An arc detection method, device and readable storage medium

Abstract

The application discloses an arc detection method and device and a readable storage medium, and relates to the field of electronic circuits. The arc detection method provided by the application introduces the concept of a training model. After an arc input signal is obtained, time-frequency domain features of the input signal are extracted, convolution features are obtained through a convolution model, and the convolution model and the time-frequency domain features are spliced into a synthetic vector. The synthetic vector is used to represent the fault probability of the arc input signal, thereby realizing detection of the arc. Compared with the traditional method of judging a threshold according to a field environment by manual operation, the application introduces the concept of the convolution model, that is, the complex environmental factors traditionally judged by technical personnel are replaced by the convolution features corresponding to the convolution model. Therefore, only the arc signal needs to be input, and the final synthetic vector can be obtained, thereby obtaining the fault probability of the arc. Therefore, the experience of technical personnel is not needed, and the application has high applicability while reducing the labor cost.

Description

Technical Field

[0001] This application relates to the field of arc detection technology, and in particular to an arc detection method, apparatus and readable storage medium. Background Technology

[0002] In recent years, with the advancement of power technology, especially the popularization of photovoltaic power generation technology, related technologies have also developed accordingly. While photovoltaic power generation technology brings convenience to people, it also poses certain dangers. Since the discharge phenomenon of photovoltaic power generation technology is much more common than that of other technologies, accurate and rapid detection and identification of the electric arc generated during the discharge of photovoltaic power generation systems has become an important research topic for the protection of photovoltaic power generation systems and is of great significance.

[0003] Existing arc detection methods mainly rely on the time-domain and / or frequency-domain characteristics of arc current and voltage for analysis and judgment. This usually involves adjusting different threshold parameters according to the environment, temperature, noise, etc. of the photovoltaic installation site. These methods depend on the experience of technicians, making them inconvenient to use and highly unstable. If the above traditional methods are used, not only is the labor cost high, but the adaptability is also poor.

[0004] Given the above-mentioned technologies, finding a more applicable arc detection method is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0005] The purpose of this application is to provide an arc detection method to address the problems of the current arc detection methods' limited applicability, the need for experienced personnel to perform the detection when environmental factors are considered, and the resulting high labor costs.

[0006] To solve the above-mentioned technical problems, this application provides an arc detection method, comprising: Acquire the electric arc input signal and extract the time-frequency domain features from the electric arc input signal; The electric arc input signal is used to generate convolutional features through a convolutional model, which is a computational model generated in advance using historical electric arc input signals and corresponding convolutional results as training data. The time-frequency domain features and the convolutional features are concatenated into a composite vector, which is used to characterize the failure probability of the arc corresponding to the arc input signal.

[0007] Preferably, the extraction of time-frequency domain features from the electric arc input signal includes: The arc input signal is subjected to wavelet transform and 2048-point Fourier transform, and the time-frequency domain features are obtained through the following calculation formula. , , , ; Among them, W k Here, is the wavelet coefficient, a is the mean of the wavelet coefficients, and F is the mean of the wavelet coefficients j The parameters are the Fourier coefficients with frequencies between 2Hz and 90kHz, arranged in descending order of amplitude, where b is the parameter after removing F. j The sum of the first 50 maximum values ​​and the last 20 minimum values, taken as the base-10 logarithm of the Fourier coefficients, is then calculated. n To calculate the difference between the previous and subsequent points for the arc input signal, n = 1, 2, ..., 2047, c is u n The difference between the maximum and minimum values, d is u n The sum of the absolute values; The parameters a, b, c, and d are used as the time-frequency domain features of the electric arc input signal.

[0008] Preferably, the establishment of the convolution model includes the following steps: The training data is calculated using the cross-entropy loss function and the batch gradient descent algorithm; The calculation results are optimized using an optimizer; Using a cosine-decreasing learning rate adjustment strategy, the optimized calculation results are updated through multiple backpropagations to obtain the convolution result, and the convolution model is established.

