AC fault arc detection method, system, equipment, medium and program product

Abstract

The present invention proposes an AC fault arc detection method, system, device, medium and program product. By using the Kan model to fit arc data and non-arc data respectively, the sign function corresponding to the arc data and the sign function corresponding to the non-arc data are obtained. The detection result of the current data to be detected is obtained by the deviation between the current data to be detected and the sign function corresponding to the arc data and the sign function corresponding to the non-arc data. The present invention improves the Kan model fitting function, adds the law of alternating current on the basis of the original model, and adds human prior knowledge to make the model result more reasonable; and the finally obtained matching function has no special requirements for deployment, and the calculation pressure is extremely low, and the existing embedded chips can basically realize deployment.

Description

Technical Field

[0001] The present invention belongs to the technical field of arc fault detection, and in particular relates to an AC fault arc detection method, system, equipment, medium and program product. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Low-voltage AC series fault arc is currently an important cause of electrical fires, causing serious casualties and property losses. Many regions and countries around the world are studying and trying to effectively reduce electrical fires by solving the problem of fault arcs. However, AC fault arcs are affected by many factors such as their own occurrence conditions and combustion conditions. Their occurrence location and time show obvious randomness, which leads to the fact that the detection and protection of fault arcs must be distributed and deployed on a large scale. Otherwise, it will be difficult to effectively avoid the occurrence of disasters. Traditional fault arc detection methods are not accurate and cannot eliminate the hazards of fault arcs. The method based on artificial intelligence is a solution with higher detection accuracy.

[0004] However, the AI ​​model is a "black box". Once a misjudgment occurs, it is impossible to find the fundamental cause of the model error. The improvement of the model is a blind attempt to adjust the parameters, which creates challenges in the application of artificial intelligence methods. This is one of the most serious problems currently faced by existing artificial intelligence-based fire-causing fault arc detection methods.

[0005] In addition, it is well known that the size of artificial intelligence models is generally very large. The existing model lightweighting is basically to reduce the overall model parameters or the amount of calculation at the expense of model accuracy. There is no understanding of the model's behavior and needs, and there is no deep integration of the inherent characteristics of the fault arc itself to make targeted improvements. It does not fundamentally achieve the lightweighting of the model. Failure to achieve lightweighting will seriously restrict the number of model parameters, making it impossible to achieve distributed deployment. This is the second most serious problem currently faced by the existing artificial intelligence-based AC fault arc detection methods.

[0006] Existing AC arc fault artificial intelligence detection models mostly rely on MLP structures such as convolution and fully connected layers, which basically lose interpretability. Model optimization can only be attempted blindly, and the ability to analyze erroneous samples is completely lost. On the basis of pursuing high precision of artificial intelligence methods, if you want to further improve the accuracy, you must understand the behavior of the model and introduce prior knowledge in the field of arc fault detection, rather than relying entirely on the model's own feature extraction capabilities to avoid the model from going off track or falling into a local optimal solution. Arc fault detection is extremely dangerous. The pursuit of extreme accuracy and reliable decision-making is a guarantee of safety and the core of arc fault detection methods.

[0007] Therefore, achieving lightweighting, improving model accuracy, and reducing misjudgment of fault arc detection without losing model accuracy are technical problems that need to be solved at present. Summary of the invention

[0008] In order to overcome the deficiencies of the above-mentioned prior art, the present invention provides an AC fault arc detection method, system, device, medium and program product. By improving the Kan model fitting function, the laws of alternating current are added to the original model, and human prior knowledge is added to make the model results more reasonable; and the finally obtained fitting function has no special requirements for deployment, the computing pressure is extremely low, and the existing embedded chips can basically realize deployment.

