Direct-current arc fault diagnosis method and device with adaptive capability

Abstract

The invention discloses a direct-current arc fault diagnosis method and device with adaptive capacity. The method comprises the following steps: 1) collecting a current alternating-current component when a current direct-current system normally works; 2) extracting time-frequency domain characteristics of a current AC component, wherein the time-frequency domain characteristics comprise a current peak-to-peak value, a standard deviation and an amplitude reconstruction harmonic power sum; 3) performing multi-time-frequency-domain feature fusion based on mahalanobis distance to obtain arc detection features; 4) calculating an arc fault judgment threshold value based on a box plot method; 5) collecting a current AC component in real time to calculate a Mahalanobis distance, comparing the Mahalanobis distance with an arc fault determination threshold, and recording an arc event; the method has the advantages that the system adaptability is high, the detection rate is high, the false alarm rate is low, and the method can be widely applied to high-voltage and low-voltage direct-current power distribution systems such as airplanes, vehicles and ships.

Description

Technical Field

[0001] The invention relates to a DC arc fault diagnosis method and device with self-adaptation capability, belonging to the field of arc fault detection. Background Art

[0002] With the development of renewable energy, electric vehicles, microgrids, and more electric aircraft, DC power is increasingly being used. However, as the use of equipment and cables increases, insulation aging of the lines and poor contact of equipment terminals can easily lead to arc faults. Unlike AC arcs, DC arc current does not have a zero crossing point, making it more likely to produce continuous arcing. Once an arc is ignited, the arc column temperature can reach tens of thousands of degrees Celsius, which can easily ignite nearby combustibles, posing a huge threat to the safe and stable operation of DC power distribution systems.

[0003] Arc fault detection methods are mainly divided into two types: based on physical phenomena and based on electrical characteristics. Among them, the arc fault detection method based on physical phenomena uses a variety of sensors to receive the sound, heat and electromagnetic radiation emitted by the fault arc to detect the fault arc. The detection effect of this method is greatly restricted by the location of the sensor and the influence of the environment, and is not suitable for arc detection on aircraft. The electrical arc fault detection method detects arc faults by detecting the distortion of current and voltage and the derived variables related to these physical quantities. Due to its low cost, small size and no restrictions on the arc position, it has become a research trend for scholars at home and abroad.

[0004] Electrical arc fault detection methods can be divided into time domain, frequency domain and time-frequency domain methods. Most of the existing literatures simply detect arc faults from the frequency domain or time domain.

[0005] The frequency domain analysis method uses the characteristics of a large number of harmonics when an arc fault occurs to make fault judgments. Currently, most literature selects harmonic characteristics within 100kHz for arc detection, but the arc harmonic frequency band is prone to overlap with the load switching frequency. When there are nonlinear loads in the system, the harmonic spikes at the switching frequency will have a great impact on the frequency domain characteristics, making it impossible to distinguish between arcs and normal conditions. Some scholars have performed manual offline analysis to remove the influence of harmonic spikes by removing the switching frequency spectrum amplitude. This method requires offline analysis for different systems and cannot automatically identify and filter harmonic spikes, making it difficult to apply.

[0006] The time domain method detects arcs based on characteristics such as the current change rate, average value, and effective value. However, since surge current (such as that caused by switching action) and arc fault current have similar time domain characteristics, there is a high misjudgment rate.

[0007] Some scholars have applied time-frequency domain analysis methods, such as wavelet transform and Hilbert-Huang transform, to decompose arc current signals into a series of components step by step. This method is more suitable for non-stationary signal analysis and more reliable for arc fault detection. However, the accuracy of the method depends on the selection of wavelet basis, and it is difficult to determine the thresholds for arc faults and normal system operations (such as load switching). In addition, this type of method is mostly targeted at specific systems with fixed thresholds. Once the system wavelet basis and threshold are changed, they may no longer be applicable, and lack system adaptive capabilities.

