Photovoltaic arc fault detection method and system based on singular value decomposition and wavelet transform

Abstract

The invention discloses a photovoltaic arc fault detection method and system based on singular value decomposition and wavelet transform, and relates to the technical field of photovoltaic power generation system arc fault detection, and the method comprises the steps: collecting current data, carrying out the filtering processing, and constructing a Hankel matrix for the filtered current data; performing singular value decomposition on the Hankel matrix to obtain a singular value sequence; constructing a differential spectrum by using the singular value sequence, carrying out de-noising processing, and carrying out SVD reconstruction on the de-noised fault current data; and performing wavelet transform based on the reconstructed current data, extracting a wavelet detail coefficient, and when the wavelet detail coefficient continuously exceeds an alarm threshold, judging an arc fault and triggering an alarm. According to the method, the singular value decomposition and wavelet transform methods are utilized, the fault current detection precision is improved, the sensitivity is high, the omission ratio is low, and the anti-interference performance is high.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation system arc fault detection, and in particular to a photovoltaic arc fault detection method and system based on singular value decomposition and wavelet transform. Background Art

[0002] In recent years, with strong national support for photovoltaic (PV) renewable energy, the penetration of PV power generation installations has significantly increased, and power generation has steadily increased. However, the widespread deployment of PV solar panels has also led to safety concerns, with PV arc faults being the most serious. PV panels contain numerous contacts between chips, between chips and guide frames, and between wiring and junction boxes. Over time, aging can lead to poor contact, making arc faults very likely to occur. Since PV power generation systems lack zero current crossings, arcs are difficult to extinguish on their own. If these arc faults are not detected promptly, they can easily develop into fires.

[0003] When an arc fault occurs in a photovoltaic power generation system, the system circuit is equivalent to a nonlinear resistor connected in series, causing the circuit voltage to increase and the current to decrease. This makes traditional low-voltage circuit breakers and fuses unable to detect and protect against the arc. Research on DC arc fault detection has been extensive in areas such as aircraft, electric vehicles, and DC low-voltage distribution cabinets. However, in photovoltaic power generation systems, arc fault detection is extremely difficult due to environmental influences such as light intensity, temperature, and high-frequency noise from inverters. Common photovoltaic arc fault detection methods include using the solitary light noise generated by the arc and using time-domain or frequency-domain methods to detect the current in the fault circuit. Traditional time-frequency domain detection methods suffer from low detection accuracy and high false operation rates, making them particularly ineffective in detecting photovoltaic micro-arcing. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is that the existing DC arc fault detection method has low detection accuracy, high false operation rate, high missed detection rate, and how to improve the operation efficiency.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: a photovoltaic arc fault detection method based on singular value decomposition and wavelet transform, comprising collecting current data for filtering processing, constructing a Hankel matrix for the filtered current data; performing singular value decomposition on the Hankel matrix to obtain a singular value sequence; constructing a differential spectrum using the singular value sequence, performing denoising processing, and performing SVD reconstruction on the denoised fault current data; performing wavelet transform based on the reconstructed current data to extract wavelet detail coefficients, and when the wavelet detail coefficients continuously exceed the alarm threshold, determining an arc fault and triggering an alarm; filtering includes using a high-pass filter to filter out low-frequency components of the collected current data and retaining high-frequency components.

[0007] As a preferred solution of the photovoltaic arc fault detection method based on singular value decomposition and wavelet transform described in the present invention, the construction of the Hankel matrix includes arranging the collected current data in sequence, the rows of the matrix are composed of continuous signal values, the data of adjacent rows slide backward in sequence, the number of matrix rows is determined by the length of the sampled data and the number of columns, and the time series characteristics of the current data are retained.

[0008] As a preferred solution of the photovoltaic arc fault detection method based on singular value decomposition and wavelet transform described in the present invention, the obtaining of the singular value sequence includes decomposing the obtained Hankel matrix into a left singular vector matrix, a right singular vector matrix and a singular value matrix, and obtaining the singular value sequence by taking the square root of the eigenvalue of the product of the decomposition matrix.

[0009] As a preferred solution of the photovoltaic arc fault detection method based on singular value decomposition and wavelet transform described in the present invention, the left singular vector matrix and the right singular vector matrix include: the left singular vector matrix is ​​composed of the eigenvectors of the product of the transposed matrix of the Hankel matrix and the singular value matrix, showing the row space characteristics of the current data.

