Fault arc diagnosis method and system based on sampling resistor voltage signal

Abstract

The invention discloses a fault arc diagnosis method and system based on a sampling resistor voltage signal, and the method comprises the steps: carrying out the sampling of voltages at two ends of a sampling resistor through which a trunk current in an AC system circuit passes, and obtaining a voltage signal; performing time-frequency domain feature extraction on the voltage signal, and performing dimension reduction processing on the extracted time-frequency domain feature to obtain feature data after dimension reduction processing; and training a Li ghtGBM model based on the feature data, and detecting the voltage signal in each time window based on the trained Li ghtGBM model so as to diagnose the fault arc. According to the invention, the problem of series fault arc diagnosis in a complex household electricity utilization environment can be solved, and the series fault arc can be identified efficiently.

Description

Technical Field

[0001] The present invention relates to the technical field of arc fault detection, and more specifically, to a method and system for diagnosing arc fault based on a sampling resistor voltage signal. Background Art

[0002] In residential and low-voltage power distribution systems, long-term use, external forces and environmental factors can cause the cable insulation layer to age and break, thereby reducing the insulation strength and causing parallel and series arc faults. Compared with conventional electrical faults, the current amplitude of a series arc fault (SAF) changes less, mainly manifested as high-frequency distortion and increased noise in the current waveform. This makes it difficult for traditional electrical protection devices such as air switches and fuses to effectively identify low-voltage AC arcs because they are mainly based on overcurrent triggering circuit breaking. And with the construction of my country's new power system, the increase in nonlinear loads and the complexity of topological circuit structures, the effective detection and prevention of main line fault arcs has become a new challenge in the field of electrical fault protection. At present, more than dozens of manufacturers at home and abroad have successively launched low-voltage arc diagnostic devices, but the accuracy of existing low-voltage arc fault diagnostic devices based on branch (main) detection needs to be improved. The signal at both ends of the sampling resistor has no bandwidth limitation, and can perfectly retain the original characteristics of the current signal in the circuit. The time domain waveform of the fault arc sampled by the sampling resistor will uniformly present high-frequency spike waveform characteristics, and the fault current signal characteristics are obvious, making it easy to identify arc faults. However, the arc characteristics produced by low current loads will be submerged in the main current.

[0003] Therefore, the sampling resistor voltage signal needs to be further processed. For this purpose, a fault arc diagnosis method based on the sampling resistor voltage signal is proposed. Summary of the invention

[0004] The present invention provides a fault arc diagnosis method and system based on a sampling resistor voltage signal to solve the problem of how to perform fault arc diagnosis efficiently.

[0005] In order to solve the above problem, according to one aspect of the present invention, a fault arc diagnosis method based on a sampling resistor voltage signal is provided, the method comprising:

[0006] Sampling the voltage across the sampling resistor through which the main current in the AC system circuit passes, to obtain a voltage signal;

[0007] Extracting time-frequency domain features of the voltage signal, and performing dimensionality reduction processing on the extracted time-frequency domain features to obtain feature data after dimensionality reduction processing;

[0008] A LightGBM model is trained based on the characteristic data, and a voltage signal in each time window is detected based on the trained LightGBM model to diagnose the fault arc.

[0009] Preferably, the method further comprises:

[0010] Before extracting the time-frequency domain features of the voltage signal, the low-frequency information in the voltage signal is reduced frequency by frequency to highlight the high-frequency information in the voltage signal.

[0011] Preferably, the time-frequency domain features include: abnormal pulse number, Pearson correlation coefficient, peak-to-peak value, rectified mean value, standard deviation, zeroed kurtosis, skewness, root mean square, waveform factor, peak factor, pulse factor, margin factor, center of gravity frequency, root mean square frequency, frequency variance, frequency standard deviation, spectral kurtosis mean, spectral kurtosis standard deviation, spectral kurtosis skewness, spectral kurtosis kurtosis, spectrum energy, spectrum mean, sum of the amplitudes of the third and fifth harmonics, total harmonic distortion, power spectrum entropy, energy entropy, fuzzy entropy, and envelope entropy.