[0008] With the rapid development of intelligent algorithms, more and more scholars are now using machine learning to assist in the detection of fault arcs. Many scholars input time-frequency domain features as feature vectors into neural network algorithms for training to obtain arc fault diagnosis models. Some scholars input the 13-dimensional feature vector of the fault arc into the support vector machine SVM and successfully realize the arc fault detection of three types of nonlinear loads. Some scholars use random forest RF to classify the features of different wavelet component energy components to improve the detection accuracy. Some scholars also directly use intelligent algorithms to extract nonlinear features for arc fault diagnosis. Some literature converts the signal into a grayscale image, performs Wiener filtering and Laplace operator sharpening, and obtains a grayscale-gradient co-occurrence feature matrix with high accuracy. The fault arc detection method based on pattern recognition learning algorithm can effectively improve the accuracy of fault arc detection, but it depends on the number and quality of samples to a certain extent, the amount of calculation is large, and the hardware requirements are high. In addition, the anti-interference ability and adaptability to the environment of this type of method cannot exceed the range of training data, and a lot of offline training and verification are required in practical applications. More importantly, once an arc occurs in a high-voltage DC system, the damage is greater and faster arc fault detection is required. The complexity of this type of algorithm is not conducive to the application of high-voltage arc detection engineering.

[0009] In summary, the shortcomings of the methods mentioned above are mainly reflected in three aspects: (1) The current arc fault detection method is prone to misjudgment in situations such as load switching and switch action, and is easily affected by system dynamic interference and has a high false alarm rate; (2) The frequency domain characteristics are easily affected by the harmonic interference of the system's inherent noise. The existing frequency domain feature extraction method cannot automatically identify the system's inherent noise spikes, and the distinction between arc and normal under nonlinear load conditions is not high; (3) The existing arc fault detection methods are mostly targeted at specific systems, and the arc fault judgment threshold is determined by artificial offline analysis and cannot adapt to system changes. The method is complex and difficult to apply. Summary of the invention

[0010] In order to improve the accuracy of arc fault diagnosis, reduce the false alarm rate during the normal dynamic operation of the system, and improve the adaptability of the algorithm to systems with different working conditions, the present invention proposes a DC arc fault diagnosis method and device with adaptive capability.

[0011] In order to achieve the above-mentioned invention object, the present invention specifically adopts the following technical solutions: The embodiment of the present application provides a DC arc fault diagnosis method with adaptive capability, characterized by comprising the following steps: Step 1) online collecting multiple groups of current AC components when the DC system is working normally; Step 2) extract the time-frequency domain characteristics of each group of current AC components, including peak-to-peak value, standard deviation and amplitude to reconstruct the harmonic power sum; Step 3) Based on the Mahalanobis distance, the peak-to-peak value, standard deviation and amplitude of each group of current AC components are fused to reconstruct the harmonic power and three time-frequency domain features to obtain the arc detection feature - Mahalanobis distance; Step 4) online calculating the Mahalanobis distance outer limit of the current AC component based on the box plot method as the arc fault judgment threshold; Step 5) Real-time acquisition of the AC component of the current in the operation of the DC system, calculation of the current time-frequency domain characteristic value, and calculation of the Mahalanobis distance value, comparison of the Mahalanobis distance value with the arc fault determination threshold, a value greater than the arc fault determination threshold is recorded as an arc event, and a value less than the arc fault determination threshold is judged as normal; Step 6) within the detection time window, count the number of independent arc events and the cumulative number of arc events. When the cumulative number of arc events is greater than the first threshold or the number of independent arc events is greater than the second threshold, it is determined that an arc fault occurs in the system, otherwise it is determined to be normal. Step 7) Execute steps 5) to 6) repeatedly until arc fault detection is completed.

[0012] The embodiment of the present application provides a DC arc fault diagnosis device with adaptive capability, characterized by comprising the following modules: Signal acquisition module: collects multiple sets of current AC component signals through current transformers and outputs them to the signal conditioning module; Signal conditioning module: Filter, amplify and isolate each group of current AC component signals through the signal conditioning circuit to obtain the conditioned current AC component signals, which are output to the signal processing module; Signal processing module: converting the conditioned current AC component signal into a digital signal through an analog-to-digital converter, executing steps 2) to 7) of the above-mentioned DC arc fault diagnosis method with adaptive capability to perform arc fault diagnosis, and determining in real time whether an arc fault has occurred in the DC system; Communication module: composed of CAN communication circuit to realize communication with the host computer; Arc fault alarm module: when an arc fault is detected, an arc fault alarm indication is issued; Arc fault reset module: clears the arc fault alarm indication.

[0013] An embodiment of the present application also provides an electronic device, including a processor and a memory, wherein the memory stores a program that can be run on the processor, and is characterized in that when the program is executed by the processor, the steps of the above-mentioned DC arc fault diagnosis method with adaptive capability are implemented.