[0009] Preferably, the optimizer is the Adam optimizer.

[0010] Preferably, after concatenating the time-frequency domain features with the convolutional features into a synthesized vector, the method further includes: Determine whether the synthesized vector meets the preset criteria; If so, then the arc is determined to be a normal arc; If not, then the arc is determined to be a fault arc.

[0011] Preferably, it further includes: Every preset time interval, the preset standard is updated using historical arc input signals and corresponding detection results.

[0012] Preferably, after determining that the arc is a fault arc, the method further includes: Control the alarm device to sound an alarm.

[0013] To address the above problems, this application also provides an arc detection device, comprising: An acquisition module is used to acquire an electric arc input signal and extract time-frequency domain features from the electric arc input signal; The convolution module is used to generate convolutional features from the electric arc input signal through a convolution model, wherein the convolution model is a model generated in advance using historical electric arc input signals and corresponding convolution results; The splicing module is used to splice the time-frequency domain features and the convolutional features into a composite vector, which is used to characterize the fault probability of the arc corresponding to the arc input signal.

[0014] Preferably, the device further includes: The calculation module is used to perform calculations on the training data using the cross-entropy loss function and the batch gradient descent algorithm; The optimization module is used to optimize the calculation results using an optimizer; A module is established to use a cosine-decreasing learning rate adjustment strategy to update the optimized calculation results through multiple backpropagations to obtain the convolution result and establish the convolution model.

[0015] Preferably, the device further includes: The judgment module is used to determine whether the synthesized vector meets the preset standard. If yes, the first confirmation module is activated; if no, the second confirmation module is activated. The first confirmation module is used to determine whether the electric arc is a normal electric arc. The second confirmation module is used to determine that the arc is a fault arc.

[0016] Preferably, the device further includes: The update module is used to update the preset standard at preset time intervals using historical arc input signals and corresponding detection results.

[0017] Preferably, the device further includes: The alarm module is used to control the alarm device to trigger an alarm.

[0018] To address the aforementioned problems, this application also provides another arc detection device, including a memory for storing a computer program; A processor is used to implement the steps of the arc detection method described above when executing the computer program.

[0019] To address the aforementioned problems, this application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the arc detection method described above.

[0020] The arc detection method provided in this application introduces the concept of a training model. After acquiring the arc input signal, it extracts the time-frequency domain features of the input signal and obtains convolutional features through a convolutional model. The convolutional model and the time-frequency domain features are then concatenated into a synthetic vector, which is used to characterize the fault probability of the arc input signal, thereby realizing the detection of the arc. Compared with the traditional method of judging thresholds manually based on the on-site environment, this method introduces the concept of a convolutional model. The complex environmental factors that are traditionally judged by technicians are replaced by the convolutional features corresponding to the convolutional model. Therefore, only the arc signal needs to be input to obtain the final synthetic vector, thereby obtaining the fault probability of the arc. Thus, no experience of technicians is required, which reduces labor costs and has high applicability.

[0021] The arc detection device and computer-readable storage medium provided in this application correspond to the arc detection method described above, and have the same beneficial effects. Attached Figure Description

[0022] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 A flowchart of an arc detection method provided in an embodiment of this application; Figure 2 This is a schematic diagram of an arc detection device provided in an embodiment of this application; Figure 3 This is a structural diagram of an arc detection device provided in another embodiment of this application. Detailed Implementation

[0024] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0025] The core of this application is to provide an arc detection method to address the problems of the current arc detection methods' limited applicability, the need for experienced personnel to perform the detection when environmental factors are considered, and the high labor costs.

