Current waveform sample generation method, apparatus, device, and medium

Abstract

The application provides a current waveform sample generation method, device, equipment and medium. Relates to the field of data generation, the current waveform sample generation method comprises: determining a first model and a second model from a pre-trained model library according to a load type of a target power load; using the first model to perform feature fitting on a low-frequency part of a misoperation waveform corresponding to the target power load to obtain a basic fitting waveform, the misoperation waveform being a current waveform when the target power load occurs a circuit breaker misoperation; performing feature fitting on a medium-frequency part and a high-frequency part of the misoperation waveform through the second model to obtain a detail fitting waveform; and using the detail fitting waveform to perform superposition adjustment on the basic fitting waveform to obtain a current waveform sample, the current waveform sample being used to simulate a current waveform generated when the target power load occurs the circuit breaker misoperation.

Description

Technical Field

[0001] This application relates to the field of data generation, and more specifically, to a method, apparatus, device, and medium for generating current waveform samples. Background Technology

[0002] With the increasing demands for electrical safety in low-voltage power distribution systems, arc fault protection technology has been widely applied in industrial, new energy, and civil sectors. Artificial intelligence-based arc fault detection models, with their powerful time-frequency feature extraction capabilities, have become the mainstream protection method in this field.

[0003] However, due to the diverse and dynamic nature of load types in actual operating conditions, normal current signals often exhibit multimodal characteristics. Existing artificial intelligence models, however, face a bottleneck in generalization ability, with their classification boundaries highly dependent on the distribution of training data. In realizing the concept of this application, the inventors discovered that in related technologies, insufficient training data samples lead to weak generalization ability of the artificial intelligence model, making it difficult to efficiently achieve the purpose of using an arc fault detection model for detection. Summary of the Invention

[0004] In view of this, this application provides a method, apparatus, device and medium for generating current waveform samples.

[0005] One aspect of this application provides a method for generating current waveform samples, comprising: determining a first model and a second model from a pre-trained model library according to the load type of a target electrical load; using the first model to perform feature fitting on the low-frequency part of the malfunction waveform corresponding to the target electrical load to obtain a basic fitted waveform, wherein the malfunction waveform is the current waveform when the circuit breaker of the target electrical load malfunctions; using the second model to perform feature fitting on the mid-frequency and high-frequency parts of the malfunction waveform to obtain a detailed fitted waveform; and using the detailed fitted waveform to superimpose and adjust the basic fitted waveform to obtain a current waveform sample, wherein the current waveform sample is used to simulate the current waveform generated when the circuit breaker of the target electrical load malfunctions.

[0006] According to an embodiment of this application, a first model is used to perform feature fitting on the low-frequency portion of the malfunction waveform corresponding to the target electrical load to obtain a basic fitted waveform. This includes: performing harmonic decomposition on the malfunction waveform using the Fourier basis functions in the first model to obtain multiple harmonic components of different frequencies; performing feature superposition on the multiple harmonic components of different frequencies based on aggregation operations to obtain basic features; and performing waveform approximation fitting on the malfunction waveform based on the basic features to obtain the basic fitted waveform.

[0007] According to an embodiment of this application, a first model is used to perform feature fitting on the low-frequency portion of the malfunction waveform corresponding to the target electrical load to obtain a basic fitted waveform. This includes: performing harmonic decomposition on the malfunction waveform using the Fourier basis functions in the first model to obtain multiple harmonic components of different frequencies; performing feature superposition on the multiple harmonic components of different frequencies based on aggregation operations to obtain basic features; and performing waveform approximation fitting on the malfunction waveform based on the basic features to obtain the basic fitted waveform.

[0008] According to an embodiment of this application, a detailed fitted waveform is obtained by performing feature fitting on the mid-frequency and high-frequency components of the erroneous waveform using a second model. This includes: determining residual waveforms representing the mid-frequency and high-frequency characteristics of the erroneous waveform based on the difference between the erroneous waveform and the basic fitted waveform; and inputting the residual waveforms into the second model for waveform approximation fitting to obtain the detailed fitted waveform corresponding to the erroneous waveform.

[0009] According to an embodiment of this application, the residual waveform is input into a second model for waveform approximation fitting to obtain a detailed fitted waveform corresponding to the erroneous waveform, including:

[0010] The erroneous waveform is decomposed into harmonics using the Fourier basis functions in the second model to obtain multiple harmonic components corresponding to the intermediate and high frequencies. The features of the multiple harmonic components are superimposed to obtain detailed features. Based on the detailed features, the residual waveform is approximated to obtain the basic fitted waveform.

[0011] According to an embodiment of this application, a current waveform sample is obtained by superimposing and adjusting a basic fitted waveform using a detailed fitted waveform, including: enhancing the details of the detailed fitted waveform using a perturbation coefficient to obtain a perturbation waveform; and superimposing the mid-frequency and high-frequency portions of the basic fitted waveform using the perturbation waveform to obtain a current waveform sample.

[0012] According to an embodiment of this application, determining a first model and a second model from a pre-trained model library based on the load type of the target electrical load includes: determining a first model set and a second model set corresponding to the load type from the pre-trained model library based on the load type of the target electrical load; randomly determining a first model and a second model from the first model set and the second model set respectively; wherein the first model set includes multiple first models, and the second model set includes multiple second models; the multiple first models are trained based on the current waveforms at multiple times after the electrical load malfunctions, and the multiple second models are trained based on the high-frequency harmonics of the current waveforms at multiple times after the electrical load malfunctions.

[0013] According to an embodiment of this application, the current waveform sample generation method further includes: determining multiple waveform feature constraints of the target electrical load based on the malfunction waveform of the target electrical load; and performing waveform screening in multiple current waveform samples based on the multiple waveform feature constraints to determine current waveform samples that meet the waveform feature constraints.

[0014] Another aspect of this application provides a current waveform sample generation apparatus, comprising: a determination module, configured to determine a first model and a second model from a pre-trained model library based on the load type of a target electrical load; a first fitting module, configured to use the first model to perform feature fitting on the low-frequency portion of the malfunction waveform corresponding to the target electrical load to obtain a basic fitted waveform, wherein the malfunction waveform is the current waveform when the circuit breaker of the target electrical load malfunctions; a second fitting module, configured to perform feature fitting on the mid-frequency and high-frequency portions of the malfunction waveform using the second model to obtain a detailed fitted waveform; and a superposition module, configured to use the detailed fitted waveform to superimpose and adjust the basic fitted waveform to obtain a current waveform sample, thereby simulating a current waveform different from the malfunction waveform generated when the circuit breaker of the target electrical load malfunctions.