[0014] The beneficial effects of the present invention are: 1. Harmonic power and characteristics based on amplitude reconstruction adopt the method of reordering the spectrum amplitude after FFT transformation to form a new amplitude sequence, which can not only retain the basic size information of the original amplitude, but also automatically filter out the system inherent noise spikes caused by changes in system conditions, thereby avoiding the influence of system condition changes on the arc frequency domain feature extraction and enhancing the adaptability of the selected feature quantity to changes in experimental conditions.

[0015] 2. The arc fault detection method based on Mahalanobis distance eliminates the correlation interference between variables, amplifies the feature discrimination, complements the advantages between features, and improves the arc fault detection capability of the algorithm.

[0016] 3. The online threshold initialization method based on the box plot solves the problems of inconsistent arc judgment thresholds in different systems and weak adaptability of arc detection algorithms, and improves the system adaptability, scope of application and portability of the algorithm.

[0017] 4. The statistical method of independent arc events and cumulative arc events improves the stability of the DC arc fault detection algorithm and reduces the impact of interference during system switching. It can effectively reduce the false alarm rate of the algorithm and avoid affecting the normal operation of the system.

[0018] 5. The method is highly versatile and is suitable for serial and parallel fault detection of DC systems with a voltage range of 28V to 600V. It has strong anti-interference capability and low misjudgment rate, and will not cause false alarms during load power-on startup, switch switching, loading and unloading, normal operation, and load switching.

[0019] The technical solution of the present invention can be widely used in low-voltage and high-voltage power systems of aircraft, ships, vehicles, etc. It has a high arc fault detection rate and a low false alarm rate, and supports the safe operation of the distribution system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a flow chart of a DC arc fault diagnosis method with adaptive capability according to an embodiment of the present invention; Figure 2 It is a principle block diagram of a DC arc fault diagnosis device with adaptive capability according to an embodiment of the present invention; Figure 3 is a schematic diagram of a box plot structure of an embodiment of the present invention; Figure 4is a schematic diagram of independent arc events and cumulative arc events according to an embodiment of the present invention; Figure 5 is a flow chart of arc fault judgment according to an embodiment of the present invention; Figure 6 is a flow chart of adaptive arc fault detection according to an embodiment of the present invention; Figure 7 is a waveform diagram of arc fault online diagnosis result according to an embodiment of the present invention; Figure 8 It is a waveform diagram of the online diagnosis result of the interference experiment according to the embodiment of the present invention. DETAILED DESCRIPTION

[0021] The technical solution of the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] Example 1

[0023] This embodiment provides a DC arc fault diagnosis method with adaptive capability, which is applicable to high voltage and low voltage, linear and nonlinear DC power distribution systems, such as Figure 1 As shown, the following steps are included: Step 1) online collect multiple groups of current AC components when the DC system is operating normally.

[0024] Specifically, this embodiment uses a current transformer to collect multiple sets of current AC component analog signals in the loop when the DC system is working normally. I 0 , sampling rate 200kHz, sampling points for one cycle N = 256 points. The collected current AC component analog signal is conditioned by high-pass filter circuit, low-pass filter circuit, amplifier circuit and isolation circuit in turn to obtain the conditioned current AC component analog signal. The high-pass filter circuit has a cutoff frequency of 100kHz, the low-pass filter circuit has a cutoff frequency of 1kHz, the amplifier circuit has an amplification factor of 5 times, and the isolation circuit is a voltage follower circuit and uses a voltage regulator to limit the amplitude to 3.3V.

[0025] Step 2) extract the time-frequency domain characteristics of each group of current AC components respectively, including: peak-to-peak value, standard deviation and amplitude to reconstruct the harmonic power sum.

[0026] Specifically, the conditioned current AC component analog signal is converted into a digital signal through an analog-to-digital converter and stored in a microcomputer, and the time-frequency domain characteristics of each group of current AC components when the current DC system is working normally are calculated in the microcomputer, including: peak-to-peak value, standard deviation and amplitude reconstruction of harmonic power sum; Peak-to-Peak I p-p The calculation formula is as follows (1): Ip-p =max( I 0 ) (1) Standard Deviation std The calculation formula is as follows (2): (2) in N is 256, mean for I 0 average value,

[0027] In order to reduce the impact of the inherent harmonic peaks of the system on the harmonic power and characteristics, the amplitudes after FFT transformation are reordered according to size based on the amplitude reconstruction method to form a new amplitude sequence. I 0 Perform FFT analysis, assuming that the number of points for each transformation is N , then its spectrum amplitude diagram is given by N / 2 points to obtain the spectrum amplitude sequence A (1) A (2) A (3)…… A ( N / 2), sort them from small to large, and get the reconstructed new amplitude sequence A’ (1) A’ (2) A’ (3)…… A’ ( N / 2), where A’ (i) The sequence amplitude satisfies the following rules A’ (1)≤ A’ (2)≤ A’ (3)……≤ A’ ( N / 2).