[0009] In order to achieve the above object, the present invention adopts the following technical solution:

[0010] In a first aspect, the present invention provides a low voltage AC fault arc detection method, comprising:

[0011] Obtain arc data and no-arc data under different loads and perform preprocessing;

[0012] The pre-processed arc data and arc-free data are used to train the corresponding improved Kan model respectively, and the sign function corresponding to the arc data and the sign function corresponding to the arc-free data are fitted; wherein, the improvement of the Kan model is: the Chebyshev polynomial is used as the fitting function of the Kan model;

[0013] The current data to be detected is obtained, and the deviations of the obtained current data to be detected and the sign function corresponding to the arc data and the sign function corresponding to the no-arc data are calculated to obtain the detection result of the current data to be detected.

[0014] Furthermore, the Kan model is improved by using Chebyshev polynomials with sin functions as fitting functions of the Kan model.

[0015] Furthermore, the obtained arcing data and non-arcing data under different loads are preprocessed, specifically: the obtained arcing data and non-arcing data under different loads are normalized.

[0016] Furthermore, the deviations between the acquired current data to be detected and the sign functions corresponding to the arc data and the no-arc data are calculated, and the arc or no-arc type corresponding to the sign function with the smallest deviation is taken as the detection result of the current data to be detected.

[0017] In a second aspect, the present invention provides a low voltage AC fault arc detection system, comprising:

[0018] A preprocessing module is configured to: obtain arc data and no-arc data under different loads and perform preprocessing;

[0019] The training module is configured to: train the corresponding improved Kan model with the preprocessed arc data and arc-free data respectively, and fit the sign function corresponding to the arc data and the sign function corresponding to the arc-free data; wherein the improvement of the Kan model is: using Chebyshev polynomial as the fitting function of the Kan model;

[0020] The detection module is configured to: obtain current data to be detected, calculate the deviation between the obtained current data to be detected and the sign function corresponding to the arc data and the sign function corresponding to the no-arc data, and obtain the detection result of the current data to be detected.

[0021] Furthermore, in the preprocessing module, the obtained arcing data and no-arc data under different loads are preprocessed, specifically: the obtained arcing data and no-arc data under different loads are normalized.

[0022] Furthermore, in the training module, the adopted Kan model is improved, and Chebyshev polynomials are used as the fitting function of the Kan model.

[0023] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method described in the first aspect is completed.

[0024] In a fourth aspect, the present invention provides a computer-readable storage medium for storing computer instructions, wherein when the computer instructions are executed by a processor, the method described in the first aspect is performed.

[0025] In a fifth aspect, the present invention provides a computer program product, comprising a computer program, which, when executed by a processor, implements the method described in the first aspect.

[0026] One or more of the above technical solutions have the following beneficial effects:

[0027] In the present invention, by using the Kan model to fit arc data and no-arc data respectively, the sign function corresponding to the arc data and the sign function corresponding to the no-arc data are obtained, and the detection result of the current data to be detected is obtained by the deviation between the current data to be detected and the sign function corresponding to the arc data and the sign function corresponding to the no-arc data. The present invention improves the Kan model fitting function, adds the law of alternating current on the basis of the original model, and adds human prior knowledge to make the model result more reasonable; and the finally obtained matching function has no special requirements for deployment, the calculation pressure is extremely low, and the existing embedded chips can basically realize deployment.

[0028] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0030] Figure 1 This is a flow chart of obtaining a symbol function based on the Kan model in the first embodiment of the present invention;

[0031] Figure 2 This is a schematic diagram of the Kan model structure corresponding to the arc data in the first embodiment of the present invention;

[0032] Figure 3 Schematic diagram of the Kan model structure corresponding to the arc-free data in the first embodiment of the present invention;

[0033] Figure 4 It is a schematic diagram of an image of arc data and an image of a fitting function of arc data in Embodiment 1 of the present invention;

[0034] Figure 5 This is a schematic diagram of a data image without arcs in Embodiment 1 of the present invention;

[0035] Figure 6 This is a flow chart of fault identification based on a compliance function in Embodiment 1 of the present invention. DETAILED DESCRIPTION

[0036] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0037] It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.