[0010] The right singular vector matrix is ​​composed of the eigenvectors of the product of the transposed Hankel matrix and the original matrix, which shows the column space characteristics of the current data.

[0011] As a preferred embodiment of the photovoltaic arc fault detection method based on singular value decomposition and wavelet transform described in the present invention, the differential spectrum construction includes performing difference processing on adjacent singular values ​​in a singular value sequence to obtain a differential sequence of the singular value change rate, obtaining a singular value sequence through the amplitude characteristics of the differential sequence, comparing the obtained singular value sequence with the singular value sequence of normal current data, setting a corresponding number of significant singular values ​​to zero, i.e., performing denoising processing, and constructing a differential spectrum that represents the characteristic changes of the current data.

[0012] As a preferred solution of the photovoltaic arc fault detection method based on singular value decomposition and wavelet transform described in the present invention, the SVD reconstruction includes recombining the denoised singular value sequence with the retained left singular vector matrix and right singular vector matrix to form a new singular value matrix, multiplying the new singular value matrix with the left singular vector matrix and the right singular vector matrix to construct a new Hankel matrix, and obtain the denoised current waveform.

[0013] As a preferred solution of the photovoltaic arc fault detection method based on singular value decomposition and wavelet transform described in the present invention, the wavelet transform includes: performing FFT transform on the denoised and reconstructed current data, determining the characteristic frequency band of the fault arc, selecting the corresponding wavelet decomposition layer number according to the characteristic frequency band, performing multi-scale decomposition on the current data using high-pass and low-pass filters, extracting the wavelet detail coefficient of the target frequency band, and comparing it with the wavelet detail coefficient of normal operating conditions to set the alarm threshold.

[0014] As a preferred solution of the photovoltaic arc fault detection method based on singular value decomposition and wavelet transform described in the present invention, the FFT transformation includes converting the time domain signal of the current data after noise elimination and reconstruction into a frequency domain signal, calculating the energy spectrum of the current data, dividing the frequency band intervals by a fixed step size, accumulating the interval energy values, obtaining the energy spectrum density of each frequency band range, and comparing the frequency band energy to determine the characteristic frequency band.

[0015] As a preferred embodiment of the photovoltaic arc fault detection method based on singular value decomposition and wavelet transform described in the present invention, the arc fault determination and alarm triggering include extracting wavelet detail coefficients from the current data after wavelet transformation, comparing the extracted wavelet detail coefficients with a preset alarm threshold, and if the wavelet detail coefficients continuously exceed the alarm threshold, the current data is determined to be a fault arc, and an alarm action is triggered.

[0016] Another object of the present invention is to provide a photovoltaic arc fault detection system based on singular value decomposition and wavelet transform, which can obtain a singular value sequence by performing singular value decomposition on the Hankel matrix, thereby solving the problem of low detection accuracy in current DC arc fault detection technology.

[0017] As a preferred solution of the photovoltaic arc fault detection system based on singular value decomposition and wavelet transform described in the present invention, it includes: a signal acquisition module, a noise reduction processing module, a feature extraction module, and a threshold judgment module; the signal acquisition module is used to perform filtering processing based on the current data collected in real time, and retain the high-frequency components of the current data; the noise reduction processing module is used to achieve interference noise separation and suppression through singular value decomposition and differential spectrum analysis, and to construct a Hankel matrix using the filtered current data, decompose the Hankel matrix using singular value decomposition, construct a singular value differential spectrum from the decomposed matrix, and use the fault current singular value sequence to make a difference with the historical normal current singular value, and re-derive the fault current by SVD. The feature extraction module is used to capture the typical features of the fault current data through time-frequency analysis, perform FFT transformation and energy spectrum calculation on the denoised current data, determine the characteristic frequency band of the fault arc, select the number of wavelet decomposition layers based on the characteristic frequency band, extract the wavelet detail coefficients by wavelet transform, compare the difference in wavelet detail coefficients, and quantify the fault feature intensity; the threshold judgment module is used to accurately judge the fault arc and trigger the alarm through dynamic comparison, set the alarm threshold based on the difference in wavelet detail coefficients between normal current data and fault current data in combination with the empirical coefficient, continuously monitor the wavelet detail coefficients of the real-time current data, and judge the arc fault and trigger the alarm action by repeatedly exceeding the alarm threshold.