[0028] The reconstructed spectrum amplitude has a stronger adaptability to the influence of the system environment, because the peaks and fluctuations caused by environmental changes are placed on the right due to the sorting, and the spectrum amplitude after sorting is only related to the size of the spectrum amplitude.

[0029] The amplitude reconstruction harmonic power and calculation formula is shown in formula (3): (3) in, A’ (i) is the reconstructed amplitude, m, n The value of A’ (i) corresponds to the upper and lower boundaries of the corresponding specified data range, representing the spectrum amplitude corresponding to the specified data range. After comparative research, this embodiment selectsN The square sum of the amplitudes in the range of 20% to 40% of the data points is used as the characteristic quantity for identifying DC arc faults.

[0030] Step 3) Based on the Mahalanobis distance, the three time-frequency domain features of each group of current AC components are integrated to obtain the arc detection feature - Mahalanobis distance. The specific steps include: Step 3-1) Based on the three time-frequency domain features of each group of current AC components collected, a (1×3)-order feature matrix is ​​obtained respectively, and based on the (1×3)-order feature matrix of multiple groups of current AC components collected, a ( Y ×3) order characteristic matrix ,in, i =1,2,…, Y , j =1,2,3; Y Indicates the number of groups of the collected current AC components. In this embodiment, 10 groups of data are collected, namely Y =10.

[0031] Step 3-2) Obtain the feature matrix The average value of the feature and standard deviation ,matrix is a (1×3)-order matrix.

[0032] Step 3-3) Use the matrix obtained in the above two steps to normalize the data and remove the influence of dimension. The calculation method is shown in formula (4). The result is ( Y ×3)-order matrix [ P ij ], which will be used to calculate the correlation coefficient matrix. The calculation method is: (4) Step 3-4) Correlation coefficient matrix C The solution is a (3×3)-order matrix. The calculation method is shown in formula (5): (5) Step 3-5) Correlation coefficient matrix C Inverse, we get C -1 . So far, the correlation coefficient matrix, mean value, and standard deviation have been obtained.

[0033] Step 3-6) Solve the Mahalanobis distance in the real-time detection stage.

[0034] Using the parameters obtained in the previous steps, the eigenvalue based on the Mahalanobis distance is obtained through formula (6). When calculating the Mahalanobis distance in real time, only the features of the sample to be tested are xj The value of is a variable, and the others have been obtained in steps 3-1) to 3-5).

[0035] (6) Step 4) Collection M = 100 cycle current AC components, 100 Mahalanobis distance values ​​are calculated respectively through step 3, and the Mahalanobis distance outer limit value of the current AC component is calculated online based on the box plot method as the arc fault judgment threshold.

[0036] The quartiles in the box plot consist of three numbers, among which: Q 1 is the 25th percentile, also known as the first quartile; Q 2 is the 50th percentile, also known as the second quartile; Q 3 The 75th percentile, also known as the third quartile.

[0037] Percentile is a common term in statistics. For a set of data, its calculation method is to first sort the data from small to large, and then p% The value of the position is called p Percentiles are often used in descriptive analysis of data to measure the location of data and provide information about the distribution of data between the minimum and maximum values.

[0038] For each feature quantity, sort the 100 feature quantity values ​​from small to large and get the value in 75 % The value of the position is the third quartile Q 3 and at 25 % The value of the position is the first quartile Q 1 .