[0038] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0039] The KAN model (Kolmogorov-Arnold Networks) is a new type of network architecture. Its biggest innovation is that it puts the activation function on the weight, that is, it applies a learnable activation function on the weight. These one-dimensional activation functions are parameterized as splines, which enables the network to process and learn complex relationships of input data in a more flexible and closer way to the Kolmogorov-Arnold representation theorem. The KAN model has good interpretability and outperforms the traditional multi-layer perceptron (MLP) on some tasks. The KAN model is intended to be a promising alternative to the multi-layer perceptron MLP. MLP is the abbreviation of multilayer perceptron, which is a basic artificial neural network model. It consists of multiple neural network layers, including input layer, hidden layer and output layer. Each layer consists of multiple neurons (or nodes), and there are connections between neurons in adjacent layers, and each connection has a weight. The working principle of MLP is to pass the input signal through multiple neural network layers, and use activation functions in each layer to perform weighted summation of the input and produce output.

[0040] The full name of KAN is Kolmogorov–Arnold Network, which pays tribute to two great late mathematicians. The core idea behind it is the Kolmogorov–Arnold Representation Theorem, or KART (Kolmogorov–Arnold Representation Theorem). The core idea of ​​KART is that any multivariate continuous function can be represented as a combination of a finite number of single-variable functions and addition. The formula is as follows:

[0041]

[0042] Where :[0,1]→R and :R→R. It is an internal function. External function.

[0043] The application scenarios of the KAN model may include various tasks that require complex function representation capabilities. Since the KAN model has stronger function representation capabilities, it may be suitable for problems that are difficult to handle with traditional MLP.

[0044] Embodiment 1

[0045] This embodiment discloses a low voltage AC fault arc detection method, comprising:

[0046] Obtain arc data and no-arc data under different loads and perform preprocessing;

[0047] The pre-processed arc data and arc-free data are used to train the corresponding improved Kan model respectively, and the sign function corresponding to the arc data and the sign function corresponding to the arc-free data are fitted; wherein, the improvement of the Kan model is: the Chebyshev polynomial is used as the fitting function of the Kan model;

[0048] The current data to be detected is obtained, and the deviations of the obtained current data to be detected and the sign function corresponding to the arc data and the sign function corresponding to the no-arc data are calculated to obtain the detection result of the current data to be detected.

[0049] There are differences in the current waveforms when a fault arc occurs and when there is no fault arc. The main differences lie in the current waveform characteristics, high-frequency energy mutations, wavelet transforms, and other aspects. ‌ The waveform of a fault arc is usually manifested as the current waveform derivative and the cumulative arc period exceeding the set threshold, while this will not happen with a normal arc. For traditional loads, such as resistors, capacitors, and inductors, there is no high-frequency energy during normal operation, and the high-frequency characteristics can be used to indicate whether such loads generate a fault arc. However, for modern power electronic equipment such as frequency converters, inverters, and electronic switches, these switching devices exhibit messy waveforms at high frequencies due to the presence of high-frequency interference noise, making it difficult to detect fault arcs through simple threshold judgments. Therefore, the Kan model is introduced in the detection of AC fault arcs in this application. On the basis of the original model, the laws of alternating current are added, and the addition of human prior knowledge makes the results of the model more reasonable, and fundamentally reduces the size of the model to achieve extreme lightweight. ‌

[0050] Take resistive load as an example, and normalize the data with and without arc, such as Figure 1 As shown, Figure 1 The yellow color is data with arcs, and the purple color is data without arcs.

[0051] The main function of the Kan model is to fit the most suitable sign function for the arc and no-arc data respectively. This embodiment transforms the Kan model structure so that the model is suitable for one-dimensional data of fault arc current. The original Kan model requires two inputs, which does not conform to the actual situation in the field of fault arc detection. By transforming the Kan model, it adapts to one-dimensional data input, so that it has the conditions for application in the field of fault arc detection.

