Arc fault detection method, device and equipment

Abstract

The invention provides an arc fault detection method which can be applied to the technical field of arc fault recognition and artificial intelligence. The method comprises the following steps: acquiring current data corresponding to a target load; inputting the current data to a target detection model, and outputting an arc fault detection result corresponding to the current data, the arc fault detection result representing that the current data is a probability value of a fault arc; the target detection model is trained by using the following operations: an initial detection model is updated based on a target function to obtain an intermediate detection model, the target function is a function obtained by updating an initial phase of an initial function, and the initial detection model is constructed based on the initial function; and training the intermediate detection model by using the sample current data and the labels corresponding to the sample current data to obtain a target detection model. The invention further provides an arc fault detection device and equipment. Through the method, the accuracy and the detection efficiency of arc fault detection are improved, and the detection cost is reduced.

Description

Technical Field

[0001] The present disclosure relates to the field of arc fault identification and artificial intelligence technology, and in particular to an arc fault detection method, device and equipment. Background Art

[0002] Arc fault identification is the core of arc fault circuit breaker (AFCI). It can monitor the changes of arc-related physical quantities such as abnormal fluctuations in voltage and current, light radiation and electromagnetic radiation in the circuit in real time through sensors and signal processing circuits to determine whether there is an arc fault and complete arc fault identification. In related technologies, arc fault identification methods mainly include physical characteristic identification method and signal processing identification method.

[0003] The inventors have discovered that the following defects exist in the related art for arc fault identification: the physical property identification method has difficulty in parameter selection due to the complex and changeable physical properties of the fault arc, the arc fault identification process is cumbersome and complicated, and the cost of the identification equipment is high; the signal processing identification method has external noise interference, which affects the quality and accuracy of the signal, resulting in low reliability and accuracy of arc fault identification, and the identification method based on signal processing technology requires specific hardware equipment to collect and process signals. These devices have low compatibility with other electrical equipment, resulting in poor universality and difficulty in meeting actual needs. Summary of the invention

[0004] In view of the above problems, the present disclosure provides an arc fault detection method, apparatus, device, medium and program product.

[0005] According to a first aspect of the present disclosure, there is provided an arc fault detection method, comprising: acquiring current data corresponding to a target load, wherein the current data includes at least one of arc data and no-arc data based on time-frequency characteristics; inputting the current data into a target detection model, and outputting an arc fault detection result corresponding to the current data, wherein the arc fault detection result characterizes a probability value that the current data is a fault arc; wherein the target detection model is trained using the following operations: updating an initial detection model based on an objective function to obtain an intermediate detection model, wherein the objective function is a function after updating an initial phase of the initial function, the initial detection model is constructed based on the initial function, and the objective function is used to fit waveform feature data of the current data; and training the intermediate detection model using sample current data and labels corresponding to the sample current data to obtain a target detection model.

[0006] A second aspect of the present disclosure provides an arc fault detection device, including: a data acquisition module, used to acquire current data corresponding to a target load, wherein the current data includes at least one of arc data and no-arc data based on time-frequency characteristics; a detection result output module, used to input the current data into a target detection model, and output an arc fault detection result corresponding to the current data, wherein the arc fault detection result characterizes the probability value that the current data is a fault arc; wherein the target detection model is trained using the following operations: updating an initial detection model based on an objective function to obtain an intermediate detection model, wherein the objective function is a function after updating an initial phase of the initial function, the initial detection model is constructed based on the initial function, and the objective function is used to fit the waveform feature data of the current data; and training the intermediate detection model using sample current data and labels corresponding to the sample current data to obtain a target detection model.

[0007] A third aspect of the present disclosure provides an electronic device, comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the above method.

[0008] The fourth aspect of the present disclosure further provides a computer-readable storage medium having a computer program or instructions stored thereon, which implements the steps of the above method when the above computer program or instructions are executed by a processor.

[0009] The fifth aspect of the present disclosure further provides a computer program product, including a computer program or instructions, which implement the steps of the above method when the above computer program or instructions are executed by a processor.

[0010] According to the arc fault detection method, device, equipment, medium and program product provided by the present invention, the arc fault detection result is automatically output by inputting current data into the target detection model. Since the phase of the objective function in the target detection model is a trainable parameter, the model can automatically identify the phase of the arc current waveform, thereby realizing rapid fitting of the waveform feature data in the current data, thereby improving the efficiency of automatic identification of fault arcs. Moreover, since the identification of waveform feature data can be achieved only based on the time-frequency characteristics of the current data, it is applicable to different types of loads, thereby improving the universality of arc fault detection. At the same time, the analysis method of time-frequency characteristics can avoid the cumbersome process and high detection cost of traditional arc physical feature analysis, further improving the economy, accuracy and efficiency of arc fault detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The above contents and other purposes, features and advantages of the present disclosure will become more apparent through the following description of the embodiments of the present disclosure with reference to the accompanying drawings, in which:

[0012] Figure 1 The application scenario diagram of the arc fault detection method, device, equipment, medium and program product according to the embodiment of the present disclosure is schematically shown;

[0013] Figure 2 A flow chart of an arc fault detection method according to an embodiment of the present disclosure is schematically shown;

[0014] Figure 3 Schematically shows a load waveform diagram of different loads according to an embodiment of the present disclosure;

[0015] Figure 4 A schematic diagram schematically shows the effect of using unupdated and updated Fourier series to perform function fitting according to an embodiment of the present disclosure;

[0016] Figure 5 The schematic diagram of the structure of the arc fault test platform according to the embodiment of the present disclosure is shown;

[0017] Figure 6 The structure diagram of the Fourier-Kolmogorov-Arnold Network model according to the embodiment of the present disclosure is schematically shown;

[0018] Figure 7 Schematically showing confusion matrices of different test results using unupdated and updated Fourier series as model activation functions according to an embodiment of the present disclosure;

[0019] Figure 8 The structure block diagram of the arc fault detection device according to the embodiment of the present disclosure is schematically shown;

[0020] Fig. 9 A block diagram of an electronic device suitable for implementing an arc fault detection method according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0021] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0022] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0023] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0024] When using expressions such as "at least one of A, B, and C, etc.", they should generally be interpreted according to the meaning of the expression commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0025] It should be noted that in the embodiments of the present application, certain software, components, models and other existing solutions in the industry may be mentioned, and they should be regarded as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of the present application, but it does not mean that the applicant has or will necessarily use the solution.

