Lightweight-class-based arc fault detection method and device, and storage medium

Abstract

The invention provides an arc fault detection method and device based on lightweight, and a storage medium, and the method comprises the steps: obtaining an arc current signal of a power distribution network load in a preset time period, determining a signal-to-noise ratio parameter, determining a dynamic window length according to the signal-to-noise ratio parameter, carrying out the Hilbert transformation of the arc current signal according to the dynamic window length, and obtaining an arc fault detection result. The method comprises the steps of obtaining m amplitude envelope data, extracting features from the m amplitude envelope data to obtain p time domain statistical features, calculating arc current signals by adopting a preset Fourier transform method to obtain arc current frequency spectrum data, processing the arc current frequency spectrum data to obtain q frequency domain features, and fusing the p time domain statistical features and the q frequency domain features based on a preset feature fusion algorithm to obtain a target arc current feature, and inputting the target arc current feature into a preset lightweight arc fault detection model to obtain an arc fault detection result. Therefore, the accuracy of arc fault detection is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power safety detection, and in particular to an arc fault detection method and device based on a lightweight and a storage medium. BACKGROUND

[0002] At present, with the acceleration of urbanization and the continuous increase of residents' electricity consumption, the load of the distribution network also increases. When the load of the distribution network is large, especially when the voltage or current of the distribution network increases sharply, arc fault of the distribution network is easily caused. In the low-voltage distribution network, current fault is prone to cause electrical fire, thereby causing economic or safety loss to users. For arc fault detection, the current detection methods for alternating current arc fault mainly include detection based on electrical parameter characteristic change, detection based on frequency spectrum analysis, and detection based on machine learning or deep learning. These methods have achieved high accuracy in arc fault detection. However, due to the existence of various normal transient events (for example, motor starting and electrical switch operation) in the user side power consumption environment, the detection may have misjudgment or omission. In addition, since the methods for detecting alternating current arc fault have large model size, slow reasoning speed, and require a large amount of computing resources, these methods limit the application on embedded electronic devices.

[0003] Therefore, when detecting arc fault of the distribution network, how to improve the accuracy of arc fault detection and reduce the computing overhead of detecting arc fault needs to be solved. SUMMARY

[0004] The embodiments of the present application provide an arc fault detection method and device based on a lightweight and a storage medium, which can improve the accuracy of arc fault detection and reduce the computing overhead of detecting arc fault when detecting arc fault of the distribution network.

[0005] In a first aspect, the embodiments of the present application provide an arc fault detection method based on a lightweight, applied to an electronic device, and the method comprises the following steps: obtaining an arc current signal of a load in a target distribution network in a preset time period; determining a signal-to-noise ratio parameter of the arc current signal; determining a dynamic window length according to the signal-to-noise ratio parameter; performing Hilbert transform on the arc current signal according to the dynamic window length to obtain m amplitude envelope data; m is a positive integer greater than 1; extracting statistical features of the arc current signal in the time domain from the m amplitude envelope data to obtain p time domain statistical features; p is greater than or equal to m; calculating the arc current signal by using a preset Fourier transform method to obtain arc current spectrum data; process the arc current spectrum data to obtain q frequency domain features; q is a positive integer greater than 1; fuse the p time domain statistical features and the q frequency domain features based on a preset feature fusion algorithm to obtain target arc current features; input the target arc current features into a preset lightweight arc fault detection model to obtain an arc fault detection result.

[0006] In a second aspect, an embodiment of the present application provides a lightweight-based arc fault detection device applied to an electronic device, and the device comprises: An acquisition module is configured to acquire an arc current signal of a load in a target power distribution network in a preset time period. A determination module is configured to determine a signal-to-noise ratio parameter of the arc current signal, determine a dynamic window length according to the signal-to-noise ratio parameter, and calculate the arc current signal by using a preset Fourier transform method to obtain arc current spectrum data. A calculation module is configured to perform Hilbert transform on the arc current signal according to the dynamic window length to obtain m amplitude envelope data, m is a positive integer greater than 1, extract statistical features of the arc current signal in a time domain from the m amplitude envelope data to obtain p time domain statistical features, p is greater than or equal to m, process the arc current spectrum data to obtain q frequency domain features, q is a positive integer greater than 1, and fuse the p time domain statistical features and the q frequency domain features based on a preset feature fusion algorithm to obtain target arc current features. A control module is configured to input the target arc current features into a preset lightweight arc fault detection model to obtain an arc fault detection result.

[0007] In a third aspect, an embodiment of the present application provides an electronic device, which comprises a processor, a memory, a communication interface, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the processor, and the programs comprise instructions for performing steps in any method of the first aspect of the embodiments of the present application.

[0008] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program for electronic data exchange, and the computer program causes a computer to perform some or all steps described in any method of the first aspect of the embodiments of the present application.

[0009] In a fifth aspect, an embodiment of the present application provides a computer program product, wherein the computer program product includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to perform part or all of the steps in any of the methods described in the first aspect of the embodiments of the present application. The computer program product can be a software installation package.

[0010] By implementing the embodiments of the present application, the following beneficial effects are achieved: The arc fault detection method, device and storage medium provided in the present application are applied to an electronic device. The method obtains an arc current signal of a load in a target power distribution network in a preset time period, determines a signal-to-noise ratio parameter of the arc current signal, determines a dynamic window length according to the signal-to-noise ratio parameter, performs Hilbert transform on the arc current signal according to the dynamic window length, obtains m amplitude envelope data, m is a positive integer greater than 1, extracts statistical features of the arc current signal in the time domain from the m amplitude envelope data, obtains p time domain statistical features, p is greater than or equal to m, calculates the arc current signal by using a preset Fourier transform method, obtains arc current spectrum data, processes the arc current spectrum data, obtains q frequency domain features, q is a positive integer greater than 1, fuses the p time domain statistical features and the q frequency domain features based on a preset feature fusion algorithm, obtains target arc current features, and inputs the target arc current features into a preset lightweight arc fault detection model to obtain an arc fault detection result. In this way, on the one hand, the current signal envelope is extracted by using Hilbert transform, statistical quantities such as envelope mean value, peak factor and kurtosis are further calculated, and indexes such as harmonic components in a specific frequency band and spectral energy ratio are extracted by using fast Fourier transform, and the time domain and frequency domain features are comprehensively predicted to improve the accuracy of arc fault detection. On the other hand, a lightweight classification model is used to detect normal arc current and fault arc current based on the extracted features, thereby reducing the calculation overhead of arc fault detection. BRIEF DESCRIPTION OF DRAWINGS

[0011] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0012] Figure 1 is a system architecture diagram of an arc fault detection method provided by an embodiment of the present application; Figure 2 is a structural schematic diagram of an electronic device provided by an embodiment of the present application; Figure 3is a flow diagram of a light-based arc fault detection method provided by an embodiment of the present application; Figure 4 is a structural diagram of a feature extraction unit provided by an embodiment of the present application; Figure 5 is a structural diagram of an alternating current arc fault detection system provided by an embodiment of the present application; Figure 6 is a flow diagram of a feature extraction and classification process method of an alternating current arc fault detection method provided by an embodiment of the present application; Figure 7 is a flow diagram of a LightGBM-based arc fault detection process provided by an embodiment of the present application; Figure 8 is a functional module composition block diagram of a light-based arc fault detection device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0013] In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0014] The terms "first", "second", and the like in the specification and claims of the present application and the above-mentioned drawings are used to distinguish different objects, not to describe a specific order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed, or can optionally include other steps or units inherent to the process, method, product or device.

[0015] It should be understood that the term "and / or" herein is only used to describe the association relationship between the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent the three cases of A alone, A and B together, and B alone. In addition, the character " / " in this paper represents that the front and rear associated objects are a "or" relationship. The "multiple" appearing in the embodiments of the present application means two or more than two.

[0016] In the embodiments of the present application, “at least one” or similar expressions refer to any combination of the items, including any combination of single item or multiple items, refer to one or more, and multiple refers to two or more. For example, at least one of a, b or c can represent the following seven cases: a, b, c, a and b, a and c, b and c, a, b and c. Wherein, each of a, b, c can be an element or a set containing one or more elements.

[0017] In the embodiments of the present application, “connection” refers to direct connection or indirect connection and various connection modes to realize communication between devices, and the embodiments of the present application do not make any limitation.

[0018] In this document, referring to “embodiments” means that the specific features, structures or characteristics described in connection with the embodiments can be included in at least one embodiment of the present application. The phrase appears at various places in the specification does not necessarily all refer to the same embodiment, nor is it necessarily mutually exclusive of other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0019] The related terms involved in the present application will be explained first as follows: Arc current signal: Arc current refers to the gas discharge phenomenon caused by current passing through insulating medium (such as air) in an electrical system, and arc current signal is the current signal of the arc that produces discharge.

[0020] Hilbert transform: Hilbert transform is a linear operator that converts a real-valued signal into a complex-valued signal. The mathematical definition of Hilbert transform is to convolve a real signal with a signal so that the signal can be more fully analyzed and processed.

[0021] For arc fault detection, the current detection methods for alternating current arc fault mainly include detection based on electrical parameter characteristic change, detection based on frequency spectrum analysis and detection based on machine learning or deep learning. These methods have achieved high accuracy in arc fault detection. However, due to the existence of various normal transient events (such as motor starting, electrical switch operation) in user side power consumption environment, the detection may appear misjudgment or omission; in addition, since the methods for detecting alternating current arc fault have large model size, slow reasoning speed and need a large amount of computing resources, these methods limit the application on embedded electronic devices.

[0022] To solve the above problems, the embodiment of the application provides a light-based arc fault detection method, device and storage medium, which is applied to an electronic device, obtains an arc current signal of a load in a target power distribution network in a preset time period, determines a signal-to-noise ratio parameter of the arc current signal, determines a dynamic window length according to the signal-to-noise ratio parameter, performs Hilbert transform on the arc current signal according to the dynamic window length, obtains m amplitude envelope data, m is a positive integer greater than 1, extracts statistical features of the arc current signal in the time domain from the m amplitude envelope data, obtains p time domain statistical features, p is greater than or equal to m, calculates the arc current signal by using a preset Fourier transform method, obtains arc current spectrum data, processes the arc current spectrum data, obtains q frequency domain features, q is a positive integer greater than 1, fuses the p time domain statistical features and the q frequency domain features based on a preset feature fusion algorithm, obtains target arc current features, inputs the target arc current features into a preset light arc fault detection model, and obtains an arc fault detection result. In this way, on the one hand, the current signal envelope is extracted by using Hilbert transform, the statistical quantities such as envelope mean value, peak factor and kurtosis are further calculated, the harmonic components in a specific frequency band and the spectral energy ratio and other indexes are extracted by using fast Fourier transform, and the time domain and frequency domain features are comprehensively predicted, thereby improving the accuracy of arc fault detection. On the other hand, a light classification model is used to detect normal arc current and fault arc current from the extracted features, thereby reducing the calculation overhead of arc fault detection.

