Feature extraction method and system for digital substation equipment

Abstract

The invention discloses a feature extraction method for digital substation equipment. The method comprises the following steps: analyzing target digital substation equipment to confirm a fault diagnosis signal; obtaining a fault diagnosis signal and carrying out noise reduction and decomposition to obtain candidate intrinsic mode function components; calculating the sample entropy of the candidate intrinsic mode function component, and comparing the sample entropy with the sample entropy of the actual signal to obtain an intrinsic mode function component; and feature vectors in the intrinsic mode function components are extracted, dimensionality reduction, inspection and analysis are carried out, and the final features of the digital substation equipment are obtained. The invention also discloses a system for realizing the feature extraction method of the digital substation equipment. According to the method, noise reduction, decomposition, sample entropy calculation, comparison and dimension reduction analysis are carried out on the obtained transformer substation equipment signal, so that feature extraction of digital transformer substation equipment is realized, the reliability is higher, and the accuracy is better.

Description

Technical Field

[0001] The invention belongs to the field of electrical automation, and in particular relates to a feature extraction method and system for digital substation equipment. Background Art

[0002] With the development of economy and technology and the improvement of people's living standards, electricity has become an indispensable secondary energy source in people's production and life, bringing endless convenience to people's production and life. Therefore, ensuring the stable and reliable supply of electricity has become one of the most important tasks of the power system.

[0003] Substations are an important part of the power system. The number of equipment in a substation is very large, including primary equipment transformers, circuit breakers, disconnectors, current transformers, voltage transformers, GIS, switch cabinets, lightning arresters, etc.; secondary equipment measurement and control, protection, security control, intelligent fault recording devices, metering, AC and DC power supplies, etc. Therefore, it is particularly important to monitor the safety of substation equipment.

[0004] At present, digital substations have gradually become popular, and it is of great significance to monitor the equipment of digital substations. In the process of equipment monitoring, it is very important to extract the characteristics of the equipment. Traditional feature extraction methods include fast Fourier transform, wavelet transform, empirical mode decomposition, and ensemble empirical mode decomposition. Fast Fourier transform can obtain the spectral distribution of the signal, but it is not suitable for non-stationary signals; wavelet transform is constrained by the selection of wavelet functions and the number of decomposition layers; there is a mode aliasing problem in the decomposition process of empirical mode decomposition (EMD); although ensemble empirical mode decomposition (EEMD) improves the mode aliasing problem, there is a lot of noise in the signal. Summary of the invention

[0005] One of the purposes of the present invention is to provide a feature extraction method for digital substation equipment with high reliability and good accuracy.

[0006] A second object of the present invention is to provide a system for implementing the feature extraction method of the digital substation equipment.

[0007] The feature extraction method of the digital substation equipment provided by the present invention comprises the following steps:

[0008] S1. Analyze the target digital substation equipment and confirm the fault diagnosis signal;

[0009] S2. According to the result obtained in step S1, the fault diagnosis signal is obtained, and noise reduction and decomposition are performed to obtain candidate intrinsic mode function components;

[0010] S3. Calculate the sample entropy of the candidate intrinsic mode function component obtained in step S2, and compare it with the sample entropy of the actual signal to obtain the intrinsic mode function component;

[0011] S4. Extract the characteristic vectors in the intrinsic mode function components obtained in step S3, and perform dimension reduction, inspection and analysis to obtain the characteristics of the final digital substation equipment.

[0012] The step S1 of analyzing the target digital substation equipment and confirming the fault diagnosis signal specifically includes the following steps:

[0013] Analyze the target digital substation equipment and confirm the fault diagnosis signal;

[0014] The confirmed fault diagnosis signals include vibration signals, communication signals, image signals, sound signals, current signals and voltage signals.

