DC arc fault detection method and device

Abstract

The present invention relates to a DC arc fault detection method and device, the method comprising: collecting loop current signals of normal working conditions and DC arc faults, performing fast Fourier transform analysis to obtain loop current spectrum signals, comparing loop current spectrum signals of normal working conditions and DC arc faults to obtain characteristic spectrums; recording parameter values ​​in each characteristic spectrum, constructing and evaluating to obtain characteristic parameters; obtaining the loop current signal to be detected, performing fast Fourier transform on it to obtain its spectrum distribution, selecting three frequency bands from the spectrum to calculate its characteristic parameters, and judging whether a DC arc fault occurs according to the characteristic parameters. Compared with the prior art, the present invention determines the characteristic spectrum and the calculation of the characteristic parameters through analysis, and judges whether it is abnormal through an isolation forest algorithm. If it is abnormal, a DC arc fault occurs, and can quickly and accurately realize detection without being misjudged.

Description

Technical Field

[0001] The present invention relates to the technical field of direct current arc fault detection, and in particular to a method and device for detecting direct current arc faults in a photovoltaic direct current system based on frequency domain eigenvalues ​​and an isolation forest algorithm. Background Art

[0002] At present, the detection methods of DC arc faults at home and abroad are roughly divided into three categories, and some fault detectors have been developed based on these three detection methods, which are briefly described as follows:

[0003] ①Analysis of physical properties of sound, light, etc.

[0004] When an electric arc burns, it is accompanied by physical phenomena such as sound, light, heat and electromagnetic radiation. Based on this principle, good arc detection effects have been achieved in equipment with narrow spaces such as combiner boxes and switch cabinets.

[0005] ②Analysis of the time domain characteristics of system voltage and current

[0006] When an arc fault occurs in a DC system, the voltage and current waveforms in the line generally change. At the initial moment of the arc, the current in the line drops and the arc voltage rises immediately from 0. Based on this principle, the arc can be detected.

[0007] ③ Comprehensive evaluation of the time-frequency characteristics of system voltage and current

[0008] The frequency characteristics are mainly analyzed through time series. Compared with the normal state, the voltage and current harmonic content increases in the arc state, which can be used to determine the occurrence of arc faults.

[0009] However, in the analysis of physical characteristics such as sound and light, arc fault detectors in the junction box are usually used for detection and judgment. However, such detectors are expensive, only applicable to small spaces, and require that the surrounding environment has no obvious shielding of the signal, so they are not applicable to open photovoltaic power stations. The principle of analyzing the time domain characteristics of system voltage and current is simple, but due to the different working points of photovoltaic modules in the photovoltaic system, the line current may not drop when an arc fault occurs. Therefore, it is difficult to judge whether an arc occurs by simply detecting a drop in line current. The comprehensive evaluation of the time-frequency domain characteristics of system voltage and current does not take into account the influence of environmental factors such as light and temperature, and its accuracy is not high. Summary of the invention

[0010] The purpose of the present invention is to provide a DC arc fault detection method and device in order to overcome the defects of the above-mentioned prior art.

[0011] The purpose of the present invention can be achieved by the following technical solutions:

[0012] According to a first aspect of the present invention, a DC arc fault detection method is provided, comprising the following steps:

[0013] S1. Collect the loop current signal of each photovoltaic string in the photovoltaic DC system multiple times, and mark the loop current signals corresponding to normal working conditions and DC arc faults;

[0014] S2. Perform fast Fourier transform analysis on the collected loop current signal to obtain a loop current spectrum signal, compare the loop current spectrum signal under normal working conditions with the loop current spectrum signal under DC arc fault, record the s spectrum ranges with the largest difference between the two, and obtain s characteristic spectrums, s>1;

[0015] S3, recording the parameter values ​​in each characteristic spectrum, selecting different parameter values ​​for different combination calculations, and obtaining multiple characteristic quantities, wherein the parameter values ​​include the maximum amplitude value, the peak-to-peak value, and the standard deviation of the amplitude, and evaluating the effectiveness of each characteristic quantity in characterizing the DC arc fault, and taking the characteristic quantity with the highest effectiveness as the characteristic parameter;

[0016] S4. Acquire the loop current signal to be detected, perform fast Fourier transform analysis on it, obtain the loop current spectrum signal to be detected, determine the parameter value of the loop current spectrum signal in each characteristic spectrum, calculate its characteristic parameter, and judge whether a DC arc fault occurs according to the characteristic parameter.