[0012] Preferably, the method performs dimensionality reduction processing on the extracted time-frequency domain features based on a random forest algorithm to obtain feature data after dimensionality reduction, including:

[0013] A data set containing all time-frequency domain features is trained through a random forest model, the importance of each feature and the model classification accuracy are calculated, and then feature screening is performed according to the importance of each feature in the model. In each cycle, a first preset number of features are removed in order of importance from small to large, and the model is retrained until the classification accuracy reaches the preset accuracy or the number of remaining features reaches the second preset number, and the feature data after dimensionality reduction is obtained.

[0014] Preferably, the training of the LightGBM model based on the feature data comprises:

[0015] Use GOSS to process the feature data, select appropriate large gradient and small gradient sample sets, merge the selected two sample sets, use the merged sample sets for training and learning, learn a new learner, and perform multiple iterations to train the LightGBM model;

[0016] Among them, the Bayesian optimization algorithm implemented by the Optuna library simultaneously optimizes multiple hyperparameters in LightGBM.

[0017] Preferably, the voltage signal in each time window is detected based on the trained LightGBM model to diagnose the fault arc, including:

[0018] The trained LightGBM model is used to detect arc signals in each time window in turn to determine whether an arc occurs in each time window, and when the number of arc faults occurring in a first preset number of consecutive time windows is greater than or equal to a second preset number, it is determined that an arc fault occurs in the AC system.

[0019] According to another aspect of the present invention, a fault arc diagnosis system based on a sampling resistor voltage signal is provided, the system comprising:

[0020] The signal sampling unit samples the voltage across the sampling resistor through which the main current in the AC system circuit passes, and obtains a voltage signal;

[0021] A feature data acquisition unit, used to extract time-frequency domain features of the voltage signal, and perform dimensionality reduction processing on the extracted time-frequency domain features to obtain feature data after dimensionality reduction processing;

[0022] The diagnostic unit is used to train a LightGBM model based on the characteristic data, and detect the voltage signal in each time window based on the trained LightGBM model to diagnose the fault arc.

[0023] Preferably, the system further comprises:

[0024] The signal amplification unit is used to reduce the low-frequency information in the voltage signal frequency by frequency before extracting the time-frequency domain features of the voltage signal, so as to highlight the high-frequency information in the voltage signal.

[0025] Preferably, the time-frequency domain features include: abnormal pulse number, Pearson correlation coefficient, peak-to-peak value, rectified mean value, standard deviation, zeroed kurtosis, skewness, root mean square, waveform factor, peak factor, pulse factor, margin factor, center of gravity frequency, root mean square frequency, frequency variance, frequency standard deviation, spectral kurtosis mean, spectral kurtosis standard deviation, spectral kurtosis skewness, spectral kurtosis kurtosis, spectrum energy, spectrum mean, sum of the amplitudes of the third and fifth harmonics, total harmonic distortion, power spectrum entropy, energy entropy, fuzzy entropy, and envelope entropy.

[0026] Preferably, the feature data acquisition unit performs dimensionality reduction processing on the extracted time-frequency domain features based on a random forest algorithm to obtain the feature data after dimensionality reduction, including:

[0027] A data set containing all time-frequency domain features is trained through a random forest model, the importance of each feature and the model classification accuracy are calculated, and then feature screening is performed according to the importance of each feature in the model. In each cycle, a first preset number of features are removed in order of importance from small to large, and the model is retrained until the classification accuracy reaches the preset accuracy or the number of remaining features reaches the second preset number, and the feature data after dimensionality reduction is obtained.

[0028] Preferably, the diagnosis unit trains the LightGBM model based on the feature data, including:

[0029] Use GOSS to process feature data, select appropriate large gradient and small gradient sample sets, merge the selected two sample sets, use the merged sample sets for training and learning, learn a new learner, and perform multiple iterations to train the LightGBM model;

[0030] Among them, the Bayesian optimization algorithm implemented by the Optuna library simultaneously optimizes multiple hyperparameters in LightGBM.

[0031] Preferably, the diagnosis unit detects the voltage signal in each time window based on the trained LightGBM model to diagnose the fault arc, including:

[0032] The trained LightGBM model is used to detect the arc signal in each time window in turn to determine whether an arc occurs in each time window. When the number of arc faults occurring in a first preset number of consecutive time windows is greater than or equal to a second preset number, it is determined that an arc fault occurs in the AC system.

