Arc fault detection method and device, electronic equipment and storage medium
The arc fault detection is generated by Markov conversion field and deep convolutional neural network model, which solves the problem of low arc fault detection accuracy and efficiency in traditional methods, and improves the accuracy of series fault detection.
Patent Information
- Application Number
- CN202410987431.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-23
- Publication Date
- 2024-11-08
AI Technical Summary
Traditional arc fault detection methods are difficult to accurately detect arc faults, especially in series faults, resulting in low detection accuracy and efficiency.
By obtaining one-dimensional current data and using Markov conversion field for image conversion, a two-dimensional current image is generated, and feature extraction is combined with a deep convolutional neural network (such as ResNet) model to improve the accuracy of arc fault detection.
It improves the accuracy and efficiency of arc fault detection, especially in series fault conditions, and enhances the ability and robustness of arc fault recognition.
Smart Images

Figure CN118914772A_ABST
Abstract
Description
Technical Field
[0001] Embodiments of the present invention relate to the technical field of electrical fault detection, and in particular to an arc fault detection method, device, electronic equipment and storage medium. Background Art
[0002] In traditional power systems, arc faults are very dangerous faults that may cause fire or damage to electrical equipment, or even personal injury or death. Therefore, arc fault detection is crucial to ensure the safe operation of power systems.
[0003] Traditional arc fault detection methods usually directly use manually set features and simple threshold judgments after obtaining time-series current data. However, these methods are difficult to accurately detect arc faults and need to be urgently solved. Summary of the invention
[0004] The embodiments of the present invention provide an arc fault detection method, device, electronic equipment and storage medium, which improve the accuracy of arc fault detection.
[0005] According to one aspect of the present invention, there is provided an arc fault detection method, which may include: acquiring an arc fault detection model, and performing current acquisition on an acquisition circuit within an acquisition time period to obtain one-dimensional current data; using a Markov transition field to perform image conversion on the one-dimensional current data to obtain a two-dimensional current image; and determining an arc fault detection result of the acquisition circuit within the acquisition time period based on the two-dimensional current image and the arc fault detection model.
[0006] According to another aspect of the present invention, an arc fault detection device is provided, which may include: a one-dimensional current data acquisition module, used to obtain an arc fault detection model, and perform current acquisition on an acquisition circuit within an acquisition time period to obtain one-dimensional current data; a two-dimensional current image acquisition module, used to perform image conversion on the one-dimensional current data using a Markov transition field to obtain a two-dimensional current image; and an arc fault detection result determination module, used to determine the arc fault detection result of the acquisition circuit within the acquisition time period based on the two-dimensional current image and the arc fault detection model.
[0007] According to another aspect of the present invention, there is provided an electronic device, which may include:
[0008] at least one processor; and
[0009] a memory communicatively connected to at least one processor; wherein,
[0010] The memory stores a computer program that can be executed by at least one processor. The computer program is executed by at least one processor so that the arc fault detection method provided by any embodiment of the present invention is implemented when the at least one processor executes the computer program.
[0011] According to another aspect of the present invention, a computer-readable storage medium is provided, on which computer instructions are stored. The computer instructions are used to enable a processor to implement the arc fault detection method provided by any embodiment of the present invention when executed.
[0012] The technical solution of the embodiment of the present invention is to obtain an arc fault detection model, and collect current for the collection circuit within the collection time period to obtain one-dimensional current data, so as to perform arc fault detection through the one-dimensional current data; use the Markov transformation field to perform image conversion on the one-dimensional current data to obtain a two-dimensional current image, so as to obtain a two-dimensional current image that can better characterize the law and information of the one-dimensional current data through the Markov transformation field, which can be subsequently processed by the arc fault detection model that can perform image feature extraction; according to the two-dimensional current image and the arc fault detection model, the arc fault detection result of the collection circuit within the collection time period is determined, so as to determine the arc fault detection with higher accuracy through the two-dimensional current image and the arc fault detection model. The above technical solution converts the one-dimensional current data into a two-dimensional current image by using the Markov transformation field, so that the arc fault detection model that can perform image feature extraction can process the two-dimensional current image that better characterizes the law and information of the one-dimensional current data, determine the arc fault detection with higher accuracy, and thus improve the accuracy of arc fault detection.