[0015] Another aspect of this application provides an electronic device comprising: one or more processors; and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method described above.

[0016] Another aspect of this application provides a computer-readable storage medium storing computer-executable instructions that, when executed, are used to implement the method described above.

[0017] Another aspect of this application provides a computer program product including computer-executable instructions that, when executed, are used to implement the methods described above.

[0018] According to embodiments of this application, the above-described scheme, through a two-stage decoupled modeling and superposition adjustment mechanism, avoids the problem of generated data deviating from the real signal in terms of harmonic structure, RMS value, and spectral distribution, compared to traditional data-driven methods based on GANs or VAEs. Simultaneously, the technique of outputting basic and detailed fitted waveforms based on the first and second models respectively can suppress the overfitting risk of a single model, balancing high-precision fitting with very few samples with the diversity of generated data. Compared to physical methods that rely on specific load mechanisms for modeling, this scheme does not require the establishment of complex parameter models for different loads, enabling it to adapt to diverse operating conditions within a unified framework, thus improving engineering adaptability and deployment efficiency. Attached Figure Description

[0019] The above and other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0020] Figure 1 An exemplary system architecture for applying a current waveform sample generation method according to an embodiment of this application is illustrated.

[0021] Figure 2 A flowchart illustrating a current waveform sample generation method according to an embodiment of this application is shown schematically.

[0022] Figure 3 The diagram illustrates a data stream generated from a current waveform sample according to an embodiment of this application.

[0023] Figure 4 The illustration shows a schematic diagram of a second model training process according to an embodiment of this application.

[0024] Figure 5(a) schematically shows a comparison of the malfunction waveform of an incandescent lamp and a current waveform sample according to an embodiment of the present application.

[0025] Figure 5(b) schematically illustrates a comparison of the waveform of a malfunctioning electric drill and a current waveform sample according to an embodiment of the present application.

[0026] Figure 5(c) schematically illustrates a comparison of the power supply malfunction waveform and current waveform sample according to an embodiment of the present application.

[0027] Figure 5(d) schematically illustrates a comparison of the malfunction waveform and current waveform sample of a halogen lamp according to an embodiment of the present application.

[0028] Figure 5(e) schematically illustrates a comparison of the malfunction waveform and current waveform sample of a microwave oven according to an embodiment of the present application.

[0029] Figure 5(f) schematically illustrates a comparison of the malfunction waveform of a fluorescent lamp and a current waveform sample according to an embodiment of the present application.

[0030] Figure 5(g) schematically illustrates a comparison of the malfunction waveform of a resistive load and a current waveform sample according to an embodiment of the present application.

[0031] Figure 5(h) schematically shows a comparison of the malfunction waveform and current waveform sample of a vacuum cleaner according to an embodiment of the present application.

[0032] Figure 6(a) schematically illustrates a parametric comparison feature radar diagram of malfunction waveforms and current waveform samples of a microwave oven, fluorescent lamp, resistive load, and vacuum cleaner according to an embodiment of the present application.

[0033] Figure 6(b) schematically illustrates a parametric comparison feature radar plot of malfunction waveforms and current waveform samples of incandescent lamps, electric drills, power supplies and halogen lamps according to embodiments of this application.

[0034] Figure 7 A block diagram of a current waveform sample generation apparatus according to an embodiment of this application is shown schematically.

[0035] Figure 8 A block diagram of an electronic device suitable for a current waveform sample generation method according to an embodiment of this application is shown schematically. Detailed Implementation

[0036] The embodiments of this application will now 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 this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be implemented without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0037] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0038] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0039] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0040] In related technologies, electrical loads operate under complex and variable conditions. When faced with unknown non-fault disturbances in new scenarios, normal current signals are easily misinterpreted as arc faults, leading to erroneous operations. These erroneous operations are extremely dangerous: in industrial sectors, they can cause unexpected shutdowns of continuous production lines, resulting in significant economic losses; in new energy systems, they can easily trigger grid connection interruptions and grid oscillations; in civilian scenarios, they can not only cause frequent tripping and severely impact user experience, but may even damage precision electrical appliances due to non-accidental power outages.

[0041] Solving the problem of malfunctions in diverse new scenarios hinges on supplementing the model with training data under these new operating conditions, but this process is constrained by physical limitations. Specifically, according to national standards, the maximum permissible tripping time of arc fault protection devices (AFDDs) under household conditions must be controlled within 0.25 seconds. In a 50Hz power frequency system, this means that once the protection device malfunctions in response to a normal disturbance, only about 12.5 cycles of normal operating waveforms can be acquired before the system forcibly cuts off the power. Therefore, limited by low acquisition efficiency and high on-site acquisition costs, the acquired data used for supplementary training under new operating conditions often suffers from a small sample size problem.

[0042] Under the aforementioned small sample conditions, relevant methods mainly improve the sample size through data generation and augmentation, including data-driven and physical modeling methods. While data-driven methods based on GANs (Generative Adversarial Networks) or VAEs (Variational Autoencoders) possess strong nonlinear modeling capabilities, they lack physical constraints, and the generated data may deviate from the true signal. Methods based on physical mechanisms require modeling for different loads, resulting in complex parameter design, poor versatility, and difficulty adapting to diverse operating conditions. Therefore, a novel normal operating condition data generation method is urgently needed for small sample conditions. This method must achieve high-quality, high-reliability extension of normal signals while strictly ensuring electrical and physical characteristics, thereby improving the detection model's ability to identify complex disturbances and reducing the risk of misjudgment.

[0043] In view of this, embodiments of this application provide a current waveform sample generation method, comprising: determining a first model and a second model from a pre-trained model library according to the load type of the target electrical load; using the first model to perform feature fitting on the low-frequency part of the malfunction waveform corresponding to the target electrical load to obtain a basic fitted waveform, wherein the malfunction waveform is the current waveform when the circuit breaker of the target electrical load malfunctions; using the second model to perform feature fitting on the mid-frequency and high-frequency parts of the malfunction waveform to obtain a detailed fitted waveform; and using the detailed fitted waveform to superimpose and adjust the basic fitted waveform to obtain a current waveform sample, thereby simulating a current waveform different from the malfunction waveform generated when the circuit breaker of the target electrical load malfunctions.

[0044] Figure 1An exemplary system architecture 100 according to embodiments of this application, to which the current waveform sample generation method can be applied, is illustrated schematically. It should be noted that... Figure 1 The examples shown are merely examples of system architectures that can be applied to the embodiments of this application, in order to help those skilled in the art understand the technical content of this application, but do not mean that the embodiments of this application cannot be used in other devices, systems, environments or scenarios.