[0033] Based on another aspect of the present invention, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any step of a fault arc diagnosis method based on a sampling resistor voltage signal.

[0034] According to another aspect of the present invention, the present invention provides an electronic device, including:

[0035] The computer-readable storage medium described above; and

[0036] One or more processors are used to execute the program in the computer-readable storage medium.

[0037] The present invention provides a fault arc diagnosis method and system based on sampling resistor voltage signal, including: sampling the voltage across the sampling resistor through which the main current in the AC system circuit passes, and obtaining a voltage signal; extracting the time-frequency domain features of the voltage signal, and performing dimensionality reduction processing on the extracted time-frequency domain features to obtain feature data after dimensionality reduction processing; training a LightGBM model based on the feature data, and detecting the voltage signal in each time window based on the trained LightGBM model to diagnose the fault arc. The method of the present invention obtains signals on the main circuit and detects the series fault arc occurring in the area to which it belongs, which can solve the problem of diagnosing series fault arcs in complex household power environments and efficiently identify series fault arcs. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] A more complete understanding of exemplary embodiments of the present invention may be obtained by referring to the following drawings:

[0039] Figure 1 Flow chart of a fault arc diagnosis method 100 based on a sampling resistor voltage signal according to an embodiment of the present invention;

[0040] Figure 2 A circuit diagram of a low-frequency attenuation amplifier circuit according to an embodiment of the present invention;

[0041] Figure 3 A signal comparison diagram of a refrigerator voltage signal before and after amplification according to an embodiment of the present invention;

[0042] Figure 4 A signal comparison diagram of a voltage signal of a laptop computer before and after amplification according to an embodiment of the present invention;

[0043] Figure 5 Schematic diagram of the structure of a fault arc diagnosis system 500 based on sampling resistor voltage signals according to an embodiment of the present invention. DETAILED DESCRIPTION

[0044] Now, exemplary embodiments of the present invention are described with reference to the accompanying drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. These embodiments are provided to disclose the present invention in detail and completely and to fully convey the scope of the present invention to those skilled in the art. The terms used in the exemplary embodiments shown in the accompanying drawings are not intended to limit the present invention. In the accompanying drawings, the same units / elements are marked with the same reference numerals.

[0045] Unless otherwise specified, the terms (including technical terms) used herein have the commonly understood meanings to those skilled in the art. In addition, it is understood that the terms defined in commonly used dictionaries should be understood to have the same meanings as those in the context of the relevant fields, and should not be understood as idealized or overly formal meanings.

[0046] Figure 1 FIG. 1 is a flow chart of a fault arc diagnosis method 100 based on a sampling resistor voltage signal according to an embodiment of the present invention. Figure 1 As shown, the arc fault diagnosis method based on the sampling resistor voltage signal provided by the embodiment of the present invention obtains the signal on the main line and detects the series arc fault occurring in the area, which can solve the problem of diagnosing the series arc fault in the complex household power environment and efficiently identify the series arc fault. The arc fault diagnosis method based on the sampling resistor voltage signal provided by the embodiment of the present invention 100 starts from step 101. In step 101, the voltage across the sampling resistor through which the main line current passes in the AC system circuit is sampled to obtain a voltage signal.

[0047] In the present invention, the voltage signal of the resistor is sampled point by point according to the set sampling frequency for the voltage at both ends of the sampling resistor through which the current in the AC system circuit flows, wherein the voltage signal includes a fault arc signal. Specifically, the current signal in the AC system circuit is sampled point by point according to the sampling frequency of 250KHz through the sampling resistor, and the sampling resistor voltage signal x(t) is recorded by a recorder. The set time window length is 20ms. The sampling resistor has no bandwidth limitation and can perfectly preserve the original characteristics of the current signal in the circuit, but also contains a large amount of noise signals, resulting in the sampled signal data having more harmonic content and high-frequency noise than the current signal extracted by the general high-frequency current transformer, and the fault arc time domain waveform sampled by the sampling resistor will uniformly present the characteristics of the high-frequency peak waveform, the fault signal characteristics are obvious, and the arc fault is easy to identify.

[0048] In step 102, time-frequency domain features are extracted from the voltage signal, and dimension reduction processing is performed on the extracted time-frequency domain features to obtain feature data after dimension reduction processing.