[0026] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0027] Figure 1 A flowchart of an arc detection method provided in this application embodiment is shown below. Figure 1 As shown, the method includes: S10: Acquire the electric arc input signal and extract the time-frequency domain features from the electric arc input signal; An electric arc is an electrical discharge phenomenon, a momentary spark generated when the insulating medium (such as air) between two electrodes breaks down. Studies have shown that this phenomenon occurs when the voltage at the breaking point exceeds 20V and the current is greater than 80-100mA. Many other factors can also generate electric arcs, such as aging and breakage of photovoltaic modules due to long-term operation, poor contact at the module connection points, moisture, corrosion, and insulation damage to wires. Once a DC arc is generated, it is very difficult to extinguish, and the arc temperature can reach over 3000℃. This not only causes the surrounding insulating materials to decompose or carbonize, losing their insulating effect, but also easily leads to spontaneous combustion of nearby materials, seriously threatening personal and property safety.

[0028] It should be noted that DC arc detection methods require a large amount of normal and arc data under different operating conditions for data mining and feature extraction. Since it is difficult to obtain DC arc data in actual power plants, many methods use laboratory simulation environments to acquire this data. This acquisition is done by building physical circuits containing arc generating fixtures and DC signal sources. However, laboratory simulation environments struggle to simulate the DC arc signals in actual power plants, which are subject to various natural interferences. When a DC arc detection method well-adapted in a laboratory simulation environment is directly applied to an actual power plant, significant false negatives and false alarms occur. Therefore, it is necessary to optimize the detection results using DC arc data generated by actual power plants. In this application, the specific method for acquiring the arc input signal is not limited; it can be understood that the arc input signal is acquired through circuits related to the arc.

[0029] The time domain describes the relationship between a mathematical function or a physical signal and time. For example, the time-domain waveform of a signal can express how the signal changes over time. If discrete time is considered, the value of the function or signal at each discrete time point is known. If continuous time is considered, the value of the function or signal at any time is known.

[0030] The frequency domain refers to a domain where the independent variable is frequency, i.e., the horizontal axis represents frequency and the vertical axis represents the amplitude of the signal at that frequency; this is commonly known as a spectrum diagram. A spectrum diagram describes the frequency structure of a signal and the relationship between frequency and the amplitude of the signal at that frequency. When performing time-domain analysis on signals, sometimes signals have the same time-domain parameters, but this does not mean that the signals are completely identical. This is because signals not only change with time but are also related to information such as frequency and phase. Therefore, it is necessary to further analyze the frequency structure of the signal and describe it in the frequency domain. The transformation of dynamic signals from the time domain to the frequency domain is mainly achieved through Fourier series and Fourier transforms. Periodic signals rely on Fourier series, while aperiodic signals rely on Fourier transforms.

[0031] The time-frequency domain features mentioned in this embodiment refer to the time-domain features and corresponding frequency-domain features of the arc input signal. In this embodiment, the specific time-frequency domain features of the arc input signal are not limited. For example, it can be the time axis of voltage change in the circuit corresponding to the arc, or the frequency change of the electric waves in the arc, etc.

[0032] S11: Generate convolutional features from the electric arc input signal through a convolutional model. The convolutional model is a computational model that is generated in advance using historical electric arc input signals and corresponding convolutional results as training data. It should be noted that in functional analysis, convolution, spin, or folding is a mathematical operator that generates a third function from two functions f and g, representing the area of ​​the overlapping region of functions f and g after flipping and translation. If one of the functions participating in the convolution is regarded as an indicator function of the interval, convolution can also be seen as a generalization of "moving average".

[0033] It is understood that the convolution application in this embodiment is to represent the application environment of the relevant arc signal as a function. In other words, the previous judgment of complex application scenarios is replaced by the trained convolution model, thereby generating convolution features to indicate the application environment of the arc input signal. In this embodiment, there are no limitations on the convolution model and its specific establishment method.

[0034] S12: The time-frequency domain features and convolutional features are concatenated into a composite vector, which is used to characterize the failure probability of the arc corresponding to the arc input signal.