[0045] like Figure 1 As shown, the system architecture 100 according to this embodiment may include a target electrical load 101 and a first terminal device 102. The target electrical load 101 and the first terminal device 102 may be connected by a wired connection.

[0046] Users can use the first terminal device 102 to collect the malfunction waveform generated by the target electrical load 101 when the circuit breaker malfunctions. Waveform processing software can be installed on the first terminal device 101 (for example only).

[0047] The first terminal device 102 can be various electronic devices with a display screen and support web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0048] Users can process the malfunction waveform collected by the first terminal device 102 and feed back the processed current waveform sample to the terminal device.

[0049] In addition, the current waveform sample generation device can be set in the first terminal device 102.

[0050] It should be understood that Figure 1 The number of terminal devices and target electrical loads shown is merely illustrative. Any number of terminal devices and target electrical loads can be included depending on implementation requirements.

[0051] Figure 2 A flowchart illustrating a current waveform sample generation method according to an embodiment of this application is shown schematically.

[0052] like Figure 2 As shown, the method includes operations S210~S240.

[0053] In operation S210, a first model and a second model are determined from a pre-trained model library based on the load type of the target electrical load.

[0054] According to embodiments of this application, the models in the pre-trained model library described above can be stored according to the load type under malfunction conditions, with each load type corresponding to a set of independently trained model parameters. Specifically, for a specific load type, the model library stores multiple first models trained from multiple cycles of current waveforms collected under normal operating conditions for that load, and multiple corresponding second models. Here, "load type" does not only refer to a single load individual, but can also refer to load categories with similar electrical characteristics, such as resistive loads, inductive loads, nonlinear rectifier loads, etc. Multiple cycle waveforms under the same category are trained to form a model set, thereby constituting a sub-library of models corresponding to that load type.

[0055] According to an embodiment of this application, the first model described above can be trained based on the malfunction waveform when the circuit breaker of the target electrical load malfunctions, and the second model described above can be trained based on the difference waveform between the result of the first model and the malfunction waveform.

[0056] In operation S220, the first model is used to perform feature fitting on the low-frequency part of the malfunction waveform corresponding to the target electrical load to obtain the basic fitted waveform. The malfunction waveform is the current waveform when the circuit breaker of the target electrical load malfunctions.

[0057] According to an embodiment of this application, the aforementioned malfunction waveform is a current signal collected when the circuit breaker misjudges under normal operating conditions. The spectrum of the malfunction waveform covers the low-frequency part determined by the inherent characteristics of the load and the medium-to-high-frequency part generated by nonlinear disturbances.

[0058] The pre-trained first model can fit the low-frequency part of the erroneous waveform using multiple harmonic components to obtain a basic fitted waveform, which can be similar to the skeleton part of the erroneous waveform.

[0059] During operation of S230, the mid-frequency and high-frequency components of the erroneous waveform are fitted using the second model to obtain the detailed fitted waveform.

[0060] According to embodiments of this application, a pre-trained second model can use multiple different harmonic components to fit the mid-frequency and high-frequency portions of the malfunction waveform. The mid-frequency and high-frequency portions of the malfunction waveform can represent the waveform details of the malfunction waveform. Compared to the skeleton portion, the mid-frequency and high-frequency portions of the malfunction waveform can serve as the detailed portions of the malfunction waveform. These detailed portions can encompass harmonic distortion caused by load nonlinear characteristics, morphological variations caused by fluctuations in operating conditions, and random disturbance components such as measurement noise.

[0061] The second model uses the difference between the real waveform and the basic fitted waveform as the training objective. It independently fits the nonlinear microstructure features in the residual signal through a deep network structure, outputting a detailed fitted waveform. This process decouples and models the individual differences in the original signal beyond the common characteristics of the load, enabling the data generation to inherit the physical commonalities of the load type while possessing the ability to express the diversity of the actual operating state.

[0062] In operation S240, the basic fitted waveform is superimposed and adjusted using the detailed fitted waveform to obtain a current waveform sample. The current waveform sample is used to simulate the current waveform generated when the circuit breaker of the target electrical load malfunctions.

[0063] According to embodiments of this application, the aforementioned superposition adjustment refers to applying the detailed fitted waveform as a dynamic correction to the basic fitted waveform, thereby generating new current waveform samples through structured recombination. During this superposition process, the basic fitted waveform provides a steady-state skeleton that conforms to the physical laws of the load type, ensuring that the generated signal maintains authenticity in core electrical properties such as fundamental energy and harmonic structure; detailed fitting enables the synthesized current waveform sample signal to reflect individual differences in different operating states.

[0064] By randomly selecting unpaired basic fitted waveforms and detailed fitted waveforms from a model sub-library of the same load type and recombining them, a cross-combination of skeleton and residual can be achieved, thereby generating a new sample that is different from the original malfunction waveform while ensuring electrical and physical consistency.

[0065] According to embodiments of this application, the above-described scheme, through a two-stage decoupled modeling and superposition adjustment mechanism, avoids the problem of generated data deviating from the real signal in terms of harmonic structure, RMS value, and spectral distribution, compared to traditional data-driven methods based on GANs or VAEs. Simultaneously, the technique of outputting basic and detailed fitted waveforms based on the first and second models respectively can suppress the overfitting risk of a single model, balancing high-precision fitting with very few samples with the diversity of generated data. Compared to physical methods that rely on specific load mechanisms for modeling, this scheme does not require the establishment of complex parameter models for different loads, enabling it to adapt to diverse operating conditions within a unified framework, thus improving engineering adaptability and deployment efficiency.

[0066] According to an embodiment of this application, a first model is used to perform feature fitting on the low-frequency portion of the malfunction waveform corresponding to the target electrical load to obtain a basic fitted waveform. This includes: performing harmonic decomposition on the malfunction waveform using the Fourier basis functions in the first model to obtain multiple harmonic components of different frequencies; performing feature superposition on the multiple harmonic components of different frequencies based on aggregation operations to obtain basic features; and performing waveform approximation fitting on the malfunction waveform based on the basic features to obtain the basic fitted waveform.