[0023] The following will be described in combination with Figure 1 The system architecture of the light-based arc fault detection method in the embodiment of the application is described, Figure 1 The system architecture of the light-based arc fault detection method in the embodiment of the application is described,

[0024] The AC arc current experiment platform 110 is used to simulate and generate arc current signals under different working conditions, so as to provide original experimental data support for the arc fault detection device 120. The AC arc current experiment platform 110 can include various load units, for example, a load A 111, a load B 112 and a load C 113. The load units can be resistive loads, inductive loads or capacitive loads, or a combination of various loads, used to simulate the occurrence characteristics of arc faults under different electrical environments. Through the experiment platform, complex current fluctuations when arcs are generated can be reproduced under controllable conditions, thereby providing real and reliable signal samples for subsequent feature extraction and classification.

[0025] The arc fault detection device 120 is connected with the alternating arc current experiment platform 110, is used for receiving the real-time collection arc current signal, and carries out processing, feature extraction and classification identification to the collected signal, and finally outputs the detection result of arc fault. The arc fault detection device 120 includes: signal acquisition module 121, preprocessing module 122, feature extraction module 123 and classification decision module 124. Among them, the signal acquisition module 121 is used for real-time acquisition of arc current signal from the alternating arc current experiment platform 110. The module can include current sensor, data acquisition card and signal conditioning circuit and other hardware units to ensure that the acquired arc current signal is complete, continuous and meets the sampling accuracy requirements. The preprocessing module 122 is used for preprocessing operation to the collected arc current signal, mainly including denoising, normalization and window interception and other steps. Through the preprocessing module 122, the common power frequency interference, background noise and measurement error in the power system can be effectively removed, and the purity and analyzability of the input signal are ensured. The feature extraction module 123 is used for extracting multi-dimensional feature parameters in the preprocessed arc current signal, including two categories of time domain statistical features and frequency domain spectrum features. The time domain statistical features can be obtained by analyzing the amplitude envelope of the arc current signal, such as mean, variance, skewness, kurtosis, etc. The frequency domain features are obtained by Fourier transform or wavelet transform, including harmonic distribution, high-frequency pulse feature and adaptive frequency band energy feature. The classification decision module 124 is used for identifying and classifying the extracted arc features, and outputting the final arc fault detection result. The module can use a lightweight machine learning model, such as LightGBM or ensemble learning algorithm. In one possible embodiment, the classification decision module 124 can also combine an adaptive weighted fusion strategy, dynamically adjust the feature weight according to the classification accuracy of the time domain features and the frequency domain features in the historical training data, so as to improve the detection accuracy while ensuring the lightweight of the model. Through the module, the arc fault can be quickly identified under low computational overhead, which is suitable for embedded devices or edge computing terminals in actual distribution network environment. Further, there is a clear input and output relationship between each module inside the arc fault detection device 120: the signal acquisition module 121 outputs the original arc current signal as the input of the preprocessing module 122; the preprocessing module 122 outputs the denoised and intercepted signal as the input of the feature extraction module 123; the feature extraction module 123 outputs multi-dimensional feature parameters as the input of the classification decision module 124; and finally the classification decision module 124 outputs the arc fault detection result as the final output result of the whole system. As can be seen, the system has high modularity and scalability, and can flexibly upgrade or replace the functional modules according to actual needs.

[0026] It can be seen that, by the system architecture of the above-mentioned lightweight arc fault detection method, accurate detection and classification of complex arc current signals can be realized. The system provides rich and real arc current signal samples through the alternating current arc current experiment platform 110, and completes signal processing and identification by the arc fault detection device 120, which can effectively reduce the missed detection rate and misjudgment rate of arc faults, and improve the operation safety and reliability of the power distribution network. At the same time, the system adopts a lightweight feature processing and classification model, which has the advantages of low computational complexity, low resource occupation and easy deployment, and is particularly suitable for application in the edge nodes or embedded terminals of the power distribution network, thereby reducing the hardware cost and operation and maintenance pressure of the detection system while ensuring the safe operation of the power system.

[0027] The following will be described in combination with Figure 2 The electronic device in the embodiment of the present application is described, Figure 2 is a structural schematic diagram of an electronic device provided by the embodiment of the present application, as Figure 2 shown, the electronic device 200 includes one or more processors 210, a memory 220, a communication interface 230, and one or more programs 221, the processor 210 is in communication connection with the memory 220, the communication interface 230 through an internal communication bus.

[0028] Among them, the processor 210 can be a central processing unit (CPU), a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It can realize or execute various exemplary logical blocks, units and circuits described in combination with the disclosure content of the present application. The processor can also be a combination of computing functions, such as one or more microprocessor combinations, combinations of DSP and microprocessor, etc. The communication unit can be a communication interface, a transceiver, a transceiver circuit, etc., and the storage unit can be a memory.

[0029] The memory 220 can be a volatile memory or a nonvolatile memory, or can include both volatile and nonvolatile memory. The nonvolatile memory can be a read-only memory (ROM), programmable ROM (PROM), erasable PROM (EPROM), electrically EPROM (EEPROM), or flash memory, for example. The volatile memory can be a random access memory (RAM) used as an external cache memory. By way of example, and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous dynamic RAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0030] The one or more programs 221 stored in the memory 220 and configured to be executed by the processor 210 include instructions for performing any of the steps of the following embodiments of a lightweight arc fault detection method.

[0031] It can be understood that the electronic device 200 can include more or fewer structural elements than those in the above structural block diagram, for example, a power module, a physical button, a Wi-Fi module, a speaker, a Bluetooth module, a sensor, a display module, and the like, which are not limited herein. It can be understood that the electronic device 200 can be equipped with a system architecture of the lightweight arc fault detection method as described above. Figure 1 The system architecture of the lightweight arc fault detection method.

[0032] After understanding the software and hardware architecture of the present application, the following will be combined with the Figure 3 The lightweight arc fault detection method provided in the embodiments of the present application is described below, Figure 3 is a flowchart of a lightweight arc fault detection method provided in the embodiments of the present application, and specifically includes the following steps: At step S310, an arc current signal of a load in the target power distribution network in a preset time period is acquired.

[0033] The arc current signal refers to a current signal flowing through a load circuit when series arc is generated in the load circuit due to poor contact, aging of a wire, or loose connection, etc. in a low-voltage alternating current power distribution network. The current signal shows characteristics significantly different from normal working condition current, and the time-domain waveform shows intermittent jitter and distortion, rich high-frequency components, and randomness and non-stationarity.

[0034] Specifically, the arc current signal is acquired by a power distribution network signal acquisition device, which includes a current transformer (CT), a high-frequency sampling device, an oscilloscope or an equivalent data acquisition unit, a signal preprocessing module, and the like. The current transformer serves as a coupling unit between primary current and the sampling device, and can realize non-contact measurement of the current signal, thereby ensuring the safety of measurement and maintaining the original waveform characteristics of the signal. The high-frequency sampling device performs high-resolution discretization processing on the acquired current signal, and the sampling frequency needs to be greater than 10 kHz to ensure that the typical high-frequency harmonics and transient components in the arc signal can be covered. The oscilloscope or data acquisition unit is used for real-time recording and storage of the arc signal, and supports subsequent time-domain and frequency-domain analysis.

[0035] In experiments and applications, in order to better simulate actual household and industrial load scenarios, resistive load, inductive load, and resistive-inductive combined load are usually selected as test objects, an experimental platform is built according to the Chinese national standard GB-31143, and alternating current arc current signals under different load types are acquired in real time. Under different load types, the arc current signal shows different forms. For example, under a purely resistive load, the arc current waveform distortion is more obvious, and the high-frequency component is prominent. Under an inductive load, the envelope characteristics of the arc signal are more complex, and delay effect and additional harmonic components may occur. By acquiring arc signals under multiple types of loads, a more representative data set can be constructed, and the adaptability of the model to different application environments can be improved.

[0036] It should be noted that the arc current signal is significantly different from the ordinary load current signal. The load current waveform under normal operating conditions is highly correlated with the power supply voltage, the waveform is stable and the harmonic component is relatively limited. When an arc fault occurs, due to the non-linear discharge phenomenon of the arc, the signal presents the following typical characteristics: the current waveform appears intermittent discontinuity in some cycles, showing obvious non-stationarity; the harmonic component is significantly enhanced, especially the abnormal change of energy distribution in the middle and high frequency band; the amplitude of the time domain signal has a sudden change, showing a short-time sharp pulse; the signal envelope shows irregular fluctuations over time. Due to the obvious difference between the frequency domain characteristics and the time domain characteristics of the arc current signal and the ordinary load current signal, the arc fault and the normal current can be identified by analyzing these characteristics.

[0037] Step S320, determining the signal-to-noise ratio parameter of the arc current signal.

[0038] Among them, the signal-to-noise ratio (Signal-to-Noise Ratio, SNR) refers to the relative intensity ratio between the effective arc information contained in the load arc current signal in the power distribution network and the background noise in a certain period of time. The arc current signal has non-stationary and random characteristics, and is easily affected by external interference and normal transient events of the power distribution system, such as motor starting, switch switching and load mutation, etc., which will introduce additional noise components. By calculating the signal-to-noise ratio of the arc current signal, the significance of the effective arc characteristics in the signal can be quantified, and then the Hilbert transform and feature extraction in the subsequent steps are provided. In other words, the determination of the signal-to-noise ratio parameter is not only a necessary link to distinguish the arc signal and noise interference, but also a prerequisite for adaptive feature extraction and classification modeling.

[0039] Specifically, first, the arc current signal is segmented. Since the arc signal shows strong transient characteristics in the time domain, its statistical characteristics change significantly over time, so the long-time signal is divided into several window intervals for analysis. The length of each window is selected according to the sampling frequency and detection requirements, generally between 0.05s to 0.2s, which can ensure that one or several power cycles are covered and the short-time characteristics of the arc signal are reflected.

[0040] Secondly, the signal power and noise power are calculated in each window. The signal power can be estimated by the square mean value of the current signal in the window, that is:

[0041] Among them, represents the current value of the i-th sampling point, N represents the number of sampling points in the window, represents the signal power.

[0042] The estimation of noise power is achieved by filtering and separation techniques. Specifically, the original current signal is first high-pass filtered or wavelet denoised to remove the components related to the power frequency and the arc main frequency, and the remaining part can be approximately regarded as a noise signal. The power calculation method is also a square mean estimation. Then, the signal-to-noise ratio parameter is calculated according to the ratio of signal power to noise power:

[0043] wherein, represents the noise power.