[0052] The basic Kan model is based on B-spline curve fitting, which requires a large number of control points and nodes to achieve relatively accurate curve fitting. It has a heavy burden on calculation and storage, especially for arc signals with non-differentiable points (near the flat shoulder). The fitting is difficult and the effect is poor.

[0053] This embodiment uses Chebyshev polynomials to replace the fitting function in the basic kan model, and the filter weighting coefficient can be reduced by controlling the polynomial, which can effectively reduce the requirements for calculation and storage and improve the operation efficiency. In addition, the Chebyshev polynomial has a flatter amplitude-frequency response curve, has better anti-noise performance, and can retain the detailed characteristics of the signal, which has better adaptability to loads with strong noise such as switching power supplies in arc fault detection.

[0054] Specifically, Chebyshev polynomials are used as the fitting function of the Kan model.

[0055] The improved Kan model is used to fit functions to the data with arc and data without arc. Through continuous debugging, the loss rate of the improved Kan model is made to be below the requirement. The optimal structure of the Kan model in the field of fault arc detection is determined to be [1,2,1], where the first 1 is the arc data input, 2 is the number of neurons, and 1 is the output of the Kan model training result.

[0056] The training process of the improved kan model is as follows Figure 1 As shown, specifically:

[0057] Step 1: Get started;

[0058] Step 2: The initial load type L=0, Ln is the load type set according to the usage scenario;

[0059] The initial number of expressions n=0;

[0060] Set the loss value Loss;

[0061] Set the regression coefficient R2;

[0062] Step 3: Create a dataset;

[0063] Step 4: Set (modify) the improved kan model structure parameters;

[0064] Step 5: Train the improved kan model;

[0065] Step 6: Determine whether Loss < 0.01? If yes, proceed to step 7, if not, return to step 4;

[0066] Step 7: Select the symbolic function fitting formula;

[0067] Step 8: Determine whether R2 ≥ 0.98? If yes, proceed to step 9; if no, return to step 7;

[0068] Step 9: Get a single expression n;

[0069] Step 10: Use data of one load type to train the improved kan model multiple times, repeat steps 3 to 9, and obtain multiple expressions n. Determine whether n ≥ 10 at this time. If yes, reset n to 0 and continue to step 11. If no, return to step 3.

[0070] Step 11: Determine whether L≥Ln? If yes, proceed to step 12; if no, proceed to the next load type training, at this time L+1, return to step 3;

[0071] Step 12: Obtain the average value of the corresponding coefficients of multiple expressions corresponding to a load type, for example, k n x+b, we get k1~k10 and b1~b10, and then take the average value of each parameter to enhance the representativeness and generalization ability of the expression;

[0072] Step 13: For Ln set load types, obtain Ln load expression formulas according to the above steps.

[0073] According to the above method, the arcing and non-arcing data sets under different loads are processed respectively, and the symbolic function expressions corresponding to the arcing and non-arcing currents under different types of loads can be obtained.

[0074] In step 3, the data needs to be normalized when making the data set. The symbolic function expression corresponding to the new type of load can also be obtained according to the above process, and the method is extensible.

[0075] According to the commonly used sampling rates and debugging results in the field of arc fault detection, the optimal grid spacing of the model is determined to be 50, and the optimal k value of the cubic spline is 3. The Kan model parameters used are: KAN (width=[1, 2, 1], grid=50, k=3).

[0076] The structure diagram of the improved Kan model corresponding to the arc data is as follows Figure 2 As shown in the figure, it is divided into three layers: input layer, middle layer and hidden layer, with the number of nodes being 1, 2 and 1. There is an activation function between the input and middle layers, and between the middle layer and the input layer. The activation function can be learned and is the core of the improved kan model.

[0077] Once the number of neurons in the model exceeds 2, the loss function will jump repeatedly during the training process. The number of neurons in the model is set to 2, the training time of the model is short and the size is small.