[0067] According to embodiments of this application, the first model decomposes the malfunction waveform using multiple harmonic components through waveform fitting based on a Fourier series activation function, yielding multiple harmonic components of different frequencies. The above steps utilize the coefficients of trainable Fourier basis functions to adaptively learn the amplitude and phase of the fundamental wave and major lower-order harmonics, ensuring that the weight configuration of each harmonic component aligns with the electrical characteristics of a specific load type. The feature superposition of these harmonic components based on aggregation operations essentially reconstructs the discrete frequency components into basic features characterizing the overall waveform shape and energy distribution. These basic features preserve the main power structure and fundamental frequency characteristics of the target load's steady-state operation, providing a physically interpretable intermediate representation for subsequent waveform approximation fitting.

[0068] According to an embodiment of this application, the waveform approximation of the malfunction waveform can then be performed based on fundamental characteristics. This means that, driven by aggregated harmonic characteristics, a basic fitted waveform reflecting the target electrical load is output through network mapping of the first model. This fitting process approximates the contour of the original waveform using an explicit functional form. The resulting basic fitted waveform, serving as the steady-state skeleton of the current signal, can reveal the core electrical properties of the target electrical load, ensuring that the generated basic fitted waveform conforms to physical laws in terms of energy level and spectral structure.

[0069] According to the embodiments of this application, the first model can be constructed using the framework of the KAN model. The B-spline function in the KAN model can use the Fourier series as the activation function, as shown in formula (1).

[0070] (1)

[0071] in, , , All of these are trainable parameters. It can represent Fourier basis functions.

[0072] Figure 3 The diagram illustrates a data stream generated from a current waveform sample according to an embodiment of this application.

[0073] According to the embodiments of this application, refer to Figure 3The first model can accurately measure and determine the best fit of the skeleton model. To prevent underfitting or overfitting by excessively absorbing high-frequency details during training, the skeleton structure can be selected by calculating the Pearson correlation coefficient and the low-frequency energy proportion coefficient. The network width structure can be set to [1, 8, 2, 1]. Inputting current waveform data for one cycle, the output is the corresponding basic fitted waveform. The first model mainly learns the overall contour of the current waveform, achieving an approximate fit of the fundamental wave and the main low-order components, and outputting the basic fitted waveform. .

[0074] According to embodiments of this application, the above steps map the time-domain current signal into discrete frequency components using Fourier basis functions. Compared to the implicit feature extraction method of traditional neural networks, this scheme utilizes the explicit expansion characteristics of the Fourier series in the KAN model to directly map the trainable parameters of each connection edge to the amplitude and phase of a specific harmonic, thereby achieving visualization and controllability of the spectral distribution at the model structure level. This model structure not only avoids the problem of spurious harmonics or broadband noise introduced by the generated data due to a lack of frequency domain constraints, but also provides directly interpretable intermediate variables for subsequent constraint verification based on electrophysical characteristics, thus improving the transparency and credibility of the data generation process.

[0075] According to an embodiment of this application, a detailed fitted waveform is obtained by performing feature fitting on the mid-frequency and high-frequency components of the erroneous waveform using a second model. This includes: determining a residual waveform representing the mid-frequency and high-frequency characteristics of the erroneous waveform based on the difference between the erroneous waveform and the basic fitted waveform; and inputting the residual waveform into the second model for waveform approximation fitting to obtain a detailed fitted waveform corresponding to the erroneous waveform.

[0076] According to embodiments of this application, this process structurally separates the steady-state main component and dynamic variation component of the current signal in the erroneous operation waveform through differential operations. Specifically, the basic fitted waveform obtained from the first model can be used as a steady-state skeleton reflecting the common laws of load types. By calculating the difference between the erroneous operation waveform and the basic fitted waveform, the residual waveform, which characterizes the degree to which the original signal deviates from the steady-state law, can be obtained. This residual waveform essentially represents the individual difference components in the original signal beyond the common characteristics of the load. The residual waveform can encompass mid-to-high frequency variation information caused by factors such as nonlinear distortion, operating condition fluctuations, and random disturbances.

[0077] According to an embodiment of this application, the second model uses the residual waveform as input and independently approximates and functionalizes its inherent nonlinear microstructure features, outputting a detailed fitted waveform corresponding to the erroneous waveform. This fitting process differs from the first model's macroscopic extraction of the global energy distribution.

[0078] According to an embodiment of this application, a detailed fitted waveform is obtained by performing feature fitting on the mid-frequency and high-frequency components of the erroneous waveform using a second model. This includes: determining a residual waveform representing the mid-frequency and high-frequency characteristics of the erroneous waveform based on the difference between the erroneous waveform and the basic fitted waveform; and inputting the residual waveform into the second model for waveform approximation fitting to obtain a detailed fitted waveform corresponding to the erroneous waveform.

[0079] According to embodiments of this application, the second model can take the residual waveform as the processing object, and use its built-in Fourier basis functions to perform harmonic decomposition on the dynamic variation components after the steady-state skeleton is stripped, obtaining multiple harmonic components corresponding to the intermediate frequency and high frequency parts. Then, by performing feature superposition on the above multiple harmonic components, detailed features reflecting the overall variation law and high frequency energy distribution of the residual signal can be obtained, and waveform approximation fitting can be performed on the residual waveform accordingly.

[0080] According to an embodiment of this application, the detailed fitting waveform output by the second model carries individual variation information outside the waveform skeleton, and is functionally decoupled from the basic fitting waveform output by the first model. The two are then superimposed and adjusted to jointly constitute a current waveform sample that has both physical fidelity and statistical diversity.

[0081] According to an embodiment of this application, the second model can also be trained based on the KAN model, and the activation function in the connection edges of the model can be set with reference to formula (1). Reference Figure 3 It can calculate the waveform of erroneous operation. With the basic fitted waveform The difference is used to construct the residual waveform of the input second model. The network width structure can be set to [1, 16, 32, 1]. The second model is used to fit the high-frequency details and nonlinear microstructure features in the residual.

[0082] Figure 4 The illustration shows a schematic diagram of a second model training process according to an embodiment of this application.

[0083] During training, refer to Figure 4 For the inter-function channels connecting input and output neurons in the KAN model, the L1 norm of all Fourier order coefficients in the channel can be calculated, and channels below the contribution threshold can be pruned by setting the mask to zero, thereby reducing the computational burden.

[0084] According to embodiments of this application, the second model described above is used to capture local variation patterns and high-frequency structural information beyond the low-frequency steady-state skeleton portion of the erroneous waveform. Furthermore, the dual-stage architecture of the first and second models avoids the feature confusion and overfitting risks that easily occur when a single network performs global unified modeling, enabling the detailed fitted waveform to accurately reflect the reasonable variation space of the original sample, laying the foundation for subsequent controllable diversity expansion through structured recombination.