[0015] The noise reduction described in step S2 specifically includes the following steps:

[0016] Use soft threshold denoising algorithm to reduce signal noise;

[0017] Select a wavelet basis to decompose the signal into components of different scales and frequencies;

[0018] Using the selected wavelet basis, the signal is decomposed into multiple scales and completed by continuous wavelet transform;

[0019] Perform threshold processing on the decomposed wavelet coefficients and retain the signal coefficients;

[0020] The threshold processing of the wavelet coefficients obtained by decomposition is specifically performed using the following formula:

[0021]

[0022] Where W j,k is the processed wavelet coefficient; sgn(a) is the sign function, a>0 then sgn(a)=1, a=0 then sgn(a)=0, a<0 then sgn(a)=-1; ω j,k is the j-th layer and k-th wavelet coefficient in discrete wavelet transform; λ is the set threshold; R is the threshold value, and σ is the standard deviation of the noise after decomposition at the jth layer, and N is the signal length.

[0023] The decomposition described in step S2 specifically includes the following steps:

[0024] The noise-reduced signal is decomposed using the variational mode decomposition method;

[0025] Construct a variational mode decomposition model, expressed as

[0026]

[0027]

[0028] Where u k (t) is the intrinsic mode function component after variational mode decomposition; ω k is the center frequency of the kth modal component; K is the preset number of decomposed modes; is the directional gradient of the modal component; δ(t) is the Dirichlet function; j is the imaginary unit; f(t) is the undetermined decomposition signal; ||||2 is the square root of the L2 norm gradient;

[0029] The decomposed intrinsic mode function components represent the frequency components in the original signal and satisfy:

[0030] (1) Within a data segment, the number of extreme points and the number of zero-crossing points must be equal or differ by no more than one;

[0031] (2) At any time, the average value of the upper envelope formed by the local maximum point and the lower envelope formed by the local minimum point is zero; the upper and lower envelopes are locally symmetrical with respect to the time axis;

[0032] The intrinsic mode function components obtained by decomposition are used as candidate intrinsic mode function components.

[0033] The sample entropy of the candidate intrinsic mode function component obtained in step S2 is calculated in step S3, and specifically includes the following steps:

[0034] Set the candidate eigenmode function component sequence to X i ={x i (1),x i (2),…,x i (N)},x i (N) is a time series point; the m-dimensional vector of the candidate intrinsic mode function component is set to X(i) = {x(i), x(i+1), ..., x(i+m-1)}, where x(i) is the time series component of the m-dimensional vector;

[0035] The threshold θ is set to θ = 0.3η, where η is the standard deviation of the sequence;

[0036] Set d|X(i),X(j)| to be the maximum distance between corresponding elements of X(i) and X(j);

[0037] Count the number of d|X(i),X(j)|<0 Num(g) and calculate the ratio for

[0038] Calculate the ratio The average value B m (θ) is

[0039] Increase the dimension m by 1 and calculate the average value B m+1 (θ);

[0040] Finally, the sample entropy SampEn is calculated as

[0041] The step S4 of extracting the characteristic vector in the intrinsic mode function component obtained in the step S3 specifically includes the following steps:

[0042] Extracting the time domain features and frequency domain features from the intrinsic mode function components obtained in step S3 and forming a feature vector;

[0043] Among them, the time domain features include maximum value, minimum value, mean value, peak-to-peak value, RMS value, RMS amplitude, standard deviation, kurtosis, skewness, pulse index, peak index and waveform index;

[0044] Frequency domain features include centroid frequency, root mean square frequency, average frequency and frequency difference.

[0045] The dimensionality reduction described in step S4 specifically includes the following steps:

[0046] The t-SNE method was used for dimensionality reduction;

[0047] Selecting parameters of the t-SNE method; the parameters include perplexity and learning rate;

[0048] Tuning: Select points in the low-dimensional space as initial values, and construct representations in the low-dimensional space based on the similarities between data points in the high-dimensional space;

[0049] Gaussian kernel function is used to calculate the similarity between data points;

[0050] Use the gradient descent optimization algorithm to map high-dimensional data to low-dimensional space, so that the data points in the low-dimensional space maintain as much similarity as possible with those in the high-dimensional space;

[0051] During the iteration process, the positions of points in the low-dimensional space are continuously adjusted until the convergence condition is reached and the dimensionality reduction is completed.