[0078] The structure diagram of the improved Kan model corresponding to the arc-free data is as follows: Figure 3 As shown in the figure, it is divided into three layers: input layer, middle layer and hidden layer, with the number of nodes being 1, 2 and 1. There is an activation function between the input and middle layers, and between the middle layer and the input layer. The activation function can be learned and is the core of the improved kan model.

[0079] For example, under resistive load, the arc data expression is:

[0080] (1)

[0081] like Figure 4 As shown, the yellow image is the original data image, and the blue image is the fitted symbolic function image.

[0082] The expression for arc-free data is:

[0083] 0.49*sin(6.28*x-9.43)+0.5(2)

[0084] like Figure 5 As shown, there is an obvious difference in the expression between the presence and absence of arc data, which is the main basis for identifying fault arcs in this embodiment.

[0085] In this embodiment, if Figure 6 As shown in the figure, all the trained arc and no-arc sign functions are deployed on the deployment platform in advance, and the computing platform is waited for to complete the current data collection of one power frequency cycle (20ms, 2000 points). After normalizing the data, a sample of one cycle is obtained, and the deviation of the sample to each sign function is calculated respectively. The one with the smallest deviation is considered to be the sign function that best matches the sample. According to the arc / no-arc type to which the sign function that best matches the sample belongs, it can be determined whether a fault arc has occurred at this time.

[0086] In addition to adjusting the model parameters, traditional artificial intelligence methods cannot manually intervene in the model training process, and the output results of the model are uncontrollable and unexplainable. Compared with traditional artificial intelligence algorithms, this embodiment forcibly restricts the sin function in the symbolic function basis, thereby forcibly adding the law of alternating current to the black-box artificial intelligence model, and adding human prior knowledge makes the model results more reasonable. In addition, the model structure is improved to make it suitable for one-dimensional arc current data, and the optimal model parameters are adjusted to ensure that the training loss value is below the limit. The symbolic function basis is forced to be restricted to ensure that the function has real physical meaning and R2 meets the requirements.

[0087] Embodiment 2

[0088] The purpose of this embodiment is to provide a low voltage AC fault arc detection system, comprising:

[0089] A preprocessing module is configured to: obtain arc data and no-arc data under different loads and perform preprocessing;

[0090] The training module is configured to: train the corresponding improved Kan model with the preprocessed arc data and arc-free data respectively, and fit the sign function corresponding to the arc data and the sign function corresponding to the arc-free data; wherein the improvement of the Kan model is: using Chebyshev polynomial as the fitting function of the Kan model;

[0091] The detection module is configured to: obtain current data to be detected, calculate the deviation between the obtained current data to be detected and the sign function corresponding to the arc data and the sign function corresponding to the no-arc data, and obtain the detection result of the current data to be detected.

[0092] In the preprocessing module, the data with and without arcs are normalized.

[0093] In the training module, the Kan model is first improved. The main function of the Kan model is to fit the most suitable sign function for arc and no arc data respectively. By modifying the Kan model structure, the model is applicable to the one-dimensional data of fault arc current. The original Kan model requires two inputs, which does not conform to the actual situation in the field of fault arc detection. By modifying the original Kan model, it is adapted to one-dimensional data input, so that it has the conditions for application in the field of fault arc detection.

[0094] The basic Kan model is based on B-spline curve fitting, which requires a large number of control points and nodes to achieve relatively accurate curve fitting. It has a heavy burden on calculation and storage, especially for arc signals with non-differentiable points (near the flat shoulder). The fitting is difficult and the effect is poor.

[0095] In this embodiment, the Chebyshev polynomial is used to replace the fitting function in the basic kan model, and the filter weighting coefficient can be reduced by controlling the polynomial, which can effectively reduce the requirements for calculation and storage and improve the operation efficiency. In addition, the Chebyshev polynomial has a flatter amplitude-frequency response curve, has better anti-noise performance, and can retain the detailed characteristics of the signal, which has better adaptability for loads with strong noise such as switching power supplies in arc fault detection.

