Arc fault detection method based on complementary ensemble empirical mode decomposition

Abstract

An arc fault detection method based on complementary ensemble empirical mode decomposition, comprising: firstly, an arc current acquisition system acquiring current data in a circuit to obtain an arc current, and performing preprocessing; using a complementary ensemble empirical mode decomposition algorithm to extract an intrinsic mode function (IMF)1 component of the arc current, and performing piecewise mean processing on the IMF1 component to obtain an IMF1 component mean; and then using an adjacent cycle correlation coefficient algorithm to calculate an IMF1 component feature value to obtain an arc determination feature value, and comparing the arc determination feature value with a preset threshold to determine whether an arc fault occurs. Compared with existing arc fault detection algorithms, the described feature value extraction method has high anti-interference capability, can adapt to accurate extraction of arc fault features in a high-noise environment, has the advantages of high detection precision and low erroneous detection rate, and can meet the arc fault detection requirements of high-voltage direct-current power supply and distribution systems.

Description

A method for arc fault detection based on complementary set empirical mode decomposition Technical Field

[0001] The present invention belongs to the field of aircraft power systems and electrical safety monitoring, and in particular relates to an arc fault detection method based on complementary set empirical mode decomposition. Background Art

[0002] With the increasing electrification of aircraft, onboard power systems have evolved through low-voltage DC, fixed-frequency AC, and finally variable-frequency AC to meet the ever-increasing power demands of power electronic loads. Furthermore, as the voltage and current levels of aircraft power supply and distribution systems gradually increase, arc faults, caused by aging wiring harnesses, wear and tear due to vibration and thermal cycling, or loose electrical connections due to other reasons, are becoming increasingly common. However, arc faults are relatively subtle, and the resulting current and voltage fluctuations are often difficult to detect by electrical protection devices in the circuits.

[0003] Therefore, onboard high-voltage DC arc fault detection is a research work that urgently needs to be carried out.

[0004] Currently, scholars both domestically and internationally have conducted research on high-voltage DC arc faults. An extensive literature review reveals that the core of arc fault detection lies in feature analysis and selection. An effective arc signature significantly improves arc fault detection rates and exhibits strong anti-interference capabilities. For example, Dalian University of Technology proposed an arc fault detection method based on a combination of EMD and SVM in the paper "Aviation Fault Arc Detection Based on EMD-SVM." However, EMD itself has two significant drawbacks. First, when EMD is decomposed, its IMF decomposition exhibits modal aliasing, resulting in a single IMF function containing features at different time scales, leading to errors in feature analysis. Second, the EMD decomposition process requires numerous iterations, resulting in different IMFs under different stopping conditions, and there is no standardized stopping condition. Therefore, it is necessary to optimize the EMD algorithm to ensure the effectiveness of arc fault signatures and improve the accuracy of arc fault detection methods.

[0005] Summary of the Invention

[0006] The present invention provides an arc fault detection method based on complementary set empirical mode decomposition to solve the problem of detecting high-voltage direct current arc faults.

[0007] The technical solution of the present invention:

[0008] The present invention provides an arc fault detection method based on complementary set empirical mode decomposition, the method comprising the following steps:

[0009] Step 1: Collect high-voltage DC arc data and pre-process it to obtain data to be processed;

[0010] Step 2: Use the complementary set empirical mode decomposition algorithm to extract the characteristic mode function component IMF1 of the data to be processed;

[0011] Step 3: Perform piecewise mean processing on the modal function component IMF1 to obtain the IMF1 mean;

[0012] Step 4: Use the Pearson correlation coefficient to extract features from the IMF1 mean;

[0013] Step 5: Identify arc faults through a pre-set threshold determination strategy.

[0014] Furthermore, the current acquisition setting in step 1 must meet the following conditions:

[0015] 1) By setting a filter, the sampling bandwidth of the high-voltage DC arc data is not higher than 200KHz;

[0016] 2) The sampling frequency of high-voltage DC arc data shall not be less than 500KHz, and the sampling time shall not be less than 10s.

[0017] Furthermore, the pretreatment method in step 1 is specifically as follows:

[0018] The high-voltage DC arc data is normalized by the steady-state working current in the line to obtain the data to be processed Iarc.