[0049] Preferably, the method further comprises:

[0050] Before extracting the time-frequency domain features of the voltage signal, the low-frequency information in the voltage signal is reduced frequency by frequency to highlight the high-frequency information in the voltage signal.

[0051] Preferably, the time-frequency domain features include: abnormal pulse number, Pearson correlation coefficient, peak-to-peak value, rectified mean value, standard deviation, zeroed kurtosis, skewness, root mean square, waveform factor, peak factor, pulse factor, margin factor, center of gravity frequency, root mean square frequency, frequency variance, frequency standard deviation, spectral kurtosis mean, spectral kurtosis standard deviation, spectral kurtosis skewness, spectral kurtosis kurtosis, spectrum energy, spectrum mean, sum of the amplitudes of the third and fifth harmonics, total harmonic distortion, power spectrum entropy, energy entropy, fuzzy entropy, and envelope entropy.

[0052] Preferably, the method performs dimensionality reduction processing on the extracted time-frequency domain features based on a random forest algorithm to obtain feature data after dimensionality reduction, including:

[0053] A data set containing all time-frequency domain features is trained through a random forest model, the importance of each feature and the model classification accuracy are calculated, and then feature screening is performed according to the importance of each feature in the model. In each cycle, a first preset number of features are removed in order of importance from small to large, and the model is retrained until the classification accuracy reaches the preset accuracy or the number of remaining features reaches the second preset number, and the feature data after dimensionality reduction is obtained.

[0054] In the present invention, the current arc information is easily submerged in the main line high current when a fault arc occurs in a small current load. Therefore, it is necessary to gradually reduce the low-frequency signal in the voltage signal based on a low-frequency attenuation amplifier to highlight the high-frequency signal of the arc fault. When the sampling resistor voltage signal x(t) is further reduced by a designed low-frequency attenuation amplifier circuit to highlight the high-frequency fault information, the circuit diagram of the low-frequency attenuation amplifier circuit is as follows: Figure 2 As shown, the specific calculation process is:

[0055]

[0056] Where: R2 = 900mΩ, R1 = 5mΩ, L1 = 33μH. It can be seen that the lower the frequency, the greater the amplitude decreases. When the frequency is close to 3KHz, the corresponding harmonic amplitude decreases by about 0.5.

[0057] In the present invention, under 250KHz sampling, arcs are generated when multiple loads (fluorescent lamps, water dispensers, vacuum cleaners, washing machines, air conditioners, refrigerators, induction cookers, microwave ovens, computers) are running at the same time with small current loads, and the arcs are attenuated after 100m of cable. The arc signal enhancement effect is as follows: Figure 3 and 4 As shown in the figure, they are the signals of refrigerator and notebook respectively. Figure 3 and 4 In the figure, the first channel is the original sampling resistor voltage signal, and the second channel is the sampled and amplified signal.

[0058] In the present invention, suitable and effective time-frequency domain features are extracted for the amplified sampling resistor voltage signal: abnormal pulse number, Pearson correlation coefficient, peak-to-peak value, rectified average value, standard deviation, zeroed kurtosis, skewness, root mean square, waveform factor, peak factor, pulse factor, margin factor, centroid frequency, root mean square frequency, frequency variance, frequency standard deviation, spectral kurtosis mean, spectral kurtosis standard deviation, skewness of spectral kurtosis, kurtosis of spectral kurtosis, spectrum energy, spectrum mean, sum of the amplitudes of the third and fifth harmonics, total harmonic distortion, power spectrum entropy, energy entropy, fuzzy entropy, envelope entropy, and then dimensionality reduction processing of the features is performed based on random forest.

[0059] In the present invention, random forest (RF) is used for feature screening to extract effective fault features. First, a data set containing all features is trained by a random forest model. Then, the contribution of each feature in the model to the model performance is evaluated according to its importance score. The features are sorted by their importance scores, and the features with low importance are gradually removed, the model is retrained, and the changes in the classification accuracy are observed. Finally, the number of features is selected to be the second preset number of 10 or the accuracy reaches the preset accuracy, so that the accuracy of the model reaches the best balance. In the present invention, the top 10 features of importance are used as the input vector of the arc diagnosis model, namely: abnormal pulse number, peak-to-peak value, zeroed kurtosis, skewness, waveform factor, spectral kurtosis standard deviation, sum of 3rd and 5th harmonic amplitudes, total harmonic distortion, power spectrum entropy and envelope entropy.