[0085] According to an embodiment of this application, a perturbation waveform is obtained by enhancing the details of the fitted waveform using a perturbation coefficient; the mid-frequency and high-frequency portions of the basic fitted waveform are then superimposed using the perturbation waveform to obtain a current waveform sample.

[0086] According to an embodiment of this application, the perturbation coefficient can apply random scaling modulation to the variation components of the mid-to-high frequency part fitted by the second model, so that the energy level and distribution pattern of the detailed fitted waveform will produce a reasonable offset within the original sample statistical interval.

[0087] According to embodiments of this application, superimposing the perturbation-enhanced waveform onto the basic fitted waveform involves structurally recombining the basic fitted waveform, which reflects the main steady-state characteristics of the load, with mid-to-high frequency details carrying dynamic variation information, thereby obtaining a complete current waveform sample. The above steps ensure that the generated data maintains consistency with the real load in macroscopic electrical properties such as fundamental energy and major harmonic structure, while the perturbation waveform imparts reasonable variations to the generated sample that conform to the statistical characteristics of the original operating condition. The resulting current waveform sample is distinct from the original malfunction waveform and strictly adheres to the electrical and physical laws of the real load, providing the arc fault detection model with amplified training samples that combine high fidelity and statistical diversity.

[0088] According to embodiments of this application, the above steps, by using perturbation coefficients to perform parametric detail enhancement on the detail-fitted waveform, enable controllable modulation of the energy level and morphological characteristics of the variation components within the original small sample statistical distribution range. This process can minimize the electrical and physical distortion problems caused by unconstrained random noise injection, overcome the homogeneity bottleneck of multiple generated current waveform samples, expand the generated current waveform in the individual difference space, and ensure that the variation amplitude of the current waveform samples always remains within the reasonable boundaries of the actual load operating characteristics.

[0089] According to an embodiment of this application, determining a first model and a second model from a pre-trained model library based on the load type of the target electrical load includes: determining a first model set and a second model set corresponding to the load type from the pre-trained model library based on the load type of the target electrical load; randomly determining a first model and a second model from the first model set and the second model set respectively; wherein the first model set includes multiple first models, and the second model set includes multiple second models; the multiple first models are trained based on the current waveforms at multiple times after the electrical load malfunctions, and the multiple second models are trained based on the high-frequency harmonics of the current waveforms at multiple times after the electrical load malfunctions.

[0090] According to embodiments of this application, multiple models in the pre-trained model library can be categorized and stored according to the load type of the target electrical load. Each load type corresponds to an independent model sub-library, which includes a first model set and a second model set. The first model set consists of multiple first models, each independently trained based on current waveforms collected at different times after a circuit breaker malfunction for that load type. Each of the multiple first models captures the low-frequency characteristics of the current signal at different times. The second model set consists of multiple second models, each independently trained based on the mid-frequency and high-frequency harmonics of the current waveform after skeleton stripping at the corresponding time. Each model captures the nonlinear microstructure characteristics of the dynamic variation components at different times.

[0091] According to embodiments of this application, during the data generation stage, the first and second models for the current generation task can be randomly selected from the first and second model sets corresponding to the same load type. This random selection mechanism forms a non-fixed cross-pairing of the skeleton model and the residual model, ensuring that the steady-state main features and dynamic variation features used in the generation originate from sampling data at different times. The aforementioned temporal decoupling and recombination method not only ensures the inheritance of the common laws of the load type in the electrical and physical essence of the generated data, but also introduces reasonable variations that conform to the statistical distribution of the original samples through non-pairing combinations. This can minimize the homogenization problem of generated samples that is easily caused by fixed model pairing, thereby improving the diversity space and engineering applicability of the augmented data while strictly ensuring the physical characteristics of the electrical signals.

[0092] According to an embodiment of this application, for the load type of the target electrical load, the first model set may include 12 first models, which are obtained by training the current waveforms at multiple times after the target electrical load malfunctions. The first model set is shown in formula (2).

[0093] (2)

[0094] The second model set may include 12 second models, which are obtained by training the high-frequency harmonics of the current waveform at multiple times after the electrical load malfunctions. The second model set is shown in formula (3).

[0095] (3)

[0096] When generating new data, a first model can be randomly selected from the first model set and a second model can be randomly selected from the second model set. At the same time, a perturbation coefficient can be introduced to enrich the diversity of the generated samples, as shown in formula (4).

[0097] (4)

[0098] in, , , It can represent the disturbance coefficient. It can represent the generated current waveform sample that is different from the erroneous waveform.

[0099] According to embodiments of this application, the random cross-fusion mechanism in the above steps can enrich the individual variability and morphological coverage of the generated samples without introducing spurious electrical characteristics, avoiding the sample homogenization problem easily caused by fixed model pairing, thus achieving an organic unity of physical fidelity and statistical diversity under small sample conditions. Furthermore, since the generated current waveform strictly revolves around the electrical and physical constraints and statistical distribution of the original small samples, the amplified data possesses both the signal essential characteristics of the real load and covers reasonable variations of normal non-fault disturbances, effectively expanding the identification boundary of the detection model for normal current signals under unknown operating conditions. Therefore, when facing dynamically changing non-fault disturbances in new scenarios, the model can accurately distinguish the characteristic differences between arc faults and normal disturbances, fundamentally reducing the risk of circuit breaker misjudgment due to insufficient training sample coverage, and improving the reliability and safety of arc fault protection devices in actual operating environments.

[0100] According to an embodiment of this application, the current waveform sample generation method further includes: determining multiple waveform feature constraints of the target electrical load based on the malfunction waveform of the target electrical load; and performing waveform screening in multiple current waveform samples based on the multiple waveform feature constraints to determine current waveform samples that meet the waveform feature constraints.

[0101] According to an embodiment of this application, firstly, based on the malfunction waveform of the target electrical load, key electrical and physical characteristics are extracted and statistically analyzed, including but not limited to RMS value, fundamental amplitude, total harmonic distortion rate, third harmonic ratio, peak power factor, and waveform coefficient. Then, based on the distribution characteristics of the original samples, reasonable fluctuation ranges or threshold boundaries for each characteristic are determined, forming multi-dimensional waveform characteristic constraints. These constraints are not artificially set empirical constants, but rather directly derived from the statistical laws of real load operation data, thus establishing a physical reference benchmark that the generated data must follow. Then, after generating multiple current waveform samples in batches, the generated samples are screened and verified one by one according to the above waveform characteristic constraints. Abnormal generated data that deviates from the statistical distribution of the original samples is eliminated, and only current waveform samples that meet all constraints are retained as the final amplified samples.