[0018] A computer device includes a memory and a processor. The memory stores a computer program. The processor executes the computer program to implement a photovoltaic arc fault detection method based on singular value decomposition and wavelet transform.

[0019] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a photovoltaic arc fault detection method based on singular value decomposition and wavelet transform.

[0020] Beneficial effects of the present invention: A photovoltaic arc fault detection method based on singular value decomposition and wavelet transform provided by the present invention collects current data for filtering processing, constructs a Hankel matrix for the filtered current data, can effectively eliminate low-frequency interference, retain high-frequency fault characteristics, and provide more accurate data, performs singular value decomposition on the Hankel matrix to obtain a singular value sequence, can accurately identify inverter noise and environmental noise, and enhance anti-interference ability, uses the singular value sequence to construct a differential spectrum, performs noise elimination processing, performs SVD reconstruction on the fault current data after noise elimination, can eliminate noise, restore the fault arc signal, increase sensitivity, and reduce the missed detection rate, perform wavelet transform based on the reconstructed current data, extract wavelet detail coefficients, and when the wavelet detail coefficients continuously exceed the threshold, determine that an arc fault is occurring and trigger an alarm, combine threshold judgment to achieve accurate detection of the fault arc, improve operating efficiency, and save time. The present invention achieves better results in detection accuracy, anti-interference, detection sensitivity, and operating efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0022] Figure 1 This is an overall flow chart of a photovoltaic arc fault detection method based on singular value decomposition and wavelet transform provided in the first embodiment of the present invention. DETAILED DESCRIPTION

[0023] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0024] Example 1, reference Figure 1 , as an embodiment of the present invention, provides a photovoltaic arc fault detection method based on singular value decomposition and wavelet transform, comprising:

[0025] S1: Collect current data, perform filtering processing, and construct a Hankel matrix for the filtered current data.

[0026] Furthermore, constructing the Hankel matrix includes arranging the collected current data in sequence, where the rows of the matrix consist of continuous signal values, and the data of adjacent rows slide backward in sequence. The number of matrix rows is determined by the length of the sampled data and the number of columns, and the timing characteristics of the current data are retained.

[0027] It should be noted that the Hankel matrix is ​​constructed after the sampling data is filtered. The constructed Hankel matrix is ​​expressed as:

[0028]

[0029] Where Λ is a matrix, x(n) is the nth element in the sequence, N is the total length of the sequence, m is the number of rows in the matrix, n is the number of columns in the matrix, and R is a set of real numbers.

[0030] It should also be noted that the collected current data is filtered to effectively remove the interference of noise in the original signal. The Hankel matrix is ​​constructed using the filtered current data to retain the time series characteristics of the current data, solving the problem of inaccurate feature extraction due to noise in traditional current data. It provides accurate data for subsequent current data analysis and fault diagnosis, and significantly improves the accuracy of current data processing.

[0031] S2: Perform singular value decomposition on the Hankel matrix to obtain a singular value sequence.

[0032] Furthermore, obtaining the singular value sequence includes decomposing the obtained Hankel matrix into a left singular vector matrix, a right singular vector matrix and a singular value matrix, and obtaining the singular value sequence by taking the square root of the eigenvalue of the product of the decomposition matrix.

[0033] The singular value decomposition is expressed as:

[0034] A=UDV T

[0035] Where A is a matrix, U is a left singular vector matrix AA T The eigenvector of A T is the transpose of the matrix, D is the diagonal matrix, V is the right singular vector matrix A T The eigenvector of A, V T is the transpose of the right singular vector matrix.

[0036] It should be noted that the left singular vector matrix and the right singular vector matrix include: the left singular vector matrix is ​​composed of eigenvectors of the product of the transposed matrix of the Hankel matrix and the singular value matrix, and shows the row space characteristics of the current data.

[0037] The corresponding eigenvalue of the left singular matrix is ​​expressed as [λ1,λ2,…,λ i..., λ q ] and the eigenvalues of the right singular matrix are expressed as [μ1, μ2,..., μ i ..., μ q ], and when the eigenvalues are not zero, the singular values are calculated by using the singular value matrix D and the eigenvalues of U and V, and are expressed as:

[0038]

[0039] where λ i is the eigenvalue of the left singular matrix, μ i is the eigenvalue of the right singular matrix, σ i is the singular value, i is the index value, and q is the total number of eigenvalues.