[0039] Interquartile range I QR is the difference between the third quartile and the first quartile, and the calculation formula is shown in formula (7): I QR = Q 3 - Q 1 (7) inner limit Q n is the outlier cutoff point, and the calculation formula is shown in formula (8): Q n = Q 3 ±1.5I QR (8) Outer Limit Q w The calculation formula of is shown in formula (9). The data represented by the points outside the inner limit are all outliers, among which mild outliers are outliers between the inner limit and the outer limit, and extreme outliers are outliers outside the outer limit. Q w = Q 3 ±3 I QR (9) Considering that in the threshold initialization process of DC arc fault detection, it is necessary to determine the upper limit of normal experimental data and eliminate abnormal values, so that the selected threshold has strong stability and anti-interference ability. Combined with the data comparison and analysis results, the outer limit value of the box plot method is selected as the reference value of the characteristic quantity under normal conditions, that is, the arc fault judgment threshold T yuzhi , as shown in formula (10): T yuzhi = Q 3 +3*( Q 3 - Q 1 ) (10) like Figure 3 The figure shows the arc fault judgment threshold result diagram of this example. The outer limit value is 6×10 as shown by the green solid line. -5 .

[0040] Step 5) Real-time acquisition of the AC component of the current in the operation of the current DC system, extraction of time-frequency domain features according to step 2), and calculation of the Mahalanobis distance value according to step 3), and comparison of the Mahalanobis distance value with the arc fault determination threshold value obtained in step 4), a value greater than the arc fault determination threshold value is recorded as an arc event, and the arc event flag position is 1, and a value less than the arc fault determination threshold value is judged as normal, and the arc event flag position is 0; Step 6) Set the detection time window T In the case of arcing, the number of independent arcing events and the number of cumulative arcing events are counted (e.g. Figure 4 As shown in the figure), when the accumulated number of arc events is greater than the set first threshold or the number of independent arc events is greater than the set second threshold, it is determined that an arc fault occurs in the system, otherwise it is determined to be normal (such as Figure 5 as shown).

[0041] The judgment criteria of arc faults need to refer to the provisions in the SAE AS6019 standard. In order to improve the stability of the DC arc fault detection results, this embodiment adds the judgment process of independent arc events and cumulative arc events.

[0042] If the Mahalanobis distance feature of the sampling data of a cycle is greater than the arc fault judgment threshold, it is recorded as an arc event. The number of independent arc events means that no matter how many arc events occur between two adjacent normal events, they are all counted as an independent arc event, and the number is increased by 1; the number of cumulative arc events means that as long as an arc event occurs, it is counted as a cumulative arc event, and the number is increased by 1. The method diagram is shown in Figure 5 When the number of accumulated arc events is greater than the set first threshold TH1 Or the number of independent arc events is greater than the set second threshold TH2 , it is judged that an arc fault occurs in the system, otherwise it is judged to be normal.

[0043] The definition of independent arc events and cumulative arc events comes from improving the stability of the DC arc fault detection algorithm, and a key role is to reduce the impact of interference during system switching, so the cumulative arc event number threshold TH1 and threshold value of number of independent arc events TH2 The value of is closely related to the duration of interference that may exist in the system. The two thresholds can be adjusted according to the duration of interference in the actual system to improve the stability of the detection algorithm.

[0044] In this example, within the 100ms time window, we get NUM = 20 Mahalanobis distance values, count the number of independent arc events and the number of cumulative arc events. After the 100ms time window ends, perform arc fault judgment. The flow chart is as follows Figure 6 As shown, it is determined whether the accumulated number of arc events exceeds the set first threshold value. TH1 =3 or the number of independent arc events exceeds the set second threshold TH2 =4, if yes, it is determined to be an arc fault, the judgment of an arc fault is completed, and the arc fault flag is set to 1. Otherwise, the arc fault flag is set to 0.

[0045] Step 7) Execute steps 5 and 6 repeatedly to update the arc fault flag until the arc fault detection is completed. Figure 6 shown.

[0046] Build a circuit to conduct arc fault detection experiment and interference experiment verification. The real-time detection results of arc fault are as follows: Figure 7 As shown in the figure, the interference test results of power switching are as follows Figure 8 As shown in (a), the interference test results of load switching are as follows Figure 8As shown in (b), the method can effectively and quickly detect arcs and avoid false alarms when the system changes dynamically.

[0047] Example 2 This embodiment provides a DC arc fault diagnosis device with adaptive capability, such as Figure 2 As shown, it includes the following modules: Signal acquisition module: collects multiple sets of current AC component signals through current transformers and outputs them to the signal conditioning module.

[0048] Specifically, the sampling rate of this embodiment is 200kHz, and the number of sampling points is one cycle. N =256 points.

[0049] Signal conditioning module: Filter, amplify and isolate each group of current AC component signals through the signal conditioning circuit to obtain the conditioned current AC component signals, which are output to the signal processing module.

