GIS equipment fault diagnosis method based on vibration and sound fusion

Abstract

The invention relates to the technical field of GIS equipment fault diagnosis, and discloses a GIS equipment fault diagnosis method based on vibration-sound fusion, and the method comprises the following steps: S1, carrying out the feature extraction of a sound-vibration signal, namely, extracting a time domain feature, a frequency domain feature, an energy domain feature and a harmonic domain feature; s2, a CatBoost classification algorithm is optimized, wherein the CatBoost is used for processing discrete features through a symmetric line sampling and target coding strategy; and S3, utilizing the CatBoost optimized by the ant lion algorithm: because the classification precision of the CatBoost algorithm is influenced by two input parameters, namely the learning rate m and the tree depth k, introducing the ant lion optimization algorithm to perform optimization calculation on the input parameters. According to the method, an optimal correction logarithmic spectrum estimation and Wiener filtering fusion algorithm is provided, a noise signal is estimated, an amplitude part in the noisy signal is processed and updated, enhanced sound is obtained, transient noise is filtered by utilizing optimal correction logarithmic spectrum estimation, and steady-state noise is filtered by utilizing Wiener filtering.

Description

Technical Field

[0001] The present invention relates to the technical field of GIS equipment fault diagnosis, and in particular to a GIS equipment fault diagnosis method based on vibration and sound fusion. Background Art

[0002] Substations are mainly composed of GIS, transformers, busbars, grounding switches, lightning arresters, transformers, capacitors, reactors, and live structures. Generally speaking, the higher the voltage level, the greater the substation noise value. The main noise sources of transformers in substations of different voltage levels are also different.

[0003] When analyzing the acoustic signals of GIS equipment, the noise signal collected by the microphone is a mixed sound signal including transformer noise, noise from other power equipment in the substation, and background noise of the substation. Therefore, it is of great significance to explore a reliable method to suppress the GIS operation background noise and extract the pure original sound signal of GIS equipment from the mixed background sound signal of the substation.

[0004] The existing patent discloses a method and system for GIS equipment mechanical fault diagnosis (publication number CN113191247A), which belongs to the field of mechanical fault diagnosis. The method includes: collecting a vibration signal time series of length N of the GIS equipment; taking L as the moving average period, performing moving average processing on the vibration signal time series, obtaining N-L+1 moving averages in sequence, and combining the N-L+1 moving averages in time sequence to obtain a moving average time series. In the technology disclosed in the patent, since the sound signal is easily interfered by external signals during the field test, the quality of the test data is affected. Summary of the invention

[0005] The main technical problem solved by the present invention is to provide a GIS equipment fault diagnosis method based on vibration and sound fusion, which solves the problems in the above-mentioned background technology.

[0006] To solve the above technical problems, according to one aspect of the present invention, more specifically, a GIS equipment fault diagnosis method based on vibration and sound fusion comprises the following steps: S1. Feature extraction of acoustic vibration signals includes extracting time domain features, frequency domain features, energy domain features and harmonic domain features; S2. Optimize the CatBoost classification algorithm, where CatBoost processes discrete features through symmetric row sampling and target encoding strategies; S3. CatBoost optimized by Antlion Algorithm: Since the classification accuracy of CatBoost algorithm is affected by two input parameters, learning rate m and tree depth k, the Antlion optimization algorithm is introduced to optimize the above input parameters.

[0007] Furthermore, the time domain features in step S1 are obtained by introducing a root mean square that can reflect the overall characteristics of the signal and a peak characteristic parameter that reflects the extreme conditions of the signal.

[0008] Furthermore, the frequency domain characteristics in step S1 select the spectral centroid, spectral diffusion, spectral skewness and spectral kurtosis that can characterize the degree of change of the frequency domain components of the GIS equipment as frequency domain characteristic parameters.

[0009] Furthermore, the energy domain characteristics in step S1 are introduced to reflect the energy entropy and permutation entropy in the energy domain.

[0010] Furthermore, the harmonic domain characteristics in step S1 introduce characteristic parameters of odd-even subharmonic ratio and frequency complexity in the harmonic domain.

[0011] Furthermore, in step S1, the mean, standard deviation and peak factor of different features are used to determine whether the GIS equipment of the vibration-acoustic fusion is abnormal. ; Where G represents the abnormal coefficient of GIS equipment fusion of vibration and sound fusion, Represents the mean value in the time domain feature, Indicates the standard deviation size in the time domain features, Represents the peak factor in the time domain characteristics; Among them, the abnormal coefficient judgment indicators for different equipment or solutions are different. At that time, it means that there is an abnormality in the fusion of the GIS equipment of vibration and sound fusion; Among them, represents the size of the abnormal coefficient of different devices or solutions.

[0012] Furthermore, the update step of CatBoost in step S2 can be expressed as: 1) Initialize the model: ; 2) For each iteration: ; Compute the residual: ; Fit a new tree: ; In the formula, is the learning rate, represents the newly fitted decision tree; 3) Final prediction results: ; In the formula, yes The original function of .