[0040] The right singular vector matrix is composed of the eigenvalues of the product of the transposed Hankel matrix and the original matrix, and exhibits the column space characteristics of the current data.

[0041] According to the characteristic equation, the eigenvalues are calculated, and then the eigenvalues [v1 v2 … v q ] are obtained, so that the right singular vector matrix is obtained, and is expressed as:

[0042] det(AA T -λE)=0

[0043] where det() is the determinant of the matrix, A is the matrix, λ is the eigenvalue, and E is the unit matrix.

[0044] It should be noted that by singular value decomposition of the Hankel matrix, the singular value sequence is extracted based on the eigenvalue relationship, the characteristics of the row space and the column space of the current data are accurately separated, the difficulty of feature extraction caused by high matrix dimension and strong noise interference in traditional current data analysis is solved, the feature recognition ability of the current data is significantly improved, and the reliability and calculation efficiency of the current data processing are enhanced.

[0045] S3: Constructing a difference spectrum by using the singular value sequence, performing noise elimination processing, and reconstructing the fault current data after noise elimination by SVD.

[0046] Further, constructing the difference spectrum includes: performing difference processing on adjacent singular values in the singular value sequence to obtain a difference sequence of the singular value change rate, obtaining the singular value sequence through the amplitude characteristics of the difference sequence, comparing the obtained singular value sequence with the singular value sequence of the normal current data, setting the corresponding number of significant singular values to zero, i.e., performing noise elimination processing, and constructing a difference spectrum that exhibits the characteristic change of the current data.

[0047] The construction of the difference spectrum is expressed as:

[0048] b k = σk -σ k+1 ,k=0,1,…,q-1

[0049] Among them, b k is the difference of singular values, σ k , σ k+1 is a singular value, k is the index of the singular value, and q is the total number of singular values ​​involved in the calculation.

[0050] It should be noted that the SVD reconstruction includes recombining the denoised singular value sequence with the retained left singular vector matrix and right singular vector matrix to form a new singular value matrix, multiplying the new singular value matrix with the left singular vector matrix and the right singular vector matrix to construct a new Hankel matrix and obtain the denoised current waveform.

[0051] It should also be noted that by constructing a singular value difference spectrum, using the difference changes of adjacent singular values ​​to identify and filter out noise interference, and combining SVD reconstruction to restore the effective signal, accurate noise reduction and feature enhancement of fault current data can be achieved, solving the signal distortion problem caused by fixed alarm thresholds or human intervention in traditional denoising methods, significantly improving denoising accuracy, and enhancing the robustness and automation level of the algorithm.

[0052] S4: Perform wavelet transform based on the reconstructed current data to extract the wavelet detail coefficient. When the wavelet detail coefficient exceeds the threshold continuously, an arc fault is determined and an alarm is triggered.

[0053] Furthermore, the wavelet transform includes performing FFT transformation on the denoised and reconstructed current data, determining the characteristic frequency band of the fault arc, selecting the corresponding wavelet decomposition layer according to the characteristic frequency band, performing multi-scale decomposition on the current data using high-pass and low-pass filters, extracting the wavelet detail coefficient of the target frequency band, and comparing it with the wavelet detail coefficient of normal operating conditions to set the alarm threshold.

[0054] Perform FFT transformation on the current data after noise elimination and reconstruction, and calculate the energy spectrum of the current data as follows:

[0055]

[0056] S[k]=|X[k]| 2

[0057] Where X[k] is the frequency domain signal, k is the frequency domain signal index value, x[n′] is the n'th sampling point, n′ is the time domain signal index value, N′ is the total number of sampling points, and S[k] is the energy spectrum of the current data.

[0058] Wavelet transform is expressed as:

[0059]

[0060] Among them, a k is the approximate coefficient, h[m] is the low-pass filter coefficient, x[2k-m] is the fault arc current data after noise elimination, d k is the detail coefficient, g[m] is the high-pass filter coefficient, and m is the index value.

[0061] It should be noted that the FFT transformation includes converting the time domain signal of the current data after noise removal and reconstruction into a frequency domain signal, calculating the energy spectrum of the current data, dividing the frequency band intervals by a fixed step size, accumulating the interval energy values, obtaining the energy spectrum density of each frequency band range, and comparing the frequency band energy to determine the characteristic frequency band.