[0102] According to embodiments of this application, the above-mentioned physical constraints may include six items, such as: effective value constraint, fundamental amplitude constraint, total harmonic distortion rate constraint, third harmonic ratio constraint, peak value constraint, and waveform coefficient constraint. The following provides further explanation of the various physical constraints.

[0103] Effective value constraints As shown in formula (5).

[0104] (5)

[0105] in, This can represent the integration period. In a 50Hz AC power frequency system, a complete power frequency cycle, i.e., 0.02 seconds, is generally taken as the time integration interval of the current waveform. It can represent The instantaneous current value at a given moment.

[0106] Effective value constraints It can reflect the overall work capacity of the current waveform. This constraint ensures that the generated data is consistent with the energy level of the actual load under steady-state operation, avoiding the generation of energy jumps or attenuations that do not conform to physical laws.

[0107] Fundamental amplitude constraint As shown in formula (6):

[0108] (6)

[0109] in, It can represent the effective value of the fundamental frequency and reflect the equivalent work capacity of the fundamental frequency component within the period.

[0110] Fundamental amplitude constraint This represents the main component of electrical energy transmission. This constraint can strictly limit the main energy distribution range of the waveform, ensuring that the core power characteristics of the generated signal are unlikely to shift.

[0111] Total Harmonic Distortion Constraint As shown in formula (7):

[0112] (7)

[0113] in, It can represent the harmonic order. It is a positive integer greater than or equal to 2. It can theoretically represent all higher harmonic components, and in practical applications, its value can be selected according to the operating conditions. It can represent the first The effective value of a subharmonic can reflect the equivalent work done by that subharmonic component within the period.

[0114] Total Harmonic Distortion Constraint It measures the overall degree of distortion of the waveform from an ideal sine wave. It can reflect the inherent distortion characteristics of the load and prevent the introduction of false broadband noise or unreal harmonic pollution during the generation process.

[0115] Third harmonic ratio constraint As shown in formula (8):

[0116] (8)

[0117] in, It can represent the effective value of the third harmonic, which can reflect the equivalent work capacity of the third harmonic component within the period.

[0118] Third harmonic ratio constraint This constraint can target the low-order frequency domain distortion characteristics of nonlinear electrical loads. By limiting the nonlinear mapping boundary of the model in the odd-order frequency band, it ensures that the generated harmonic structure conforms to the actual power grid characteristics.

[0119] Peak coefficient constraint As shown in formula (9):

[0120] (9)

[0121] in, It can represent the peak current. It can represent the effective value of the current and reflect the overall work capacity of the current waveform. It can be used as a normalization benchmark.

[0122] Peak coefficient constraint It can characterize the severity of waveform spikes. This constraint can suppress extreme spikes in the generated data that are detached from the physical characteristics of the load, preventing the model from generating spurious transient jumps during feature reconstruction.

[0123] Waveform coefficient constraints As shown in formula (10):

[0124] (10)

[0125] Waveform coefficient constraints It can describe the macroscopic contour and morphological distribution of current waveforms. By controlling the waveform coefficients, it can be ensured that the overall shape and structural richness of the amplified data in the time domain closely match the original sample.

[0126] According to embodiments of this application, the generated current waveform samples are filtered through the aforementioned constraints. High-quality data augmentation, achieving both high physical fidelity and statistical diversity, can be achieved with a very small number of real malfunctioning samples through these constraints.

[0127] According to embodiments of this application, the screening mechanism can make the generated current waveform conform to physical consistency as much as possible, which can prevent the generation model from introducing false waveforms that do not conform to the electrical characteristics of the real load due to random disturbances or network interpolation. This ensures that the data finally used for model training has both statistical diversity and strictly follows the original sample in terms of core electrical properties such as energy level, spectral structure, and waveform morphology, thereby ensuring the high physical fidelity and engineering usability of the amplified data.

[0128] Figure 5(a) schematically shows a comparison of the malfunction waveform of an incandescent lamp and a current waveform sample according to an embodiment of the present application.

[0129] Figure 5(b) schematically illustrates a comparison of the waveform of a malfunctioning electric drill and a current waveform sample according to an embodiment of the present application.

[0130] Figure 5(c) schematically illustrates a comparison of the power supply malfunction waveform and current waveform sample according to an embodiment of the present application.

[0131] Figure 5(d) schematically illustrates a comparison of the malfunction waveform and current waveform sample of a halogen lamp according to an embodiment of the present application.

[0132] Figure 5(e) schematically illustrates a comparison of the malfunction waveform and current waveform sample of a microwave oven according to an embodiment of the present application.

[0133] Figure 5(f) schematically illustrates a comparison of the malfunction waveform of a fluorescent lamp and a current waveform sample according to an embodiment of the present application.

[0134] Figure 5(g) schematically illustrates a comparison of the malfunction waveform of a resistive load and a current waveform sample according to an embodiment of the present application.

[0135] Figure 5(h) schematically shows a comparison of the malfunction waveform and current waveform sample of a vacuum cleaner according to an embodiment of the present application.

[0136] According to embodiments of this application, referring to Figures 5(a) to 5(h), waveform comparisons of incandescent lamps, electric drills, power supplies, halogen lamps, microwave ovens, fluorescent lamps, resistive loads, and vacuum cleaners are shown. The input waveform represents the malfunction waveforms of the aforementioned electrical loads, and the fitted waveform represents the current waveform samples of the aforementioned electrical loads. As can be seen from Figures 5(a) to 5(h), the fitted waveform is smoother than the original input waveform, with high-frequency glitches and random noise components effectively suppressed, while the macroscopic shape and key inflection points of the waveform are accurately preserved. It can be seen that the first model focuses on extracting low-frequency waveforms, stripping away high-frequency variations caused by nonlinear disturbances and measurement noise, thus ensuring that the fitting results inherit the physical commonalities of the loads while avoiding the risk of overfitting. Therefore, the basic fitted waveform output by the first model can provide reliable underlying support conforming to the laws of electrophysics for subsequent detail enhancement and cross-combination generation, ensuring that the amplified data does not deviate from the true operating essence of the load in terms of energy level and spectral structure.

[0137] Figure 6(a) schematically illustrates a parametric comparison feature radar diagram of malfunction waveforms and current waveform samples of a microwave oven, fluorescent lamp, resistive load, and vacuum cleaner according to an embodiment of the present application.