[0013] Furthermore, the specific implementation steps in step S3 are as follows: 1) Determine the number of ants and ant lions M, N and the maximum number of iterations, initialize their positions and calculate the corresponding fitness values; 2) Select the antlion with the smallest fitness value in the initialized antlion population as the elite antlion; 3) Use the roulette strategy to match ants with ant lions one by one, iteratively update the relevant values ​​according to the position of the ant lion, and make the ants randomly walk between the ant lion and the elite ant lion according to a certain rule, and finally take the average value as the ant position coordinate; 4) After each iteration, the fitness of the ant and ant lion is calculated respectively. When the position and fitness of the ant are known, the position of the ant lion is updated. The position with the minimum fitness is the new position of the ant lion after iteration. 5) Determine whether the iteration termination condition is met. If the current iteration number is met, stop the iteration and output {m, k}. Otherwise, repeat steps 3 to 5.

[0014] Beneficial effects:

[0015] The algorithm of the optimal modified logarithmic spectrum estimation and Wiener filtering fusion proposed in the present invention estimates the noise signal and processes and updates the amplitude part of the noisy signal to obtain an enhanced sound. The optimal modified logarithmic spectrum estimation is used to filter the transient noise, and the Wiener filtering is used to filter the steady-state noise. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of the iterative convergence curve in the present invention; Figure 2 This is a schematic diagram of the order of importance of feature values ​​in the present invention; Figure 3 Schematic diagram of the confusion matrix in the present invention. DETAILED DESCRIPTION

[0017] In order to make the technical solution of the present invention clearer, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] Example 1

[0019] like Figure 1 - 3, according to one aspect of the present invention, a GIS equipment fault diagnosis method based on vibration-acoustic fusion is provided, comprising the following steps: Step 1: Feature extraction of acoustic vibration signals includes extraction of time domain features, frequency domain features, energy domain features, and harmonic domain features. By introducing the root mean square that can reflect the overall characteristics of the signal and the peak characteristic parameters that reflect the extreme conditions of the signal. The spectral center of gravity, spectral diffusion, spectral skewness, and spectral kurtosis that can characterize the degree of change of the frequency domain components of GIS equipment are selected as frequency domain feature parameters. Introduce energy entropy and permutation entropy that can reflect the energy domain. Introduce the even-odd subharmonic ratio and frequency complexity feature parameters in the harmonic domain.

[0020] Step 2: Optimize the CatBoost classification algorithm, where CatBoost processes discrete features through symmetric row sampling and target encoding strategies; Step 3: CatBoost optimized with the Ant Lion algorithm: Since the classification accuracy of the CatBoost algorithm is affected by two input parameters: learning rate m and tree depth k, the Ant Lion optimization algorithm is introduced to optimize the above input parameters.

[0021] Example 2

[0022] The update step of CatBoost can be expressed as: 1) Initialize the model: ; 2) For each iteration: ; Compute the residual: ; Fit a new tree: ; In the formula, is the learning rate, represents the newly fitted decision tree; 3) Final prediction results: ; In the formula, yes The original function of .

[0023] Example 3

[0024] The specific implementation steps are as follows: 1) Determine the number of ants and ant lions M, N and the maximum number of iterations, initialize their positions and calculate the corresponding fitness values; 2) Select the antlion with the smallest fitness value in the initialized antlion population as the elite antlion; 3) Use the roulette strategy to match ants with ant lions one by one, iteratively update the relevant values ​​according to the position of the ant lion, and make the ants randomly walk between the ant lion and the elite ant lion according to a certain rule, and finally take the average value as the ant position coordinate; 4) After each iteration, the fitness of the ant and ant lion is calculated respectively. When the position and fitness of the ant are known, the position of the ant lion is updated. The position with the minimum fitness is the new position of the ant lion after iteration. 5) Determine whether the iteration termination condition is met. If the current number of iterations If yes, the iteration stops and {m, k} is output. Otherwise, steps 3 to 5 are repeated.

[0025] The ant lion optimization algorithm is used to optimize the learning rate m and the tree depth k. The initial input parameters of the ant lion optimization algorithm are set as: the number of ants M = 200, the number of ant lions N = 200, the variable dimension D = 100, and the maximum number of iterations tmax = 100, and the population position is randomly initialized.

[0026] like Figure 1 As shown in the figure, it is the iterative convergence curve obtained by using the ant lion optimization algorithm. It can be seen that the curve of the fitness value changing with the number of iterations converges after 70 iterations and reaches a stable value. At the same time, the optimal parameters can be obtained: m = 0.001, k = 8.

[0027] like Figure 2 As shown in the figure, the eigenvalue importance ranking obtained by CatBoost is as follows: odd-even harmonic ratio>frequency complexity>root mean square>spectral kurtosis>spectral entropy>spectral centroid>spectral diffusion. After clarifying the importance of each eigenvalue, subsequent work such as threshold warning can be carried out based on it, which provides a certain basis for inspection and identification of operation and maintenance personnel.