[0017] Furthermore, in step S3, the effectiveness of evaluating a characteristic value in characterizing a DC arc fault is specifically as follows:

[0018] Acquire multiple loop current signals, calculate the characteristic quantity of each loop current signal, and obtain a data set, wherein any data in the data set is a characteristic quantity of a loop current signal, and the data corresponding to the normal working condition in the data set is more than the data corresponding to the DC arc fault;

[0019] Use the isolation forest algorithm to judge the data set and identify the outliers in the data set;

[0020] The abnormal point is compared with the data corresponding to the DC arc fault in the data set. The higher the coverage and effectiveness of the abnormal point on the data corresponding to the DC arc fault in the data set, the more effective the feature quantity is in representing the DC arc fault.

[0021] Further, in step S4, the loop current signal to be detected is the loop current signal of multiple photovoltaic strings in the photovoltaic DC system, and the characteristic parameter of each loop current signal is calculated to obtain the data set to be detected, and any data in the data set to be detected is a characteristic parameter of a loop current signal;

[0022] The isolation forest algorithm is used to judge the data set to be detected, identify the abnormal points in the data set to be detected, and mark the photovoltaic strings corresponding to the abnormal points as having DC arc faults.

[0023] Further, in step S4, a plurality of loop current signals corresponding to normal working conditions are obtained, characteristic parameters of the plurality of loop current signals corresponding to normal working conditions and the loop current signal to be detected are respectively calculated to obtain a data set to be detected, wherein any data in the data set to be detected is a characteristic parameter of a loop current signal;

[0024] The isolation forest algorithm is used to judge the data set to be detected and identify the abnormal points in the data set to be detected. If the abnormal point corresponds to the loop current signal to be detected, a DC arc fault is marked, otherwise, no fault occurs.

[0025] Furthermore, in step S2, s is set to 3, and three characteristic frequency spectra are obtained, which are 10-20 kHz, 30-50 kHz and 60-75 kHz respectively.

[0026] Furthermore, in step S3, the most effective feature quantity is k=0.4I max +0.3I peak +0.3I std , where I max ,I peak ,I std They represent the maximum amplitude, peak-to-peak value, and standard deviation of the amplitude respectively.

[0027] A DC arc fault detection device, based on the DC arc fault detection method according to the first aspect of the present invention, comprises:

[0028] A current acquisition module, used for acquiring a loop current signal to be detected;

[0029] A preprocessing module is used to analyze and determine the characteristic spectrum and the calculation formula of the characteristic parameter. The characteristic spectrum is analyzed and determined as follows: comparing the loop current spectrum signal under normal working conditions and the loop current spectrum signal under DC arc fault, recording the s spectrum ranges with the largest difference between the two, and obtaining s characteristic spectrums, where s>1; the calculation formula for analyzing and determining the characteristic parameter is as follows: recording the parameter values ​​in each characteristic spectrum, selecting different parameter values ​​for different combination calculations, and obtaining multiple characteristic quantities, wherein the parameter values ​​include the maximum amplitude value, the peak-to-peak value, and the standard deviation of the amplitude, and respectively evaluating the effectiveness of each characteristic quantity in characterizing the DC arc fault, and taking the characteristic quantity with the highest effectiveness as the characteristic parameter;

[0030] The characteristic parameter module is used to perform fast Fourier transform analysis on the loop current signal to obtain the loop current spectrum signal, and calculate its characteristic parameters according to the calculation formula of the characteristic spectrum and characteristic parameters;

[0031] The detection module is used to determine whether a DC arc fault occurs based on characteristic parameters of the loop current signal to be detected.