[0035] It should be noted that the splicing mentioned in this embodiment refers to the synthesis vector, which involves modeling the data in the time-frequency domain features and the corresponding convolutional features, and using the data of the two features as the horizontal and vertical axis vectors respectively. After the splicing is completed, the fault probability of the electric arc can be obtained based on the previous historical data. That is, the fault probability corresponding to different synthesis vectors can be obtained through the previous test data. In this embodiment, there are no specific limitations on the specific type of synthesis vector and the corresponding splicing method.

[0036] The arc detection method provided in this application introduces the concept of a training model. After acquiring the arc input signal, it extracts the time-frequency domain features of the input signal and obtains convolutional features through a convolutional model. The convolutional model and the time-frequency domain features are then concatenated into a synthetic vector, which is used to characterize the fault probability of the arc input signal, thereby realizing the detection of the arc. Compared with the traditional method of judging thresholds manually based on the on-site environment, this method introduces the concept of a convolutional model. The complex environmental factors that are traditionally judged by technicians are replaced by the convolutional features corresponding to the convolutional model. Therefore, only the arc signal needs to be input to obtain the final synthetic vector, thereby obtaining the fault probability of the arc. Thus, no experience of technicians is required, which reduces labor costs and has high applicability.

[0037] The above embodiments do not limit the specific method for obtaining the time-frequency domain features of the arc input signal. Here, a preferred scheme is proposed, in which the extraction of time-frequency domain features from the arc input signal includes: Wavelet transform and 2048-point Fourier transform were performed on the electric arc input signal, and the time-frequency domain features were obtained using the following calculation formula. , , , ; Among them, W k Here, is the wavelet coefficient, a is the mean of the wavelet coefficients, and F is the mean of the wavelet coefficients j The parameters are Fourier coefficients with frequencies between 2Hz and 90kHz, arranged in descending order of amplitude. 'b' represents the parameter after removing the F... j The sum of the first 50 maximum values ​​and the last 20 minimum values ​​is obtained by taking the base-10 logarithm of the Fourier coefficients. n Let c be the difference between the previous and subsequent points of the arc input signal, where n = 1, 2, ..., 2047, and c is u. n The difference between the maximum and minimum values, d is u n The sum of the absolute values; The parameters a, b, c, and d are used as the time-frequency domain characteristics of the electric arc input signal.

[0038] It should be noted that performing wavelet transform on a given signal involves expanding the signal according to a family of wavelet functions, representing the signal as a linear combination of wavelet functions at different scales and time shifts. The coefficients of each term are called wavelet coefficients, and the linear combination of all wavelet functions at the same scale and different time shifts is called the wavelet component of the signal at that scale. Fourier transform can represent a function satisfying certain conditions as a linear combination of trigonometric functions (sine and / or cosine functions) or their integrals. In different research fields, Fourier transform has various variants, such as continuous Fourier transform and discrete Fourier transform. Initially, Fourier analysis was proposed as a tool for analytical analysis of thermal processes. In this embodiment, the Fourier coefficients and the specific parameters after transforming the arc input signal are presented. This embodiment proposes a specific method for obtaining time-frequency domain characteristics. Through this method, relatively complete time-frequency domain data of the arc input signal can be obtained, and the obtained parameters can more completely reflect the time and frequency domain information of the entire arc input signal, thereby increasing the accuracy of detection.

[0039] The above embodiments do not specify the method for establishing the convolutional model. A preferred solution is provided here, which includes the following steps: The training data is calculated using the cross-entropy loss function and the batch gradient descent algorithm; The calculation results are optimized using an optimizer; Using a cosine-decreasing learning rate adjustment strategy, the optimized calculation results are updated through multiple backpropagations to obtain the convolution result, and a convolution model is established.

[0040] Cross-entropy, a key concept in information theory, is used to measure the difference between two probability distributions. Batch gradient descent is its most basic form, where the gradient is updated using all samples when updating each parameter. It is obtained by recursively applying the convolution expression to m training data and then differentiating the cross-entropy loss function.