[0013] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0015] Figure 1 is a flow chart of an arc fault detection method provided according to an embodiment of the present invention;
[0016] Figure 2 is an example diagram of the structure of an arc fault detection model in an arc fault detection method provided according to an embodiment of the present invention;
[0017] Figure 3 is an example diagram of a series fault arc current waveform in an arc fault detection method provided by an embodiment of the present invention;
[0018] Figure 4 is a flow chart of another arc fault detection method provided according to an embodiment of the present invention;
[0019] Figure 5 is an example diagram of a high-pass filter in another arc fault detection method provided by an embodiment of the present invention;
[0020] Figure 6 is a flow chart of another arc fault detection method provided according to an embodiment of the present invention;
[0021] Figure 7 is a flow chart of obtaining an arc fault detection model through training in another arc fault detection method provided in an embodiment of the present invention;
[0022] Figure 8 is a flow chart of an optional example of another arc fault detection method provided according to an embodiment of the present invention;
[0023] Fig. 9 is a structural block diagram of an arc fault detection device provided according to an embodiment of the present invention;
[0024] Fig.10 It is a structural schematic diagram of an electronic device for implementing the arc fault detection method according to an embodiment of the present invention. DETAILED DESCRIPTION
[0025] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.
[0026] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. The situations of "target", "original", etc. are similar and will not be repeated here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0027] Before introducing the embodiments of the present invention, the implementation process of the scheme currently adopted for arc fault detection and the reasons why it is difficult to accurately detect arc faults are exemplified to better understand why the scheme proposed in the embodiments of the present invention improves the accuracy of arc fault detection.
[0028] For example, there is a situation where the fault arc current is not much different from the normal current, so the effective detection of current is one of the keys to arc fault detection; the main goal of current current sampling technology is to obtain accurate data related to the acquisition circuit state and current load to ensure the reliability, safety and efficiency of arc fault detection. However, the current current sampling technology has some limitations, especially the extraction of characteristic signals of current is not accurate enough. Specifically, the current current sampling technology usually uses current transformers or bus clamp sensors to measure current, obtains the current size in the acquisition circuit at different time points, and uses these current data to detect the operating state of the acquisition circuit and detect potential faults or abnormal conditions. However, especially in series fault arcs, there may be a situation where the fault arc current is not much different from the normal current. It is difficult to extract characteristic information only from the time series current data, which makes it very important to effectively extract the characteristic information of the fault arc in arc fault detection. At present, it is often used to directly collect one-dimensional current data and then input it into the trained traditional convolutional neural network model to extract feature signals, but it is difficult to obtain sufficiently accurate feature information. For example, arc fault detection based on back propagation (BP) neural network achieves the purpose of arc fault detection by extracting the transformation of the frequency domain and identifying the characteristics of normal operation and fault arc. However, for unstable signals such as fault arcs, a lot of training is often required to achieve the expected effect, and the work efficiency is low. Therefore, this method is difficult to accurately detect arc faults, especially for series fault arcs.
[0029] In addition, there is currently an arc fault detection method based on a combination of wavelet transform and singular value decomposition, but its computational complexity is large, and since there are no specific rules for the selection of wavelet mother functions and fundamental waves, it can only rely on people's experience, so this method is also difficult to accurately detect arc faults.
[0030] In view of this, the embodiment of the present invention converts one-dimensional current data into a two-dimensional current image by using a Markov transition field, so that the arc fault detection model that can extract image features can process the two-dimensional current image that better characterizes the rules and information of the one-dimensional current data, determine arc fault detection with higher accuracy, and thus improve the accuracy of arc fault detection. This will be explained in detail below.
[0031] Figure 1 1 is a flow chart of an arc fault detection method provided in an embodiment of the present invention. This embodiment is applicable to arc fault detection. The method can be executed by an arc fault detection device provided in an embodiment of the present invention, which can be implemented by software and / or hardware, and can be integrated in an electronic device, which can be various user terminals or servers.
[0032] See also Figure 1 The method of the embodiment of the present invention specifically includes the following steps:
[0033] S110, obtaining an arc fault detection model, and collecting current from a collection circuit within a collection time period to obtain one-dimensional current data.