[0032] Furthermore, in the preprocessing module, the effectiveness of evaluating a characteristic quantity in characterizing a DC arc fault is specifically as follows:

[0033] Acquire multiple loop current signals, calculate the characteristic quantity of each loop current signal, and obtain a data set, wherein any data in the data set is a characteristic quantity of a loop current signal, and the data corresponding to the normal working condition in the data set is more than the data corresponding to the DC arc fault;

[0034] Use the isolation forest algorithm to judge the data set and identify the outliers in the data set;

[0035] The abnormal point is compared with the data corresponding to the DC arc fault in the data set. The higher the coverage and effectiveness of the abnormal point on the data corresponding to the DC arc fault in the data set, the more effective the feature quantity is in representing the DC arc fault.

[0036] Furthermore, the loop current signal to be detected is the loop current signal of multiple photovoltaic strings in the photovoltaic DC system. In the detection module, the characteristic parameter of each loop current signal is calculated to obtain the data set to be detected. Any data in the data set to be detected is a characteristic parameter of a loop current signal.

[0037] The isolation forest algorithm is used to judge the data set to be detected, identify the abnormal points in the data set to be detected, and mark the photovoltaic strings corresponding to the abnormal points as having DC arc faults.

[0038] Furthermore, in the detection module, a plurality of loop current signals corresponding to normal working conditions are obtained, characteristic parameters of the plurality of loop current signals corresponding to normal working conditions and the loop current signal to be detected are respectively calculated to obtain a data set to be detected, and any data in the data set to be detected is a characteristic parameter of a loop current signal;

[0039] The isolation forest algorithm is used to judge the data set to be detected and identify the abnormal points in the data set to be detected. If the abnormal point corresponds to the loop current signal to be detected, a DC arc fault is marked, otherwise, no fault occurs.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] The characteristic spectrum and characteristic parameters are determined by analysis, and the spectrum distribution of the collected high-frequency current signal is obtained by fast Fourier transform. Three frequency bands are selected from the spectrum to calculate its characteristic parameters. The isolation forest algorithm is used to determine whether it is abnormal. If it is abnormal, a DC arc fault has occurred. The detection can be achieved quickly and accurately, and it is not easy to misjudge. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic diagram of the structure of the present invention. DETAILED DESCRIPTION

[0043] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0044] Embodiment 1:

[0045] The present invention provides a DC arc fault detection method, such as Figure 1 As shown, the following steps are included:

[0046] S1. Collect the loop current signal of each photovoltaic string in the photovoltaic DC system multiple times, and mark the loop current signals corresponding to normal working conditions and DC arc faults;

[0047] S2. Perform fast Fourier transform analysis on the collected loop current signal to obtain a loop current spectrum signal, compare the loop current spectrum signal under normal working conditions with the loop current spectrum signal under DC arc fault, record the s spectrum ranges with the largest difference between the two, and obtain s characteristic spectrums, s>1;

[0048] In this embodiment, s is taken as 3, and three characteristic spectra are obtained, namely 10-20kHz, 30-50kHz and 60-75kHz. Within these three spectrum ranges, the loop current spectrum signal of the DC arc extinguishing fault varies greatly, which can better characterize the difference in the frequency domain between the loop current under normal working conditions and the DC arc fault. In other implementation methods, several spectrum ranges can be selected as characteristic spectra according to the specific situation and the working parameters of the photovoltaic DC system.