[0026] In the technical solution of the present disclosure, the user information (including but not limited to user personal information, user image information, user device information, such as location information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved are all information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data comply with relevant laws, regulations and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0027] In the process of conceiving the present invention, the inventors found that in the related art, for arc fault identification, the physical property identification method is difficult to select parameters due to the complex and changeable physical properties of the fault arc, the arc fault identification process is cumbersome and complicated, and the cost of the identification equipment is high; the signal processing identification method has external noise interference, which affects the quality and accuracy of the signal, resulting in low reliability and accuracy of arc fault identification, and the identification method based on signal processing technology requires specific hardware equipment to collect and process signals. These devices have low compatibility with other electrical equipment, resulting in poor universality and difficulty in meeting actual needs.

[0028] In view of this, the present invention automatically outputs arc fault detection results by inputting current data into a target detection model. Since the phase of the objective function in the target detection model is a trainable parameter, the model can automatically identify the phase of the arc current waveform, thereby realizing rapid fitting of the waveform feature data in the current data, thereby improving the efficiency of automatic identification of fault arcs. Moreover, since the identification of waveform feature data can be achieved only based on the time-frequency characteristics of the current data, it is applicable to different types of loads, thereby improving the universality of arc fault detection. At the same time, the analysis method of time-frequency characteristics can avoid the cumbersome process and high detection cost of traditional arc physical feature analysis, thereby further improving the economy, accuracy and efficiency of arc fault detection.

[0029] Embodiments of the present disclosure provide an arc fault detection method, device, equipment, medium and program product. The method includes obtaining current data corresponding to a target load, wherein the current data includes at least one of arc data and no-arc data based on time-frequency characteristics; inputting the current data into a target detection model, and outputting an arc fault detection result corresponding to the current data, wherein the arc fault detection result characterizes the probability value of the current data being a fault arc; wherein the target detection model is trained using the following operations: updating an initial detection model based on an objective function to obtain an intermediate detection model, wherein the objective function is a function after updating the initial phase of the initial function, the initial detection model is constructed based on the initial function, and the objective function is used to fit the waveform feature data of the current data; and training the intermediate detection model using sample current data and labels corresponding to the sample current data to obtain a target detection model.

[0030] Figure 1 The application scenario diagram of the arc fault detection method, device, equipment, medium and program product according to the embodiments of the present disclosure is schematically shown.

[0031] like Figure 1 As shown, the application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, a server 105, and a monitoring device 106. The network 104 is used to provide a medium for a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or optical fiber cables, etc.

[0032] The monitoring device 106 can be used to collect and process current data, including but not limited to current transformers, Hall sensors, shunt resistors, data acquisition cards, and intelligent ammeters. The monitoring device 106 can be connected to a variety of loads (such as resistive loads, household appliances) to collect and process the current data of the loads in real time. The monitoring device 106 can be connected to the server 105 or a server cluster through the network 104 to transmit the acquired current data to the server for processing.

[0033] The user can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only for example).

[0034] The first terminal device 101, the second terminal device 102, and the third terminal device 103 may be various electronic devices having display screens and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0035] The server 105 may be a server that provides various services, such as a background management server (only as an example) that provides support for websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103. The background management server may analyze and process the received data such as user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0036] It should be noted that the arc fault detection method provided in the embodiment of the present disclosure can generally be executed by the server 105. Accordingly, the arc fault detection device provided in the embodiment of the present disclosure can generally be set in the server 105. The arc fault detection method provided in the embodiment of the present disclosure can also be executed by a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105. Accordingly, the arc fault detection device provided in the embodiment of the present disclosure can also be set in a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105.

[0037] It should be understood that Figure 1The number of terminal devices, networks, monitoring devices and servers in the embodiment is only for illustration. Any number of terminal devices, networks, monitoring devices and servers may be provided according to implementation requirements.

[0038] The following will be based on Figure 1 The scene described by Figure 2~Figure 7 The arc fault detection method of the disclosed embodiment is described in detail.

[0039] Figure 2 The flowchart of the arc fault detection method according to the embodiment of the present disclosure is schematically shown.

[0040] like Figure 2 As shown, the arc fault detection method of this embodiment includes operations S210 to S220.

[0041] In operation S210, current data corresponding to a target load is acquired, wherein the current data includes at least one of arc data and no-arc data based on a time-frequency feature.

[0042] In an embodiment of the present disclosure, the target load may include AC loads of various types or scenarios. When the target load is in a normal current flow condition, no arc data is generated. However, when the wire is in poor contact, damaged, or the electrical equipment fails, the current may jump through the air or an irregular path to form an arc, thereby collecting current data through a monitoring device, and based on time-frequency characteristics (time domain characteristics and frequency domain characteristics), the current data may be distinguished into arc data and no arc data, including using time domain analysis to extract the peak value, average value, fluctuation amplitude and other characteristics of the current waveform. If there are sudden high-amplitude fluctuations in the current waveform, and waveform distortion or pulse characteristics frequently appear, it can be judged that the fault is caused by an arc; or the time domain current signal is converted into a frequency domain signal through Fourier transform, and its spectrum is analyzed to determine whether there is an arc or no arc. The current data may be stored in a processor connected to the monitoring device or in a specific database.