[0019] Furthermore, the step 2 specifically includes the following steps:

[0020] 1) Create Gaussian white noise w0(t), and perform positive modulation and reverse modulation on the data to be processed Iarc to obtain a current signal Iarc(0+) mixed with positive white noise and a current signal Iarc(0-) mixed with negative white noise, respectively;

[0021] Iarc(0+)=Iarc+w0(t);

[0022] Iarc(0-)=Iarc-w0(t);

[0023] 2) Using the empirical mode decomposition algorithm, the current signal Iarc(0+) mixed with positive white noise and the current signal Iarc(0-) mixed with negative white noise are decomposed to obtain the intrinsic mode function components C0+ and C0-;

[0024] 3) Repeat the above two steps to add different noises and perform empirical mode decomposition to obtain the corresponding intrinsic mode function components Ci+ and Ci-, where i = 1, 2, 3, ..., N;

[0025] 4) Averaging all obtained intrinsic mode function components to obtain the final decomposition result;

[0026] Finally, the characteristic mode function component IMF1=(S++S-) / 2 is obtained, where N represents the number of times white noise is added.

[0027] Furthermore, the step 3 specifically includes the following steps:

[0028] 1) Divide the modal function component IMF1 into n segments according to the time span of 100 μs, and discard the data segment less than 100 μs;

[0029] 2) Find the extreme points and their coordinates in each divided segment, and record them as the maximum value Max(k) and the maximum value coordinate Maxzb(k), the minimum value Min(k) and the minimum value coordinate Minzb(k), k = 1, 2, 3, ..., n;

[0030] 3) Discard the extreme points in the modal function component IMF1 to obtain the reconstructed modal function component IMF2;

[0031] 4) Perform mean processing on the reconstructed modal function component IMF2. For the adjacent position points of the extreme point in the reconstructed modal function component IMF2, take the adjacent j points of the adjacent position point for mean processing; for other non-extreme points in the reconstructed modal function component IMF2, take the adjacent three points of the non-extreme point for mean processing, and j is the set value.

[0032] Furthermore, the step 4 specifically includes the following steps:

[0033] 1) Divide the mean of the modal function component IMF1 into m segments with a time span of 100 μs, and obtain the component set imf1, imf2, …, imfm, where m is an odd number greater than 5;

[0034] 2) Select two adjacent segments and calculate their Pearson correlation coefficient to obtain the arc characteristic value ρ. The calculation formula is as follows:

[0035] Where E represents expectation, μ represents standard deviation, σ represents variance, and p=1, ..., m-1.

[0036] Furthermore, the step 5 specifically includes the following steps:

[0037] 1) Compare each element of the arc characteristic value ρ with the first arc judgment threshold YZ0. If the arc characteristic value ρ is less than the judgment threshold YZ0, it is considered that an arc fault has occurred. Otherwise, proceed to the next step of judgment.

[0038] 2) Compare the arc characteristic value ρ with the second arc judgment threshold value YZ1. If the arc characteristic value ρ is less than the judgment threshold value YZ1, count and accumulate; the first judgment threshold value YZ0 is less than the second judgment threshold value YZ1;

[0039] 3) Determine whether an arc fault occurs based on the accumulated count value.

[0040] Furthermore, determining whether an arc fault occurs according to the accumulated count value is specifically as follows:

[0041] An arc fault is considered to have occurred when the count is greater than 6 times within 10 consecutive segments; or an arc fault is considered to have occurred when the count is greater than 1 time and less than or equal to 6 times within 10 consecutive segments and occurs more than 4 times within a 100ms time window.

[0042] The technical solution of the present invention uses a stable working current in the line to perform normalized preprocessing on the arc current, reducing the influence of the current amplitude on the signal characteristics, and solving the problem of large changes in arc identification parameters under different rated working currents; the mode function component IMF1 obtained by the complementary set empirical mode decomposition algorithm in this solution has high noise resistance and can effectively avoid noise interference on algorithm recognition during the arc identification process. DETAILED DESCRIPTION

[0043] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments of the present invention. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0044] The specific steps of the present invention are as follows:

[0045] Step 1: High-voltage DC arc fault data collection; and pre-processing of arc current data;

[0046] A section of high-voltage DC arc data is collected through a current data acquisition device, where the arc data acquisition device meets the following conditions:

[0047] 1) The sampling bandwidth is 200KHz;

[0048] 2) The sampling frequency is 500KHz,

[0049] 3) The sampling time is 10s;

[0050] The arc current data is normalized and pre-processed using the steady-state working current. In this embodiment, the steady-state working current of the arc current is 3 A. The normalized pre-processed arc current data Iarc is obtained by dividing the arc current data by the steady-state current.