[0049] S3. Record the parameter values ​​in each characteristic spectrum, select different parameter values ​​for different combination calculations, and obtain multiple characteristic quantities, including the maximum amplitude value, peak-to-peak value, standard deviation of the amplitude, effective value, instantaneous value, etc., and evaluate the effectiveness of each characteristic quantity in characterizing the DC arc fault, and take the characteristic quantity with the highest effectiveness as the characteristic parameter;

[0050] Among them, the effectiveness of evaluating a characteristic quantity in characterizing a DC arc fault is specifically as follows:

[0051] S31, acquiring multiple loop current signals, calculating a characteristic value of each loop current signal, and obtaining a data set, wherein any data in the data set is a characteristic value of a loop current signal, and the data corresponding to normal working conditions in the data set is more than the data corresponding to DC arc faults;

[0052] S32, using the isolation forest algorithm to judge the data set and identify the abnormal points in the data set;

[0053] S33, comparing the abnormal point with the data corresponding to the DC arc fault in the data set, the higher the coverage and effectiveness of the abnormal point on the data corresponding to the DC arc fault in the data set, the higher the effectiveness of the feature value in representing the DC arc fault;

[0054] The coverage ratio is q1 / q2, where q1 represents the number of data corresponding to DC arc faults in the abnormal point, and q2 represents the number of data corresponding to DC arc faults in the entire data set;

[0055] The validity is q1 / q3, q1 represents the number of data corresponding to DC arc faults in the abnormal points, and q3 represents the number of abnormal points.

[0056] Obviously, the higher the coverage and effectiveness, the better.

[0057] In order to more intuitively represent the difference between the loop current in the frequency domain under normal working conditions and DC arc faults, a characteristic quantity is proposed. Multiple characteristic quantities can be constructed through mathematical formulas such as k1=aI by using elementary functions such as power functions and exponential functions and constants through a finite number of rational operations (addition, subtraction, multiplication, division, etc.). max +bI peak +cI std +dI pk , a, b, c, d are constant coefficients to be determined, k2=aI max +bI peak , k3=…,I max ,I peak ,I std ,I pk They represent the maximum amplitude, peak-to-peak value, standard deviation of the amplitude, and the difference between the maximum positive value and the minimum negative value respectively.

[0058] In this embodiment, multiple characteristic quantities are constructed, and the characteristic quantity with the highest effectiveness is found as the characteristic parameter, which is used as a parameter for judging whether a DC arc fault occurs. The formula is as follows: k = 0.4I max +0.3I peak +0.3I std .

[0059] S4. Acquire the loop current signal to be detected, perform fast Fourier transform analysis on it, obtain the loop current spectrum signal to be detected, determine the parameter value of the loop current spectrum signal in each characteristic spectrum, calculate its characteristic parameter, and judge whether a DC arc fault occurs according to the characteristic parameter.

[0060] It can be understood that steps S1-S3 are performed in advance to determine the characteristic spectrum and the calculation formula of the characteristic parameters, and step S4 is the actual detection process.

[0061] ① In one implementation mode, the loop current signal to be detected is the loop current signal of multiple or all photovoltaic strings in the photovoltaic DC system. Generally speaking, most photovoltaic strings in the entire photovoltaic DC system are operating normally, and occasionally individual photovoltaic strings may have DC arc faults. Therefore, the characteristic parameters of each loop current signal are calculated to obtain the data set to be detected. Any data in the data set to be detected is a characteristic parameter of a loop current signal, and it should be ensured that the data corresponding to normal working conditions is much larger than the data corresponding to DC arc faults. The isolation forest algorithm is used to judge the data set to be detected, identify the abnormal points in the data set to be detected, and mark the photovoltaic strings corresponding to the abnormal points as having DC arc faults.

[0062] ② In one implementation mode, considering that the proportion of the loop current signal corresponding to the DC arc fault in the loop current signal to be detected is uncertain, a plurality of loop current signals corresponding to normal operating conditions are obtained, and characteristic parameters of the plurality of loop current signals corresponding to normal operating conditions and the loop current signal to be detected are calculated respectively to obtain a data set to be detected, and any data in the data set to be detected is a characteristic parameter of a loop current signal. To ensure that the data corresponding to the normal operating condition is far more than the data corresponding to the DC arc fault, the loop current signal corresponding to the normal operating condition is far more than the loop current signal to be detected; an isolation forest algorithm is used to judge the data set to be detected, and an abnormal point in the data set to be detected is identified. If the abnormal point corresponds to the loop current signal to be detected, a DC arc fault is marked; otherwise, no fault occurs.