[0096] There are two main types of Chebyshev polynomials: Chebyshev polynomials of the first kind and Chebyshev polynomials of the second kind. Chebyshev polynomials of the first kind are usually denoted as Tn(x), while Chebyshev polynomials of the second kind are denoted as Un(x). Chebyshev polynomials of the first kind can be defined by a generating function, while Chebyshev polynomials of the second kind also have their own specific generating function representation. These sequences of polynomials are defined recursively and have many important mathematical properties, such as orthogonality and recurrence relations. ‌

[0097] The improved Kan model is fitted with functions for arc data and no-arc data. Through continuous debugging, the loss rate of the improved Kan model is reduced to below the requirement, and the optimal structure of the Kan model in the field of fault arc detection is determined.

[0098] Among them, the training of the improved Kan model includes:

[0099] Step 1: Get started;

[0100] Step 2: The initial load type L=0, Ln is the load type set according to the usage scenario;

[0101] The initial number of expressions n=0;

[0102] Set the loss value Loss;

[0103] Set the regression coefficient R2;

[0104] Step 3: Create a dataset;

[0105] Step 4: Set (modify) the improved kan model structure parameters;

[0106] Step 5: Train the improved kan model;

[0107] Step 6: Determine whether Loss < 0.01? If yes, proceed to step 7, if not, return to step 4;

[0108] Step 7: Select the symbolic function fitting formula;

[0109] Step 8: Determine whether R2 ≥ 0.98? If yes, proceed to step 9; if no, return to step 7;

[0110] Step 9: Get a single expression n;

[0111] Step 10: Use data of one load type to train the improved kan model multiple times, repeat steps 3 to 9, and obtain multiple expressions n. Determine whether n ≥ 10 at this time. If yes, reset n to 0 and continue to step 11. If no, return to step 3.

[0112] Step 11: Determine whether L≥Ln? If yes, proceed to step 12; if no, proceed to the next load type training, at this time L+1, return to step 3;

[0113] Step 12: Obtain the average value of the corresponding coefficients of multiple expressions corresponding to a load type, for example, k n x+b, we get k1~k10 and b1~b10, and then take the average value of each parameter to enhance the representativeness and generalization ability of the expression;

[0114] Step 13: For Ln set load types, obtain Ln load expression formulas according to the above steps.

[0115] According to the above method, the arcing and non-arcing data sets under different loads are processed respectively, and the symbolic function expressions corresponding to the arcing and non-arcing currents under different types of loads can be obtained.

[0116] The improved Kan model structure corresponding to arc data is divided into three layers: input layer, middle layer and hidden layer, with the number of nodes being 1, 2, and 1. There is an activation function between the input and middle layers, and between the middle layer and the input layer. The activation function can be learned and is the core of the improved Kan model.

[0117] The improved Kan model structure corresponding to arc-free data is divided into three layers: input layer, middle layer and hidden layer, with the number of nodes being 1, 2, and 1. There is an activation function between the input and middle layers, and between the middle layer and the input layer. The activation function can be learned and is the core of the improved Kan model.

[0118] Most of the existing artificial intelligence methods in the field of arc fault detection are large in size and basically cannot be deployed on embedded platforms, and their distributed deployment capabilities are very poor. Even if there are a few deployable lightweight models, the flash size and main frequency requirements of the distributed computing platform are still very high, and there are even mandatory requirements for the chip model. The final output of the artificial intelligence model in this solution is a symbolic function, which has no special requirements for deployment, has extremely low computing pressure, and can basically be deployed on existing embedded chips.

[0119] It should be noted that: the system provided in the above embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the fault detection verification device and the fault detection verification method embodiment provided in the above embodiment belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0120] A schematic diagram of the structure of a server provided by an exemplary embodiment of the present application. Specifically:

[0121] The server includes a central processing unit (CPU), a system memory including a random access memory (RAM) and a read-only memory (ROM), and a system bus connecting the system memory and the central processing unit. The server also includes a large-capacity storage device for storing an operating system, application programs, and other program modules.

