Electric arc detection method based on high-precision sampling circuit and empirical mode decomposition

Abstract

The invention provides an arc detection method based on a high-precision sampling circuit and empirical mode decomposition, and belongs to the field of fault signal sampling and detection.The designed high-precision sampling circuit is adopted for sampling arc fault signals, empirical mode decomposition is conducted on the obtained sampling signals, and the sampling signals are decomposed into a series of intrinsic mode functions; and analyzing the intrinsic mode functions, extracting characteristic values, and accurately judging whether the fault arc occurs or not according to a comparison result of a plurality of characteristic values and corresponding threshold values. Compared with the prior art, the arc fault detection method has higher adaptability and transient characteristic capture capability, is more convenient for analysis and processing of nonlinear and non-stable arc fault signals, is simple in detection and judgment method, is convenient to popularize and is wide in application range, and the extracted characteristic values are only required to be compared with a set threshold value.

Description

Technical Field

[0001] The invention relates to an arc detection method based on a high-precision sampling circuit and empirical mode decomposition, and belongs to the field of fault signal sampling and detection. Background Art

[0002] With the rapid development of the national economy, the rapid progress of power electronics and new energy technologies, the electrification level of the whole society is getting higher and higher, the popularity of new energy vehicles and their charging facilities is also getting higher and higher, the frequency of electrical fires and electric vehicle fires is getting higher and higher, and the proportion of fires in the country is increasing. The loss of life and economy is also difficult to estimate. Among them, fault arc is one of the main causes of electrical fires and electric vehicle power battery fires. The collection and detection of fault arc is the primary problem faced by the field of fault arc detection. At present, traditional arc fault detection technology mainly relies on overcurrent or overvoltage protection devices, but these methods have problems such as insufficient sensitivity, high false alarm rate and inability to effectively identify early weak arc signals, which makes it difficult to meet the requirements of modern power systems for fast and accurate detection of fault arc.

[0003] Existing feature extraction and signal analysis methods, such as the short-time Fourier transform (STFT) method, cannot analyze the sudden change signals of the waveform, and local detail features are easily ignored. At the same time, due to the complex current frequency components of the fault arc, the detection effect is limited, and the false alarm and missed alarm rates are high, which makes it difficult to meet the requirements of electrical fires and battery fires for fast and accurate detection of fault arcs. Summary of the invention

[0004] In view of the above problems, the purpose of the present invention is to provide an arc detection method based on a high-precision sampling circuit and empirical mode decomposition, which adopts a designed high-precision sampling circuit to sample arc fault signals, performs empirical mode decomposition on the obtained sampled signals, decomposes them into a series of intrinsic mode functions, and analyzes these intrinsic mode functions to extract eigenvalues. By comparing the results of multiple eigenvalues ​​with corresponding thresholds, it is accurately determined whether a fault arc occurs.

[0005] In order to realize the above technical features, the object of the present invention is achieved as follows: an arc detection method based on a high-precision sampling circuit and empirical mode decomposition comprises the following steps: Step 1, signal acquisition: Establish power supply and different load connection lines under standard working conditions, simulate signal waveforms under various abnormal conditions, set sampling points in the circuit and use high-precision sampling circuits to collect current or voltage signals when arc faults occur; Step 2, signal preprocessing: Preprocess the amplified and denoised arc fault signal to remove the linear or nonlinear trend in the signal, make the signal fluctuate around the zero mean, and then normalize the signal amplitude to unify the signal amplitude range; Step 3, perform empirical mode decomposition (EMD): Apply the EMD algorithm to the preprocessed signal to decompose it into a series of intrinsic mode functions (IMFs) and residuals, where each IMF represents a component in a specific frequency band of the signal; Step 4, Identify IMFs: Analyze the obtained IMFs, perform fast Fourier transform (FFT) on each IMF to obtain the corresponding spectrum, and identify the specific IMFs related to the arc fault through the high-frequency components and peaks in the specific frequency band in the spectrum; Step 5, feature extraction: Extract key eigenvalues ​​from the identified IMFs, including energy entropy , Number of singular points , Kurtosis ; Step 6, arc fault judgment: The extracted eigenvalues ​​are compared with the set thresholds, and the corresponding error threshold coefficient is set for each eigenvalue. The corresponding threshold of energy entropy is , the error threshold coefficient is ; The corresponding threshold value of the number of singular points is ; The error threshold coefficient is ; The corresponding threshold of kurtosis is k , the error threshold coefficient is ;when , , When the three conditions are met at the same time, it is determined that an arc fault has occurred.