[0138] Figure 6(b) schematically illustrates a parametric comparison feature radar plot of malfunction waveforms and current waveform samples of incandescent lamps, electric drills, power supplies and halogen lamps according to embodiments of this application.

[0139] According to an embodiment of this application, the original waveforms in Figures 6(a) and 6(b) represent malfunction waveforms of incandescent lamps, electric drills, power supplies, halogen lamps, microwave ovens, fluorescent lamps, resistive loads, and vacuum cleaners, and the generated waveforms refer to current waveform samples of incandescent lamps, electric drills, power supplies, halogen lamps, microwave ovens, fluorescent lamps, resistive loads, and vacuum cleaners.

[0140] As shown in Figures 6(a) and 6(b), the parametric comparison radar charts of the eight loads reveal that the generated waveforms are consistent with the original waveforms across six core electrical characteristic dimensions. For microwave ovens, fluorescent lamps, and resistive loads, the effective value, fundamental amplitude, total harmonic distortion (THD), third harmonic ratio, peak value, and waveform coefficient of the generated data closely follow the distribution of the original samples, with relative deviations of each characteristic generally controlled within a very small range, and the radar chart outlines almost completely overlap. Although the generated data for vacuum cleaner loads shows slightly larger deviations in a few dimensions, it still remains within the statistical constraints of the original samples overall, and no significant distortion or out-of-bounds phenomena are observed in the six characteristics. These results demonstrate that the aforementioned current waveform sample generation method can achieve high-quality amplification while strictly ensuring the physical characteristics of the electrical signals, and the generated data maintains statistical consistency with the real load in core attributes such as energy level, spectral structure, and waveform morphology.

[0141] Further analysis reveals that different load types exhibit distinct characteristic distributions on the radar chart, and the generated data accurately reproduces the inherent characteristics of each load type. The six characteristic distributions of microwave ovens and resistive loads are relatively uniform and symmetrical, with the generated waveforms demonstrating particularly high accuracy in approximating these regular loads. Fluorescent lamps and vacuum cleaners, as nonlinear or strongly disturbed loads, exhibit specific distortion patterns in their original waveforms regarding harmonic distortion and peak characteristics. The generated data still captures the core variation patterns of these loads, without introducing false broadband noise or unrealistic harmonic contamination due to random disturbances. Therefore, the aforementioned six electrical characteristic constraints play an effective physical consistency barrier role in the posterior screening of the generated data, ensuring that the amplified samples possess statistical diversity distinct from the original waveforms while strictly adhering to the operational patterns of real loads in an electrophysical essence. This provides high-fidelity data support for the anti-maloperation training of the arc fault detection model.

[0142] Figure 7 A block diagram of a current waveform sample generation apparatus according to an embodiment of this application is shown schematically.

[0143] like Figure 4 As shown, the current waveform sample generation device 700 includes a determination module 710, a first fitting module 720, a second fitting module 730, and a superposition module 740.

[0144] The determination module 710 is used to determine the first model and the second model from a pre-trained model library based on the load type of the target electrical load.

[0145] The first fitting module 720 is used to perform feature fitting on the low-frequency part of the malfunction waveform corresponding to the target electrical load using the first model to obtain the basic fitting waveform. The malfunction waveform is the current waveform when the circuit breaker of the target electrical load malfunctions.

[0146] The second fitting module 730 is used to perform feature fitting on the mid-frequency and high-frequency components of the erroneous waveform using the second model to obtain the detailed fitted waveform.

[0147] The overlay module 740 is used to overlay and adjust the basic fitted waveform using the detailed fitted waveform to obtain a current waveform sample. The current waveform sample is used to simulate the current waveform generated when the circuit breaker of the target electrical load malfunctions.

[0148] Any one or more of the modules, submodules, units, and subunits according to the embodiments of this application, or at least part of the functions of any one or more of them, can be implemented in one module. Any one or more of the modules, submodules, units, and subunits according to the embodiments of this application can be implemented by dividing them into multiple modules. Any one or more of the modules, submodules, units, and subunits according to the embodiments of this application can be at least partially implemented as hardware circuits, such as field-programmable gate arrays (FPGAs), programmable logic arrays (PLAs), systems-on-a-chip, systems-on-a-substrate, systems-on-package, application-specific integrated circuits (ASICs), or implemented by hardware or firmware in any other reasonable manner by integrating or packaging circuits, or implemented in any one of software, hardware, and firmware, or in a suitable combination of any of these. Alternatively, one or more of the modules, submodules, units, and subunits according to the embodiments of this application can be at least partially implemented as computer program modules, which, when run, can perform corresponding functions.

[0149] For example, any plurality of the determining module 710, the first fitting module 720, the second fitting module 730, and the superposition module 740 can be combined into one module / unit / subunit, or any one of these modules / units / subunits can be split into multiple modules / units / subunits. Alternatively, at least part of the functionality of one or more of these modules / units / subunits can be combined with at least part of the functionality of other modules / units / subunits and implemented in one module / unit / subunit. According to embodiments of this application, at least one of the determining module 710, the first fitting module 720, the second fitting module 730, and the superposition module 740 can be at least partially implemented as hardware circuitry, 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-package, an application-specific integrated circuit (ASIC), or any other reasonable means of integrating or packaging the circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, at least one of the determining module 710, the first fitting module 720, the second fitting module 730, and the superposition module 740 may be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is run.

[0150] It should be noted that the data processing system part in the embodiments of this application corresponds to the data processing method part in the embodiments of this application. The specific description of the data processing system part is referred to in the data processing method part, and will not be repeated here.

[0151] Figure 8 A block diagram of an electronic device suitable for implementing the methods described above, according to an embodiment of this application, is illustrated schematically. Figure 8 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0152] like Figure 8 As shown, an electronic device 800 according to an embodiment of this application includes a processor 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage portion 808 into a random access memory (RAM) 803. The processor 801 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 801 may also include onboard memory for caching purposes. The processor 801 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of this application.

[0153] RAM 803 stores various programs and data required for the operation of electronic device 800. Processor 801, ROM 802, and RAM 803 are interconnected via bus 804. Processor 801 executes various operations of the method flow according to embodiments of this application by executing programs in ROM 802 and / or RAM 803. It should be noted that the programs may also be stored in one or more memories other than ROM 802 and RAM 803. Processor 801 may also execute various operations of the method flow according to embodiments of this application by executing programs stored in said one or more memories.