[0034] Among them, the arc fault detection model is a model for detecting arc faults in the collected current within the collection time period; the arc fault detection model can be, for example, a residual network (Residual Network, ResNet) model. In the embodiment of the present invention, the structure and parameters of the arc fault detection model are not specifically limited. The collection time period is a time period for current data; the collection time period can be determined, for example, based on the collection cycle of the current data, and the collection time period can be determined, for example, based on the demand for arc fault detection of the collected current, and so on. The collection circuit is a circuit that requires arc fault detection. One-dimensional current data is the current data of a one-dimensional collection circuit; the one-dimensional current data can be, for example, each current signal with time sequence obtained by collecting current for the collection circuit within the collection time period.
[0035] For example, see Figure 2The arc fault detection model includes an input layer, a convolutional batch normalization ReLU activation function layer (Cov BN ReLU), a maximum pooling (Max Pooling, MaxPool) layer, a stacked residual block (Stacked ResidualBlocks), a convolutional average pooling layer (Cov AvgPool), a flattening layer (Flatten), a softmax layer, a fully connected layer (Fully Connect) and an output layer. The two-dimensional current image can be input into the arc fault detection model. The two-dimensional current image can be passed through multiple convolutional layers and pooling layers to obtain a feature map representing the image feature information; the feature map is reduced in dimension through an adaptive convolution average pooling layer, and the image after dimension reduction is flattened, which is the role of the Flatten layer. It can ensure that the model can handle inputs of any size without worrying about the specific shape of the data; the Softmax layer represents a layer for classification operations, and its role is to convert the image after the flattening operation in the previous step into a probability distribution, so that the predicted probability of each category is between 0 and 1; finally, the weights and biases of the input data are learned through the fully connected layer, and the original features of the input are converted into higher-level feature representations. In addition, each output neuron of the fully connected layer can obtain information from all neurons in the input layer or the previous layer, integrate this information, extract features, and learn parameters, and finally output whether an arc fault has occurred.
[0036] S120, using a Markov transition field to perform image conversion on the one-dimensional current data to obtain a two-dimensional current image.
[0037] The two-dimensional current image is a two-dimensional image obtained by performing image conversion on one-dimensional current data using a Markov transformation field.
[0038] It can be understood that Markov Transition Field (MTF) is a method of converting time sequence data into a two-dimensional image. Considering that directly processing one-dimensional current data through an arc fault detection model will result in a low accuracy of arc estimation detection, a Markov transition field can be used to convert the one-dimensional current data of normal and fault arcs into a two-dimensional current image which is a spatial image by using fuzzy kernel aggregation, so as to retain the discretized time domain dynamic information and capture the time dependence and dynamic changes on different time scales in the one-dimensional current data in the form of an image, so that the arc fault detection model can process the two-dimensional current image and improve the accuracy of arc estimation detection.
[0039] For example, the one-dimensional current data X(t)=(x t,t=1,2,...,T) is divided into Q bins, the serial numbers of the Q bins are marked as 1, 2,..., Q, and the amount of data in each bin is the same; each current signal in the one-dimensional current data is changed to its corresponding bin serial number; a transfer matrix is constructed Among them, w ij It represents the probability of the transfer of bin i to bin j. That is, the dynamic transfer probability of the one-dimensional current data can be calculated, and the properties of each bin in the two-dimensional current image can be obtained through the dynamic transfer probability; construct the Markov transition field A two-dimensional current image is obtained through the Markov transition field.
[0040] S130: Determine arc fault detection results of the acquisition circuit within an acquisition time period according to the two-dimensional current image and the arc fault detection model.
[0041] The arc fault detection result is a result that characterizes whether an arc fault occurs in the acquisition circuit; the arc fault detection result may include at least one of whether an arc fault occurs in the acquisition circuit, the type of arc fault, and the target load type of the acquisition circuit.
[0042] In the embodiment of the present invention, for example, the two-dimensional current image can be input into the arc fault detection model, and the arc fault detection model can perform complex transformation on the input two-dimensional current image, extract effective feature information, generate feature signals, and then obtain the output result of the arc fault detection model, and determine the arc fault detection result of the acquisition circuit according to the output result. In the embodiment of the present invention, the method of determining the arc fault detection result of the acquisition circuit within the acquisition time period according to the two-dimensional current image and the arc fault detection model is not specifically limited.