[0122] The mass storage device is connected to the central processing unit through a mass storage controller (not shown) connected to the system bus. The mass storage device and its associated computer readable medium provide non-volatile storage for the server. That is, the mass storage device may include a computer readable medium (not shown) such as a hard disk or a compact disc read-only memory (CD-ROM) drive.

[0123] Without loss of generality, computer-readable media may include computer storage media and communication media. Computer storage media include volatile and non-volatile, removable and non-removable media implemented by any method or technology for storing information such as computer-readable instructions, data structures, program modules or other data. Computer storage media include RAM, ROM, Erasable Programmable Read Only Memory (EPROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other solid-state storage technology, CD-ROM, Digital Versatile Disc (DVD) or other optical storage, cassettes, magnetic tapes, disk storage or other magnetic storage devices. Of course, those skilled in the art will appreciate that computer storage media are not limited to the above. The above-mentioned system memory and mass storage devices may be collectively referred to as memory.

[0124] According to various embodiments of the present application, the server can also be connected to a remote computer on the network through a network such as the Internet. That is, the server can be connected to the network through a network interface unit connected to the system bus, or the network interface unit can be used to connect to other types of networks or remote computer systems (not shown).

[0125] The above-mentioned memory also includes one or more programs, and the one or more programs are stored in the memory and configured to be executed by the CPU.

[0126] An embodiment of the present application also provides a computer device, which includes a processor and a memory, in which at least one instruction, at least one program, a code set or an instruction set is stored, and the at least one instruction, at least one program, a code set or an instruction set is loaded and executed by the processor to implement the fault detection verification method provided by the above-mentioned method embodiments.

[0127] An embodiment of the present application also provides a computer-readable storage medium, on which is stored at least one instruction, at least one program, code set or instruction set, and the at least one instruction, at least one program, code set or instruction set is loaded and executed by a processor to implement the fault detection verification method provided by the above-mentioned method embodiments.

[0128] The embodiment of the present application also provides a computer program product, the computer program product or computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the verification method of fault detection described in any of the above embodiments.

[0129] Optionally, the computer readable storage medium may include: a read-only memory (ROM), a random access memory (RAM), a solid state drive (SSD), or an optical disk. Among them, the random access memory may include a resistive random access memory (ReRAM) and a dynamic random access memory (DRAM). The serial numbers of the above embodiments of the present application are only for description and do not represent the advantages and disadvantages of the embodiments.

[0130] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0131] The above description is only an optional embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0132] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

Claims

1. A low voltage AC fault arc detection method, characterized in that: include: Obtain arc data and no-arc data under different loads and perform preprocessing; The pre-processed arc data and arc-free data are used to train the corresponding improved Kan model respectively, and the sign function corresponding to the arc data and the sign function corresponding to the arc-free data are fitted; wherein, the improvement of the Kan model is: the Chebyshev polynomial is used as the fitting function of the Kan model; The improved Kan model training process is as follows: Step 1: Get started; Step 2: The initial load type L=0, Ln is the load type set according to the usage scenario; The initial number of expressions n=0; Set the loss value Loss; Set the regression coefficient R2; Step 3: Create a dataset; Step 4: Set the structural parameters of the improved Kan model; Step 5: Train the improved Kan model; Step 6: Determine whether Loss < 0.01, if so, proceed to step 7, if not, return to step 4; Step 7: Select the symbolic function fitting formula; Step 8: Determine whether R2 ≥ 0.