[0041] In this embodiment, the training data is not specifically limited. It can be understood that the training data refers to the previous arc input signal and the corresponding calculation results after being processed by the model. The type and number of training data are not limited in this embodiment, nor is the optimizer used for algorithm optimization. The cosine reduction in this embodiment utilizes the calculation method of cosine function reduction to accelerate model building. Simultaneously, the corresponding convolutional results are obtained after updating through backpropagation. The specific model building constraints proposed in this embodiment are based on the network layers in the convolutional model. That is, the layers of the convolutional model in this embodiment differ from other existing models. Because of the adoption of the above model, a relatively lightweight and fast convolutional model can be built.

[0042] In the above embodiments, the type of optimizer used in building the convolution model is not limited. Here, a preferred option is proposed: the optimizer is the Adam optimizer.

[0043] The Adaptive Moment Estimation (Adam) optimizer addresses the machine learning task of optimizing parameters to achieve optimal values, which in turn minimizes the loss function. The loss function is the difference between the objective function and the true value, essentially the function of the parameters to be optimized. The optimizer's task is to calculate the gradient of the loss function in each epoch and update the parameters accordingly. Therefore, the Adam optimizer allows for different learning rates to be adjusted for different parameters, updating frequently changing parameters with smaller step sizes and sparse parameters with larger step sizes.

[0044] Considering that the arc input signal needs to be judged based on the result of the synthesized vector, a preferred scheme is provided here. After concatenating the time-frequency domain features and convolutional features into a synthesized vector, the following steps are also included: Determine whether the synthesized vector meets the preset criteria; If so, then the arc is determined to be a normal arc; If not, then the arc is determined to be a fault arc.

[0045] It should be noted that the preset standard in this embodiment is the standard for judging whether the arc is normal in the training model generated by the pre-set arc input signal and its corresponding fault result. In this solution, it can be understood as judging the composite vector based on the coordinate axis of the composite vector, where one part is arc fault and the other part is arc normal. In this embodiment, the preset standard is not specifically limited, and the addition of the judgment step clarifies whether the arc is normal, which is beneficial to the subsequent handling by the staff.

[0046] Considering the learning capabilities of the electric arc system and the model, a preferred approach is proposed, which further includes: At preset intervals, the preset standards are updated using historical arc input signals and corresponding detection results.

[0047] It should be noted that the historical arc input signal in this embodiment refers to all arc input signals collected in the model and judged by the model. The corresponding detection results can be the results judged based on the above method. However, it is understood that since the accuracy of the method provided in this solution is not 100%, after the staff conducts a specific investigation, they can judge whether the arc is faulty on site and modify the corresponding detection results accordingly, based on the above arc input signal and the corresponding detection results.

[0048] Because arc faults pose significant safety hazards, a preferred solution is provided here, which, after confirming that the arc is a faulty arc, also includes: Control the alarm device to sound an alarm.

[0049] It should be noted that the specific type of alarm device is not limited in this embodiment. It can be a buzzer or a signal light, or it can send a signal directly to the relevant terminal to notify the staff to handle it in a timely manner.

[0050] In the above embodiments, the arc detection method has been described in detail. This application also provides embodiments corresponding to the arc detection device. It should be noted that this application describes the embodiments of the device from two perspectives: one is based on the functional modules, and the other is based on the hardware.

[0051] Figure 2 This is a schematic diagram of an arc detection device provided in an embodiment of this application, as shown below. Figure 2 The device shown includes: The acquisition module 10 is used to acquire the electric arc input signal and extract the time-frequency domain features from the electric arc input signal; Convolution module 11 is used to generate convolution features from the electric arc input signal through a convolution model. The convolution model is a model generated in advance using historical electric arc input signals and corresponding convolution results. The splicing module 12 is used to splice the time-frequency domain features and convolutional features into a synthetic vector, which is used to characterize the fault probability of the arc corresponding to the arc input signal.