[0044] Finally, the obtained SNR is in decibels (dB), and the larger the value, the more prominent the arc characteristic component in the signal and the weaker the noise interference; when the value is small, it indicates that the signal is severely masked by noise and needs to be enhanced. The signal-to-noise ratio parameter of the arc current signal is not only used to measure the signal quality, but also provides a basis for the adjustment of the subsequent adaptive Hilbert transform. In addition, the signal-to-noise ratio parameter can also be used for quality control in the arc detection process. For example, when the SNR is lower than a certain threshold, the system can automatically trigger secondary sampling or enhancement processing to ensure that the acquired data can meet the needs of feature extraction. In large-scale deployment scenarios, this mechanism helps to improve the automation and intelligence level of the detection system and reduces manual intervention. Further, to improve the real-time and adaptability of the calculation, the present application also introduces a sliding window mechanism in the process of determining the signal-to-noise ratio parameter. That is, an overlapping sliding window is used to segment the signal on the time axis, and the SNR parameter is calculated and updated in real time.

[0045] Step S330, determining a dynamic window length according to the signal-to-noise ratio parameter.

[0046] wherein, the dynamic window length refers to the length of the analysis time interval that is adaptively adjusted according to the signal-to-noise ratio when performing Hilbert transform and feature extraction on the arc current signal. The reasonable selection of the window length is crucial for capturing the arc signal features: if the window is too short, it may result in insufficient frequency domain resolution, making it difficult to fully reflect the statistical characteristics of the arc signal; if the window is too long, it may smooth out the transient characteristics of the arc signal, reducing the sensitivity to the suddenness of the arc fault. Therefore, by dynamically determining the window length according to the signal-to-noise ratio parameter, a balance can be achieved between the noise environment and the signal clarity, thereby improving the accuracy and robustness of the arc feature extraction.

[0047] In one possible embodiment, the determination of the dynamic window length according to the signal-to-noise ratio parameter specifically includes the following steps: 331, obtaining a preset low signal-to-noise ratio threshold, a high signal-to-noise ratio threshold, a minimum window length and a maximum window length; 332. Intercept the arc current signal corresponding to the minimum window length from the arc current signal to obtain a first arc current signal set; 333. Intercept the arc current signal corresponding to the maximum window length from the arc current signal to obtain a second arc current signal set; 334. Calculate all first arc current signals in the first arc current signal set based on a preset root mean square estimation method to obtain a first signal power and a first noise power. 335. Calculate, based on the root mean square estimation method, a second signal power and a second noise power for all second arc current signals in the second arc current signal set; 336. Determine a first signal-to-noise ratio parameter according to the first signal power and the first noise power; 337. Determine a second signal-to-noise ratio parameter according to the second signal power and the second noise power; 338. Determine a reference signal-to-noise ratio parameter according to the first signal-to-noise ratio parameter and the second signal-to-noise ratio parameter; 339. If the signal-to-noise ratio parameter is less than or equal to the low signal-to-noise ratio threshold, determine the dynamic window length according to the maximum window length, the reference signal-to-noise ratio parameter, and the signal-to-noise ratio parameter; 3310. If the signal-to-noise ratio parameter is greater than or equal to the high signal-to-noise ratio threshold, determine the dynamic window length according to the minimum window length, the high signal-to-noise ratio threshold, and the signal-to-noise ratio parameter. 3311. If the signal-to-noise ratio parameter is greater than the low signal-to-noise ratio threshold and less than the high signal-to-noise ratio threshold, determining the dynamic window length based on a preset interpolation formula, the low signal-to-noise ratio threshold, the high signal-to-noise ratio threshold, the minimum window length, the maximum window length, and the signal-to-noise ratio parameter.

[0048] Among them, the low signal-to-noise ratio threshold and the high signal-to-noise ratio threshold are used to distinguish different intervals of signal-to-noise ratio levels, and are boundary parameters for achieving adaptive adjustment. The low signal-to-noise ratio threshold corresponds to a situation where the arc signal is interfered with by strong noise, and the high signal-to-noise ratio threshold corresponds to a situation where the arc signal has high clarity. By setting thresholds in different intervals, the dynamic window length can be adjusted differently according to the actual signal environment, thereby ensuring the accuracy and robustness of feature extraction. The minimum window length and the maximum window length are the time range boundaries determined according to the sampling frequency and the characteristics of the arc signal. The minimum window length ensures sensitivity to the sudden characteristics of the arc signal, while the maximum window length ensures the ability to smooth and suppress noise under low signal-to-noise ratio conditions. By dynamically selecting the window length, a balance between time domain resolution and frequency domain resolution can be taken into account.

[0049] Specifically, by intercepting signal segments of corresponding lengths in the original arc current signal, two signal sets of different scales are constructed. The first arc current signal set retains short-time arc characteristics and is suitable for analyzing arc burstiness. The second arc current signal set covers a longer time interval, which is helpful for smoothing noise interference and extracting overall characteristics. Then, the root mean square estimation method is used to calculate the power of the two types of signal sets. The root mean square estimation method can accurately reflect the energy level of signals and noise within a limited number of sampling points, and its results are applicable to different window lengths. By comparing the power distribution under different windows, the signal-to-noise ratio characteristics of the arc signal at different scales can be more accurately reflected. The first signal-to-noise ratio parameter and the second signal-to-noise ratio parameter correspond to the signal-to-noise ratio estimates under the minimum window length and the maximum window length, respectively, and can provide comparison information of the arc signal and noise at different scales. Then, the reference signal-to-noise ratio parameter is determined by the first signal-to-noise ratio parameter and the second signal-to-noise ratio parameter. The reference signal-to-noise ratio can be obtained by weighted averaging or other fusion methods and is used to provide a reference value in the window adjustment process. By introducing the reference signal-to-noise ratio, the influence of instantaneous fluctuations caused by signal mutations on window length selection can be reduced, and the stability of the adaptive mechanism can be improved. According to the interval of the signal-to-noise ratio level, different window adjustment strategies are used: when the signal-to-noise ratio is lower than the low threshold, the arc signal is severely disturbed by noise, and a longer window needs to be selected to enhance the signal smoothing effect. At this time, the dynamic window length is determined in the range close to the maximum window, and is adjusted in combination with the reference signal-to-noise ratio to ensure sufficient statistical reliability; when the signal-to-noise ratio is higher than the high threshold, the signal clarity is high, and the window can be shortened to improve the ability to capture arc details. The dynamic window length is determined in the range close to the minimum window, and is optimized in combination with the high signal-to-noise ratio threshold to achieve high-resolution analysis of signal details; when the signal-to-noise ratio is between the two thresholds, the noise level is moderate, and therefore the window length is dynamically adjusted between the minimum window length and the maximum window length through an interpolation formula. This interpolation method ensures that the window length changes continuously and smoothly with the signal-to-noise ratio, avoiding instability caused by parameter mutations. Specifically, the window length can be represented by the following formula:

[0050] wherein, L represents the window length, represents the maximum window length, represents the minimum window length, represents the low signal-to-noise ratio threshold, represents the high signal-to-noise ratio threshold, represents the reference signal-to-noise ratio, and k represents the adjustment coefficient.

[0051] It should be noted that the determination of the dynamic window length not only depends on the signal-to-noise ratio parameter, but also is closely related to the actual sampling conditions of the arc signal. Under different sampling frequencies, there is a conversion relationship between the time scale of the window length and the number of sampling points. For example, when the sampling frequency is 10 kHz, the minimum window length can be set to 0.05 seconds (corresponding to 500 sampling points), and the maximum window length can be set to 0.2 seconds (corresponding to 2000 sampling points). Through reasonable parameter configuration, the extracted arc features can not only reflect transient changes, but also have strong noise immunity.

[0052] In step S340, a Hilbert transform is performed on the arc current signal according to the dynamic window length to obtain m amplitude envelope data; m is a positive integer greater than 1.

[0053] The principle of the Hilbert transform is to map the original real signal to a complex analytic signal, thereby extracting the instantaneous amplitude and instantaneous phase information of the signal. In the processing of the arc current signal, the Hilbert transform can effectively obtain the amplitude envelope curve of the signal, which is used to represent the dynamic characteristics of the energy change of the current signal in the arc discharge process. Compared with traditional time or frequency domain statistical methods, the Hilbert transform has a significant advantage in capturing the transient characteristics of non-stationary signals, and is particularly suitable for electrical phenomena such as arc faults, which have strong randomness and obvious intermittence.

[0054] The amplitude envelope data is a signal instantaneous amplitude sequence calculated by performing a Hilbert transform on the arc current signal, and can directly reflect the fluctuation of the arc signal energy over time. For the load current under normal working conditions, the amplitude envelope is relatively stable and presents an approximately periodic smooth change; while under arc fault working conditions, the amplitude envelope shows sharp fluctuations and irregular fluctuations, often accompanied by sharp peaks or abrupt points. These amplitude envelope characteristics provide an important basis for subsequent statistical quantity extraction and machine learning modeling.

[0055] In one possible embodiment, the Hilbert transform of the arc current signal according to the dynamic window length to obtain m amplitude envelope data specifically includes the following steps: 341. The dynamic window length corresponding to the arc current signal is cut from the arc current signal to obtain m arc current signal segments; the time interval of each arc current signal segment in the m arc current signal segments is t ; 342. Each arc current signal segment in the m arc current signal segments is calculated based on a preset Hilbert transform formula to obtain m arc current data; wherein the Hilbert transform formula is:

[0056] wherein,x_hat(t) is the arc current data at time point t, t is the integral variable, and P is the Cauchy principal value integral, is the arc current signal segment, is the time variable; t 343、determine the amplitude envelope in each of the m arc current data, to obtain the m amplitude envelope data. wherein the arc current signal segment is a time series data segment intercepted according to a dynamic window length. The dynamic window length is adaptively adjusted according to a signal-to-noise ratio parameter, so that in a low signal-to-noise ratio scenario, the segment time interval t is relatively long to ensure smoothing of noise and enhancement of statistical stability; in a high signal-to-noise ratio scenario, the segment time interval t is relatively short to ensure that the arc transient characteristics are not eliminated by smoothing. The Hilbert transform is performed on the signal segment x(t) with a window length of L, and the calculation result is a complex signal. Through the modulus value of the complex signal, the amplitude envelope can be calculated:

[0057]

[0058] wherein,

[0059] represents the amplitude envelope, represents the original signal, and x_hat(t) represents the arc current data at time point t, is the integral variable, and P is the Cauchy principal value integral, is the arc current signal segment, is the time variable; t Specifically, the amplitude envelope data has the characteristics that under normal load current, the amplitude envelope curve presents regularity and smoothness, and the change amplitude is small; under arc fault state, the amplitude envelope curve will appear sharp peaks, mutations, irregular fluctuations and high-frequency fluctuations. These significant differences provide a basis for subsequent statistical feature extraction. For example, based on the m amplitude envelope data, envelope mean, variance, kurtosis, peak factor and other indicators can be further calculated to distinguish arc fault signals from normal load signals. It should be noted that the amplitude envelope data not only reflects the characteristics of the arc signal in the time domain, but also can complement the frequency domain characteristics. For example, the envelope curve is subjected to fast Fourier transform, and the energy distribution of the arc signal at different frequencies can be obtained. By combining the amplitude envelope characteristics and the frequency energy characteristics, a more comprehensive time-frequency domain feature vector can be formed, thereby improving the performance of the LightGBM-based classifier in arc fault recognition.