[0052] The inspection described in step S4 specifically includes the following steps:

[0053] A cross-validation scheme was used to evaluate the performance of dimensionality reduction.

[0054] The analysis described in step S4 specifically includes the following steps:

[0055] In the dimension reduction and inspection data, cluster analysis is used to form a low-dimensional feature clustering image and obtain the final characteristics of the digital substation equipment.

[0056] The present invention also provides a system for implementing the feature extraction method of the digital substation equipment, comprising an equipment analysis module, a noise reduction decomposition module, a component calculation module and a feature extraction module; the equipment analysis module, the noise reduction decomposition module, the component calculation module and the feature extraction module are connected in series in sequence; the equipment analysis module is used to analyze the target digital substation equipment, confirm the fault diagnosis signal, and upload the data information to the noise reduction decomposition module; the noise reduction decomposition module is used to obtain the fault diagnosis signal according to the received data information and the obtained results, and perform noise reduction and decomposition to obtain candidate intrinsic mode function components, and upload the data information to the component calculation module; the component calculation module is used to calculate the sample entropy of the candidate intrinsic mode function components according to the received data information, compare it with the sample entropy of the actual signal, obtain the intrinsic mode function components, and upload the data information to the feature extraction module; the feature extraction module is used to extract the characteristic vector in the intrinsic mode function components obtained according to the received data information, and perform dimensionality reduction, inspection and analysis to obtain the final characteristics of the digital substation equipment.

[0057] The feature extraction method and system for digital substation equipment provided by the present invention not only realizes the feature extraction of digital substation equipment, but also has higher reliability and better accuracy by performing noise reduction, decomposition, sample entropy calculation, comparison and dimensionality reduction analysis on the acquired substation equipment signals. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 The figure is a schematic diagram of the method flow of the present invention.

[0059] Figure 2 Schematic diagram of the functional modules of the system of the present invention. DETAILED DESCRIPTION

[0060] like Figure 1 The method flow chart of the method of the present invention is shown as follows: The feature extraction method of the digital substation equipment disclosed in the present invention comprises the following steps:

[0061] S1. Analyze the target digital substation equipment and confirm the fault diagnosis signal; specifically include the following steps:

[0062] Analyze the target digital substation equipment and confirm the fault diagnosis signal;

[0063] The confirmed fault diagnosis signals include vibration signals, communication signals, image signals, sound signals, current signals and voltage signals;

[0064] S2. According to the result obtained in step S1, the fault diagnosis signal is obtained, and noise reduction and decomposition are performed to obtain candidate intrinsic mode function components;

[0065] In specific implementation, noise reduction includes the following steps:

[0066] Use soft threshold denoising algorithm to reduce signal noise;

[0067] Select a wavelet basis to decompose the signal into components of different scales and frequencies;

[0068] Using the selected wavelet basis, the signal is decomposed into multiple scales and completed by continuous wavelet transform;

[0069] Perform threshold processing on the decomposed wavelet coefficients and retain the signal coefficients;

[0070] The threshold processing of the wavelet coefficients obtained by decomposition is specifically performed using the following formula:

[0071]

[0072] Where W j,k is the processed wavelet coefficient; sgn(a) is the sign function, a>0 then sgn(a)=1, a=0 then sgn(a)=0, a<0 then sgn(a)=-1; ω j,k is the j-th layer and k-th wavelet coefficient in discrete wavelet transform; λ is the set threshold, preferably R; R is the threshold value, and σ is the standard deviation of the noise after decomposition at the jth layer, and N is the signal length;

[0073] The decomposition includes the following steps:

[0074] The noise-reduced signal is decomposed using the variational mode decomposition method;

[0075] Construct a variational mode decomposition model, expressed as

[0076]

[0077]

[0078] Where u k (t) is the intrinsic mode function component after variational mode decomposition; ω k is the center frequency of the kth modal component; K is the preset number of decomposed modes; is the directional gradient of the modal component; δ(t) is the Dirichlet function; j is the imaginary unit; f(t) is the undetermined decomposition signal; ||||2 is the square root of the L2 norm gradient;