[0063] The isolation forest algorithm no longer describes normal sample points, but isolates abnormal points. In the feature space, sparsely distributed areas indicate that the probability of an event occurring in that area is very low, so the data falling in these areas can be considered abnormal. In the isolation forest, the data set is recursively and randomly segmented until all sample points are isolated. The collected continuous current signal and the collected high-frequency current signal are subjected to fast Fourier transform to obtain their spectral distribution. Three frequency bands are selected from the spectrum to calculate their characteristic parameters. The isolation forest algorithm is used to determine whether it is abnormal. If it is abnormal, a DC arc fault has occurred.

[0064] It should also be noted that in the above steps S1-S4, the collected loop current signal only has a high-frequency part, which is less likely to cause misjudgment.

[0065] A DC arc fault detection device, based on the DC arc fault detection method as described above, comprises:

[0066] (1) a current acquisition module, used to acquire the loop current signal to be detected;

[0067] (2) A preprocessing module, which is used to analyze and determine the characteristic spectrum and the calculation formula of the characteristic parameter. The characteristic spectrum is analyzed and determined as follows: comparing the loop current spectrum signal under normal working conditions with the loop current spectrum signal under DC arc fault, recording the s spectrum ranges with the largest difference between the two, and obtaining s characteristic spectrums, where s>1; the calculation formula for analyzing and determining the characteristic parameter is as follows: recording the parameter values ​​in each characteristic spectrum, selecting different parameter values ​​for different combination calculations, and obtaining multiple characteristic quantities, wherein the parameter values ​​include the maximum amplitude value, the peak-to-peak value, and the standard deviation of the amplitude, and respectively evaluating the effectiveness of each characteristic quantity in characterizing the DC arc fault, and taking the characteristic quantity with the highest effectiveness as the characteristic parameter;

[0068] Among them, in the preprocessing module, the effectiveness of evaluating a characteristic quantity in characterizing a DC arc fault is specifically as follows:

[0069] Acquire multiple loop current signals, calculate the characteristic quantity of each loop current signal, and obtain a data set, wherein any data in the data set is a characteristic quantity of a loop current signal, and the data corresponding to normal working conditions in the data set are more than the data corresponding to DC arc faults; use the isolation forest algorithm to judge the data set and identify abnormal points in the data set; compare the abnormal points with the data corresponding to the DC arc fault in the data set, and the higher the coverage and effectiveness of the abnormal points to the data corresponding to the DC arc fault in the data set, the more effective the characteristic quantity is in characterizing the DC arc fault.

[0070] (3) a characteristic parameter module, which is used to perform fast Fourier transform analysis on the loop current signal to obtain the loop current spectrum signal, and calculate its characteristic parameters according to the calculation formula of the characteristic spectrum and characteristic parameters;

[0071] (4) A detection module, used to determine whether a DC arc fault occurs based on characteristic parameters of the loop current signal to be detected.

[0072] ① In one implementation mode, the loop current signal to be detected is the loop current signal of multiple photovoltaic strings in a photovoltaic DC system. In the detection module, the characteristic parameter of each loop current signal is calculated to obtain a data set to be detected, and any data in the data set to be detected is a characteristic parameter of a loop current signal; the isolation forest algorithm is used to judge the data set to be detected, identify abnormal points in the data set to be detected, and mark the photovoltaic strings corresponding to the abnormal points as having a DC arc fault.