[0060]

[0061] ​Step S350, extracting statistical features of the arc current signal in the time domain from the m amplitude envelope data, obtaining p time domain statistical features; p is greater than or equal to m.

[0062] The time domain statistical features refer to a set of numerical quantification indexes calculated based on the variation law of the arc current signal amplitude envelope in the time domain. These statistical features can reflect the mean value, volatility, peak abnormality degree, and non-Gaussianity of the arc current signal in different time periods. Compared with the original current signal, the amplitude envelope data can more intuitively reveal the fluctuation law under the arc fault state, and therefore, extracting the time domain features on this basis can effectively enhance the robustness and accuracy of arc fault recognition.

[0063] The time domain statistical features include: mean (Mean), used to describe the average level of arc current signal energy in a period of time, which can reflect the stability of the overall power of the arc signal; variance (Variance), reflecting the amplitude of signal fluctuation, the arc signal usually shows greater variance characteristics; kurtosis (Kurtosis), used to measure the degree of signal distribution peak, in the arc signal, the kurtosis value is usually high, indicating that there is an abnormal peak in the waveform; skewness (Skewness), used to describe the symmetry of signal distribution, the arc signal may appear waveform offset, resulting in a difference in skewness characteristics from normal signals; crest factor (Crest Factor), the ratio of peak value to root mean square value, is an important feature for measuring the instantaneous peak energy in the arc signal; form factor (Form Factor), the ratio of root mean square value to average value, used to represent the overall waveform characteristics of the signal; impulse factor (Impulse Factor), the ratio of signal peak value to average absolute value, used to reflect the strength of abnormal pulses in the arc signal. These statistical features can reflect the non-stationarity and randomness of the arc current signal from different dimensions, so that the subsequent classification model can better distinguish between normal current and arc current.

[0064] Step S360, calculating the arc current signal by using a preset Fourier transform method to obtain arc current spectrum data.

[0065] The Fourier transform is a method of decomposing a time-domain signal into a superposition of sinusoidal and cosine components of different frequencies, thereby characterizing the distribution of signal energy in the frequency domain. Due to the strong non-stationarity and randomness of arc current signals in the time domain, it is difficult to fully characterize the arc characteristics by relying solely on time-domain statistical characteristics. In the frequency domain, arc signals often exhibit rich high-frequency components and harmonic characteristics. Therefore, the frequency spectrum data obtained by using the Fourier transform method to calculate the arc current signal can intuitively reflect the frequency distribution law of the arc signal, providing an important basis for arc fault detection and classification. The arc current spectrum data refers to the energy distribution sequence of the arc current signal at different frequencies. Compared with normal load current, the spectral characteristics of arc current have obvious differences: under normal working conditions, the spectral energy is mainly concentrated near the power frequency and its low-order harmonics, and the high-frequency energy is relatively limited; while under arc fault working conditions, the spectral energy abnormally increases in the medium and high frequency region, accompanied by irregular harmonics and broadband noise.

[0066] Specifically, the preset Fourier transform calculation process is as follows: first, the arc current signal is preprocessed. The preprocessing content includes removing the direct current component to eliminate the influence of baseline drift on spectral analysis and performing normalization processing to make signals of different amplitude levels comparable in the frequency domain; in addition, windowing methods such as Hanning window and Hamming window can be used to reduce the spectral leakage caused by window function effects in the Fourier transform process. Next, the preprocessed arc current signal is calculated using fast Fourier transform. Through power spectral density (PSD) analysis, the energy distribution of the signal at different frequencies can be further quantified. The calculation formula of power spectral density is:

[0067] where N is the signal intensity, f is the frequency, is the signal after Fourier transform, is the power spectral density.

[0068] Finally, the spectral data is sorted and output to obtain the arc current spectrum data. This data can be represented as a two-dimensional sequence of frequency-amplitude or frequency-energy, which can be used to directly observe the frequency domain characteristics of the arc signal, and also as input for subsequent feature extraction and classification modeling.

[0069] Step S370, processing the arc current spectrum data to obtain q frequency domain features; q is a positive integer greater than 1.

[0070] The frequency domain features are extracted based on the distribution law of the arc current spectrum data in different frequency intervals, and are used to depict the energy distribution, harmonic distortion degree, and spectral energy concentration of the arc current signal in the frequency domain. The frequency domain features can depict the essential characteristics of the arc signal from the frequency domain. Since the arc fault signal shows significant high-frequency components and irregular harmonics in the frequency domain, the extract of the frequency domain features can significantly improve the distinguishability of the arc fault and the normal load state. The frequency domain features include: frequency band energy distribution features, which divide the frequency spectrum into several sub-bands, calculate the energy proportion of each frequency band, and reflect the energy distribution of the arc signal in different frequency intervals; the ratio of the amplitude of the main frequency point or the harmonic point to the amplitude of the fundamental wave, which quantifies the harmonic distortion degree of the arc signal; spectral entropy, which uses the definition of information entropy to measure the uniformity of the frequency spectrum energy distribution, and the higher the spectral entropy value, the more dispersed the frequency spectrum, which can reflect the complexity of the arc signal; the center frequency and the spectral shift, which reflect the overall shift of the frequency spectrum distribution by calculating the weighted average frequency of the frequency spectrum energy; and the bandwidth feature, which calculates the effective bandwidth of the energy distribution in the frequency spectrum to depict the frequency domain expansion degree of the arc signal. These frequency domain features not only reflect the difference between the arc signal and the normal current, but also provide high-distinguishability feature input for subsequent machine learning models.

[0071] In one possible embodiment, the processing of the arc current spectrum data to obtain q frequency domain features specifically includes the following steps: 371. Obtain historical arc current fault signals; 372. Extract high-frequency pulse features from the historical arc current fault signals; 373. Transform the historical arc current fault signals using the Fourier transform method, and extract harmonic distribution features to obtain the harmonic distribution features; 374. Determine the historical arc current energy distribution features corresponding to the historical arc current fault signals according to the high-frequency pulse features and the harmonic distribution features; 375. Determine the historical power spectral density corresponding to the historical arc current energy distribution features; 376. Determine adaptive frequency band parameters based on a preset significant frequency band detection method, a preset energy threshold, and the historical power spectral density; 377. Divide the arc current spectrum data based on the adaptive frequency band parameters to obtain q first arc current spectrum data; 378. Extract frequency domain features from the q first arc current data to obtain the q frequency domain features.

[0072] The historical arc current fault signal refers to a plurality of groups of representative arc fault current signals collected and stored by a monitoring device during long-term operation of a power system. The historical arc current fault signal has typical high-frequency pulse and harmonic distortion characteristics, and thus can be used as a priori sample for frequency domain feature extraction. The high-frequency pulse feature is a transient high-frequency component generated during the intermittent process of arc discharge. Specifically, the high-frequency pulse feature can be extracted by a band-pass filter, short-time Fourier transform or wavelet packet decomposition. The main contents include the amplitude, occurrence frequency and duration of the pulse, which reflect the burstiness and instability of high-frequency energy in the arc fault process. The harmonic distribution feature refers to the amplitude distribution and proportion of each order harmonic in the current frequency spectrum obtained by Fourier transform. The arc fault signal usually shows obvious fundamental wave distortion and significant increase in high-order harmonic content, so the harmonic distribution feature can directly reflect the difference between the arc signal and the normal current signal. Specifically, the harmonic distribution feature includes total harmonic distortion (THD), odd harmonic proportion, even harmonic proportion, 3rd / 5th harmonic energy proportion, etc.

[0073] The energy distribution feature refers to the regularity index of the energy distribution of the signal in the frequency spectrum range. The transient energy corresponding to the high-frequency pulse feature and the steady-state energy corresponding to the harmonic distribution feature can be weighted and fused to obtain the energy distribution curve of the historical arc signal. This curve can reflect the energy proportion relationship of the arc signal in the low-frequency region (fundamental wave and low-order harmonic), the medium-frequency region (medium-order harmonic) and the high-frequency region (high-frequency pulse). PSD is used to represent the power distribution of the signal in a unit frequency bandwidth. By calculating the frequency spectrum power of the historical arc current energy distribution feature, the power spectrum density function can be obtained. The power spectrum density can quantify the concentration degree and extension range of the signal energy at different frequencies, and usually shows that the arc signal has an energy peak in a certain frequency band.

[0074] Specifically, after obtaining the above features, they are combined to form a q-dimensional frequency domain feature vector, which provides a basis for the input of subsequent arc fault detection. It should be noted that the size of q can be adjusted according to actual needs, either by selecting fewer frequency domain features to reduce computational complexity, or by selecting more frequency domain features to improve detection accuracy and robustness.

[0075] In one possible embodiment, the dividing the arc current spectrum data based on the adaptive frequency band parameter to obtain q first arc current spectrum data specifically includes the following steps: 3771、determining the frequency range in the adaptive frequency band parameter; 3772、extracting the spectrum data corresponding to the frequency range from the arc current spectrum data to obtain a candidate arc current data set; 3773、performing energy distribution calculation on each candidate arc current data in the candidate arc current data set to obtain a candidate arc current energy set; 3774、counting the energy parameters of all candidate arc currents in the candidate arc current energy set to obtain a candidate arc current energy parameter; 3775、determining the energy parameter in the arc current spectrum data to obtain a first arc current energy parameter; 3776、determining an energy proportion parameter according to the candidate arc current energy parameter and the first arc current energy parameter; 3777、determining a spectrum entropy parameter based on the candidate arc current energy parameter; 3778、adjusting the frequency range according to the energy proportion parameter and the spectrum entropy parameter to obtain a target frequency range; 3779、extracting arc current spectrum data of the target frequency range from the arc current spectrum data to obtain the q first arc current spectrum data.

[0076] wherein the adaptive frequency band parameter is obtained by frequency energy analysis and significant frequency band detection method of historical arc current signals, and mainly includes information such as start frequency, end frequency and center frequency. The adaptive frequency band parameter can dynamically depict the energy concentration area of the arc signal in the frequency domain.