[0043] In operation S220, the current data is input into a target detection model, and an arc fault detection result corresponding to the current data is output, wherein the arc fault detection result represents a probability value that the current data is a fault arc.

[0044] In the embodiments of the present disclosure, the target detection model can be a model for characterizing and analyzing nonlinear and complex structures, and can represent a specific type of multivariable continuous function through a weighted sum of a finite number of variables or a nonlinear transformation. The arc fault detection result output by the target detection model can be a probability value of a fault arc corresponding to the input current data. When the probability value is greater than or equal to a probability threshold, it can be determined that the current data has an arc fault, and corresponding warnings and processing can be performed.

[0045] Preferably, the target detection model is trained using the following operations, including operation S221 to operation S222: in operation S221, the initial detection model is updated based on the objective function to obtain an intermediate detection model, wherein the objective function is a function after updating the initial phase of the initial function, the initial detection model is constructed based on the initial function, and the objective function is used to fit the waveform feature data of the current data; in operation S222, the intermediate detection model is trained using sample current data and labels corresponding to the sample current data to obtain a target detection model.

[0046] In an embodiment of the present disclosure, the initial detection model may be a detection model before the activation function is updated. The initial activation function of the initial detection model may be a mathematical function for smoothing curves and surfaces. The present disclosure takes into account the fact that the initial activation function has a long training time and a low accuracy rate in the recognition of arc data, and updates the initial activation function of the initial detection model and replaces it with the initial function. The initial function may have the feature of being superimposed by multiple sinusoidal functions, so as to quickly fit the part with sinusoidal features in the current data, and meet the requirement of high fitting performance during model training.

[0047] In an embodiment of the present disclosure, the intermediate detection model may be a detection model obtained by replacing the initial activation function with the initial function. The initial function may be a Fourier series, and the objective function may represent the improved Fourier series. In practical applications, considering that the phase of the current waveform characteristics in the collected current data is random, the phases of the sine and cosine functions in the traditional Fourier series are all zero, and when fitting the arc current data, it is impossible to distinguish the correspondence between the main body of the arc current in the function and the high-frequency burrs. The present disclosure converts the phase of the sine and cosine functions in the traditional Fourier series from zero to a non-zero parameter (trainable parameter), so that the model can automatically identify the phase of the current arc waveform according to the time-frequency characteristics of the signal.

[0048] According to the embodiments of the present disclosure, the arc fault detection result is automatically output by inputting the current data into the target detection model. Since the phase of the objective function in the target detection model is a trainable parameter, the model can automatically identify the phase of the arc current waveform, thereby realizing rapid fitting of the waveform feature data in the current data, thereby improving the efficiency of automatic identification of fault arcs. Moreover, since the identification of waveform feature data can be achieved only based on the time-frequency characteristics of the current data, it is applicable to different types of loads, thereby improving the universality of arc fault detection. At the same time, the analysis method of time-frequency characteristics can avoid the cumbersome process and high detection cost of traditional arc physical feature analysis, thereby further improving the economy, accuracy and efficiency of arc fault detection.

[0049] It can be understood that the above has described how to determine the arc fault detection result and the training of the initial detection model, and the process of determining the arc fault detection result will be further described below.

[0050] According to an embodiment of the present disclosure, current data is input into a target detection model, and an arc fault detection result corresponding to the current data is output, including: extracting features of the current data corresponding to a preset period to obtain multiple target features; and obtaining an arc fault detection result based on the multiple target features.

[0051] In the embodiments of the present disclosure, the current data for feature extraction may be data obtained by segmenting the overall arc data based on a preset period. For example, the overall arc signal data may be segmented according to each 50 Hz power frequency period so as to extract features from the arc data in each period. For a 50 Hz signal, there are 50 periods per second, so the time length of each period is 20 milliseconds. It is understood that the preset period may be determined according to actual needs, and is not specifically determined here.

[0052] In the embodiments of the present disclosure, the target detection model may be a neural network model based on nonlinear system theory (e.g., Kolmogorov-Arnold Network, KAN), including an input layer, a hidden layer, and an output layer, and the number of hidden layers may be set to multiple according to actual needs. The input current data may be feature extracted through multiple hidden layers, and when the target detection model is a neural network model, the hidden layer may include a feature extraction layer of multiple neurons.

[0053] For example, in the hidden layer of the target detection model, high-dimensional features in the current data can be extracted through nonlinear mapping between layers. These high-dimensional features can include the change pattern, periodic fluctuations, mutations, etc. of the current waveform; for arc data, the target detection model can capture the mutation, violent fluctuations, periodic disturbances and other features in the current waveform; for normal current data without arc, the target detection model can extract a stable and relatively smooth current change pattern. Then, according to the target features of different levels extracted by the hidden layer, they are used for subsequent classification tasks, and the probability result of judging whether the input current data is arc fault data is determined.

[0054] According to the embodiments of the present disclosure, a target detection model is used to capture complex time series data through nonlinear mapping, so that when processing current data, complex fluctuations and mutations occurring in arc phenomena can be identified, especially for scenarios where nonlinear changes and noise interference may exist in current signals, these abnormal features can be effectively extracted from current data, and the occurrence of arcs can be better detected compared with conventional linear models; and the input data is mapped to a high-dimensional space, so that richer and more complex features can be extracted, especially for small fluctuations, periodic changes and mutation changes in current signals, which can be effectively extracted through the high-dimensional representation of the target detection model, thereby improving the detection accuracy of arc phenomena.

[0055] It can be understood that an example of how to determine an arc fault detection result has been described above, and the process of determining the objective function will be described below.

[0056] According to an embodiment of the present disclosure, the method further includes: updating an initial activation function of the initial detection model to determine an initial function; and updating an initial phase of the initial function based on a waveform corresponding to the current data to obtain a target function.