[0079] After determining the parameters of the variational mode decomposition model, the constructed variational mode decomposition model is used to iteratively solve and obtain the intrinsic mode function components;

[0080] The decomposed intrinsic mode function components represent the frequency components in the original signal and satisfy:

[0081] (1) Within a data segment, the number of extreme points and the number of zero-crossing points must be equal or differ by no more than one;

[0082] (2) At any time, the average value of the upper envelope formed by the local maximum point and the lower envelope formed by the local minimum point is zero; the upper and lower envelopes are locally symmetrical with respect to the time axis;

[0083] The intrinsic mode function components obtained by decomposition are used as candidate intrinsic mode function components;

[0084] S3. Calculate the sample entropy of the candidate intrinsic mode function component obtained in step S2, and compare it with the sample entropy of the actual signal to obtain the intrinsic mode function component; specifically comprising the following steps:

[0085] Sample entropy is a measure of the complexity of a sequence;

[0086] Set the candidate eigenmode function component sequence to X i ={x i (1),x i (2),…,x i (N)},x i (N) is a time series point; the m-dimensional vector of the candidate intrinsic mode function component is set to X(i) = {x(i), x(i+1), ..., x(i+m-1)}, where x(i) is the time series component of the m-dimensional vector;

[0087] The threshold θ is set to θ = 0.3η, where η is the standard deviation of the sequence;

[0088] Set d|X(i),X(j)| to be the maximum distance between corresponding elements of X(i) and X(j);

[0089] Count the number of d|X(i),X(j)|<0 Num(g) and calculate the ratio for

[0090] Calculate the ratio The average value B m (θ) is

[0091] Increase the dimension m by 1 and calculate the average value B m+1 (θ);

[0092] Finally, the sample entropy SampEn is calculated as

[0093] The calculated sample entropy SampEn is compared with the sample entropy threshold of the real signal to select the intrinsic mode function components that can present the characteristics of the actual signal; the selected intrinsic mode function components are subjected to feature extraction to form a high-dimensional feature data set;

[0094] S4. Extracting the eigenvectors in the intrinsic mode function components obtained in step S3, and performing dimensionality reduction, inspection and analysis to obtain the characteristics of the final digital substation equipment; specifically comprising the following steps:

[0095] Extracting the time domain features and frequency domain features from the intrinsic mode function components obtained in step S3 and forming a feature vector;

[0096] Among them, the time domain features include maximum value, minimum value, mean value, peak-to-peak value, RMS value, RMS amplitude, standard deviation, kurtosis, skewness, pulse index, peak index and waveform index;

[0097] Frequency domain features include centroid frequency, root mean square frequency, average frequency, and frequency difference;

[0098] The t-SNE method was used for dimensionality reduction;

[0099] Selecting parameters of the t-SNE method; the parameters include perplexity and learning rate;

[0100] Tuning: Select points in the low-dimensional space as initial values, and construct representations in the low-dimensional space based on the similarities between data points in the high-dimensional space;

[0101] Gaussian kernel function is used to calculate the similarity between data points;

[0102] Use the gradient descent optimization algorithm to map high-dimensional data to low-dimensional space, so that the data points in the low-dimensional space maintain as much similarity as possible with those in the high-dimensional space;

[0103] In the iterative process, the positions of the points in the low-dimensional space are continuously adjusted until the convergence condition is reached and the dimensionality reduction is completed;

[0104] Use a cross-validation scheme to evaluate the performance of dimensionality reduction;

[0105] In the dimension reduction and inspection data, cluster analysis is used to form a low-dimensional feature clustering image and obtain the final characteristics of the digital substation equipment.

[0106] The method of the present invention can be applied to various types of equipment in a digital substation, including primary equipment: transformers, circuit breakers, disconnectors, current transformers, voltage transformers, GIS, switch cabinets, lightning arresters; secondary equipment: measurement and control equipment, protection equipment, security control equipment, intelligent fault recording devices, metering equipment, AC and DC power supplies, and other equipment.