[0006] The high-precision sampling circuit in step 1 uses a low-noise amplifier chip to amplify the arc signal. The external resistor in the high-precision sampling circuit Control signal amplification factor ; ; A filter chip is selected for filtering, and the two amplifiers inside the filter chip are used for high-pass filtering and low-pass filtering respectively; the effective frequency distribution of the arc signal is 40kHz~100kHz.

[0007] The high-pass filtering is performed by setting appropriate capacitor and resistor values ​​in the high-precision sampling circuit so that signals below 40kHz are filtered out and signals above 40kHz are passed; The low-pass filtering means that by setting appropriate capacitor and resistor values, the signals higher than 100 kHz are filtered out, while the signals lower than 100 kHz are passed, and unnecessary noise and interference are removed.

[0008] In step 2, the signal Detrending gives: ; Then apply the filter chip to denoise to reduce the noise interference in the signal, and then normalize the signal amplitude to unify the signal amplitude range, and get: ; in, and are the mean and standard deviation respectively. After the above preprocessing, the normalized signal is obtained. .

[0009] In step 4, IMFs are the basic units of EMD decomposition and have two characteristics: local symmetry and zero mean. That is, at any time, the number of local maximum points and local minimum points of IMF differs by at most 1. At any time, the local mean of IMF is , that is, the mean of the upper envelope and the lower envelope is 0. Based on the above two properties, the signal is decomposed by the EMD algorithm, and then screened to extract the IMF.

[0010] The specific extraction process of IMF: Find extreme points: determine all local maximum and local minimum points in the signal; Generate envelope: Use spline interpolation to connect all maximum and minimum points to get the upper envelope and lower envelope ; Calculate mean: Calculate the mean of the upper and lower envelopes ; Eliminate the mean: get a new signal ; Determining IMF conditions: Check Does it meet the IMF definition? If so, extract As an IMF; otherwise, As a new signal, repeat the above steps; Iterative decomposition: remove the extracted IMF from the original signal to obtain the residual signal Continue to perform EMD decomposition on the residual signal until the residual signal is a monotonic function or close to zero.

[0011] In step 5, the calculation formula of energy entropy is: ; In the formula, It is The energy of an IMF component accounts for the percentage of the total signal energy. , is the total number of IMFs.

[0012] In step 5, the number of singular points is obtained by summing the first-order derivatives of IMFs, and the kurtosis is obtained by the following calculation formula: .

[0013] The present invention has the following beneficial effects: The present invention provides an arc detection method based on a high-precision sampling circuit and empirical mode decomposition. The arc fault signal is sampled by the high-precision sampling circuit, the obtained sampled signal is detrended and normalized, the pre-processed signal is subjected to empirical mode decomposition, and the intrinsic mode function obtained after the decomposition is analyzed to extract the characteristic value, and the obtained characteristic value is compared with a set threshold value to judge whether an arc fault occurs. The method uses a high-precision sampling circuit based on a chip INA163UA and a chip ADA4898 to sample the fault arc signal, which has higher accuracy and precision than an ordinary sampling circuit, and also uses empirical mode decomposition for signal processing. Compared with ordinary processing methods such as wavelet transform, the method has higher adaptability and transient feature capture ability, and is more convenient for analyzing and processing such nonlinear and non-stationary arc fault signals. The detection and judgment method is simple, and only the extracted characteristic value needs to be compared with the set threshold value, which is easy to promote and has a wide range of applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0015] Figure 1 It is a schematic diagram of the high-precision sampling circuit of the present invention.

[0016] Figure 2 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0017] The embodiments of the present invention are further described below in conjunction with the accompanying drawings.

[0018] like Figure 2 As shown, an arc detection method based on a high-precision sampling circuit and empirical mode decomposition, the detection steps include: Step 1: Signal acquisition: establish power supply and different load connection lines under standard working conditions, simulate signal waveforms under various abnormal conditions, set sampling points in the circuit and use high-precision sampling circuits to collect current or voltage signals when arc faults occur; Step 2: Signal preprocessing: preprocess the amplified and denoised arc fault signal to remove the linear or nonlinear trend in the signal and make the signal fluctuate around the zero mean. Then normalize the signal amplitude to unify the signal amplitude range; Step 3: Perform empirical mode decomposition (EMD) and apply the EMD algorithm to the preprocessed signal to decompose it into a series of intrinsic mode functions (IMFs) and residuals. Each IMF represents a component of a specific frequency band of the signal; Step 4: Identify IMFs, analyze the obtained IMFs, and perform fast Fourier transform (FFT) on each IMF to obtain their spectrum. Specific IMFs related to arc faults are identified through high-frequency components in the spectrum and peaks in specific frequency bands; Step 5: Feature extraction: extract key feature values ​​from the identified IMFs, including energy entropy values , Number of singular points , Kurtosis ; Step 6, arc fault judgment, compare the extracted eigenvalue with the set threshold, and set the corresponding error threshold coefficient for each eigenvalue. The energy entropy corresponding threshold is , the error threshold coefficient is ; The corresponding threshold value of the number of singular points is ; The error threshold coefficient is ; The corresponding threshold of kurtosis is k , the error threshold coefficient is ;when , , When the three conditions are met at the same time, it is determined that an arc fault has occurred.