[0057] In the embodiments of the present disclosure, the initial activation function may be a conventional mathematical function for smoothing curves and surfaces, such as a B-spline function. For AC current, the present disclosure takes into account the fact that the training time is increased and the accuracy is low when fitting arc waveforms with partial sine wave characteristics using B-splines. The present disclosure uses an initial function instead of an initial activation function to optimize the initial detection model.

[0058] In an embodiment of the present disclosure, the initial function may be a combination of a complex periodic function decomposed into simple sine and cosine functions, such as a Fourier series. However, the initial phases of the sine and cosine functions in the traditional Fourier series are both zero, and when the arc current is fitted using the Fourier series, it is impossible to distinguish the correspondence between the main body of the arc current and the high-frequency burrs (noise) in the function. Therefore, the present disclosure updates the initial phase of the initial function to a trainable non-zero parameter, so that the model can automatically identify the phase of the arc current waveform. For example, the initial phase can be updated by introducing a phase offset in the sine and cosine functions, thereby obtaining the target function.

[0059] It can be understood that an example of how to determine the objective function has been described above, and the process of obtaining the current data will be described below.

[0060] According to an embodiment of the present disclosure, obtaining current data corresponding to a target load includes: obtaining initial current data corresponding to the target load using a monitoring device; and processing the initial current data based on a preset processing strategy to obtain current data.

[0061] In the embodiments of the present disclosure, the monitoring device can be used to collect and process current data, such as a current transformer, a Hall sensor, etc. The target load can include various load types, such as a resistive load, an incandescent lamp, a fluorescent lamp, a switching power supply, a washing machine, a halogen lamp, a refrigerator, an air conditioner, a vacuum cleaner, and a water dispenser, etc.

[0062] Figure 3 A schematic diagram of load waveforms of different loads according to an embodiment of the present disclosure is schematically shown.

[0063] like Figure 3 As shown, there is a load waveform 301 corresponding to a resistive load, a load waveform 302 corresponding to an incandescent lamp, a load waveform 303 corresponding to a fluorescent lamp, a load waveform 304 corresponding to a switching power supply, a load waveform 305 corresponding to a washing machine, a load waveform 306 corresponding to a halogen lamp, a load waveform 307 corresponding to a refrigerator, a load waveform 308 corresponding to an air conditioner, a load waveform 309 corresponding to a vacuum cleaner, and a load waveform 310 corresponding to a water dispenser.

[0064] In an embodiment of the present disclosure, the preset processing strategy may include a segmentation strategy and a conversion strategy. The segmentation strategy is used to segment the initial current data according to the power frequency (for example, segmenting according to every 2000 data as a cycle) to obtain multiple groups of current data after segmentation. The conversion strategy can be used to normalize or standardize the multiple groups of current data after segmentation to obtain current data.

[0065] It can be understood that an example of how to obtain current data has been described above, and the process of obtaining current data will be further described below.

[0066] According to an embodiment of the present disclosure, the preset processing strategy includes a segmentation strategy; the initial current data is processed based on the preset processing strategy to obtain current data, including: the initial current data is segmented based on the industrial frequency or preset time length corresponding to the initial current data to obtain multiple current data, wherein the current data includes multiple sub-data.

[0067] In an embodiment of the present disclosure, the initial current data can be segmented according to the power frequency or the preset duration. It can be understood that the power frequency refers to the frequency in the power system, such as 50Hz or 60Hz. When the sampling frequency is met, the preset duration of each cycle can be determined as T = 1 / f, where f is the power frequency (when f is 50Hz, T = 1 / 50 = 0.02 seconds, or 20 milliseconds), and the initial current data can be segmented according to 2000 data points per cycle, and each cycle can correspond to a signal (current data) with a time length of 20 milliseconds.

[0068] For example, an array or list containing all current data points can be obtained through the acquisition system (which stores the initial current data corresponding to the target load), with a length of N; with every 2000 data points as a cycle, the data can be cut according to every 2000 data points according to the sampling rate; if the total data length N cannot be divided by 2000, the remaining data can be added to a separate cycle. It is understandable that the remaining data can be processed according to actual needs. For example, incomplete cycles can be discarded, filled with zeros, or otherwise processed.

[0069] According to the embodiments of the present disclosure, by dividing the overall initial current data into multiple grouped current data for processing, each cycle of the arc signal can be effectively analyzed independently, and dividing the data according to the power frequency cycle can effectively isolate the influence of power frequency noise, especially in the power system, power frequency noise is a common interference; at the same time, the periodicity recognition ability of the signal is improved: by cutting the data according to the cycle, the changes of the arc in each cycle can be captured more accurately, thereby analyzing the fluctuation characteristics of the arc under the power frequency.

[0070] It can be understood that an example of how to obtain current data has been described above, and the process of obtaining current data will be further described below.

[0071] According to an embodiment of the present disclosure, the preset processing strategy also includes a conversion strategy; processing the initial current data based on the preset processing strategy to obtain current data also includes: determining peak data and valley data from the initial current data; and processing the initial current data based on the peak data, valley data and the conversion strategy to obtain current data.

[0072] In the embodiment of the present disclosure, the conversion strategy may include normalization processing and standardization processing. The conversion strategy may be selected according to the characteristics of the initial current actually taken.

[0073] For example, if the initial current data has a large range difference, Min-Max normalization can be used to make the initial current data within the same scale. For example, if the neural network model is used for data processing later, Z-score normalization can be used to set the feature data on the same scale, thereby improving the output accuracy of the model.

[0074] For example, a method for processing the initial current data using Min-Max normalization may include: determining the minimum value and the maximum value in the current data set; and normalizing the initial current data to scale it to the interval of [-1, 1].

[0075] It can be understood that an example of how to obtain current data has been described above, and the process of how to obtain the intermediate detection model will be described below.