[0050] Specifically, the signal conditioning circuit of this embodiment includes a high-pass filter circuit, a low-pass filter circuit, an amplifier circuit, and an isolation circuit. The collected current AC component analog signal is sequentially conditioned through the high-pass filter circuit, the low-pass filter circuit, the amplifier circuit, and the isolation circuit. The cutoff frequency of the high-pass filter circuit is 100kHz, the cutoff frequency of the low-pass filter circuit is 1kHz, the amplification factor of the amplifier circuit is 5 times, and the isolation circuit is a voltage follower circuit and uses a voltage regulator to limit the amplitude to 3.3V.

[0051] Signal processing module: The output signal of the signal conditioning module is converted into a digital signal through an analog-to-digital converter, and an arc fault detection algorithm is processed in a microcomputer, that is, steps 2) to 7) of the DC arc fault diagnosis method with adaptive capability provided in the above embodiment are executed to perform arc fault diagnosis and determine in real time whether an arc fault has occurred in the DC system.

[0052] Communication module: It is composed of CAN communication circuit to realize the communication between microcomputer and host computer.

[0053] Arc fault alarm module: When an arc fault is detected, an arc fault alarm indication is issued and the alarm indicator light is on.

[0054] Arc fault reset module: used to clear the arc fault alarm indication.

[0055] Example 3 This embodiment provides an electronic device, including a processor and a memory connected to the processor for communication, wherein the memory stores a program executable on the processor, and when the program is executed by the processor, the steps of the DC arc fault diagnosis method with adaptive capability provided in the above embodiment are implemented.

[0056] The above program can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on the remote computer or server.

[0057] The DC arc fault diagnosis method with adaptive capability of the present invention eliminates the influence of the harmonic peak of the inherent noise of the system on the arc detection characteristics under different working conditions based on the amplitude reconstruction method; performs threshold initialization based on the box plot method, eliminates the interference of abnormal large values ​​on the arc judgment threshold, and solves the problem of adaptive arc detection threshold of different systems; performs multi-feature fusion based on Mahalanobis distance, complements the advantages between features and eliminates the influence of the dimension of the feature value, thereby improving the arc fault detection rate; finally, based on the statistics of independent and cumulative arc events, reduces the influence of system dynamic interference and reduces the false alarm rate. The present invention has the characteristics of strong system adaptability, high detection rate and low false alarm rate, and can be widely used in high and low voltage DC power distribution systems such as aircraft, vehicles and ships.

[0058] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A DC arc fault diagnosis method with adaptive capability, characterized in that: The following steps are involved: Step 1) online collecting multiple groups of current AC components when the DC system is working normally; Step 2) extract the time-frequency domain characteristics of each group of current AC components, including peak-to-peak value, standard deviation and amplitude to reconstruct the harmonic power sum; Step 3) Based on the Mahalanobis distance, the peak-to-peak value, standard deviation and amplitude of each group of current AC components are fused to reconstruct the harmonic power and three time-frequency domain features to obtain the arc detection feature - Mahalanobis distance; Step 4) online calculating the Mahalanobis distance outer limit of the current AC component based on the box plot method as the arc fault judgment threshold; Step 5) Real-time acquisition of the AC component of the current in the operation of the DC system, calculation of the current time-frequency domain characteristic value, and calculation of the Mahalanobis distance value, comparison of the Mahalanobis distance value with the arc fault determination threshold, a value greater than the arc fault determination threshold is recorded as an arc event, and a value less than the arc fault determination threshold is judged as normal; Step 6) within the detection time window, count the number of independent arc events and the cumulative number of arc events. When the cumulative number of arc events is greater than the first threshold or the number of independent arc events is greater than the second threshold, it is determined that an arc fault occurs in the system, otherwise it is determined to be normal. Step 7) Execute steps 5) to 6) repeatedly until arc fault detection is completed.

2. The DC arc fault diagnosis method with adaptive capability according to claim 1, characterized in that: In step 2), the amplitude of the AC component of the current is used to reconstruct the harmonic power and the extraction method is as follows: Assume that the AC component of the current passes through N The spectrum amplitude sequence obtained by point FFT transformation is: A (1)， A (2)， A (3)…… A ( N / 2) The above amplitude sequence is reconstructed according to size, ignoring the influence of frequency on the position of spectrum amplitude. The reconstructed new amplitude sequence is: A’ (1)， A’ (2)， A’ (3)…… A’ ( N / 2) in, A’ (1)≤ A’ (2)≤ A’ (3)……≤ A’ ( N / 2) ; The amplitude reconstruction harmonic power and calculation formula is shown in formula (3): （3） in, A’ (i) is the reconstructed amplitude, and the values ​​of m and n are A’ (i) The corresponding upper and lower boundaries of the specified data range correspond.