[0060] In step 103, a LightGBM model is trained based on the characteristic data, and the voltage signal in each time window is detected based on the trained LightGBM model to diagnose the fault arc.

[0061] Preferably, the step of training the LightGBM model based on the feature data comprises:

[0062] Use GOSS to process feature data, select appropriate large gradient and small gradient sample sets, merge the selected two sample sets, use the merged sample sets for training and learning, learn a new learner, and perform multiple iterations to train the LightGBM model;

[0063] Among them, the Bayesian optimization algorithm implemented by the Optuna library simultaneously optimizes multiple hyperparameters in LightGBM.

[0064] Preferably, the voltage signal in each time window is detected based on the trained LightGBM model to diagnose the fault arc, including:

[0065] The trained LightGBM model is used to detect the arc signal in each time window in turn to determine whether an arc occurs in each time window. When the number of arc faults occurring in a first preset number of consecutive time windows is greater than or equal to a second preset number, it is determined that an arc fault occurs in the AC system.

[0066] In the present invention, the LightGBM model is trained according to the feature data after dimensionality reduction, and the hyperparameters in the model are optimized by the Bayesian optimization method. The LightGBM model process is: first, the feature data is processed according to GOSS, and a suitable large gradient and small gradient sample set is selected, and then the selected two parts of the sample are merged, and the formed samples are used for training and learning, and a new learner is learned, and the detection accuracy is reached after multiple iterations. Among them, there are multiple hyperparameters in the LightGBM model, and the Bayesian optimization algorithm implemented by the Optuna library is selected to simultaneously optimize multiple hyperparameters in the LightGBM to improve the accuracy of model classification and recognition.

[0067] In the present invention, the trained LightGBM model is used to detect whether an arc occurs in each time window. Specifically, the predictions of five consecutive time windows can be counted. If the number of arc faults output is not less than three, it is considered that an arc fault has indeed occurred in the system, an alarm is issued, and the protection device is driven to cut off the fault circuit. Otherwise, the detection continues.

[0068] The method of the present invention can solve the problem of diagnosing series fault arcs in the current complex household power environment, and can efficiently identify series fault arcs occurring in the main circuit.

[0069] Figure 5 FIG. 5 is a schematic diagram of a fault arc diagnosis system 500 based on a sampling resistor voltage signal according to an embodiment of the present invention. Figure 5 As shown, the arc fault diagnosis system 500 based on sampling resistor voltage signal provided in the embodiment of the present invention includes: a signal sampling unit 501 , a characteristic data acquisition unit 502 and a diagnosis unit 503 .

[0070] Preferably, the signal sampling unit 501 is used to sample the voltage across a sampling resistor through which the main current in the AC system circuit passes, to obtain a voltage signal.

[0071] Preferably, the feature data acquisition unit 502 is used to extract time-frequency domain features of the voltage signal, and perform dimensionality reduction processing on the extracted time-frequency domain features to obtain feature data after dimensionality reduction processing.

[0072] Preferably, the system further comprises:

[0073] The signal amplification unit is used to reduce the low-frequency information in the voltage signal frequency by frequency before extracting the time-frequency domain features of the voltage signal, so as to highlight the high-frequency information in the voltage signal.

[0074] Preferably, the time-frequency domain features include: abnormal pulse number, Pearson correlation coefficient, peak-to-peak value, rectified mean value, standard deviation, zeroed kurtosis, skewness, root mean square, waveform factor, peak factor, pulse factor, margin factor, center of gravity frequency, root mean square frequency, frequency variance, frequency standard deviation, spectral kurtosis mean, spectral kurtosis standard deviation, spectral kurtosis skewness, spectral kurtosis kurtosis, spectrum energy, spectrum mean, sum of the amplitudes of the third and fifth harmonics, total harmonic distortion, power spectrum entropy, energy entropy, fuzzy entropy, and envelope entropy.