[0062] It should also be noted that determining an arc fault and triggering an alarm includes extracting wavelet detail coefficients from the current data after wavelet transformation, and comparing the extracted wavelet detail coefficients with a preset alarm threshold. If the wavelet detail coefficients continuously exceed the alarm threshold, the current data is determined to be a fault arc, and an alarm action is triggered.

[0063] The preset threshold is expressed as:

[0064] T dk =K·(|d-d'|)+d'

[0065] Among them, T dk is the alarm threshold, K is the threshold coefficient, d is the wavelet detail coefficient of the fault arc current, and d′ is the wavelet detail coefficient of the normal arc current.

[0066] It should also be noted that by performing FFT transformation on the current data after noise elimination and reconstruction to determine the characteristic frequency band, and using wavelet transform to extract the wavelet detail coefficients, the problems of high false alarm rate and insufficient sensitivity caused by noise interference in traditional arc detection methods are solved, and the fault arc characteristics are accurately extracted and quickly identified, which significantly improves the accuracy and real-time performance of arc fault detection.

[0067] Example 2 is an embodiment of the present invention, which provides a photovoltaic arc fault detection system based on singular value decomposition and wavelet transform, including a signal acquisition module, a noise reduction processing module, a feature extraction module, and a threshold determination module.

[0068] The signal acquisition module is used to perform filtering processing based on the current data collected in real time, and retain the high-frequency components of the current data.

[0069] The noise reduction processing module is used to separate and suppress interference noise through singular value decomposition and differential spectrum analysis. The Hankel matrix is ​​constructed using the filtered current data, and the Hankel matrix is ​​decomposed using singular value decomposition. The singular value difference spectrum of the decomposed matrix is ​​constructed by subtracting the fault current singular value sequence from the historical normal current singular value, and the denoised current data is reconstructed and restored through SVD.

[0070] The feature extraction module is used to capture the typical characteristics of fault current data through time-frequency analysis, perform FFT transformation and energy spectrum calculation on the denoised current data, determine the characteristic frequency band of the fault arc, select the number of wavelet decomposition layers based on the characteristic frequency band, use wavelet transform to extract wavelet detail coefficients, compare the differences in wavelet detail coefficients, and quantify the fault feature intensity.

[0071] The threshold judgment module is used to accurately judge the fault arc and trigger the alarm through dynamic comparison. The alarm threshold is set based on the difference in wavelet detail coefficients between normal current data and fault current data, combined with the empirical coefficient. The wavelet detail coefficient of real-time current data is continuously monitored, and arc faults are judged and alarm actions are triggered by repeatedly exceeding the alarm threshold.

Claims

1. A photovoltaic arc fault detection method based on singular value decomposition and wavelet transform, characterized in that: include: Collect current data, perform filtering processing, and construct a Hankel matrix for the filtered current data; Performing singular value decomposition on the Hankel matrix to obtain a singular value sequence; The difference spectrum is constructed using the singular value sequence to perform denoising, and the SVD reconstruction is performed on the denoised fault current data; Perform wavelet transform based on the reconstructed current data to extract the wavelet detail coefficient. When the wavelet detail coefficient exceeds the alarm threshold continuously, an arc fault is determined and an alarm is triggered. Filtering involves using a high-pass filter to remove low-frequency components from the collected current data and retain high-frequency components.

2. The photovoltaic arc fault detection method based on singular value decomposition and wavelet transform according to claim 1, characterized in that: The construction of the Hankel matrix includes, The collected current data are arranged in sequence. The rows of the matrix consist of continuous signal values. The data of adjacent rows slide backward in sequence. The number of matrix rows is determined by the length of the sampled data and the number of columns to retain the timing characteristics of the current data.

3. The photovoltaic arc fault detection method based on singular value decomposition and wavelet transform according to claim 1, characterized in that: The obtaining of the singular value sequence comprises: The obtained Hankel matrix is ​​decomposed into a left singular vector matrix, a right singular vector matrix and a singular value matrix. According to the product eigenvalues ​​of the decomposition matrix, the square root of the eigenvalue is taken to obtain the singular value sequence.

4. The photovoltaic arc fault detection method based on singular value decomposition and wavelet transform according to claim 3, characterized in that: The left singular vector matrix and the right singular vector matrix include: The left singular vector matrix is ​​composed of the eigenvectors of the product of the Hankel matrix and the transposed matrix of the singular value matrix, which shows the row space characteristics of the current data; The right singular vector matrix is ​​composed of the eigenvectors of the product of the transposed Hankel matrix and the original matrix, which shows the column space characteristics of the current data.