[0073] ② In one implementation mode, the loop current signal to be detected is the loop current signal of multiple photovoltaic strings in a photovoltaic DC system. In the detection module, multiple loop current signals corresponding to normal operating conditions are obtained, and characteristic parameters of the multiple loop current signals corresponding to normal operating conditions and the loop current signal to be detected are calculated respectively to obtain a data set to be detected, and any data in the data set to be detected is a characteristic parameter of a loop current signal; an isolation forest algorithm is used to judge the data set to be detected, and an abnormal point in the data set to be detected is identified. If the abnormal point corresponds to the loop current signal to be detected, a DC arc fault is marked, otherwise, no fault occurs.

[0074] In this embodiment, the solid-state relay with IGBT structure can effectively avoid the problem that the mechanical relay may generate arc at the moment of electric contact separation during the disconnection process; the high-frequency signal of the loop current is collected in real time through the high-frequency current transformer; when testing photovoltaic field, the system is connected in series to the incoming line end of the combiner box, which can accurately and quickly realize the alarm function, and the detection time is less than 100ms, which is much shorter than the UL1699B requirement that the arc detection device has an arcing time of less than 2S. At the same time, factors such as shadow shielding and light intensity will not cause the device to malfunction.

[0075] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A DC arc fault detection method, characterized in that: The following steps are involved: S1. Collect the loop current signal of each photovoltaic string in the photovoltaic DC system multiple times, and mark the loop current signals corresponding to normal working conditions and DC arc faults; S2. Perform fast Fourier transform analysis on the collected loop current signal to obtain a loop current spectrum signal, compare the loop current spectrum signal under normal working conditions with the loop current spectrum signal under DC arc fault, record the s spectrum ranges with the largest difference between the two, and obtain s characteristic spectrums, s>1; S3, recording the parameter values ​​in each characteristic spectrum, selecting different parameter values ​​for different combination calculations, and obtaining multiple characteristic quantities, wherein the parameter values ​​include the maximum amplitude value, the peak-to-peak value, and the standard deviation of the amplitude, and evaluating the effectiveness of each characteristic quantity in characterizing the DC arc fault, and taking the characteristic quantity with the highest effectiveness as the characteristic parameter; S4. Acquire the loop current signal to be detected, perform fast Fourier transform analysis on it, obtain the loop current spectrum signal to be detected, determine the parameter value of the loop current spectrum signal in each characteristic spectrum, calculate its characteristic parameter, and judge whether a DC arc fault occurs according to the characteristic parameter.

2. A DC arc fault detection method according to claim 1, characterized in that: In step S3, the effectiveness of evaluating a characteristic value in characterizing a DC arc fault is specifically as follows: Acquire multiple loop current signals, calculate the characteristic quantity of each loop current signal, and obtain a data set, wherein any data in the data set is a characteristic quantity of a loop current signal, and the data corresponding to the normal working condition in the data set is more than the data corresponding to the DC arc fault; Use the isolation forest algorithm to judge the data set and identify the outliers in the data set; The abnormal point is compared with the data corresponding to the DC arc fault in the data set. The higher the coverage and effectiveness of the abnormal point on the data corresponding to the DC arc fault in the data set, the more effective the feature quantity is in representing the DC arc fault.

3. A DC arc fault detection method according to claim 2, characterized in that: In step S4, the loop current signal to be detected is the loop current signal of multiple photovoltaic strings in the photovoltaic DC system, and the characteristic parameter of each loop current signal is calculated to obtain a data set to be detected, and any data in the data set to be detected is a characteristic parameter of a loop current signal; The isolation forest algorithm is used to judge the data set to be detected, identify the abnormal points in the data set to be detected, and mark the photovoltaic strings corresponding to the abnormal points as having DC arc faults.

4. A DC arc fault detection method according to claim 2, characterized in that: In step S4, a plurality of loop current signals corresponding to normal working conditions are obtained, characteristic parameters of the plurality of loop current signals corresponding to normal working conditions and the loop current signal to be detected are respectively calculated to obtain a data set to be detected, wherein any data in the data set to be detected is a characteristic parameter of a loop current signal; The isolation forest algorithm is used to judge the data set to be detected and identify the abnormal points in the data set to be detected. If the abnormal point corresponds to the loop current signal to be detected, a DC arc fault is marked, otherwise, no fault occurs.