[0077] Specifically, the candidate arc current dataset refers to the subset of spectral data within the original spectrum that falls within the adaptive frequency band. This is achieved through bandpass filtering, index filtering, or frequency bin cropping. The resulting candidate arc current dataset is significantly smaller than the original spectral data, reducing data redundancy and computational complexity while retaining core information containing arc characteristics. Energy distribution calculation is achieved by integrating or squaring the power spectral density at each frequency point, characterizing the energy distribution characteristics of the arc signal within a specific frequency bin. The candidate arc current energy set contains multiple energy values ​​that reflect the energy concentration in different frequency bins. For example, when an arc occurs, significant energy peaks often appear in certain high-frequency bins, which are important indicators for subsequent feature identification. Candidate arc current energy parameters include total energy, mean, variance, energy peak value, and its corresponding frequency location. These statistics reflect the overall shape and volatility of the energy distribution within the candidate frequency band. If the energy is concentrated in a few frequency points, it indicates a strong correlation between that bin and the arc characteristics; if the energy is more dispersed, the contribution of that bin is relatively small. The first arc current energy parameter is an overall energy indicator statistically calculated across the entire frequency spectrum and serves as a benchmark for comparing candidate interval energies. This parameter is calculated by integrating the power spectral density over the entire frequency range. The principle of frequency range adjustment is as follows: when the energy percentage is too low, the frequency band lacks energy and can be eliminated; when the spectral entropy is too high, the energy is too dispersed and the frequency band can be narrowed to concentrate on the main energy region; when the energy percentage is high and the spectral entropy is low, the frequency band is retained and appropriately expanded to enhance the expressiveness of arc characteristics. By jointly constraining the energy percentage and spectral entropy, the frequency band boundaries can be dynamically corrected, resulting in a more accurate target frequency range. The resulting q first arc current spectrum data correspond to several optimized sub-bands, each of which retains the frequency domain information most relevant to the arc characteristics. Compared to fixed frequency band division, the q first arc current spectrum data generated by this method have higher information density and stronger feature correlation, significantly improving the accuracy and robustness of subsequent frequency domain feature extraction and classification model training.

[0078] In a possible embodiment, performing frequency domain feature extraction on the q first arc current data to obtain the q frequency domain features specifically includes the following steps: 3781. Normalize the q first arc current spectrum data to obtain q second arc current spectrum data; 3782. Extracting a main harmonic amplitude and a subharmonic amplitude from each of the q second arc current spectrum data to obtain q main harmonic amplitudes and q subharmonic amplitudes; 3783、determining a proportion of each of the q harmonic amplitudes in each of the q second arc current data, obtaining q harmonic proportions; 3784、determining the q frequency domain features based on the q main harmonic amplitudes and the q harmonic proportions.

[0079] wherein the normalization processing is used to eliminate the differences of different arc current signals in sampling amplitude, spectral energy size, and ensure the comparability between different samples. The normalization method can adopt amplitude normalization, energy normalization or maximum and minimum value standardization, etc., which is not limited here. The main harmonic amplitude refers to the amplitude of the frequency component closest to the fundamental frequency and its integer multiple frequency, which is the component with the highest energy proportion and the strongest stability in the arc current spectrum. The harmonic amplitude refers to the amplitude of the significant secondary frequency component other than the main harmonic, which often appears in the arc discharge process and shows nonlinear and high-frequency pulse characteristics. Extracting the main harmonic and harmonic amplitude helps to construct the unique spectral fingerprint of the arc. By ratio operation of the harmonic amplitude and the main harmonic amplitude or the total amplitude, the harmonic proportion is obtained. The harmonic proportion can reflect the energy proportion of the arc current signal at high frequency components. When the arc occurs, the harmonic component is significantly enhanced compared to the normal working condition signal, resulting in an increase in the harmonic proportion. The harmonic proportion is an important parameter for characterizing the arc fault characteristics. After the main harmonic amplitude and the harmonic proportion are extracted, they are further combined in a vectorization manner to form a complete set of frequency domain features. The q frequency domain features contain both the low-frequency stable component (main harmonic) and the high-frequency disturbance component (harmonic proportion), which can completely characterize the energy distribution characteristics of the arc signal in the frequency domain space.

[0080] Specifically, in order to ensure the robustness and real-time performance of the frequency domain feature extraction, first, the main harmonic detection method is optimized, and the fast Fourier transform combined with the peak value search algorithm is adopted to quickly locate the amplitude corresponding to the main harmonic frequency; when there is a power grid frequency drift, an adaptive filtering method can be introduced to dynamically track the main harmonic frequency. Next, the harmonic recognition method is optimized, and the threshold detection and energy clustering are combined to accurately identify the harmonic component in multiple frequency components, avoiding the interference of normal background noise; then, the proportion calculation is standardized, in order to eliminate the influence of different sampling window sizes on the harmonic proportion, energy normalization processing can be introduced to make the calculation results consistent across working conditions; finally, the feature vector is constructed, and the q frequency domain feature vectors are finally obtained, which can reflect the main energy distribution of the arc current spectrum and reveal the high-frequency disturbance characteristics when the arc occurs, which is represented as:

[0081] wherein, denotes the corresponding main harmonic proportion, denotes the corresponding sub-harmonic proportion, F denotes the frequency domain feature.

[0082] Step S380, based on the preset feature fusion algorithm, the p time domain statistical features and the q frequency domain features are fused to obtain a target arc current feature.

[0083] The purpose of the fusion processing is to uniformly model the multi-dimensional information from the time domain and the frequency domain, and form a feature vector with higher discrimination. The arc current signal is irregular in the time domain, the mean value is offset, and the fluctuation is enhanced. In the frequency domain, the main harmonic and sub-harmonic features are prominent, and the energy distribution is uneven. Therefore, the joint representation of time domain and frequency domain features helps to fully characterize the dynamic characteristics and spectral characteristics of the arc fault signal, thereby providing high-quality input for subsequent detection and recognition based on machine learning models. The feature fusion algorithm can be a simple feature-level splicing method, or a fusion method based on a weighting strategy, or a fusion method realized through a dimension reduction algorithm (such as principal component analysis PCA, linear discriminant analysis LDA) or a deep learning feature encoding method (such as autoencoder, attention mechanism), which is not limited here. The target arc current feature is a comprehensive feature vector after fusion, which contains p time domain statistical features (such as mean, variance, kurtosis, skewness, short-time energy, etc.) and q frequency domain features (such as main harmonic amplitude, sub-harmonic proportion, spectral energy proportion, spectral entropy, etc.), and can fully reflect the performance of the arc signal in different domains.

[0084] In one possible embodiment, the preset feature fusion algorithm is used to fuse the p time domain statistical features and the q frequency domain features to obtain a target arc current feature, which specifically includes the following steps: 381, obtaining historical arc current data; 382, extracting time domain statistical features from the historical arc current data to obtain a plurality of historical time domain statistical features; 383, performing feature frequency domain extraction on the historical arc current data to obtain a plurality of historical frequency domain features; 384, constructing a historical arc current data set according to the plurality of historical time domain statistical features, the plurality of historical frequency domain features, and the historical arc current data; 385, inputting the historical arc current data set into a preset dimension reduction model to obtain a dimension reduction projection matrix; 386、inputting the plurality of historical time domain statistical features and the plurality of historical frequency domain features into a preset lightweight machine learning classification model respectively to obtain a first classification accuracy and a second classification accuracy; the first classification accuracy is a classification accuracy output by the lightweight machine learning classification model according to the plurality of historical time domain statistical features; the second classification accuracy is a classification accuracy output by the lightweight machine learning classification model according to the plurality of historical frequency domain features; 387、determining a time domain feature weight and a frequency domain feature weight according to the first classification accuracy and the second classification accuracy; 388、determining a first arc current feature according to the p time domain statistical features, the q frequency domain features, the time domain feature weight and the frequency domain feature weight; 389、projecting the first arc current feature based on the dimension reduction projection matrix to obtain the target arc current feature.

[0085] wherein the historical arc current data is a data set obtained by long-term collection of current waveforms with arc fault or similar load disturbance in a power system, the data containing a large number of current time series signals under different working conditions, including both normal load fluctuation current and abnormal current signal generated by arc discharge. In these data, the arc fault current usually has the characteristics of short-time burst, irregular amplitude and spectrum diffusion, and is therefore very suitable as a basic sample for training and verification. The dimension reduction model refers to a mathematical model capable of extracting the most representative information from a high-dimensional feature space, which can be a statistical method based on principal component analysis (PCA), linear discriminant analysis (LDA), or a deep learning method based on autoencoder, convolutional neural network, etc., without limitation. The dimension reduction model maps the high-dimensional time domain and frequency domain features to a low-dimensional subspace by generating a projection matrix, thereby reducing feature redundancy and improving the compactness of feature expression and the generalization performance of the classifier. The lightweight machine learning classification model can be LightGBM, XGBoost, random forest or support vector machine, etc. capable of maintaining high efficiency under limited computing resources. In the arc fault detection scenario, LightGBM can reduce the computational complexity while ensuring the classification accuracy due to its efficient decision tree construction mechanism based on gradient boosting, and is therefore particularly suitable for embedded or edge computing devices. The time domain feature weight and the frequency domain feature weight are quantitative coefficients obtained by comparing the classification accuracy of the classification model when using a single feature source.

[0086] Specifically, first, historical arc current data is generated by field collection or simulation platform, and preprocessed, such as denoising, normalization and slicing, to ensure the consistency and availability of the data. Time domain statistical features are extracted from the preprocessed arc current data, and then the same batch of historical arc current data is analyzed in the frequency domain, and frequency domain features are extracted by using fast Fourier transform (FFT), wavelet packet decomposition and other methods. Then, the extracted historical time domain statistical features, historical frequency domain features and corresponding original arc current data are constructed into a complete historical arc current data set. The data set contains input features and corresponding labels (normal state or arc fault state), which is used for subsequent model training and evaluation, and the historical arc current data set is input into a dimension reduction model to obtain a dimension reduction projection matrix. For example, if the PCA method is used, the projection matrix is composed of the first several eigenvectors of the data covariance matrix.

[0087] Then, the historical time domain statistical features and the historical frequency domain features are used as inputs respectively to train a lightweight machine learning classification model, and the classification accuracy on the validation set is calculated. The first classification accuracy and the second classification accuracy respectively reflect the contribution of the time domain and frequency domain features to the arc fault classification, and the weights of the time domain features and the weights of the frequency domain features are determined by using a weighted average method according to the first classification accuracy and the second classification accuracy. The p time domain statistical features and the q frequency domain features are weighted to obtain the first arc current feature. The feature vector numerically integrates the time domain and frequency domain information and adjusts the importance of the two through the weight. Its weight allocation formula can be expressed as:

[0088] represents the accuracy obtained by inputting only the time domain features into the lightweight classification model, represents the accuracy obtained by inputting only the frequency domain features into the lightweight classification model.

[0089] Finally, the first arc current feature is input into the dimension reduction projection matrix to complete the feature space mapping and obtain the low-dimensional target arc current feature. The feature not only retains the discriminative information of the original feature, but also reduces the redundancy, improves the detection efficiency and the generalization performance of the model.

[0090] Step S390, inputting the target arc current feature into a preset lightweight arc fault detection model to obtain an arc fault detection result.