[0019] Furthermore, the high-precision sampling circuit in step 1 uses a low-noise amplifier chip INA163UA to amplify the arc signal. The schematic diagram of the amplifier circuit is shown in Figure 1 . External resistors in high-precision sampling circuits Control signal amplification factor ; ; The chip ADA4898 is selected for filtering. The two amplifiers inside the chip are used for high-pass filtering and low-pass filtering respectively. The effective frequency of the arc signal is distributed between 40kHz and 100kHz, so the high-pass filter cutoff frequency is set at about 40kHz and the low-pass filter cutoff frequency is set at about 100kHz. and are low-pass cutoff frequency and high-pass cutoff frequency respectively, calculate Figure 1 The corresponding capacitance C 7. C8. C 11 , C 12 and resistor R 4. R 5. R 6. R 7, namely: ; ; After passing through the above high-precision circuit, the signal is obtained .

[0020] In step 1, high-pass filtering means setting appropriate capacitor and resistor values ​​so that signals below 40kHz are filtered out and signals above 40kHz are passed. Low-pass filtering means setting appropriate capacitor and resistor values ​​so that signals above 100kHz are filtered out and signals below 100kHz are passed. Therefore, the combination of high-pass filtering and low-pass filtering can retain the effective frequency range of the arc signal (40kHz~100kHz) while removing unnecessary noise and interference.

[0021] In step 1, the accuracy and precision of the current or voltage signal collected when an arc fault occurs through a high-precision sampling circuit are higher, which is more convenient for analyzing and processing nonlinear and non-stationary arc fault signals.

[0022] In the above step 2, in order to improve the signal-to-noise ratio of the signal, eliminate noise interference, unify the dimension and range of the signal, and provide an accurate and reliable signal basis for subsequent feature extraction and analysis, the collected arc signal is pre-processed by amplification, denoising, normalization, etc.

[0023] In step 2, the linear or nonlinear trend in the signal may interfere with the accurate analysis of the arc fault signal. By removing the trend, the signal can be made more stable, which is convenient for the subsequent empirical mode decomposition and feature extraction steps, so as to more accurately determine whether the arc fault occurs. Detrending gives: ; Then apply Figure 1 The filter in denoising is used to reduce the noise interference in the signal, and then the signal amplitude is normalized to unify the amplitude range of the signal to obtain: ; in, and are the mean and standard deviation respectively. After the above preprocessing, the normalized signal is obtained. .

[0024] In step 2, normalization eliminates the amplitude difference of the signal, making the signal comparable at different times and under different conditions. Detrending, denoising and normalization preprocessing can reduce false positives and false negatives. Therefore, the signal obtained after the above preprocessing can more accurately and comprehensively cover arc detection under different conditions.

[0025] In step 3 above, the EMD algorithm is a data-driven decomposition method. Its basic principle is to decompose complex signals into a series of simple oscillation modes (IMFs) and a low-frequency residual signal. Each IMF represents a specific frequency component in the signal, allowing us to more clearly analyze the different frequencies and dynamic changes of the signal.

[0026] In step 3, the reason for using the EMD algorithm is that it is adaptive. The EMD method completely relies on the characteristics of the signal itself for decomposition, and does not require any preset basis functions, which makes EMD highly adaptable when processing complex nonlinear and non-stationary signals. The time domain local characteristics of the EMD algorithm enable it to capture the local characteristics of the signal in the time domain, and each IMF corresponds to a different frequency component of the signal. This locality makes EMD particularly suitable for analyzing instantaneous changes and mutations in signals. In addition, the EMD algorithm does not require a priori basis functions. Unlike traditional Fourier transforms or wavelet transforms, EMD does not rely on any fixed basis functions, such as sine waves or wavelet bases. EMD adaptively constructs basis functions in a data-driven manner, so that the decomposition results are more in line with the actual characteristics of the signal.