5. A DC arc fault detection method according to claim 1, characterized in that: In step S2, s is set to 3, and three characteristic frequency spectra are obtained, namely 10-20 kHz, 30-50 kHz and 60-75 kHz.

6. A DC arc fault detection method according to claim 5, characterized in that: In step S3, the most effective feature quantity is k=0.4I max +0.3I peak +0.3I std , where I max ,I peak ,I std They represent the maximum amplitude, peak-to-peak value, and standard deviation of the amplitude respectively.

7. A DC arc fault detection device, characterized in that: A DC arc fault detection method according to any one of claims 1 to 6, comprising: A current acquisition module, used for acquiring a loop current signal to be detected; A preprocessing module is used to analyze and determine the characteristic spectrum and the calculation formula of the characteristic parameter. The characteristic spectrum is analyzed and determined as follows: comparing the loop current spectrum signal under normal working conditions and the loop current spectrum signal under DC arc fault, recording the s spectrum ranges with the largest difference between the two, and obtaining s characteristic spectrums, where s>1; the calculation formula for analyzing and determining the characteristic parameter is as follows: recording the parameter values ​​in each characteristic spectrum, selecting different parameter values ​​for different combination calculations, and obtaining multiple characteristic quantities, wherein the parameter values ​​include the maximum amplitude value, the peak-to-peak value, and the standard deviation of the amplitude, and respectively evaluating the effectiveness of each characteristic quantity in characterizing the DC arc fault, and taking the characteristic quantity with the highest effectiveness as the characteristic parameter; The characteristic parameter module is used to perform fast Fourier transform analysis on the loop current signal to obtain the loop current spectrum signal, and calculate its characteristic parameters according to the calculation formula of the characteristic spectrum and characteristic parameters; The detection module is used to determine whether a DC arc fault occurs based on characteristic parameters of the loop current signal to be detected.

8. A DC arc fault detection device according to claim 7, characterized in that: In the preprocessing module, the effectiveness of a feature quantity in characterizing a DC arc fault is evaluated as follows: Acquire multiple loop current signals, calculate the characteristic quantity of each loop current signal, and obtain a data set, wherein any data in the data set is a characteristic quantity of a loop current signal, and the data corresponding to the normal working condition in the data set is more than the data corresponding to the DC arc fault; Use the isolation forest algorithm to judge the data set and identify the outliers in the data set; The abnormal point is compared with the data corresponding to the DC arc fault in the data set. The higher the coverage and effectiveness of the abnormal point on the data corresponding to the DC arc fault in the data set, the more effective the feature quantity is in representing the DC arc fault.

9. A DC arc fault detection device according to claim 8, characterized in that: The loop current signal to be detected is the loop current signal of multiple photovoltaic strings in the photovoltaic DC system. In the detection module, the characteristic parameter of each loop current signal is calculated to obtain the data set to be detected. Any data in the data set to be detected is a characteristic parameter of a loop current signal. The isolation forest algorithm is used to judge the data set to be detected, identify the abnormal points in the data set to be detected, and mark the photovoltaic strings corresponding to the abnormal points as having DC arc faults.

10. A DC arc fault detection device according to claim 8, characterized in that: In the detection module, a plurality of loop current signals corresponding to normal working conditions are obtained, characteristic parameters of the plurality of loop current signals corresponding to normal working conditions and the loop current signal to be detected are respectively calculated to obtain a data set to be detected, wherein any data in the data set to be detected is a characteristic parameter of a loop current signal; The isolation forest algorithm is used to judge the data set to be detected and identify the abnormal points in the data set to be detected. If the abnormal point corresponds to the loop current signal to be detected, a DC arc fault is marked, otherwise, no fault occurs.

Images