[0028] The confusion matrix is ​​introduced to quantitatively evaluate the classification effect, such as Figure 3 The figure shows the calculation result of the confusion matrix obtained by the proposed method. The diagonal numbers of the confusion matrix in the figure represent the number of samples that correctly identify each state using the proposed method. Figure 3 It can be seen that only 4 of the 200 test sets were misclassified, and the classification accuracy of the test sample set was 98.00%. Figure 3 It can be seen that the average state recognition accuracy, recall rate and F1 value of the proposed method are 98.05%, 98.00% and 98.01% respectively. The results show that the proposed method has good recognition effect, proving the feasibility and effectiveness of the proposed method.

[0029] Example 4

[0030] The mean, standard deviation and peak factor of different features are used to judge whether the GIS equipment with vibration-acoustic fusion is abnormal. ; Where G represents the abnormal coefficient of GIS equipment fusion of vibration and sound fusion, Represents the mean value in the time domain feature, Indicates the standard deviation size in the time domain features, Represents the peak factor in the time domain characteristics.

[0031] Among them, the abnormal coefficient judgment indicators for different equipment or solutions are different. At that time, it means that there is an abnormality in the fusion of the GIS equipment of vibration and sound fusion; Among them, represents the size of the abnormal coefficient of different devices or solutions.

[0032] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A GIS equipment fault diagnosis method based on vibration and sound fusion, characterized in that: The following steps are involved: S1. Feature extraction of acoustic vibration signals includes extracting time domain features, frequency domain features, energy domain features and harmonic domain features; S2. Optimize the CatBoost classification algorithm, where CatBoost processes discrete features through symmetric row sampling and target encoding strategies; S3. CatBoost optimized by Antlion Algorithm: Since the classification accuracy of CatBoost algorithm is affected by two input parameters, learning rate m and tree depth k, the Antlion optimization algorithm is introduced to optimize the above input parameters.

2. The GIS equipment fault diagnosis method based on vibration and acoustic fusion according to claim 1 is characterized in that: The time domain characteristics in step S1 are obtained by introducing a root mean square characteristic that can reflect the overall characteristics of the signal and a peak characteristic parameter that reflects the extreme conditions of the signal; Frequency domain characteristics, the spectral center of gravity, spectral diffusion, spectral skewness and spectral kurtosis that can characterize the degree of change of frequency domain components of GIS equipment are selected as frequency domain characteristic parameters.

3. The GIS equipment fault diagnosis method based on vibration and acoustic fusion according to claim 1 is characterized in that: The energy domain features in step S1 are introduced to reflect the energy entropy and permutation entropy in the energy domain; Harmonic domain characteristics, introducing the odd-even subharmonic ratio and frequency complexity characteristic parameters in the harmonic domain.

4. The GIS equipment fault diagnosis method based on vibration and acoustic fusion according to claim 1 is characterized in that: In step S1, the mean, standard deviation and peak factor of different features are used to determine whether the GIS equipment with vibration-acoustic fusion is abnormal. ； Where G represents the abnormal coefficient of GIS equipment fusion of vibration and sound fusion, Indicates the mean value in the time domain feature, Indicates the standard deviation of the time domain features. Represents the peak factor in the time domain characteristics.

5. The GIS equipment fault diagnosis method based on vibration and acoustic fusion according to claim 4 is characterized in that: The abnormal coefficient judgment indicators for different devices or solutions are different. , it means that there is an abnormality in the fusion of GIS equipment of vibration and sound fusion; in, Indicates the size of the abnormal coefficient of different devices or solutions.

6. The GIS equipment fault diagnosis method based on vibration and acoustic fusion according to claim 1 is characterized in that: The update step of CatBoost in step S2 can be expressed as: 1) Initialize the model: ； 2) For each iteration: ; Compute the residual: ； Fit a new tree: ； In the formula, is the learning rate, represents the newly fitted decision tree; 3) Final prediction results: ； In the formula, yes The original function of .

7. The GIS equipment fault diagnosis method based on vibration and acoustic fusion according to claim 1 is characterized in that: The specific implementation steps in step S3 are as follows: 1) Determine the number of ants and ant lions M, N and the maximum number of iterations , initialize its position and calculate the corresponding fitness value; 2) Select the antlion with the smallest fitness value in the initialized antlion population as the elite antlion; 3) Use the roulette strategy to match ants with ant lions one by one, iteratively update the relevant values ​​according to the position of the ant lion, and make the ants randomly walk between the ant lion and the elite ant lion according to a certain rule, and finally take the average value as the ant position coordinate; 4) After each iteration, the fitness of the ant and ant lion is calculated respectively. When the position and fitness of the ant are known, the position of the ant lion is updated. The position with the minimum fitness is the new position of the ant lion after iteration. 5) Determine whether the iteration termination condition is met. If the current iteration number is met, stop the iteration and output {m, k}. Otherwise, repeat steps 3 to 5.

Images