[0043] In the embodiment of the present invention, considering that the ResNet model is a deep convolutional neural network that can accept images as input, the network architecture of the ResNet model mainly includes an input layer, an initial convolutional layer, a stacked residual block, a batch normalization layer, an activation function layer, a global average pooling layer, a fully connected layer and an output layer. These layers are arranged in sequence to form a learning model for extracting information features of a two-dimensional current image. The introduction of this structure can effectively solve the common gradient vanishing problem in deep network training, so that the ResNet model can learn more complex feature representations, especially for the recognition of details and edge parts of the image. It is crucial; the feature extraction of the ResNet model is composed of an initial convolutional layer, a stacked residual block, a batch normalization layer, an activation function layer and a global average pooling layer; the initial convolutional layer is mainly used for feature extraction, usually starting with a larger convolution kernel, which is used to reduce the spatial resolution of the input image and extract low-level features, aiming to reduce the size and complexity of the image; the activation function layer adopts the ReLU function to accelerate the convergence of the network and is used in processing deep networks. The convolution layer of the ResNet model is used for feature extraction, and the core part is multiple stacked residual blocks. Each block contains jump connections, which can effectively solve the problem of gradient disappearance. The mathematical model of the residual block can be H(x)=x+F(x), where x represents the input data of the residual block, F(x) represents the residual function, that is, the mapping relationship of the two-dimensional current image learned by the residual block, and H(x) represents the output of the residual block, that is, the feature representation of the two-dimensional current image; The ResNet model ends with a global average pooling layer, which converts the output of the last residual block into a single value of each feature map, retaining important information, followed by a fully connected layer for classification, which is responsible for reducing the matrix data dimension of the pooling layer and extracting feature samples, that is, the extracted feature signal of the fault arc current, and using the Softmax loss function as the fully connected layer conversion function; Each residual block is followed by a batch normalization layer and a ReLU activation function, which can improve the training speed and generalization ability. Therefore, the ResNet model can be used as an arc fault detection model. The two-dimensional current image obtained by Markov transition field conversion effectively reduces the complex extraction of the arc fault detection model. The ResNet model can use local perception to effectively reduce complex extraction, and can merge the extracted local image information into global image representation features, avoiding the perception of each pixel point and reducing the complexity of feature extraction. Parameter sharing based on the same kernel further reduces the parameter scale of the ResNet model and enhances the generalization performance of the ResNet model. It can still have a good recognition effect on images after scaling, tilting and translation.In an embodiment of the present invention, at the input stage of the ResNet model, an attention mechanism can be introduced to enhance the learning and recognition capabilities of important features. Specifically, first, local key information is obtained through global scanning, and secondly, effective information is enhanced and invalid information is suppressed. For example, for a two-dimensional current image, there are n input information X=[X1, X2,..., X. N ], indicating the attention distribution According to the attention distribution calculation, we can get Among them, Softmax is the normalized exponential function, q is the query vector, s(X i ,q) is the attention evaluation function, using the dot product model, is the output result, Att(X,q) is the attention mechanism function, and the image processed by the attention mechanism can make the model pay more attention to the key parts of arc fault detection during the learning process, such as zero-break phenomenon in arc current, slight changes in amplitude, etc.; the attention mechanism can also be combined with the network structure of the model, for example, introducing attention weighting at the input of each residual block in the model, which can make the model more effectively focus on different levels and improve the accuracy and robustness of arc fault detection; it should be noted that in the model training stage, the parameters of the attention mechanism usually need to be optimized together with other parameters of the ResNet model. By backpropagating and updating parameters on the training samples, the model can learn how to better classify the current during normal operation and the current during fault arc according to the attention-weighted features.
[0044] It is understandable that in real life, when there is an arc fault in the acquisition circuit, its current is usually lower than the normal operating current, and the series fault arc is hidden. Conventional overcurrent protection devices cannot detect such faults. The solution of the embodiment of the present invention can use the Markov transition field to perform image conversion on the one-dimensional current data, describe the normal current and arc current from multiple time scales, and obtain a fine aggregate image of the dynamic transfer probability as a two-dimensional current image. Then, the arc fault detection model is fully utilized to identify the structural data feature information of the two-dimensional current image, and the classification of the current during normal operation and the current when a fault arc occurs is realized, thereby determining the arc fault detection result of the acquisition circuit within the acquisition time period, thereby improving the accuracy of arc fault detection.