[0076] According to an embodiment of the present disclosure, the initial detection model is updated based on the objective function to obtain an intermediate detection model, including: updating the initial phase based on a phase offset strategy to obtain a target phase; and updating the initial detection model using the target phase to obtain an intermediate detection model.

[0077] In an embodiment of the present disclosure, in the process of training the initial detection model, the initial zero phase of the initial detection model can be replaced by a non-zero phase. The intermediate detection model can be a model before training with sample current data. For example, a Kolmogorov-Arnold Network model (initial detection model) with strong interpretability and waveform fitting ability can be used, and the B-spline function (initial activation function) in the KAN model can be updated, and a Fourier series (objective function) that can train both amplitude and phase is proposed and added to the initial detection model to obtain an intermediate detection model.

[0078] For example, the B-spline function in the Kolmogorov-Arnold Network model can be replaced by the Fourier series as the activation function. The Fourier series can decompose a periodic signal into the sum of a series of sine and cosine functions. The coefficient of each function represents the intensity of the signal at different frequency components. These coefficients reveal the frequency characteristics of the signal, that is, the energy distribution of the signal at different frequencies. The specific formula is as follows (1):

[0079] (1);

[0080] Among them, Fourier(x) represents the target periodic function, k, a0, a k , b k 、c k With d k These are all trainable parameters.

[0081] Figure 4 The schematic diagram schematically shows the effect of function fitting using unupdated and updated Fourier series according to an embodiment of the present disclosure.

[0082] like Figure 4 As shown, in Figure 4In Figure (a), the initial function corresponding to the intermediate detection model is the unimproved Fourier series Kolmogorov-Arnold Network model. The initial phase of the initial function is zero. When fitting the arc current, it is impossible to distinguish the corresponding relationship between the main body of the arc current and the high-frequency burr. It can be verified through testing that the Pearson correlation coefficient between the unimproved fitting result and the original data is 0.93. Figure 4 In Figure (b), by performing a phase shift on the Fourier series, the phase of the improved Fourier series is updated to a trainable parameter, and then the objective function is obtained. It can be verified through testing that the Pearson correlation coefficient between the improved fitting result and the original data is 0.98.

[0083] It can be understood that an example of how to obtain an intermediate detection model has been described above, and the training process of the target detection model will be described below.

[0084] According to an embodiment of the present disclosure, an intermediate detection model is trained using sample current data and labels corresponding to the sample current data to obtain a target detection model, including: based on an iteration strategy, a preset learning rate and a preset attenuation rate, training the intermediate detection model using sample current data and labels; and when the detection accuracy of the intermediate detection model is greater than or equal to a preset threshold, determining the intermediate detection model as the target detection model.

[0085] In an embodiment of the present disclosure, the label may be a label corresponding to the sample current data, and may include two labels, the label of the arc-free data is 0, and the label of the arc-containing data is 1. The label may be associated with the normalized sample current data, and after random arrangement, 70% may be divided into a training set and 30% may be divided into a test set.

[0086] In an embodiment of the present disclosure, the step of training an intermediate detection model to obtain a target detection model may include: using the normalized sample current data as input data of a Fourier-Kolmogorov-Arnold Network (F-KAN) model, and the number of layers of the model, the number of neurons in each layer, and the activation function may be set according to actual needs; and using a loss function (mean square error) and an optimizer (such as Adam or SGD) to adjust the model weights to minimize the loss function; and setting a reasonable learning rate to ensure stable model training.

[0087] For example, based on the defined F-KAN model, initialize the network weights; iterate multiple times, take a batch of data from the training set each time, get the prediction results through the network forward propagation, calculate the loss function value, and then update the network weights through back propagation; adjust the learning rate at the end of each round of training according to the preset learning rate decay strategy; during the training process, regularly evaluate the model performance on the validation set, and adjust the hyperparameters (such as learning rate, network architecture, etc.) as needed. After the training is completed, the accuracy or Area Under the Curve (AUC) indicator can be used to evaluate the performance of the model. Furthermore, the model's hyperparameters, such as learning rate, batch size, number of model layers, etc., can be tuned through cross-validation or grid search methods to obtain the best training results, so as to obtain a target detection model that meets the preset detection accuracy.

[0088] It should be noted that after obtaining the sample current data in the model training, the sample current data can be segmented, and the sample current data can be divided into arc data and non-arc data according to various methods and current waveforms.

[0089] For example, the waveform feature extraction method can be used to determine whether there is an arc in the current. The method may include: detecting mutations by calculating the maximum value and rate of change (such as the first-order difference) of the current waveform; when there is an arc, the current waveform will usually have a large, rapid current peak or mutation, and there may be frequent instantaneous large fluctuations in the current waveform; when there is no arc, the current waveform is usually relatively stable, with small changes and no obvious mutations, thereby determining whether there is arc data or no arc data.

[0090] For example, by setting a threshold, it is determined whether the current waveform meets the characteristics of an arc. Parameters such as the amplitude of the current value, the rate of change, or the intensity of high-frequency noise can be set to determine whether there is an arc. Specifically, it includes: obtaining current waveform data within the target time period according to demand; calculating waveform characteristics, which can calculate the amplitude, fluctuation amplitude, change rate, etc. of the current; setting one or more thresholds based on experience or historical data, for example, setting that when the current fluctuation exceeds a certain amplitude or the change rate exceeds a certain value, it is considered that an arc exists; comparing the calculated characteristics with the threshold, and judging whether the current waveform has an arc based on the comparison result.

[0091] In a feasible embodiment, the training method of the Fourier-Kolmogorov-Arnold Network model may include operations S510 to S540.

[0092] In operation S510 , sample current data corresponding to a sample load is acquired.

[0093] Figure 5The structural diagram of an arc fault test platform according to an embodiment of the present disclosure is schematically shown.

