode functions; Calculating normalized mutual information of the GIS equipment mechanical vibration signals and a plurality of eigenvalue functions; removing false eigenmode functions in the eigenmode functions based on the normalized mutual information to obtain target eigenmode functions; linearly superposing a plurality of target eigen mode functions to obtain a mechanical vibration signal of the reconstructed GIS equipment; Extracting the characteristics of the mechanical vibration signals of the reconstructed GIS equipment and a plurality of eigenvalue functions, and constructing a characteristic matrix based on the characteristics; and training based on the feature matrix to obtain a defect detection model. Optionally, the extracting the characteristics of the mechanical vibration signal of the reconstructed GIS device and the plurality of eigenmode functions includes: Extracting the amplitude, fundamental frequency amplitude, skewness index, kurtosis index, parity response ratio and vibration entropy of the mechanical vibration signal of the reconstructed GIS equipment; And extracting the modal energy ratio and the modal gravity center ratio of a plurality of the eigenmode functions. Optionally, the training based on the feature matrix to obtain a defect detection model includes: training a random forest model based on the data set by taking the feature matrix as the data set; and determining the trained random forest model as the defect detection model. Optionally, the training random forest model based on the data set includes: training the random forest model based on the data set, and searching optimized model parameters through a firefly algorithm in the training process. Optionally, the training based on the feature matrix to obtain a defect detection model includes: Performing data dimension reduction on the feature matrix to obtain a dimension reduction feature matrix; and training based on the dimension reduction feature matrix to obtain the defect detection model. Optionally, the step of performing data dimension reduction on the feature matrix to obtain a dimension-reduced feature matrix includes: and performing PCA dimension reduction on the feature matrix to obtain the dimen