[0052] Preferably, the device further includes: The computation module is used to perform calculations on the training data using the cross-entropy loss function and the batch gradient descent algorithm. The optimization module is used to optimize the calculation results using an optimizer; A module is established to use a cosine-decreasing learning rate adjustment strategy to obtain the convolution result after multiple backpropagations of the optimized calculation result, and to build a convolution model.

[0053] Preferably, the device further includes: The judgment module is used to determine whether the synthesized vector meets the preset standard. If yes, the first confirmation module is activated; otherwise, the second confirmation module is activated. The first confirmation module is used to determine whether the electric arc is a normal electric arc. The second confirmation module is used to determine that the arc is a faulty arc.

[0054] Preferably, the device further includes: The update module is used to update the preset standard at preset intervals using historical arc input signals and corresponding detection results.

[0055] Preferably, the device further includes: The alarm module is used to control the alarm device to trigger an alarm.

[0056] Since the embodiments of the apparatus and the embodiments of the method correspond to each other, please refer to the description of the embodiments of the method for the embodiments of the apparatus and their corresponding beneficial effects, which will not be repeated here.

[0057] Figure 3 This is a structural diagram of an arc detection device provided in another embodiment of this application, as shown below. Figure 3 As shown, the arc detection device includes: a memory 20 for storing computer programs; The processor 21 is used to execute a computer program to implement the steps of the arc detection method mentioned in the above embodiments.

[0058] The arc detection device provided in this embodiment may include, but is not limited to, smartphones, tablets, laptops, or desktop computers.

[0059] The processor 21 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 21 may be implemented using at least one hardware form selected from Digital Signal Processor (DSP), Field-Programmable Gate Array (FPGA), and Programmable Logic Array (PLA). The processor 21 may also include a main processor and a coprocessor. The main processor, also known as the Central Processing Unit (CPU), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 21 may integrate a Graphics Processing Unit (GPU), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 21 may also include an Artificial Intelligence (AI) processor, which handles computational operations related to machine learning.

[0060] The memory 20 may include one or more computer-readable storage media, which may be non-transitory. The memory 20 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In this embodiment, the memory 20 is used to store at least the following computer program 201, which, after being loaded and executed by the processor 21, is capable of implementing the relevant steps of the arc detection method disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 20 may also include an operating system 202 and data 203, and the storage method may be temporary or permanent storage. The operating system 202 may include Windows, Unix, Linux, etc. The data 203 may include, but is not limited to, the data involved in the above-described arc detection method.

[0061] In some embodiments, the arc detection device may further include a display screen 22, an input / output interface 23, a communication interface 24, a power supply 25, and a communication bus 26.

[0062] Those skilled in the art will understand that Figure 3 The structure shown does not constitute a limitation on the arc detection device and may include more or fewer components than illustrated.

[0063] The arc detection device provided in this application includes a memory and a processor. When the processor executes the program stored in the memory, it can implement the following method: the arc detection method involved in the above embodiments.

[0064] Since the embodiments of the apparatus and the embodiments of the method correspond to each other, please refer to the description of the embodiments of the method for the embodiments of the apparatus and their corresponding beneficial effects, which will not be repeated here.

[0065] Finally, this application also provides an embodiment corresponding to a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps described in the above method embodiments.

[0066] It is understood that if the methods in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and executes all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0067] Since the embodiments of the readable storage medium portion correspond to the embodiments of the method portion, the embodiments of the apparatus portion and their corresponding beneficial effects are described in the description of the embodiments of the method portion, and will not be repeated here.

[0068] The arc detection device, method, and readable storage medium provided in this application have been described in detail above. The various embodiments in the specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the device disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section. It should be noted that those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of the claims of this application.