[0094] like Figure 5 As shown, an arc fault experiment is designed, and an arc fault test platform can be built. The arc fault test platform may include an arc generator 501, a current transformer 502, and a load 503. The arc generator 501 can simulate an arc fault, and by generating an arc discharge phenomenon, provide a fault current and voltage waveform, and simulate a real arc fault environment; the current transformer 502 can obtain sample current data corresponding to different loads 503, and by sensing the change of the current, especially when an arc fault occurs, detect the fluctuation of the current and convert it into an easy-to-measure signal, which is sent to a server or a server cluster for subsequent analysis and processing; the load 503 can simulate the load in the actual power system, and provide a current path when an arc occurs, so that the arc generator can be tested under specific load conditions to ensure the authenticity and effectiveness of the fault simulation.

[0095] Among them, operation S510 may include operation S511 to operation S513.

[0096] In operation S511, an arc fault test platform may be constructed according to relevant specifications or standards, for example, to simulate a scenario in which an arc is caused by loose cable joints and circuit insulation failure.

[0097] In operation S512, a transformer is used to collect sample current data of an arc fault test at a frequency of 100 kHz and store the data in a device.

[0098] In operation S513, the collected sample arc data may be divided according to the power frequency (divided into 2000 data as one cycle) to obtain multiple groups of sample data, and the presence of arc and the absence of arc may be distinguished according to the current waveform.

[0099] In operation S520, the initial activation function in the KAN model (initial detection model) is replaced by the improved Fourier series to build a F-KAN model.

[0100] Among them, operation S520 may include operation S521 to operation S522.

[0101] In operation S521 , the B-spline function in the KAN model (initial detection model) may be changed to an improved Fourier series (Fourier) as an activation function to obtain an F-KAN model.

[0102] In operation S522 , the current data of a single cycle may be input to the input layer of the model, the current data features may be processed through a plurality of intermediate layers (hidden layers), and the prediction result may be output through the output layer.

[0103] Figure 6 The structural diagram of the Fourier-Kolmogorov-Arnold Network model according to an embodiment of the present disclosure is schematically shown.

[0104] like Figure 6 As shown, the current data of a single cycle can be used as the model input dimension (2000), and the current data is processed by the first intermediate layer 610 (first hidden layer), and the first neuron (first characteristic data) can be obtained to be 8, so as to input the second intermediate layer 620 (second hidden layer) to process the first neuron, and the second neuron (second characteristic data) can be obtained to be 4. By extracting features of the current data with arc and without arc, the dimension of the output layer 630 is 2, and the data is classified into two categories as data with arc or data without arc. The prediction result can represent the probability value of the current data being data with arc or data without arc.

[0105] In operation S530, the arc data and the non-arc data in S510 are normalized, and after being shuffled, the data set is divided into 70% of the training set and 30% of the test set.

[0106] Among them, operation S530 may include operation S531 to operation S533.

[0107] In operation S531, each set of sample data may be subjected to Min-Max normalization to scale the data to the interval [-1, 1].

[0108] In operation S532 , a label corresponding to the sample data is created, where the label of the data without arcs is 0 and the label of the data with arcs is 1.

[0109] In operation S533 , the labels are associated with the normalized sample current data, and after random arrangement, 70% are divided into a training set and 30% are divided into a test set.

[0110] In operation S540, the normalized data is input into the F-KAN model. The model can be set to train 100 times with a batch size of 10 batches, and the model is trained with a learning rate of 0.0001 and a weight decay rate of 0.001, and then the model accuracy is determined using the test set.

[0111] Figure 7 The confusion matrices of different test results using the unupdated and updated Fourier series as the model activation function according to an embodiment of the present disclosure are schematically shown.

[0112] like Figure 7As shown, Figure (a) is the confusion matrix of the test results when the unupdated Fourier series is used as the model activation function, and Figure (b) is the confusion matrix of the test results when the updated Fourier series is used as the model activation function. By comparison, it can be seen that the model using the unimproved Fourier series has an accuracy of 99.67% in arc recognition of arc data of ten load types; the model using the improved Fourier series has an accuracy of 99.82% in arc recognition of arc data of ten load types. Compared with the unimproved Fourier KAN model, the recognition accuracy of the F-KAN model disclosed in the present invention is improved by 0.15%, and arc faults can be identified more accurately.

[0113] According to the embodiments of the present disclosure, the updated Fourier series can perform efficient calculations in the frequency domain by decomposing the signal into a set of sine and cosine wave components. For signals containing periodic fluctuations, the Fourier series can describe the signal with fewer frequency components, thereby reducing the amount of calculation. Compared with the B-spline function, the improved Fourier series can speed up the calculation process, especially in large-scale data or real-time fault detection, providing a faster response speed. At the same time, the improved Fourier series has good spectral decomposition capabilities, and can still effectively extract sinusoidal features from the fault current in the presence of noise or interference. By more accurately decomposing the periodic components in the fault current, the improved activation function improves the robustness of the system and can work more stably in a complex power environment.

[0114] Based on the above arc fault detection method, the present disclosure also provides an arc fault detection device. Figure 8 The device is described in detail.

[0115] Figure 8 The structure block diagram of the arc fault detection device according to the embodiment of the present disclosure is schematically shown.

[0116] like Figure 8 As shown, the arc fault detection device of this embodiment includes a data acquisition module 810 and a detection result output module 820 .

[0117] The data acquisition module 810 is used to acquire current data corresponding to the target load, wherein the current data includes at least one of arc data and no-arc data based on the time-frequency characteristics. In one embodiment, the data acquisition module 810 can be used to perform the operation S210 described above, which will not be repeated here.

[0118] The detection result output module 820 is used to input the current data into the target detection model and output the arc fault detection result corresponding to the current data, wherein the arc fault detection result represents the probability value that the current data is a fault arc. In one embodiment, the detection result output module 820 can be used to perform the operation S220 described above, which will not be repeated here.