[0091] ​The lightweight arc fault detection model refers to a machine learning or deep learning model that can operate under low computational complexity, low storage overhead, and high real-time requirements while ensuring arc fault detection accuracy. The model can be a LightGBM model based on gradient boosting decision trees, an XGBoost model, an improved support vector machine model, or a lightweight convolutional neural network model suitable for embedded environments. Preferably, the LightGBM model is used as the lightweight arc fault detection model. This model has faster training speed and lower resource consumption, and can meet the application requirements of rapid detection of arc faults in power distribution systems.

[0092] Specifically, first, the obtained target arc current features are input as input vectors into the lightweight arc fault detection model. The target arc current features integrate the weighted fusion information of time domain statistical features and frequency domain features and are processed by dimension reduction to ensure the discriminability and compactness of the feature vectors, thereby improving the effectiveness of model input. In the model inference stage, when a new target arc current feature is input into the model, the model will perform feature matching and classification calculation based on its internal decision tree structure or classification network structure, and finally output the corresponding arc fault detection result. The detection result can be a binary classification result (i.e., arc fault / non-arc fault), or a multi-class classification result (e.g., series arc fault, parallel arc fault, transient disturbance, etc.). Further, the arc fault detection result can include a probability value of arc fault occurrence. The model can not only output whether the current signal at a certain time exists arc fault, but also give the confidence score of the detection result, thereby providing a reference for subsequent protection strategy or scheduling decision.

[0093] For ease of understanding, please refer to Figure 4 , Figure 4 is a structural diagram of a feature extraction unit provided by an embodiment of the present application. As can be seen, Figure 4The internal composition of the feature extraction unit and the hierarchical relationship of each module are presented, the complete processing flow from current signal input to time-frequency domain feature extraction is shown, and the division of labor and cooperation mechanism of time domain and frequency domain feature extraction are clarified. Specifically, the feature extraction unit 400 as the core processing unit is responsible for the extraction process of time domain and frequency domain features, and it contains two parallel processing sub-units: the Hilbert transform unit 410 and the fast Fourier transform unit 420. Among them, the Hilbert transform unit 410 is time domain feature extraction, which is composed of three functional modules: the signal-to-noise ratio estimation module 411, which is used for segmented processing of the input current signal, and the signal-to-noise ratio (SNR) is calculated to provide a basis for subsequent window length adjustment; the dynamic window adjustment module 412 dynamically selects the window length according to the signal-to-noise ratio estimation result (a longer window length is used to smooth the noise when the signal-to-noise ratio is low, a shorter window length is used to retain local features when the signal-to-noise ratio is high, and the window length is determined by linear interpolation in the middle region); the Hilbert transform calculation module 413 performs Hilbert transform on the signal based on the adjusted window length, and the analytic signal is constructed through fast Fourier transform (FFT), and finally the amplitude envelope of the current signal is extracted. The fast Fourier transform unit 420 is used for frequency domain feature extraction, which includes four functional modules: the signal preprocessing module 421, which performs mean removal and normalization processing on the input current signal to reduce the influence of baseline drift and amplitude difference on subsequent analysis; the Fourier calculation module 422 performs fast Fourier transform (FFT) on the preprocessed signal to obtain the frequency spectrum distribution of the signal; the dynamic frequency band segmentation module 423 adaptively divides the significant frequency band according to the power spectrum density (PSD) calculation result through the energy threshold (determined based on the empirical coefficient of the power spectrum mean and standard deviation); the frequency domain feature extraction module 424 calculates the normalized energy, main harmonic amplitude and sub-harmonic ratio of each frequency band based on the divided frequency band, and provides key input for the subsequent classification model. Through the cooperative work of the Hilbert transform unit and the fast Fourier transform unit, the efficient extraction of time-frequency domain features of the current signal is realized, which provides a comprehensive feature input basis for the accurate discrimination of the LightGBM classification model.

[0094] For ease of understanding, please refer to Figure 5 , Figure 5 is a structural schematic diagram of an alternating current arc fault detection system provided by an embodiment of the present application. As can be seen, Figure 5The complete circuit structure of arc generation and detection from the AC power supply to the load end is clearly presented, including the connection mode of the signal source, the load, the measurement module, and the arc generation unit, and the process of arc signal acquisition, processing and analysis is presented. Specifically, the AC arc fault detection system includes: an AC power supply (220V / 50Hz power supply), controllable switches (S1 and S2), an arc generation device (point contact arc generator), a detection device (current transformer), a display device (RIGOL MSO5000 oscilloscope), and a load. The AC power supply serves as the signal input end of the system and is used to provide stable AC current to simulate the power supply process in the actual power distribution network. The arc generation device is connected in series between the power supply and the load. The device can intentionally cause arc phenomenon by controlling the on-off of the switch or the gap breakdown, thereby forming a typical arc fault environment for detection and analysis. At the output end of the arc generation device, a detection module is connected. The detection module can be a current transformer, a voltage transformer, or a high-precision sampling circuit, and its function is to collect and preprocess the arc current signal in real time. The original current signal collected is transmitted to the display device and the subsequent data processing unit after being filtered and amplified to ensure signal quality and the accuracy of subsequent feature extraction. Further, the display device uses a digital oscilloscope or a power analyzer to display the arc current waveform, spectral distribution, and amplitude characteristics in real time, which facilitates the observation of the change law of the arc signal at different times. By monitoring the display results, researchers can intuitively determine whether an arc occurs, its duration, and its characteristic strength. In addition, the display device can also output digitized arc current data to provide experimental data support for the subsequent arc fault detection method based on Hilbert transform and LightGBM algorithm. At the load end, a variable impedance load or a typical household appliance model is connected to simulate the performance characteristics of arc faults under different operating scenarios. For example, when the load impedance is small, the high-frequency component of the arc signal is significantly enhanced; when the load impedance is large, the arc signal exhibits obvious intermittent characteristics. By adjusting the load parameters, different working conditions of arc faults in the actual power grid can be simulated to provide diverse data samples for algorithm training and model verification. In addition, the arc generation device is internally provided with controllable switches or mechanical contact devices to periodically or randomly create arc discharge processes. By adjusting the switch action frequency and contact resistance, strong arc, weak arc, or intermittent arc signals can be obtained.

[0095] For ease of understanding, please refer to Figure 6 , Figure 6The flowchart of the feature extraction and classification process of the alternating current arc fault detection method provided by the embodiment of the application is shown. Specifically, the arc current signal is taken as input data, which first enters the signal processing link. In this process, the arc current features are processed through two feature channels: first, the Hilbert transform is used to extract the envelope features of the arc signal, and the instantaneous amplitude and instantaneous energy distribution of the arc current are obtained; second, the Fourier transform is used for frequency spectrum analysis of the arc current, and the amplitude features and harmonic components of the arc current at different frequency bands are obtained. The Hilbert transform can effectively reveal the instability and intermittency characteristics of the arc signal in the time domain, while the Fourier transform can depict the harmonic distortion characteristics of the arc signal in the frequency domain. The combination of the two can comprehensively represent the multi-dimensional features of the arc fault. After the transformation is completed, all time domain and frequency domain features are input to the feature extraction module for feature extraction. Feature extraction normalizes, statistically calculates, and selects features from the above data to ensure that the obtained feature vector has high discriminability and robustness. For example, it can extract the main harmonic amplitude, sub-harmonic ratio, spectral entropy parameter, root mean square value, kurtosis, skewness, and other time-frequency domain statistical features of the arc current. Through the comprehensive expression of these features, the arc fault state and the normal working state can be effectively distinguished, and the limitation that a single feature cannot fully reflect the characteristics of the arc is overcome. However, the feature vector after feature extraction is input into the LightGBM (LGBM) lightweight machine learning model. LightGBM is based on the idea of gradient boosting decision tree, and has the advantages of efficient training, low memory occupation, and strong generalization ability, especially suitable for realizing rapid arc fault detection in limited hardware resource scenarios. In the model inference stage, LGBM traverses the decision tree layer by layer according to the input feature vector, and finally outputs the detection result. If the model judges that the signal features are highly consistent with the normal working condition features, it outputs “Normal”; if the features have a high matching degree with the arc working condition features, it outputs “Arc”, thereby realizing accurate identification of the arc fault. The detection result not only can provide real-time early warning for power grid operation, but also can provide decision basis for subsequent protection control actions.

[0096] For ease of understanding, please refer to Figure 7 , Figure 7is a LightGBM-based arc fault detection process schematic diagram provided by an embodiment of the present application. As can be seen, first, the alternating current signal is collected in real time, and the collected alternating current signal is input to a preprocessing module for processing. Among them, the collected arc current signal is a waveform sequence, and is input to the preprocessing module for processing. The main role of preprocessing is to remove the common background noise, power frequency interference and random disturbance in the power system, so as to ensure the accuracy and stability of subsequent feature extraction. Then, through the method of adaptive Hilbert transform, the signal-to-noise ratio is analyzed, and the window is dynamically adjusted to optimize the extraction of amplitude envelope. Specifically, based on the adaptive frequency band parameter, the arc current signal is divided, the energy features and spectral entropy features in different frequency bands are extracted, and the preliminary multi-dimensional feature vector is constructed. Through this adaptive division method, the high-frequency feature information related to the arc can be highlighted while reducing redundant features, which helps to improve the discrimination of feature representation. Then, the real-time collected alternating current signal is analyzed by using fast Fourier transform, and through the dynamic frequency band segmentation technology, the analysis frequency band is adaptively selected according to the characteristics of the arc fault signal to extract the frequency domain features, and the time domain and frequency domain statistical features are combined to extract the feature vector. Specifically, after completing the preliminary feature extraction, the feature fusion and optimization mechanism is further introduced. First, the time domain statistical features and frequency domain features are combined by using the weighted fusion strategy, and the balance and complementarity between different features are realized by reasonably allocating weight parameters; then, the principal component analysis (PCA) method is used to reduce the dimension of the fused high-dimensional features, and the main component features are retained to reduce the calculation overhead and avoid the interference of feature correlation. Through a series of steps, the optimized target arc current features are finally obtained. Finally, the LightGBM lightweight gradient boosting decision tree model is used to classify and train the feature vectors, and the discrimination results of arc fault or normal state are output. Specifically, the optimized features are input to the classification module for modeling and identification, and the lightweight gradient boosting tree (LightGBM) model is used as the classifier, which can significantly reduce the calculation complexity while maintaining high classification accuracy, suitable for deployment scenarios of embedded or edge computing devices, and improves the detection accuracy.