[0027] In step 4 above, IMFs are the basic units of EMD decomposition and have two characteristics: local symmetry and zero mean. That is, at any time, the number of local maximum and local minimum points of IMF differs by at most 1. At any time, the local mean of IMF is , that is, the mean of the upper envelope and the lower envelope is 0. Based on the above two properties, the signal is decomposed by the EMD algorithm, and then screened to extract the IMF.

[0028] The specific extraction process of IMF: Find extreme points: determine all local maximum and local minimum points in the signal; Generate envelope: Use spline interpolation to connect all maximum and minimum points to get the upper envelope and lower envelope ; Calculate mean: Calculate the mean of the upper and lower envelopes ; Eliminate the mean: get a new signal ; Determining IMF conditions: Check Does it meet the IMF definition? If so, extract As an IMF; otherwise, As a new signal, repeat the above steps; Iterative decomposition: remove the extracted IMF from the original signal to obtain the residual signal Continue to perform EMD decomposition on the residual signal until the residual signal is a monotonic function or close to zero.

[0029] In step 4, IMFs are converted from the time domain to the frequency domain through fast Fourier transform (FFT), and the spectrum characteristics of IMFs can be clearly seen. On the spectrum graph, the energy distribution of IMFs can be observed, thereby extracting the key feature of energy entropy. The shape and characteristics of the spectrum graph can also help extract the number of singular points and kurtosis eigenvalues, which are crucial for subsequent judgment of arc faults.

[0030] In step 5, the calculation formula of energy entropy is: ; In the formula, It is The energy of an IMF component accounts for the percentage of the total signal energy. , is the total number of IMFs.

[0031] In step 5, the number of singular points is obtained by summing the first-order derivatives of IMFs, and the kurtosis is obtained by the following calculation formula: .

[0032] In step 5, the size of the energy entropy value reflects the distribution of different frequency components in the signal. When the energy of a certain frequency component in the signal accounts for a large proportion, the energy entropy value will decrease accordingly; and when the energy distribution of each frequency component in the signal is relatively uniform, the energy entropy value will increase. This feature is of great significance for judging arc faults, because arc faults often cause changes in the energy of specific frequency components in the signal. By calculating the energy entropy value, this change can be quantified, thereby assisting in judging whether an arc fault occurs or not. Singular points usually represent mutations or abnormal points in the signal, and their number can reflect the instability and complexity of the signal. In arc fault detection, an increase in the number of singular points may mean the occurrence of an arc fault. The size of the kurtosis value can reflect the impact component or abnormal fluctuation in the signal, and the change in kurtosis may indicate the occurrence of an arc fault, because arc faults are often accompanied by mutations or abnormal fluctuations in the signal.

[0033] The principle of arc fault detection in step 6 is: through a large amount of experimental data analysis and experience accumulation, a threshold that can reflect the characteristics of arc fault is determined, and it is used as the set threshold, compared with the extracted characteristic value, and the corresponding error threshold coefficient is set for each characteristic value. Through experiments, the energy entropy corresponding threshold is , the error threshold coefficient is ;The corresponding threshold value of the number of singular points is , the error threshold coefficient is ; Kurtosis corresponding threshold is , the error threshold coefficient is ; In the case of a fault arc, the extracted eigenvalue energy entropy Greater than and The product of, the number of eigenvalue singular points Greater than and The product of, eigenvalue, kurtosis Greater than and By comparing, if and only if the above three conditions are met at the same time, it can be determined that the arc signal has a fault.

[0034] like Figure 2 As shown in the figure, in the whole process of arc fault detection, step six is ​​the key link to determine whether an arc fault occurs. This step is based on the arc fault signal collected by the high-precision sampling circuit, and extracts eigenvalues ​​from the intrinsic mode functions (IMFs) obtained after signal preprocessing and empirical mode decomposition (EMD). The eigenvalues ​​mainly include energy entropy, the number of singular points and kurtosis, which can reflect the characteristics of the arc fault signal. In order to improve the accuracy and robustness of detection, a corresponding error threshold coefficient is set for each eigenvalue. An arc fault is determined to have occurred only when and only when the three eigenvalues ​​of energy entropy, number of singular points and kurtosis meet their corresponding threshold conditions (that is, they all fall within the allowable range of the set threshold, considering the error threshold coefficient). This method makes full use of the advantages of high-precision sampling circuits and empirical mode decomposition, can accurately capture the transient characteristics of arc fault signals, and realize effective detection of arc faults.