[0075] Preferably, the feature data acquisition unit 502 performs dimensionality reduction processing on the extracted time-frequency domain features based on random forest to obtain the feature data after dimensionality reduction, including:

[0076] A data set containing all time-frequency domain features is trained through a random forest model, the importance of each feature and the model classification accuracy are calculated, and then feature screening is performed according to the importance of each feature in the model. In each cycle, a first preset number of features are removed in order of importance from small to large, and the model is retrained until the classification accuracy reaches the preset accuracy or the number of remaining features reaches the second preset number, and the feature data after dimensionality reduction is obtained.

[0077] Preferably, the diagnosis unit 503 is used to train a LightGBM model based on the characteristic data, and detect the voltage signal in each time window based on the trained LightGBM model to diagnose the fault arc.

[0078] Preferably, the diagnosis unit 503 trains a LightGBM model based on the feature data, including:

[0079] Use GOSS to process the feature data, select appropriate large gradient and small gradient sample sets, merge the selected two sample sets, use the merged sample sets for training and learning, learn a new learner, and perform multiple iterations to train the LightGBM model;

[0080] Among them, the Bayesian optimization algorithm implemented by the Optuna library simultaneously optimizes multiple hyperparameters in LightGBM.

[0081] Preferably, the diagnosis unit detects the voltage signal in each time window based on the trained LightGBM model to diagnose the fault arc, including:

[0082] The trained LightGBM model is used to detect arc signals in each time window in turn to determine whether an arc occurs in each time window, and when the number of arc faults occurring in a first preset number of consecutive time windows is greater than or equal to a second preset number, it is determined that an arc fault occurs in the AC system.

[0083] The arc fault diagnosis system 500 based on the sampling resistor voltage signal of the embodiment of the present invention corresponds to the arc fault diagnosis method 100 based on the sampling resistor voltage signal of another embodiment of the present invention, which will not be described in detail herein.

[0084] Based on another aspect of the present invention, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any step of a fault arc diagnosis method based on a sampling resistor voltage signal.

[0085] According to another aspect of the present invention, the present invention provides an electronic device, including:

[0086] The computer-readable storage medium described above; and

[0087] One or more processors are used to execute the program in the computer-readable storage medium.

[0088] The invention has been described above with reference to a few embodiments. However, it is readily apparent to a person skilled in the art that other embodiments than the ones disclosed above are equally within the scope of the invention, as defined by the appended patent claims.

[0089] Generally, all terms used in the claims are to be interpreted according to their ordinary meaning in the technical field, unless explicitly defined otherwise therein. All references to "a / said / the [means, components, etc.]" are to be openly interpreted as at least one instance of the means, components, etc., unless explicitly stated otherwise. The steps of any method disclosed herein do not necessarily have to be performed in the exact order disclosed, unless explicitly stated otherwise.

[0090] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0091] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0092] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0093] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0094] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A fault arc diagnosis method based on sampling resistor voltage signal, characterized in that: The method comprises: Sampling the voltage across the sampling resistor through which the main current in the AC system circuit passes, to obtain a voltage signal; Extracting time-frequency domain features of the voltage signal, and performing dimensionality reduction processing on the extracted time-frequency domain features to obtain feature data after dimensionality reduction processing; The LightGBM model is trained based on the characteristic data, and the voltage signal in each time window is detected based on the trained LightGBM model to diagnose the fault arc.

2. The method according to claim 1, characterized in that The method further comprises: Before extracting the time-frequency domain features of the voltage signal, the low-frequency information in the voltage signal is reduced frequency by frequency to highlight the high-frequency information in the voltage signal.

3. The method according to claim 1, characterized in that The time-frequency domain features include: abnormal pulse number, Pearson correlation coefficient, peak-to-peak value, rectified mean value, standard deviation, zeroed kurtosis, skewness, root mean square, waveform factor, peak factor, pulse factor, margin factor, center of gravity frequency, root mean square frequency, frequency variance, frequency standard deviation, spectral kurtosis mean, spectral kurtosis standard deviation, spectral kurtosis skewness, spectral kurtosis kurtosis, spectrum energy, spectrum mean, sum of the amplitudes of the third and fifth harmonics, total harmonic distortion, power spectrum entropy, energy entropy, fuzzy entropy, and envelope entropy.