[0097] It can be seen that the arc fault detection method based on lightweight described in the application is applied to an electronic device. The method obtains an arc current signal of a load in a target power distribution network in a preset time period, determines a signal-to-noise ratio parameter of the arc current signal, determines a dynamic window length according to the signal-to-noise ratio parameter, performs Hilbert transform on the arc current signal according to the dynamic window length, obtains m amplitude envelope data, m is a positive integer greater than 1, extracts statistical features of the arc current signal in the time domain from the m amplitude envelope data, obtains p time domain statistical features, p is greater than or equal to m, calculates the arc current signal by using a preset Fourier transform method, obtains arc current spectrum data, processes the arc current spectrum data, obtains q frequency domain features, q is a positive integer greater than 1, fuses the p time domain statistical features and the q frequency domain features based on a preset feature fusion algorithm, obtains a target arc current feature, inputs the target arc current feature into a preset lightweight arc fault detection model, and obtains an arc fault detection result. In this way, on the one hand, the current signal envelope is extracted by using Hilbert transform, statistical quantities such as envelope mean value, peak factor, kurtosis are further calculated, and indexes such as harmonic components in a specific frequency band and spectral energy ratio are extracted by using fast Fourier transform, and time domain and frequency domain features are comprehensively predicted to improve the accuracy of arc fault detection. On the other hand, a lightweight classification model is used to detect normal arc current and fault arc current from the extracted features, thereby reducing the calculation overhead of detecting arc faults.

[0098] The above describes the scheme of the embodiments of the application mainly from the perspective of the execution process of the method. It can be understood that the electronic device includes a hardware structure and / or a software module corresponding to the execution of each function to implement the above functions. Those skilled in the art should easily realize that, in combination with the unit and algorithm steps of each example described in the embodiments provided herein, the application can be realized in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the application.

[0099] The embodiments of the application can divide the functional units of the electronic device according to the above method examples. For example, each functional unit can be divided according to each function, or two or more functions can be integrated in one processing unit. The integrated unit can be realized in the form of hardware or software functional unit. It should be noted that the division of units in the embodiments of the application is illustrative, and is only a logical function division. There can be another division method when actually implemented.

[0100] In the case of dividing each functional module according to each function, Figure 8 is a functional module composition block diagram of an arc fault detection device based on lightness provided by an embodiment of the present application. The arc fault detection device 800 based on lightness comprises: The acquisition module 810 is configured to acquire an arc current signal of a load in a target power distribution network within a preset time period. The determination module 820 is configured to determine a signal-to-noise ratio parameter of the arc current signal, determine a dynamic window length according to the signal-to-noise ratio parameter, and calculate the arc current signal by using a preset Fourier transform method to obtain arc current spectrum data. The calculation module 830 is configured to perform Hilbert transform on the arc current signal according to the dynamic window length to obtain m amplitude envelope data, where m is a positive integer greater than 1; extract statistical features of the arc current signal in the time domain from the m amplitude envelope data to obtain p time domain statistical features, where p is greater than or equal to m; process the arc current spectrum data to obtain q frequency domain features, where q is a positive integer greater than 1; and fuse the p time domain statistical features and the q frequency domain features based on a preset feature fusion algorithm to obtain target arc current features. The control module 840 is configured to input the target arc current features into a preset light arc fault detection model to obtain an arc fault detection result.

[0101] In a possible embodiment, the determination module 820, in terms of determining a dynamic window length according to a signal-to-noise ratio parameter, is specifically configured to: acquire a preset low signal-to-noise ratio threshold, a preset high signal-to-noise ratio threshold, a minimum window length, and a maximum window length; cut the arc current signal corresponding to the minimum window length from the arc current signal to obtain a first arc current signal set; cut the arc current signal corresponding to the maximum window length from the arc current signal to obtain a second arc current signal set; calculate all first arc current signals in the first arc current signal set based on a preset root mean square estimation method to obtain a first signal power and a first noise power; calculate all second arc current signals in the second arc current signal set based on the root mean square estimation method to obtain a second signal power and a second noise power; determine a first signal-to-noise ratio parameter according to the first signal power and the first noise power; determine a second signal-to-noise ratio parameter according to the second signal power and the second noise power; determine a reference signal-to-noise ratio parameter according to the first signal-to-noise ratio parameter and the second signal-to-noise ratio parameter; If the signal-to-noise ratio parameter is less than or equal to the low signal-to-noise ratio threshold, determining the dynamic window length according to the maximum window length, the reference signal-to-noise ratio parameter, and the signal-to-noise ratio parameter; If the signal-to-noise ratio parameter is greater than or equal to the high signal-to-noise ratio threshold, determining the dynamic window length according to the minimum window length, the high signal-to-noise ratio threshold, and the signal-to-noise ratio parameter; If the signal-to-noise ratio parameter is greater than the low signal-to-noise ratio threshold and less than the high signal-to-noise ratio threshold, the dynamic window length is determined based on a preset interpolation formula, the low signal-to-noise ratio threshold, the high signal-to-noise ratio threshold, the minimum window length, the maximum window length, and the signal-to-noise ratio parameter.

[0102] In a possible embodiment, the calculation module 830, in performing Hilbert transform on the arc current signal according to the dynamic window length to obtain m amplitude envelope data, is specifically configured to: The arc current signal corresponding to the dynamic window length is intercepted from the arc current signal to obtain m arc current signal segments; the time interval of each arc current signal segment in the m arc current signal segments is t ; Each of the m arc current signal segments is calculated based on a preset Hilbert transform formula to obtain m arc current data; wherein the Hilbert transform formula is:

[0103] in, x_hat(t) For time point t Arc current data, is the integral variable, P is the Cauchy principal value integral, is the arc current signal segment, t is a time variable; An amplitude envelope of each of the m arc current data is determined to obtain the m amplitude envelope data.

[0104] In a possible embodiment, the calculation module 830, in processing the arc current spectrum data to obtain q frequency domain features, is specifically configured to: Obtain historical arc current fault signals; Extracting high-frequency pulse features from the historical arc current fault signal to obtain high-frequency pulse features; The historical arc current fault signal is transformed by using the Fourier transform method, and the frequency domain harmonic distribution characteristics are extracted to obtain the harmonic distribution characteristics; determine a historical arc current energy distribution feature corresponding to the historical arc current fault signal according to the high-frequency pulse feature and the harmonic distribution feature; determine a historical power spectral density corresponding to the historical arc current energy distribution feature; determine an adaptive frequency band parameter based on a preset significant frequency band detection method, a preset energy threshold, and the historical power spectral density; divide the arc current spectrum data based on the adaptive frequency band parameter to obtain q first arc current spectrum data; extract frequency domain features from the q first arc current data to obtain q frequency domain features.

[0105] In a possible implementation, the computing module 830, in the aspect of dividing the arc current spectrum data based on the adaptive frequency band parameter to obtain q first arc current spectrum data, is specifically configured to: determine a frequency range in the adaptive frequency band parameter; extract spectrum data corresponding to the frequency range from the arc current spectrum data to obtain a candidate arc current data set; perform energy distribution calculation on each candidate arc current data in the candidate arc current data set to obtain a candidate arc current energy set; statistically determine energy parameters of all candidate arc currents in the candidate arc current energy set to obtain a candidate arc current energy parameter; determine an energy parameter in the arc current spectrum data to obtain a first arc current energy parameter; determine an energy proportion parameter according to the candidate arc current energy parameter and the first arc current energy parameter; determine a spectral entropy parameter based on the candidate arc current energy parameter; adjust the frequency range according to the energy proportion parameter and the spectral entropy parameter to obtain a target frequency range; extract arc current spectrum data of the target frequency range from the arc current spectrum data to obtain the q first arc current spectrum data.

[0106] In a possible implementation, the computing module 830, in the aspect of fusing the p time domain statistical features and the q frequency domain features based on the preset feature fusion algorithm to obtain target arc current features, is specifically configured to: obtain historical arc current data; extract time domain statistical features from the historical arc current data to obtain a plurality of historical time domain statistical features; perform feature frequency domain extraction on the historical arc current data to obtain a plurality of historical frequency domain features; construct a historical arc current dataset according to the plurality of historical time domain statistical features, the plurality of historical frequency domain features, and the historical arc current data; input the historical arc current dataset into a preset dimension reduction model to obtain a dimension reduction projection matrix; input the plurality of historical time domain statistical features and the plurality of historical frequency domain features into preset lightweight machine learning classification models respectively to obtain a first classification accuracy and a second classification accuracy; the first classification accuracy is a classification accuracy output by the lightweight machine learning classification model according to the plurality of historical time domain statistical features; the second classification accuracy is a classification accuracy output by the lightweight machine learning classification model according to the plurality of historical frequency domain features; determine a time domain feature weight and a frequency domain feature weight according to the first classification accuracy and the second classification accuracy; determine a first arc current feature according to the p time domain statistical features, the q frequency domain features, the time domain feature weight, and the frequency domain feature weight; project the first arc current feature based on the dimension reduction projection matrix to obtain the target arc current feature.

[0107] It can be seen that the embodiment of the present application describes an arc fault detection device based on lightweight. The envelope of the current signal is extracted by using Hilbert transform, and further statistical quantities such as envelope mean value, peak factor, kurtosis are calculated, and indicators such as harmonic component in a specific frequency band and spectral energy ratio are extracted by using fast Fourier transform, and time domain and frequency domain features are comprehensively input into a lightweight classification prediction model, thereby improving the accuracy of arc fault detection and reducing the calculation overhead of arc fault detection.

[0108] The embodiment of the present application further provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program for electronic data exchange, and the computer program causes a computer to execute part or all steps of any method described in the above method embodiment, and the computer includes an electronic device.

[0109] The embodiment of the present application further provides a computer program product, and the computer program product includes a non-transitory computer readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute part or all steps of any method described in the above method embodiment. The computer program product can be a software installation package, and the computer includes an electronic device.

[0110] It should be noted that, for the above-mentioned various embodiments, in order to simply describe, all are expressed as a series of action combinations. Those skilled in the art should know that the present application is not limited to the action sequence described, because some steps in the embodiments of the present application can be performed in other sequences or simultaneously. In addition, those skilled in the art should know that the embodiments described in the specification all belong to preferred embodiments, and the actions, steps, modules or units involved are not necessarily required in the embodiments of the present application.

[0111] In the above embodiments, the description of each embodiment of the embodiments of the present application has its own focus, and the parts not described in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0112] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by a computer program instructing relevant hardware to complete, and the program can be stored in a computer readable storage medium. When the program is executed, it can include the processes of the above-mentioned method embodiments. The storage medium mentioned above includes: ROM or random access memory (RAM), magnetic disk or optical disk and various program code storage media.

[0113] The steps of the method or algorithm described in the embodiments of the present application can be implemented in the form of hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, and the software modules can be stored in RAM, flash memory, ROM, EPROM, electrically EPROM (EEPROM), register, hard disk, mobile hard disk, read-only optical disk (CD-ROM) or any other form of storage medium well known in the art. An exemplary storage medium is coupled to the processor, so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be an integral part of the processor. The processor and the storage medium can be located in an ASIC. In addition, the ASIC can be located in a terminal device or a management device. Of course, the processor and the storage medium can also exist as discrete components in the terminal device or the management device.