[0107] like Figure 2 The functional module schematic diagram of the system of the present invention is shown as follows: the system for realizing the feature extraction method of the digital substation equipment disclosed in the present invention comprises an equipment analysis module, a noise reduction decomposition module, a component calculation module and a feature extraction module; the equipment analysis module, the noise reduction decomposition module, the component calculation module and the feature extraction module are connected in series in sequence; the equipment analysis module is used to analyze the target digital substation equipment, confirm the fault diagnosis signal, and upload the data information to the noise reduction decomposition module; the noise reduction decomposition module is used to obtain the fault diagnosis signal according to the received data information and the obtained result, and perform noise reduction and decomposition to obtain the candidate intrinsic mode function component, and upload the data information to the component calculation module; the component calculation module is used to calculate the sample entropy of the candidate intrinsic mode function component according to the received data information, compare it with the sample entropy of the actual signal, obtain the intrinsic mode function component, and upload the data information to the feature extraction module; the feature extraction module is used to extract the characteristic vector in the obtained intrinsic mode function component according to the received data information, and perform dimensionality reduction, inspection and analysis to obtain the final characteristics of the digital substation equipment.

Claims

1. A feature extraction method for digital substation equipment, comprising the following steps: S1. Analyze the target digital substation equipment and confirm the fault diagnosis signal; S2. According to the result obtained in step S1, the fault diagnosis signal is obtained, and noise reduction and decomposition are performed to obtain candidate intrinsic mode function components; S3. Calculate the sample entropy of the candidate intrinsic mode function component obtained in step S2, and compare it with the sample entropy of the actual signal to obtain the intrinsic mode function component; S4. Extract the characteristic vectors in the intrinsic mode function components obtained in step S3, and perform dimension reduction, inspection and analysis to obtain the characteristics of the final digital substation equipment.

2. The feature extraction method of digital substation equipment according to claim 1 is characterized in that The step S1 of analyzing the target digital substation equipment and confirming the fault diagnosis signal specifically includes the following steps: Analyze the target digital substation equipment and confirm the fault diagnosis signal; The confirmed fault diagnosis signals include vibration signals, communication signals, image signals, sound signals, current signals and voltage signals.

3. The feature extraction method of digital substation equipment according to claim 2 is characterized in that The noise reduction described in step S2 specifically includes the following steps: Use soft threshold denoising algorithm to reduce signal noise; Select a wavelet basis to decompose the signal into components of different scales and frequencies; Using the selected wavelet basis, the signal is decomposed into multiple scales and completed by continuous wavelet transform; Perform threshold processing on the decomposed wavelet coefficients and retain the signal coefficients; The threshold processing of the wavelet coefficients obtained by decomposition is specifically performed using the following formula: Where W j,k is the processed wavelet coefficient; sgn(a) is the sign function, a>0 then sgn(a)=1, a=0 then sgn(a)=0, a<0 then sgn(a)=-1; ω j,k is the j-th layer and k-th wavelet coefficient in discrete wavelet transform; λ is the set threshold; R is the threshold value, and σ is the standard deviation of the noise after decomposition at the jth layer, and N is the signal length.

4. The feature extraction method of digital substation equipment according to claim 3 is characterized in that The decomposition described in step S2 specifically includes the following steps: The noise-reduced signal is decomposed using the variational mode decomposition method; Construct a variational mode decomposition model, expressed as Where u k (t) is the intrinsic mode function component after variational mode decomposition; ω k is the center frequency of the kth modal component; K is the preset number of decomposed modes; is the directional gradient of the modal component; δ(t) is the Dirichlet function; j is the imaginary unit; f(t) is the undetermined decomposition signal; ||||2 is the square root of the L2 norm gradient; The decomposed intrinsic mode function components represent the frequency components in the original signal and satisfy: (1) Within a data segment, the number of extreme points and the number of zero-crossing points must be equal or differ by no more than one; (2) At any time, the average value of the upper envelope formed by the local maximum point and the lower envelope formed by the local minimum point is zero; the upper and lower envelopes are locally symmetrical with respect to the time axis; The intrinsic mode function components obtained by decomposition are used as candidate intrinsic mode function components.