[0119] Preferably, the target detection model is trained using the following operations: updating the initial detection model based on the objective function to obtain an intermediate detection model, wherein the objective function is a function after updating the initial phase of the initial function, the initial detection model is constructed based on the initial function, and the objective function is used to fit the waveform feature data of the current data; and training the intermediate detection model using sample current data and labels corresponding to the sample current data to obtain the target detection model.

[0120] According to the embodiments of the present disclosure, through the data acquisition module 810 and the detection result output module 820 in the arc fault detection device, the current data is input into the target detection model to automatically output the arc fault detection result. Since the phase of the objective function in the target detection model is a trainable parameter, the model can automatically identify the phase of the arc current waveform, thereby realizing rapid fitting of the waveform feature data in the current data, thereby improving the efficiency of automatic identification of fault arcs. Moreover, since the identification of waveform feature data can be achieved only based on the time-frequency characteristics of the current data, it is applicable to different types of loads, thereby improving the universality of arc fault detection. At the same time, the analysis method of time-frequency characteristics can avoid the cumbersome process and high detection cost of traditional arc physical feature analysis, thereby further improving the economy, accuracy and efficiency of arc fault detection.

[0121] According to an embodiment of the present disclosure, the detection result output module 820 includes: a feature extraction submodule and a detection result acquisition module. The feature extraction submodule is used to extract features from the current data corresponding to the preset period to obtain multiple target features. The detection result acquisition module is used to obtain arc fault detection results based on the multiple target features.

[0122] According to an embodiment of the present disclosure, the device further includes: an initial function determination module and a phase update module. The initial function determination module is used to update the initial activation function of the initial detection model to determine the initial function. The phase update module is used to update the initial phase of the initial function based on the waveform corresponding to the current data to obtain the target function.

[0123] According to an embodiment of the present disclosure, the data acquisition module 810 includes: an initial current data acquisition submodule and an initial current data processing submodule. The initial current data acquisition submodule is used to acquire initial current data corresponding to the target load using a monitoring device. The initial current data processing submodule is used to process the initial current data based on a preset processing strategy to obtain current data.

[0124] According to an embodiment of the present disclosure, the preset processing strategy includes a segmentation strategy; the initial current data processing submodule includes: a segmentation unit, used to segment the initial current data based on the industrial frequency or preset duration corresponding to the initial current data to obtain multiple current data, wherein the current data includes multiple sub-data.

[0125] According to an embodiment of the present disclosure, the preset processing strategy also includes a conversion strategy; the initial current data processing submodule also includes: a data determination unit and an initial current data processing unit. The data determination unit is used to determine peak data and valley data from the initial current data. The initial current data processing unit is used to process the initial current data based on the peak data, valley data and the conversion strategy to obtain current data.

[0126] According to an embodiment of the present disclosure, the initial detection model is updated based on the objective function to obtain an intermediate detection model, including: updating the initial phase based on a phase offset strategy to obtain a target phase; and updating the initial detection model using the target phase to obtain an intermediate detection model.

[0127] According to an embodiment of the present disclosure, an intermediate detection model is trained using sample current data and labels corresponding to the sample current data to obtain a target detection model, including: based on an iteration strategy, a preset learning rate and a preset attenuation rate, training the intermediate detection model using sample current data and labels; and when the detection accuracy of the intermediate detection model is greater than or equal to a preset threshold, determining the intermediate detection model as the target detection model.

[0128] According to an embodiment of the present disclosure, any multiple modules in the data acquisition module 810 and the detection result output module 820 can be combined in one module for implementation, or any one of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present disclosure, at least one of the data acquisition module 810 and the detection result output module 820 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application-specific integrated circuit (ASIC), or can be implemented by hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or implemented in any one of the three implementation methods of software, hardware and firmware or in any appropriate combination of any of them. Alternatively, at least one of the data acquisition module 810 and the detection result output module 820 can be at least partially implemented as a computer program module, and when the computer program module is run, the corresponding function can be executed.

[0129] Fig. 9 A block diagram of an electronic device suitable for implementing an arc fault detection method according to an embodiment of the present disclosure is schematically shown.

[0130] like Fig. 9 As shown, the electronic device according to the embodiment of the present disclosure includes a processor 901, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 902 or the program loaded from the storage part 908 to the random access memory (RAM) 903. The processor 901 may include, for example, a general-purpose microprocessor (such as a CPU), an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (for example, an application-specific integrated circuit (ASIC)), etc. The processor 901 may also include an onboard memory for caching purposes. The processor 901 may include a single processing unit or multiple processing units for performing different actions of the method flow according to the embodiment of the present disclosure.

[0131] In RAM 903, various programs and data required for the operation of the electronic device are stored. The processor 901, ROM 902, and RAM 903 are connected to each other via a bus 904. The processor 901 performs various operations of the method flow according to the embodiment of the present disclosure by executing the program in ROM 902 and / or RAM 903. It should be noted that the program may also be stored in one or more memories other than ROM 902 and RAM 903. The processor 901 may also perform various operations of the method flow according to the embodiment of the present disclosure by executing the program stored in the one or more memories.

[0132] According to an embodiment of the present disclosure, the electronic device may further include an input / output (I / O) interface 905, which is also connected to the bus 904. The electronic device may further include one or more of the following components connected to the input / output (I / O) interface 905: an input portion 906 including a keyboard, a mouse, etc.; an output portion 907 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 908 including a hard disk, etc.; and a communication portion 909 including a network interface card such as a LAN card, a modem, etc. The communication portion 909 performs communication processing via a network such as the Internet. A drive 910 is also connected to the input / output (I / O) interface 905 as needed. A removable medium 911, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 910 as needed, so that a computer program read therefrom is installed into the storage portion 908 as needed.