[0114] Those skilled in the art should be able to understand that, in one or more of the above examples, the functions described in the embodiments of the present application can be implemented wholly or partially by software, hardware, firmware, or any combination thereof. When implemented by software, it can be implemented in the form of a computer program product wholly or partially. The computer program product includes one or more computer instructions. When loaded and executed on a computer, the computer instructions wholly or partially generate the processes or functions described in the embodiments of the present application. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable apparatus. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transferred from one website, computer, electronic device, or data center to another website, computer, electronic device, or data center through a wired (for example, coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (for example, infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium accessible by a computer or data storage device, such as an electronic device, data center, etc. that includes one or more available media.

[0115] Each module / unit included in each device / product described in the above embodiments can be a software module / unit or a hardware module / unit, or part of a software module / unit and part of a hardware module / unit. For example, for each device / product applied to or integrated in a chip, each module / unit included therein can be implemented in the form of hardware such as a circuit, or at least part of the modules / units can be implemented in the form of a software program running on a processor integrated in the chip, and the remaining (if any) part of the modules / units can be implemented in the form of hardware such as a circuit; for each device / product applied to or integrated in a chip module, each module / unit included therein can be implemented in the form of hardware such as a circuit, and different modules / units can be located in the same component (e.g., a chip, a circuit module, etc.) or different components of the chip module, or at least part of the modules / units can be implemented in the form of a software program running on a processor integrated in the chip module, and the remaining (if any) part of the modules / units can be implemented in the form of hardware such as a circuit; for each device / product applied to or integrated in a terminal device, each module / unit included therein can be implemented in the form of hardware such as a circuit, and different modules / units can be located in the same component (e.g., a chip, a circuit module, etc.) or different components of the terminal device, or at least part of the modules / units can be implemented in the form of a software program running on a processor integrated in the terminal device, and the remaining (if any) part of the modules / units can be implemented in the form of hardware such as a circuit.

[0116] The above detailed description has further explained the purposes, technical solutions and beneficial effects of the embodiments of the present application. It should be understood that the above description is only a specific implementation of the embodiments of the present application and is not intended to limit the protection scope of the embodiments of the present application. Any modification, equivalent replacement, improvement, etc. made on the basis of the technical solutions of the embodiments of the present application shall be included in the protection scope of the embodiments of the present application.

Claims

1. A lightweight arc fault detection method, characterized in that: Applied to electronic equipment, the method includes: Obtaining arc current signals of loads in a target distribution network within a preset time period; determining a signal-to-noise ratio parameter of the arc current signal; Determining a dynamic window length according to the signal-to-noise ratio parameter; Performing Hilbert transform on the arc current signal according to the dynamic window length to obtain m amplitude envelope data, where m is a positive integer greater than 1; Extracting statistical features of the arc current signal in the time domain from the m amplitude envelope data to obtain p time domain statistical features; p is greater than or equal to m; Calculating the arc current signal using a preset Fourier transform method to obtain arc current spectrum data; Processing the arc current spectrum data to obtain q frequency domain features, where q is a positive integer greater than 1; The p time-domain statistical features and the q frequency-domain features are fused based on a preset feature fusion algorithm to obtain a target arc current feature; The target arc current characteristics are input into a preset lightweight arc fault detection model to obtain an arc fault detection result.

2. The method according to claim 1, wherein The determining the dynamic window length according to the signal-to-noise ratio parameter includes: Get the preset low signal-to-noise ratio threshold, high signal-to-noise ratio threshold, minimum window length, and maximum window length; intercepting the arc current signal corresponding to the minimum window length from the arc current signal to obtain a first arc current signal set; intercepting the arc current signal corresponding to the maximum window length from the arc current signal to obtain a second arc current signal set; Calculating all first arc current signals in the first arc current signal set based on a preset root mean square estimation method to obtain a first signal power and a first noise power; calculating, based on the root mean square estimation method, a second signal power and a second noise power for all second arc current signals in the second arc current signal set; determining a first signal-to-noise ratio parameter according to the first signal power and the first noise power; determining a second signal-to-noise ratio parameter according to the second signal power and the second noise power; Determine a reference signal-to-noise ratio parameter according to the first signal-to-noise ratio parameter and the second signal-to-noise ratio parameter; If the signal-to-noise ratio parameter is less than or equal to the low signal-to-noise ratio threshold, determining the dynamic window length according to the maximum window length, the reference signal-to-noise ratio parameter, and the signal-to-noise ratio parameter; If the signal-to-noise ratio parameter is greater than or equal to the high signal-to-noise ratio threshold, determining the dynamic window length according to the minimum window length, the high signal-to-noise ratio threshold, and the signal-to-noise ratio parameter; If the signal-to-noise ratio parameter is greater than the low signal-to-noise ratio threshold and less than the high signal-to-noise ratio threshold, the dynamic window length is determined based on a preset interpolation formula, the low signal-to-noise ratio threshold, the high signal-to-noise ratio threshold, the minimum window length, the maximum window length, and the signal-to-noise ratio parameter.

3. The method according to claim 1, wherein The Hilbert transform is performed on the arc current signal according to the dynamic window length to obtain m amplitude envelope data, including: The arc current signal corresponding to the dynamic window length is intercepted from the arc current signal to obtain m arc current signal segments; the time interval of each arc current signal segment in the m arc current signal segments is t ; Each of the m arc current signal segments is calculated based on a preset Hilbert transform formula to obtain m arc current data; wherein the Hilbert transform formula is: in, x_hat(t) For time point t Arc current data, is the integral variable, P is the Cauchy principal value integral, is the arc current signal segment, t is a time variable; An amplitude envelope of each of the m arc current data is determined to obtain the m amplitude envelope data.

4. The method according to any one of claims 1 to 3, wherein The arc current spectrum data is processed to obtain q frequency domain features, including: Obtain historical arc current fault signals; Extracting high-frequency pulse features from the historical arc current fault signal to obtain high-frequency pulse features; The historical arc current fault signal is transformed by using the Fourier transform method, and the frequency domain harmonic distribution characteristics are extracted to obtain the harmonic distribution characteristics; Determining a historical arc current energy distribution feature corresponding to the historical arc current fault signal according to the high-frequency pulse feature and the harmonic distribution feature; Determining a historical power spectrum density corresponding to the historical arc current energy distribution characteristics; Determining adaptive frequency band parameters based on a preset significant frequency band detection method, a preset energy threshold, and the historical power spectrum density; Dividing the arc current spectrum data based on the adaptive frequency band parameter to obtain q first arc current spectrum data; Frequency domain features are extracted from the q first arc current spectrum data to obtain the q frequency domain features.

5. The method according to claim 4, wherein The dividing the arc current spectrum data based on the adaptive frequency band parameter to obtain q first arc current spectrum data includes: determining a frequency range in the adaptive frequency band parameter; extracting spectrum data corresponding to the frequency range from the arc current spectrum data to obtain a candidate arc current data set; performing energy distribution calculation on each candidate arc current data in the candidate arc current data set to obtain a candidate arc current energy set; Counting energy parameters of all candidate arc currents in the candidate arc current energy set to obtain candidate arc current energy parameters; determining an energy parameter in the arc current spectrum data to obtain a first arc current energy parameter; determining an energy proportion parameter according to the candidate arc current energy parameter and the first arc current energy parameter; determining a spectral entropy parameter based on the candidate arc current energy parameter; Adjusting the frequency range according to the energy proportion parameter and the spectral entropy parameter to obtain a target frequency range; The arc current spectrum data in the target frequency range is extracted from the arc current spectrum data to obtain the q first arc current spectrum data.

6. The method according to claim 4, wherein The extracting frequency domain features of the q first arc current data to obtain the q frequency domain features includes: Normalizing the q first arc current spectrum data to obtain q second arc current spectrum data; Extracting a main harmonic amplitude and a subharmonic amplitude from each of the q second arc current spectrum data to obtain q main harmonic amplitudes and q subharmonic amplitudes; determining a ratio of each of the q subharmonic amplitudes in each of the q second arc current spectrum data to obtain q subharmonic ratios; The q frequency domain features are determined based on the q main harmonic amplitudes and the q subharmonic ratios.

7. The method according to any one of claims 1 to 3, wherein: The method of fusing the p time-domain statistical features and the q frequency-domain features based on a preset feature fusion algorithm to obtain a target arc current feature includes: Obtain historical arc current data; extracting time-domain statistical features from the historical arc current data to obtain a plurality of historical time-domain statistical features; Performing feature frequency domain extraction on the historical arc current data to obtain a plurality of historical frequency domain features; Constructing a historical arc current data set according to the multiple historical time-domain statistical features, the multiple historical frequency-domain features, and the historical arc current data; Inputting the historical arc current data set into a preset dimensionality reduction model to obtain a dimensionality reduction projection matrix; The multiple historical time-domain statistical features and the multiple historical frequency-domain features are respectively input into a preset lightweight machine learning classification model to obtain a first classification accuracy rate and a second classification accuracy rate; the first classification accuracy rate is the classification accuracy rate output by the lightweight machine learning classification model according to the multiple historical time-domain statistical features; the second classification accuracy rate is the classification accuracy rate output by the lightweight machine learning classification model according to the multiple historical frequency-domain features; Determine a time domain feature weight and a frequency domain feature weight according to the first classification accuracy and the second classification accuracy; Determine a first arc current feature according to the p time-domain statistical features, the q frequency-domain features, the time-domain feature weight, and the frequency-domain feature weight; The first arc current feature is projected based on the dimension reduction projection matrix to obtain the target arc current feature.

8. A lightweight arc fault detection device, characterized in that: Applied to electronic equipment, the device comprises: An acquisition module is used to acquire arc current signals of loads in a target distribution network within a preset time period; a determination module, configured to determine a signal-to-noise ratio parameter of the arc current signal; determine a dynamic window length according to the signal-to-noise ratio parameter; and calculate the arc current signal using a preset Fourier transform method to obtain arc current spectrum data; A calculation module is configured to perform a Hilbert transform on the arc current signal according to the dynamic window length to obtain m amplitude envelope data; m is a positive integer greater than 1; extract statistical features of the arc current signal in the time domain from the m amplitude envelope data to obtain p time domain statistical features; p is greater than or equal to m; process the arc current spectrum data to obtain q frequency domain features; q is a positive integer greater than 1; and fuse the p time domain statistical features and the q frequency domain features based on a preset feature fusion algorithm to obtain a target arc current feature; The control module is used to input the target arc current characteristics into a preset lightweight arc fault detection model to obtain an arc fault detection result.

9. An electronic device, characterized in that: include: a processor, a memory, a communication interface, and one or more programs; The one or more programs are stored in the memory and configured to be executed by the processor, wherein the programs include instructions for executing the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor is caused to perform the method according to any one of claims 1 to 7.

Images