[0069] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. An arc detection method, characterized in that, include: Acquire the electric arc input signal and extract the time-frequency domain features from the electric arc input signal; The electric arc input signal is used to generate convolutional features through a convolutional model, which is a computational model generated in advance using historical electric arc input signals and corresponding convolutional results as training data. The time-frequency domain features and the convolutional features are concatenated to form a composite vector, which is used to characterize the fault probability of the arc corresponding to the arc input signal; wherein: The extraction of time-frequency domain features from the electric arc input signal includes: The arc input signal is subjected to wavelet transform and 2048-point Fourier transform, and the time-frequency domain features are obtained through the following calculation formula. , , , ; Among them, W k Here, is the wavelet coefficient, a is the mean of the wavelet coefficients, and F is the mean of the wavelet coefficients j The parameters are the Fourier coefficients with frequencies between 2Hz and 90kHz, arranged in descending order of amplitude, where b is the parameter after removing F. j The sum of the first 50 maximum values ​​and the last 20 minimum values, taken as the base-10 logarithm of the Fourier coefficients, is then calculated. n To calculate the difference between the previous and subsequent points for the arc input signal, n = 1, 2, ..., 2047, c is u n The difference between the maximum and minimum values, d is u n The sum of the absolute values; The parameters a, b, c, and d are used as the time-frequency domain features of the electric arc input signal; The establishment of the convolutional model includes the following steps: The training data is calculated using the cross-entropy loss function and the batch gradient descent algorithm; The calculation results are optimized using an optimizer; Using a cosine-decreasing learning rate adjustment strategy, the optimized calculation results are updated through multiple backpropagations to obtain the convolution result, and the convolution model is established.

2. The arc detection method according to claim 1, characterized in that, The optimizer is the Adam optimizer.

3. The arc detection method according to any one of claims 1 to 2, characterized in that, After concatenating the time-frequency domain features with the convolutional features into a synthesized vector, the method further includes: Determine whether the synthesized vector meets the preset criteria; If so, then the arc is determined to be a normal arc; If not, then the arc is determined to be a fault arc.

4. The arc detection method according to claim 3, characterized in that, Also includes: Every preset time interval, the preset standard is updated using historical arc input signals and corresponding detection results.

5. The arc detection method according to claim 4, characterized in that, After determining that the arc is a fault arc, the process further includes: Control the alarm device to sound an alarm.

6. An arc detection device, characterized in that, include: The acquisition module is used to acquire the electric arc input signal and extract the time-frequency domain features from the electric arc input signal; The convolution module is used to generate convolutional features from the electric arc input signal through a convolution model. The convolution model is a model that is generated in advance using historical electric arc input signals and corresponding convolution results as training data. The concatenation module is used to concatenate the time-frequency domain features and the convolutional features into a composite vector, which is used to characterize the fault probability of the arc corresponding to the arc input signal; wherein: The acquisition module is used for: The arc input signal is subjected to wavelet transform and 2048-point Fourier transform, and the time-frequency domain features are obtained through the following calculation formula. , , , ; Among them, W k Here, is the wavelet coefficient, a is the mean of the wavelet coefficients, and F is the mean of the wavelet coefficients j The parameters are the Fourier coefficients with frequencies between 2Hz and 90kHz, arranged in descending order of amplitude, where b is the parameter after removing F. j The sum of the first 50 maximum values ​​and the last 20 minimum values, taken as the base-10 logarithm of the Fourier coefficients, is then calculated. n To calculate the difference between the previous and subsequent points for the arc input signal, n = 1, 2, ..., 2047, c is u n The difference between the maximum and minimum values, d is u n The sum of the absolute values; The parameters a, b, c, and d are used as the time-frequency domain features of the electric arc input signal. Also includes: The calculation module is used to perform calculations on the training data using the cross-entropy loss function and the batch gradient descent algorithm; The optimization module is used to optimize the calculation results using an optimizer; A module is established to use a cosine-decreasing learning rate adjustment strategy to update the optimized calculation results through multiple backpropagations to obtain the convolution result and establish the convolution model.

7. An arc detection device, characterized in that, Includes memory used to store computer programs; A processor, configured to implement the steps of the arc detection method as described in any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the arc detection method as described in any one of claims 1 to 5.

Images