5. The feature extraction method for digital substation equipment according to claim 4 is characterized in that The sample entropy of the candidate intrinsic mode function component obtained in step S2 is calculated in step S3, and specifically includes the following steps: Set the candidate eigenmode function component sequence to X i ={x i (1),x i (2),…,x i (N)},x i (N) is a time series point; The m-dimensional vector of the candidate intrinsic mode function component is set to X(i) = {x(i), x(i+1), ..., x(i+m-1)}, where x(i) is the time series component of the m-dimensional vector; The threshold θ is set to θ = 0.3η, where η is the standard deviation of the sequence; Set d|X(i),X(j)| to be the maximum distance between corresponding elements of X(i) and X(j); Count the number of d|X(i),X(j)|<0 Num(g) and calculate the ratio B i m (θ) is Calculate the ratio The average value B m (θ) is Increase the dimension m by 1 and calculate the average value B m+1 (θ); Finally, the sample entropy SampEn is calculated as 6. The feature extraction method of digital substation equipment according to claim 5 is characterized in that The step S4 of extracting the characteristic vector in the intrinsic mode function component obtained in the step S3 specifically includes the following steps: Extracting the time domain features and frequency domain features from the intrinsic mode function components obtained in step S3 and forming a feature vector; Among them, the time domain features include maximum value, minimum value, mean value, peak-to-peak value, RMS value, RMS amplitude, standard deviation, kurtosis, skewness, pulse index, peak index and waveform index; Frequency domain features include centroid frequency, root mean square frequency, average frequency and frequency difference.

7. The feature extraction method of digital substation equipment according to claim 6 is characterized in that The dimensionality reduction described in step S4 specifically includes the following steps: The t-SNE method was used for dimensionality reduction; Selecting parameters of the t-SNE method; the parameters include perplexity and learning rate; Tuning: Select points in the low-dimensional space as initial values, and construct representations in the low-dimensional space based on the similarities between data points in the high-dimensional space; Gaussian kernel function is used to calculate the similarity between data points; Use the gradient descent optimization algorithm to map high-dimensional data to low-dimensional space, so that the data points in the low-dimensional space maintain as much similarity as possible with those in the high-dimensional space; During the iteration process, the positions of points in the low-dimensional space are continuously adjusted until the convergence condition is reached and the dimensionality reduction is completed.

8. The feature extraction method for digital substation equipment according to claim 7 is characterized in that The inspection described in step S4 specifically includes the following steps: A cross-validation scheme was used to evaluate the performance of dimensionality reduction.

9. The feature extraction method for digital substation equipment according to claim 8 is characterized in that The analysis described in step S4 specifically includes the following steps: In the dimension reduction and inspection data, cluster analysis is used to form a low-dimensional feature clustering image and obtain the final characteristics of the digital substation equipment.

10. A system for implementing the feature extraction method of digital substation equipment according to any one of claims 1 to 9, characterized in that It includes an equipment analysis module, a noise reduction decomposition module, a component calculation module and a feature extraction module; the equipment analysis module, the noise reduction decomposition module, the component calculation module and the feature extraction module are connected in series in sequence; the equipment analysis module is used to analyze the target digital substation equipment, confirm the fault diagnosis signal, and upload the data information to the noise reduction decomposition module; The noise reduction and decomposition module is used to obtain the fault diagnosis signal according to the received data information and the obtained results, and perform noise reduction and decomposition to obtain the candidate intrinsic mode function components, and upload the data information to the component calculation module; The component calculation module is used to calculate the sample entropy of the candidate intrinsic mode function component according to the received data information, compare it with the sample entropy of the actual signal, obtain the intrinsic mode function component, and upload the data information to the feature extraction module; The feature extraction module is used to extract the feature vectors in the intrinsic mode function components according to the received data information, and perform dimension reduction, inspection and analysis to obtain the final features of the digital substation equipment.

Images