[0051] Step 2: Use the complementary set empirical mode decomposition algorithm to extract the characteristic mode function component IMF1. The specific implementation steps are as follows;

[0052] 1) Create Gaussian white noise w0(t) and perform positive modulation and reverse modulation on the arc current signal to obtain the current signal Iarc(0+) mixed with positive white noise and the current signal Iarc(0-) mixed with negative white noise respectively;

[0053] Iarc(0+)=Iarc+w0(t);

[0054] Iarc(0-)=Iarc-w0(t);

[0055] 2) The empirical mode decomposition algorithm is used to decompose the arc currents Iarc(0+) and Iarc(0-) after adding Gaussian white noise to obtain the intrinsic mode function components C0+ and C0-;

[0056] 3) Repeat the above two steps to add different noises Ci (i = 1, 2, 3...) and perform empirical mode decomposition;

[0057] 4) Averaging all obtained intrinsic modal components to obtain the final decomposition result;

[0058] Finally, the modal function component IMF1=(S++S-) / 2 is obtained;

[0059] Step 3: Perform piecewise mean processing on the modal function component IMF1 to obtain the IMF1 mean;

[0060] 1) Divide the modal function component IMF1 into n segments according to the time span of 100 μs, and discard the data segment less than 100 μs;

[0061] 2) Find the extreme points and their coordinates in each segment, namely the maximum value Max(i) and the maximum value coordinate Maxzb(i), the minimum value Min(i) and the minimum value coordinate Minzb(i), i = 1, 2, 3, ..., n;

[0062] 3) Reconstruct the modal function component IMF1 according to the extreme value coordinates and discard the extreme value points in IMF1;

[0063] 4) Perform mean processing on the reconstructed IMF1. If the calculated modal function component IMF1 is not an extreme point, take the three adjacent points for mean processing; if the calculated modal function component IMF1 is a point adjacent to an extreme point, take the seven adjacent points for mean processing;

[0064] Step 4: Use the Pearson correlation coefficient to extract features from the IMF1 mean;

[0065] 1) Divide the modal function component IMF1 into m segments with a time span of 100 μs, and obtain the component set imf1, imf2, …, imfm, where m is an odd number greater than 5;

[0066] 0068.2) Select two adjacent segments and calculate their Pearson correlation coefficient to obtain the arc characteristic value ρ;

[0067] Where E represents expectation, μ represents standard deviation, σ represents variance, and p=1, ..., m-1.

[0068] Step 5: Identify arc faults using a pre-set threshold determination strategy;

[0069] 1) Compare each element of the arc characteristic value ρ with the first arc judgment threshold YZ0. If the arc characteristic value ρ is less than the judgment threshold YZ0, it is considered that an arc fault has occurred. Otherwise, proceed to the next step of judgment.

[0070] 2) Compare the arc characteristic value ρ with the second arc judgment threshold value YZ1. If the arc characteristic value ρ is less than the judgment threshold value YZ1, count and accumulate; the first judgment threshold value YZ0 is less than the second judgment threshold value YZ1;

[0071] 3) When the count is greater than 6 times within 10 consecutive segments, an arc fault is considered to have occurred; or when the count is greater than 1 time and less than or equal to 6 times within 10 consecutive segments and occurs more than 4 times within a time window of 100ms, an arc fault is considered to have occurred.

[0072] By adopting the above technical solution, the present invention has the following advantages:

[0073] This solution uses the stable working current in the line to perform normalized preprocessing on the arc current, reducing the impact of the current amplitude on the signal characteristics and solving the problem of large variations in arc identification parameters under different rated working currents.

[0074] The scheme adopts the complementary set empirical mode decomposition algorithm to obtain the modal function component IMF1, which has high noise resistance and can effectively avoid the interference of noise on algorithm recognition during the arc recognition process.

Claims

1. An arc fault detection method based on complementary set empirical mode decomposition, characterized in that: The method comprises the following steps: Step 1: Collect high-voltage DC arc data and perform preprocessing to obtain data to be processed; Step 2: Use the complementary set empirical mode decomposition algorithm to extract the characteristic mode function component IMF1 of the data to be processed; Step 3: Perform piecewise mean processing on the modal function component IMF1 to obtain the IMF1 mean; Step 4: Use the Pearson correlation coefficient to extract features from the IMF1 mean; Step 5: Identify arc faults through a pre-set threshold determination strategy.