4. The method according to claim 1, characterized in that: The method performs feature screening based on random forest to obtain feature data after dimensionality reduction, including: A data set containing all time-frequency domain features is trained through a random forest model, the importance of each feature and the model classification accuracy are calculated, and then feature screening is performed according to the importance of each feature in the model. In each cycle, a first preset number of features are removed in order of importance from small to large, and the model is retrained until the classification accuracy reaches the preset accuracy or the number of remaining features reaches the second preset number, and the feature data after dimensionality reduction is obtained.

5. The method according to claim 1, characterized in that The training of the LightGBM model based on the feature data includes: Use GOSS to process feature data, select appropriate large gradient and small gradient sample sets, merge the selected two sample sets, use the merged sample sets for training and learning, learn a new learner, and perform multiple iterations to train the LightGBM model; Among them, the Bayesian optimization algorithm implemented by the Optuna library simultaneously optimizes multiple hyperparameters in LightGBM.

6. The method according to claim 1, characterized in that Based on the trained LightGBM model, the voltage signal in each time window is detected to diagnose the fault arc, including: The trained LightGBM model is used to detect the arc signal in each time window in turn to determine whether an arc occurs in each time window. When the number of arc faults occurring in a first preset number of consecutive time windows is greater than or equal to a second preset number, it is determined that an arc fault occurs in the AC system.

7. A fault arc diagnosis system based on sampling resistor voltage signal, characterized in that: The system comprises: A signal sampling unit is used to sample the voltage across the sampling resistor through which the main current in the AC system circuit passes, and obtain a voltage signal; A feature data acquisition unit, used to extract time-frequency domain features of the voltage signal, and perform dimensionality reduction processing on the extracted time-frequency domain features to obtain feature data after dimensionality reduction processing; The diagnostic unit is used to train a LightGBM model based on the characteristic data, and detect the voltage signal in each time window based on the trained LightGBM model to diagnose the fault arc.

8. The system according to claim 7, characterized in that The system further comprises: The signal amplification unit is used to reduce the low-frequency information in the voltage signal frequency by frequency before extracting the time-frequency domain features of the voltage signal, so as to highlight the high-frequency information in the voltage signal.

9. The system according to claim 7, characterized in that The time-frequency domain features include: abnormal pulse number, Pearson correlation coefficient, peak-to-peak value, rectified mean value, standard deviation, zeroed kurtosis, skewness, root mean square, waveform factor, peak factor, pulse factor, margin factor, center of gravity frequency, root mean square frequency, frequency variance, frequency standard deviation, spectral kurtosis mean, spectral kurtosis standard deviation, spectral kurtosis skewness, spectral kurtosis kurtosis, spectrum energy, spectrum mean, sum of the amplitudes of the third and fifth harmonics, total harmonic distortion, power spectrum entropy, energy entropy, fuzzy entropy, and envelope entropy.

10. The system according to claim 7, characterized in that The feature data acquisition unit performs dimensionality reduction processing on the extracted time-frequency domain features based on the random forest algorithm to obtain the feature data after dimensionality reduction, including: A data set containing all time-frequency domain features is trained through a random forest model, the importance of each feature and the model classification accuracy are calculated, and then feature screening is performed according to the importance of each feature in the model. In each cycle, a first preset number of features are removed in order of importance from small to large, and the model is retrained until the classification accuracy reaches the preset accuracy or the number of remaining features reaches the second preset number, and the feature data after dimensionality reduction is obtained.

11. The system according to claim 7, characterized in that The diagnosis unit trains the LightGBM model based on the feature data, including: Use GOSS to process feature data, select appropriate large gradient and small gradient sample sets, merge the selected two sample sets, use the merged sample sets for training and learning, learn a new learner, and perform multiple iterations to train the LightGBM model; Among them, the Bayesian optimization algorithm implemented by the Optuna library simultaneously optimizes multiple hyperparameters in LightGBM.

12. The system according to claim 7, characterized in that The diagnosis unit detects the voltage signal in each time window based on the trained LightGBM model to diagnose the fault arc, including: The trained LightGBM model is used to detect the arc signal in each time window in turn to determine whether an arc occurs in each time window. When the number of arc faults occurring in a first preset number of consecutive time windows is greater than or equal to a second preset number, it is determined that an arc fault occurs in the AC system.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

14. An electronic device, characterized in that: include: The computer readable storage medium as claimed in claim 13; as well as One or more processors are used to execute the program in the computer-readable storage medium.

Images