[0133] The present disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present disclosure is implemented.

[0134] According to an embodiment of the present disclosure, a computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, may include but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, an apparatus or a device. For example, according to an embodiment of the present disclosure, a computer-readable storage medium may include the ROM 902 and / or RAM 903 described above and / or one or more memories other than ROM 902 and RAM 903.

[0135] The embodiment of the present disclosure also includes a computer program product, which includes a computer program, and the computer program contains program code for executing the method shown in the flowchart. When the computer program product is run in a computer system, the program code is used to enable the computer system to implement the arc fault detection method provided by the embodiment of the present disclosure.

[0136] The above functions defined in the system / device of the embodiment of the present disclosure are performed when the computer program is executed by the processor 901. According to the embodiment of the present disclosure, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0137] In one embodiment, the computer program may be based on a tangible storage medium such as an optical storage device, a magnetic storage device, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of a signal on a network medium, and downloaded and installed through the communication part 909, and / or installed from a removable medium 911. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0138] In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 909, and / or installed from the removable medium 911. When the computer program is executed by the processor 901, the above functions defined in the system of the embodiment of the present disclosure are performed. According to the embodiment of the present disclosure, the system, device, means, module, unit, etc. described above can be implemented by a computer program module.

[0139] According to an embodiment of the present disclosure, the program code for executing the computer program provided by the embodiment of the present disclosure can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, Java, C++, python, "C" language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect through the Internet).

[0140] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0141] It will be appreciated by those skilled in the art that the features described in the various embodiments of the present disclosure may be combined and / or combined in a variety of ways, even if such combinations or combinations are not explicitly described in the present disclosure. In particular, without departing from the spirit and teachings of the present disclosure, the features described in the various embodiments of the present disclosure may be combined and / or combined in a variety of ways. All of these combinations and / or combinations fall within the scope of the present disclosure.

[0142] The embodiments of the present disclosure are described above. However, these embodiments are only for illustrative purposes and are not intended to limit the scope of the present disclosure. Although the embodiments are described above, this does not mean that the measures in the various embodiments cannot be used in combination to advantage. Without departing from the scope of the present disclosure, those skilled in the art may make a variety of substitutions and modifications, which should all fall within the scope of the present disclosure.

Claims

1. A method for detecting an arc fault, characterized in that: The method comprises: Acquire current data corresponding to the target load, wherein the current data includes at least one of arc data and no-arc data based on time-frequency characteristics; Inputting the current data into a target detection model, and outputting an arc fault detection result corresponding to the current data, wherein the arc fault detection result represents a probability value that the current data is a fault arc; The target detection model is trained using the following operations: updating the initial detection model based on the objective function to obtain an intermediate detection model, wherein the objective function is a function after updating the initial phase of the initial function, the initial detection model is constructed based on the initial function, and the objective function is used to fit the waveform characteristic data of the current data; and The intermediate detection model is trained using the sample current data and labels corresponding to the sample current data to obtain a target detection model.

2. The detection method according to claim 1, characterized in that: Inputting the current data into a target detection model, and outputting arc fault detection results corresponding to the current data, including: Extracting features from the current data corresponding to the preset period to obtain multiple target features; and Based on the multiple target features, the arc fault detection result is obtained.

3. The detection method according to claim 1, characterized in that: The method further comprises: Updating the initial activation function of the initial detection model to determine the initial function; and The initial phase of the initial function is updated based on a waveform corresponding to the current data to obtain the target function.

4. The detection method according to claim 1, characterized in that: Obtain current data corresponding to the target load, including: Acquiring initial current data corresponding to the target load using a monitoring device; and The initial current data is processed based on a preset processing strategy to obtain the current data.

5. The detection method according to claim 4, characterized in that: The preset processing strategy includes a segmentation strategy; Processing the initial current data based on a preset processing strategy to obtain the current data includes: The initial current data is divided based on the power frequency or the preset duration corresponding to the initial current data to obtain a plurality of the current data, wherein the current data includes a plurality of sub-data.

6. The detection method according to claim 5, characterized in that: The preset processing strategy also includes a conversion strategy; Processing the initial current data based on a preset processing strategy to obtain the current data also includes: determining peak data and valley data from the initial current data; as well as The initial current data is processed based on the peak data, the valley data and the conversion strategy to obtain the current data.

7. The detection method according to claim 1, characterized in that: The initial detection model is updated based on the objective function to obtain an intermediate detection model, including: Updating the initial phase based on a phase shift strategy to obtain a target phase; and The initial detection model is updated using the target phase to obtain the intermediate detection model.

8. The detection method according to claim 1, characterized in that: The intermediate detection model is trained using the sample current data and the labels corresponding to the sample current data to obtain a target detection model, including: Based on an iteration strategy, a preset learning rate, and a preset decay rate, training the intermediate detection model using the sample current data and the label; and When the detection accuracy of the intermediate detection model is greater than or equal to a preset threshold, the intermediate detection model is determined as the target detection model.

9. An arc fault detection device, characterized in that: The device comprises: A data acquisition module, used to acquire current data corresponding to the target load, wherein the current data includes at least one of arc data and no-arc data based on time-frequency characteristics; a detection result output module, configured to input the current data into a target detection model, and output an arc fault detection result corresponding to the current data, wherein the arc fault detection result represents a probability value that the current data is a fault arc; The target detection model is trained using the following operations: updating the initial detection model based on the objective function to obtain an intermediate detection model, wherein the objective function is a function after updating the initial phase of the initial function, the initial detection model is constructed based on the initial function, and the objective function is used to fit the waveform characteristic data of the current data; and The intermediate detection model is trained using the sample current data and labels corresponding to the sample current data to obtain a target detection model.

10. An electronic device comprising: one or more processors; a memory for storing one or more computer programs, It is characterized in that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 8.

Images