2. The arc fault detection method based on complementary set empirical mode decomposition according to claim 1, characterized in that: The current acquisition setting in step 1 must meet the following conditions: 1) The sampling bandwidth of the high-voltage DC arc data is set to be no higher than 200KHz by setting a filter; 2) The sampling frequency of high-voltage DC arc data shall not be less than 500KHz, and the sampling time shall not be less than 10s.

3. The arc fault detection method based on complementary set empirical mode decomposition according to claim 1, characterized in that: The pretreatment method in step 1 is specifically: The high voltage DC arc data is processed in a normalized manner using the steady-state working current in the line to obtain the data to be processed Iarc.

4. The arc fault detection method based on complementary set empirical mode decomposition according to claim 1, characterized in that: The step 2 specifically comprises the following steps: 1) Create Gaussian white noise w0(t), and perform positive modulation and reverse modulation on the data to be processed Iarc to obtain a current signal Iarc(0+) mixed with positive white noise and a current signal Iarc(0-) mixed with negative white noise respectively; Iarc(0+)=Iarc+w0(t); Iarc(0-)=Iarc-w0(t); 2) The empirical mode decomposition algorithm is used to analyze the current signal Iarc(0+) mixed with positive white noise and the current signal Iarc(0+) mixed with positive white noise. The current signal Iarc(0-) with negative white noise is decomposed to obtain the intrinsic mode function components C0+ and C0-; 3) Repeat the above two steps to add different noises, and perform empirical mode decomposition to obtain the corresponding intrinsic mode function components Ci+ and Ci-, i = 1, 2, 3, ..., N; 4) Averaging all obtained intrinsic mode function components to obtain the final decomposition result; Finally, the characteristic mode function component IMF1=(S++S-) / 2 is obtained, where N represents the number of times white noise is added.

5. The arc fault detection method based on complementary set empirical mode decomposition according to claim 1, characterized in that: The step 3 specifically comprises the following steps: 1) Divide the modal function component IMF1 into n segments according to the time span of 100 μs, and discard the data segment less than 100 μs; 2) Find the extreme points and their coordinates in each segment, and record them as the maximum value Max(k) and the maximum value coordinate Maxzb(k), the minimum value Min(k) and the minimum value coordinate Minzb(k), k = 1, 2, 3, ..., n; 3) Discard the extreme points in the modal function component IMF1 to obtain the reconstructed modal function component IMF2; 4) Perform mean processing on the reconstructed modal function component IMF2. For the adjacent position points of the extreme point in the reconstructed modal function component IMF2, take the adjacent j points of the adjacent position points for mean processing; for other non-extreme points in the reconstructed modal function component IMF2, take the adjacent three points of the non-extreme point for mean processing, and j is the set value.

6. The arc fault detection method based on complementary set empirical mode decomposition according to claim 5, characterized in that: The step 4 specifically comprises the following steps: 1) Divide the mean of the modal function component IMF1 into m segments according to the time span of 100 μs, and obtain the component set imf1, imf2, ..., imfm, where m is an odd number greater than 5; 2) Select two adjacent segments to calculate their Pearson correlation coefficient to obtain the arc characteristic value ρ, and the calculation formula is as follows: Where E represents expectation, μ represents standard deviation, σ represents variance, and p=1,...,m-1.

7. The arc fault detection method based on complementary set empirical mode decomposition according to claim 1, characterized in that: The step 5 specifically comprises the following steps: 1) Compare each element of the arc characteristic value ρ with the first arc judgment threshold value YZ0. If the arc characteristic value ρ is less than the judgment threshold value YZ0, it is considered that an arc fault has occurred, otherwise, proceed to the next step of judgment; 2) Compare the arc characteristic value ρ with the second arc judgment threshold value YZ1, and if the arc characteristic value ρ is less than the judgment threshold value YZ1, count and accumulate; The first determination threshold value YZ0 is less than the second determination threshold value YZ1; 3) Determine whether an arc fault occurs according to the count accumulated value.

8. The arc fault detection method based on complementary set empirical mode decomposition according to claim 7, characterized in that: Determining whether an arc fault occurs according to the count accumulation value is specifically as follows: When the count is greater than 6 times within 10 consecutive segments, an arc fault is considered to have occurred; or if the count is greater than 1 time and less than or equal to 6 times within 10 consecutive segments and occurs more than 4 times within a 100ms time window, an arc fault is considered to have occurred.

Images