Claims

1. An arc detection method based on high-precision sampling circuit and empirical mode decomposition, characterized in that: The following steps are involved: Step 1, signal acquisition: Establish power supply and different load connection lines under standard working conditions, simulate signal waveforms under various abnormal conditions, set sampling points in the circuit and use high-precision sampling circuits to collect current or voltage signals when arc faults occur; Step 2, signal preprocessing: Preprocess the amplified and denoised arc fault signal to remove the linear or nonlinear trend in the signal, make the signal fluctuate around the zero mean, and then normalize the signal amplitude to unify the signal amplitude range; Step 3, perform empirical mode decomposition (EMD): Apply the EMD algorithm to the preprocessed signal to decompose it into a series of intrinsic mode functions (IMFs) and residuals, where each IMF represents a component in a specific frequency band of the signal; Step 4, Identify IMFs: Analyze the obtained IMFs, perform fast Fourier transform (FFT) on each IMF to obtain the corresponding spectrum, and identify the specific IMFs related to the arc fault through the high-frequency components and peaks in the specific frequency band in the spectrum; Step 5, feature extraction: Extract key eigenvalues ​​from the identified IMFs, including energy entropy , Number of singular points , Kurtosis ; Step 6, arc fault judgment: The extracted eigenvalues ​​are compared with the set thresholds, and the corresponding error threshold coefficient is set for each eigenvalue. The corresponding threshold of energy entropy is , the error threshold coefficient is ; The corresponding threshold value of the number of singular points is ; The error threshold coefficient is ; The corresponding threshold of kurtosis is k , the error threshold coefficient is ;when , , When the three conditions are met at the same time, it is determined that an arc fault has occurred.

2. The arc detection method based on high-precision sampling circuit and empirical mode decomposition according to claim 1, characterized in that: The high-precision sampling circuit in step 1 uses a low-noise amplifier chip to amplify the arc signal. The external resistor in the high-precision sampling circuit Control signal amplification factor ; ； A filter chip is selected for filtering, and the two amplifiers inside the filter chip are used for high-pass filtering and low-pass filtering respectively; the effective frequency distribution of the arc signal is 40kHz~100kHz.

3. The arc detection method based on high-precision sampling circuit and empirical mode decomposition according to claim 2, characterized in that: The high-pass filtering is performed by setting appropriate capacitor and resistor values ​​in the high-precision sampling circuit so that signals below 40kHz are filtered out and signals above 40kHz are passed; The low-pass filtering means that by setting appropriate capacitor and resistor values, the signals higher than 100 kHz are filtered out, while the signals lower than 100 kHz are passed, and unnecessary noise and interference are removed.

4. The arc detection method based on high-precision sampling circuit and empirical mode decomposition according to claim 2, characterized in that: In step 2, the signal Detrending gives: ； Then apply the filter chip to denoise to reduce the noise interference in the signal, and then normalize the signal amplitude to unify the signal amplitude range, and get: ； in, and are the mean and standard deviation respectively. After the above preprocessing, the normalized signal is obtained. .

5. The arc detection method based on high-precision sampling circuit and empirical mode decomposition according to claim 2, characterized in that: In step 4, IMFs are the basic units of EMD decomposition and have two characteristics: local symmetry and zero mean. That is, at any time, the number of local maximum points and local minimum points of IMF differs by at most 1. At any time, the local mean of IMF is , that is, the mean of the upper envelope and the lower envelope is 0. Based on the above two properties, the signal is decomposed by the EMD algorithm, and then screened to extract the IMF.

6. The arc detection method based on high-precision sampling circuit and empirical mode decomposition according to claim 5, characterized in that: The specific extraction process of IMF: Find extreme points: determine all local maximum and local minimum points in the signal; Generate envelope: Use spline interpolation to connect all maximum and minimum points to get the upper envelope and lower envelope ; Calculate mean: Calculate the mean of the upper and lower envelopes ; Eliminate the mean: get a new signal ; Determining IMF conditions: Check Does it meet the IMF definition? If so, extract As an IMF; otherwise, As a new signal, repeat the above steps; Iterative decomposition: remove the extracted IMF from the original signal to obtain the residual signal Continue to perform EMD decomposition on the residual signal until the residual signal is a monotonic function or close to zero.

7. The arc detection method based on high-precision sampling circuit and empirical mode decomposition according to claim 6, characterized in that: In step 5, the calculation formula of energy entropy is: ； In the formula, It is The energy of an IMF component accounts for the percentage of the total signal energy. , is the total number of IMFs.

8. The arc detection method based on high-precision sampling circuit and empirical mode decomposition according to claim 6, characterized in that: In step 5, the number of singular points is obtained by summing the first-order derivatives of IMFs, and the kurtosis is obtained by the following calculation formula: 。

Images