3. The DC arc fault diagnosis method with adaptive capability according to claim 1, characterized in that: In step 3), the method of fusing three time-frequency domain features based on Mahalanobis distance is as follows: Step 3-1) Based on the collected time-frequency characteristics of multiple sets of current AC components, ( Y ×3) order characteristic matrix ,in, i =1,2,…, Y , j =1,2,3; Y Indicates the number of groups of the collected current AC components; Step 3-2) Obtain the feature matrix The average value of the characteristic and standard deviation ; Step 3-3) Perform normalization calculation on the data to remove the influence of dimension and obtain ( Y ×3)-order matrix [ P ij ], as shown in formula (4): （4） Step 3-4) Solve the correlation coefficient matrix C , the calculation method is shown in formula (5): （5） Step 3-5) Correlation coefficient matrix C Inverse, we get C -1 ; Steps 3-6) Utilization C -1 ,average value and standard deviation The Mahalanobis distance is calculated in real time. The calculation method is shown in formula (6): （6） in, is the characteristic of the sample to be tested.

4. The DC arc fault diagnosis method with adaptive capability according to claim 1, characterized in that: In step 4), the method of online calculating the Mahalanobis distance outer limit of the current AC component based on the box plot method as the arc fault judgment threshold is as follows: Calculating the interquartile range for box plots I QR , as shown in formula (7): I QR = Q 3 - Q 1 （7） in, Q 1 is the first quartile in the box plot, Q 3 It is the third quartile in the box plot; Calculate the outer limits of the box plot method Q w , as shown in formula (9); Q w = Q 3 ±3 I QR （9） The outer limit value of the box plot method is selected as the reference value of the characteristic quantity under normal conditions, that is, the arc fault judgment threshold, as shown in formula (10): T yuzhi = Q 3 +3*( Q 3 - Q 1 ) （10） in, T yuzhi Indicates the arc fault judgment threshold.

5. The DC arc fault diagnosis method with adaptive capability according to claim 1, characterized in that: In step 6), the definition and statistical method of independent arc events and cumulative arc events and arc fault judgment decision are as follows: If the Mahalanobis distance feature of the sampling data of one cycle is greater than the threshold, it is recorded as an arc event; an independent arc event means that when sampling multiple cycles, no matter how many arc events occur between two adjacent normal events, they are all counted as an independent arc event; a cumulative arc event means that as long as one arc event occurs, it is counted as a cumulative arc event; by judging whether the number of cumulative arc events exceeds the first threshold or whether the number of independent arc events exceeds the second threshold, if so, it is determined to be an arc fault, and the judgment of an arc fault is completed.

6. The DC arc fault diagnosis method with adaptive capability as claimed in claim 5, characterized in that: In step 6), the first threshold and the second threshold are adjusted according to the interference duration in the actual system to improve the stability of the detection algorithm.

7. A DC arc fault diagnosis device with adaptive capability, characterized in that: Includes the following modules: Signal acquisition module: collects multiple sets of current AC component signals through current transformers and outputs them to the signal conditioning module; Signal conditioning module: Filter, amplify and isolate each group of current AC component signals through the signal conditioning circuit to obtain the conditioned current AC component signals, which are output to the signal processing module; Signal processing module: converting the conditioned current AC component signal into a digital signal through an analog-to-digital converter, executing steps 2) to 7) of the DC arc fault diagnosis method with adaptive capability described in any one of claims 1 to 6 to perform arc fault diagnosis, and determining in real time whether an arc fault has occurred in the DC system; Communication module: composed of CAN communication circuit to realize communication with the host computer; Arc fault alarm module: when an arc fault is detected, an arc fault alarm indication is issued; Arc fault reset module: clears the arc fault alarm indication.

8. An electronic device comprising a processor and a memory, wherein the memory stores a program that can be run on the processor, characterized in that: When the program is executed by the processor, the steps of the DC arc fault diagnosis method with adaptive capability described in any one of claims 1 to 6 are implemented.

Images