5. The photovoltaic arc fault detection method based on singular value decomposition and wavelet transform according to claim 1, characterized in that: The constructing difference spectrum includes: The adjacent singular values ​​in the singular value sequence are subjected to difference processing to obtain the difference sequence of the singular value change rate. The singular value sequence is obtained through the amplitude characteristics of the difference sequence. The obtained singular value sequence is compared with the singular value sequence of normal current data. The corresponding number of significant singular values ​​are set to zero, that is, noise removal is performed, and a differential spectrum that shows the characteristic changes of the current data is constructed.

6. The photovoltaic arc fault detection method based on singular value decomposition and wavelet transform according to claim 1, characterized in that: The SVD reconstruction comprises: The denoised singular value sequence is recombined with the retained left singular vector matrix and right singular vector matrix to form a new singular value matrix. The new singular value matrix is ​​multiplied with the left singular vector matrix and the right singular vector matrix to construct a new Hankel matrix and obtain the denoised current waveform.

7. The photovoltaic arc fault detection method based on singular value decomposition and wavelet transform according to claim 1, characterized in that: The wavelet transform includes: The denoised and reconstructed current data is subjected to FFT transformation to determine the characteristic frequency band of the fault arc. The corresponding wavelet decomposition layer number is selected according to the characteristic frequency band. The current data is decomposed at multiple scales using high-pass and low-pass filters. The wavelet detail coefficient of the target frequency band is extracted and compared with the wavelet detail coefficient of the normal operating condition to set the alarm threshold.

8. The photovoltaic arc fault detection method based on singular value decomposition and wavelet transform according to claim 7, characterized in that: The FFT transformation includes: The time domain signal of the denoised and reconstructed current data is converted into a frequency domain signal, and the energy spectrum of the current data is calculated. The frequency band intervals are divided by a fixed step size, and the interval energy values ​​are accumulated to obtain the energy spectrum density of each frequency band range. The frequency band energies are compared to determine the characteristic frequency band.

9. The photovoltaic arc fault detection method based on singular value decomposition and wavelet transform according to claim 1, characterized in that: Determining an arc fault and triggering an alarm includes: The wavelet detail coefficients are extracted from the current data after wavelet transformation, and the extracted wavelet detail coefficients are compared with the preset alarm threshold. If the wavelet detail coefficients exceed the alarm threshold continuously, the current data is determined to be a fault arc and the alarm action is triggered.

10. A photovoltaic arc fault detection system based on singular value decomposition and wavelet transform, characterized by: It includes a signal acquisition module (100), a noise elimination processing module (200), a feature extraction module (300), and a threshold determination module (300); The signal acquisition module (100) is used to perform filtering processing based on the current data collected in real time, and retain the high-frequency components of the current data; The noise elimination processing module (200) is used to separate and suppress interference noise through singular value decomposition and differential spectrum analysis, construct a Hankel matrix using the filtered current data, decompose the Hankel matrix using singular value decomposition, construct a singular value differential spectrum from the decomposed matrix, perform a difference between the fault current singular value sequence and the historical normal current singular value, and reconstruct and restore the noise-eliminated current data through SVD; The feature extraction module (300) is used to capture typical features of fault current data through time-frequency analysis, perform FFT transformation and energy spectrum calculation on the de-noised current data, determine the characteristic frequency band of the fault arc, select the number of wavelet decomposition layers based on the characteristic frequency band, extract wavelet detail coefficients using wavelet transformation, compare the differences in wavelet detail coefficients, and quantify the fault feature intensity; The threshold determination module (400) is used for accurately determining fault arcs and triggering alarms through dynamic comparison, setting alarm thresholds based on the difference in wavelet detail coefficients between normal current data and fault current data in combination with empirical coefficients, continuously monitoring the wavelet detail coefficients of real-time current data, and determining arc faults and triggering alarm actions by repeatedly exceeding the alarm threshold.

11. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of a photovoltaic arc fault detection method based on singular value decomposition and wavelet transform according to any one of claims 1 to 9 are implemented.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a photovoltaic arc fault detection method based on singular value decomposition and wavelet transform according to any one of claims 1 to 9 are implemented.

Images