98. If so, proceed to step 9. If not, return to step 7. Step 9: Get a single expression n; Step 10: Use data of one load type to train the improved Kan model multiple times, repeat steps 3 to 9, obtain multiple expressions n, and determine whether n ≥ 10 at this time. If so, reset n to 0 and continue to step 11. If not, return to step 3; Step 11: Determine whether L≥Ln. If so, proceed to step 12. If not, perform training of the next load type, at this time L+1, and return to step 3. Step 12: Take the average value of the corresponding coefficients of multiple expressions corresponding to one load type to obtain the expression formula of the load type; Step 13: For Ln types of set load types, obtain Ln load expression formulas according to the above steps; The current data to be detected is obtained, and the deviations of the obtained current data to be detected and the sign function corresponding to the arc data and the sign function corresponding to the no-arc data are calculated to obtain the detection result of the current data to be detected.

2. A low voltage AC fault arc detection method as claimed in claim 1, characterized in that: The Kan model is improved and Chebyshev polynomials with sin function are used as the fitting function of the Kan model.

3. A low voltage AC fault arc detection method as claimed in claim 1, characterized in that: The obtained arc data and no-arc data under different loads are preprocessed, specifically: the obtained arc data and no-arc data under different loads are normalized.

4. A low voltage AC fault arc detection method as claimed in claim 1, characterized in that: Calculate the deviations between the acquired current data to be detected and the sign function corresponding to the arc data and the sign function corresponding to the no-arc data, and take the arc or no-arc type corresponding to the sign function with the smallest deviation as the detection result of the current data to be detected.

5. A low voltage AC fault arc detection system, characterized in that: include: A preprocessing module is configured to: obtain arc data and no-arc data under different loads and perform preprocessing; The training module is configured to: train the corresponding improved Kan model with the preprocessed arc data and arc-free data respectively, and fit the sign function corresponding to the arc data and the sign function corresponding to the arc-free data; wherein the improvement of the Kan model is: using Chebyshev polynomial as the fitting function of the Kan model; The improved Kan model training process is as follows: Step 1: Get started; Step 2: The initial load type L=0, Ln is the load type set according to the usage scenario; The initial number of expressions n=0; Set the loss value Loss; Set the regression coefficient R2; Step 3: Create a dataset; Step 4: Set the structural parameters of the improved Kan model; Step 5: Train the improved Kan model; Step 6: Determine whether Loss < 0.01, if so, proceed to step 7, if not, return to step 4; Step 7: Select the symbolic function fitting formula; Step 8: Determine whether R2 ≥ 0.

98. If so, proceed to step 9. If not, return to step 7. Step 9: Get a single expression n; Step 10: Use data of one load type to train the improved Kan model multiple times, repeat steps 3 to 9, obtain multiple expressions n, and determine whether n ≥ 10 at this time. If so, reset n to 0 and continue to step 11. If not, return to step 3; Step 11: Determine whether L≥Ln. If so, proceed to step 12. If not, perform training of the next load type, at this time L+1, and return to step 3. Step 12: Take the average value of the corresponding coefficients of multiple expressions corresponding to one load type to obtain the expression formula of the load type; Step 13: For Ln types of set load types, obtain Ln load expression formulas according to the above steps; The detection module is configured to: obtain current data to be detected, calculate the deviation between the obtained current data to be detected and the sign function corresponding to the arc data and the sign function corresponding to the no-arc data, and obtain the detection result of the current data to be detected.

6. A low voltage AC fault arc detection system as claimed in claim 5, characterized in that: In the preprocessing module, the acquired arcing data and no-arc data under different loads are preprocessed, specifically: the acquired arcing data and no-arc data under different loads are normalized.

7. A low voltage AC fault arc detection system as claimed in claim 5, characterized in that: In the training module, the adopted Kan model is improved, and Chebyshev polynomials are used as the fitting function of the Kan model.

8. An electronic device, characterized in that: The method comprises a memory and a processor and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method according to any one of claims 1 to 4 is completed.

9. A computer-readable storage medium, characterized in that: Used to store computer instructions, which, when executed by a processor, complete the method described in any one of claims 1 to 4.

10. A computer program product, characterized in that The invention comprises a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 4.

Images