[0154] According to embodiments of this application, the electronic device 800 may further include an input / output (I / O) interface 805, which is also connected to a bus 804. The electronic device 800 may also include one or more of the following components connected to the input / output (I / O) interface 805: an input section 806 including a keyboard, mouse, etc.; an output section 807 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the input / output (I / O) interface 805 as needed. A removable medium 811, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 810 as needed so that computer programs read from it can be installed into the storage section 808 as needed.

[0155] According to embodiments of this application, the method flow according to embodiments of this application can be implemented as a computer software program. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable storage medium, the computer program containing program code for performing the method shown in the flowchart. In such embodiments, the computer program can be downloaded and installed from a network via communication section 809, and / or installed from removable medium 811. When the computer program is executed by processor 801, it performs the functions defined in the system of embodiments of this application. According to embodiments of this application, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0156] This application also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of this application.

[0157] According to embodiments of this application, the computer-readable storage medium can be a non-volatile computer-readable storage medium. Examples include, but are not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0158] For example, according to embodiments of this application, a computer-readable storage medium may include the ROM 802 and / or RAM 803 described above and / or one or more memories other than ROM 802 and RAM 803.

[0159] Embodiments of this application also include a computer program product comprising a computer program containing program code for performing the methods provided in the embodiments of this application. When the computer program product is run on an electronic device, the program code is used to enable the electronic device to implement the current waveform sample generation method provided in the embodiments of this application.

[0160] When the computer program is executed by the processor 801, it performs the functions defined in the system / apparatus of this application embodiment. According to the embodiments of this application, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0161] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 809, and / or installed from a removable medium 811. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0162] According to embodiments of this application, program code for executing the computer programs provided in the embodiments of this application can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0163] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions. Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations are not explicitly described in this application. In particular, without departing from the spirit and teachings of this application, the features described in the various embodiments of this application can be combined and / or combined in various ways. All such combinations and / or combinations fall within the scope of this application.

[0164] The embodiments of this application have been described above. However, these embodiments are merely illustrative and not intended to limit the scope of this application. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. This application does not depart from its scope, and those skilled in the art can make various substitutions and modifications, all of which should fall within the scope of this application.

Claims

1. A method for generating current waveform samples, characterized in that, include: Based on the load type of the target electrical load, determine the first and second models from the pre-trained model library; The first model is used to perform feature fitting on the low-frequency part of the malfunction waveform corresponding to the target electrical load to obtain the basic fitted waveform, wherein the malfunction waveform is the current waveform when the circuit breaker of the target electrical load malfunctions. The mid-frequency and high-frequency components of the erroneous waveform are fitted using the second model to obtain the detailed fitted waveform. The basic fitted waveform is superimposed and adjusted using detailed fitted waveforms to obtain current waveform samples, which are used to simulate the current waveform generated when the circuit breaker of the target electrical load malfunctions.

2. The current waveform sample generation method according to claim 1, characterized in that, The first model is used to perform feature fitting on the low-frequency portion of the malfunction waveform corresponding to the target electrical load to obtain a basic fitted waveform, including: Harmonic decomposition of the erroneous waveform is performed using the Fourier basis functions in the first model to obtain multiple harmonic components of different frequencies. The basic features are obtained by superimposing the features of multiple harmonic components of different frequencies based on the aggregation operation. Based on the aforementioned basic characteristics, the waveform of the erroneous action is approximated and fitted to obtain a basic fitted waveform.

3. The current waveform sample generation method according to claim 1, characterized in that, The step of performing feature fitting on the mid-frequency and high-frequency components of the erroneous waveform using the second model to obtain a detailed fitted waveform includes: Based on the difference between the erroneous waveform and the basic fitted waveform, a residual waveform representing the mid-frequency and high-frequency characteristics of the erroneous waveform is determined. The residual waveform is input into the second model for waveform approximation fitting to obtain the detailed fitted waveform corresponding to the erroneous waveform.

4. The current waveform sample generation method according to claim 3, characterized in that, The step of inputting the residual waveform into the second model for waveform approximation fitting to obtain the detailed fitted waveform corresponding to the erroneous waveform includes: The erroneous waveform is decomposed using the Fourier basis functions in the second model to obtain multiple harmonic components corresponding to the intermediate and high frequency components. By superimposing features from multiple harmonic components, detailed features can be obtained. Based on the detailed features, the residual waveform is approximated to obtain the detailed fitted waveform.

5. The current waveform sample generation method according to claim 1, characterized in that, The process of superimposing and adjusting the basic fitted waveform using detailed fitted waveforms to obtain current waveform samples includes: The perturbation waveform is obtained by enhancing the details of the fitted waveform using perturbation coefficients. The mid-frequency and high-frequency portions of the basic fitted waveform are superimposed using a perturbation waveform to obtain a current waveform sample.

6. The current waveform sample generation method according to claim 1, characterized in that, The step of determining the first model and the second model from a pre-trained model library based on the load type of the target electrical load includes: Based on the load type of the target electrical load, a first model set and a second model set corresponding to the load type are determined from the pre-trained model library; The first model and the second model are randomly selected from the first model set and the second model set, respectively; The first model set includes multiple first models, and the second model set includes multiple second models. The multiple first models are trained based on the current waveforms at multiple times after the electrical load malfunctions, and the multiple second models are trained based on the high-frequency harmonics of the current waveforms at multiple times after the electrical load malfunctions.

7. The current waveform sample generation method according to claim 1, characterized in that, Also includes: Based on the malfunction waveform of the target electrical load, determine multiple waveform characteristic constraints of the target electrical load; Based on multiple waveform characteristic constraints, waveforms are screened from multiple current waveform samples to determine current waveform samples that meet the waveform characteristic constraints.

8. A current waveform sample generation device, characterized in that, include: The determination module is used to determine the first model and the second model from a pre-trained model library based on the load type of the target electrical load. The first fitting module is used to perform feature fitting on the low-frequency part of the malfunction waveform corresponding to the target electrical load using the first model to obtain a basic fitted waveform, wherein the malfunction waveform is the current waveform when the circuit breaker of the target electrical load malfunctions. The second fitting module is used to perform feature fitting on the mid-frequency and high-frequency components of the erroneous waveform using the second model to obtain a detailed fitted waveform. as well as The overlay module is used to overlay and adjust the basic fitted waveform using the detailed fitted waveform to obtain a current waveform sample. The current waveform sample is used to simulate the current waveform generated when the circuit breaker of the target electrical load malfunctions.

9. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs. Wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the method of any one of claims 1 to 7.

10. A computer-readable storage medium having executable instructions stored thereon, characterized in that, When the instructions are executed by the processor, the processor performs the method according to any one of claims 1 to 7.

Images