[0045] It is understandable that in actual application, since there is a series relationship between the fault arc and the load in the acquisition circuit, the typical series fault arc current waveform is as follows: Figure 3As shown, the current on the acquisition circuit is reduced and is still within the normal current range. This phenomenon is a prominent feature of a series arc fault. There are two main reasons for a series fault arc. The first reason is that the contact between the electrodes is poor, the connection is loose, and the vibration causes the contact to be disconnected, and then an intermittent arc is formed. This is a point contact series arc fault. When an intermittent arc appears at the contact point of the two electrodes, the intermittent burning of the arc will cause heat to accumulate. If the heat cannot be dissipated in time, the conductor will be in an overheated state, which may ignite the external insulation material of the conductor and cause an electrical fire. The second reason is that the distribution line has been in disrepair for a long time, resulting in corrosion and oxidation of the line, carbonization of the cable insulation sheath, or rust on the connection terminals. Usually, a so-called carbonization channel is formed on the insulation sheath, and an arc is burned on this basis, which is a series fault arc in the form of a so-called carbonization path. The series fault arc current and its voltage signal usually have the following basic characteristics: Characteristic 1: When a series fault arc occurs, high-order harmonics will appear in the arc voltage signal and the current signal in the line; Characteristic 2: There is a voltage drop along the arc direction; Characteristic 3: When a series arc fault occurs on a line without taking specific compensation measures, the fault current signal is reduced compared to the normal current signal; Characteristic 4: When a series arc fault occurs, the rate of change of the current signal increases after the zero point; Characteristic 5: When a series arc fault occurs, due to the extinction and re-ignition of the arc, a flat shoulder area will appear in the current waveform at the zero point, that is, the zero rest phenomenon appears once every half cycle of the current signal; Characteristic 6: The voltage waveform of the series arc fault is no longer a sine wave, but is similar to a rectangular wave. Taking the above characteristics into consideration, the accuracy of the technical solution of the embodiment of the present invention for the fault detection of series fault arcs is particularly improved.
[0046] The technical solution of the embodiment of the present invention is to obtain an arc fault detection model, and collect current for the collection circuit within the collection time period to obtain one-dimensional current data, so as to perform arc fault detection through the one-dimensional current data; use the Markov transformation field to perform image conversion on the one-dimensional current data to obtain a two-dimensional current image, so as to obtain a two-dimensional current image that can better characterize the law and information of the one-dimensional current data through the Markov transformation field, which can be subsequently processed by the arc fault detection model that can perform image feature extraction; according to the two-dimensional current image and the arc fault detection model, the arc fault detection result of the collection circuit within the collection time period is determined, so as to determine the arc fault detection with higher accuracy through the two-dimensional current image and the arc fault detection model. The above technical solution converts the one-dimensional current data into a two-dimensional current image by using the Markov transformation field, so that the arc fault detection model that can perform image feature extraction can process the two-dimensional current image that better characterizes the law and information of the one-dimensional current data, determine the arc fault detection with higher accuracy, and thus improve the accuracy of arc fault detection.
[0047] An optional technical solution, during a collection time period, current is collected for a collection circuit to obtain one-dimensional current data, including: determining the minimum resistor with the smallest resistance in the collection circuit; during the collection time period, voltage is collected for the minimum resistor to obtain a collection voltage, and a voltage division algorithm is used to obtain one-dimensional current data based on the collection voltage and the resistance value of the minimum resistor.
[0048] Among them, the minimum resistance is the resistance with the smallest resistance in the acquisition circuit. The acquisition voltage is the voltage obtained by voltage acquisition for the minimum resistance. The resistance value is the resistance value of the minimum resistance.
[0049] In the embodiment of the present invention, the signal output end of the current transformer can be connected to the minimum resistor, and during the collection time period, the voltage is collected for the minimum resistor to obtain the collected voltage, and the voltage division algorithm of the resistor is used to obtain one-dimensional current data from the resistance.
[0050] In the embodiment of the present invention, the voltage can be collected for the minimum resistance to obtain the collected voltage, and then the one-dimensional current data can be obtained, which can improve the accuracy of the obtained one-dimensional current data.
[0051] Figure 4It is a flow chart of another arc fault detection method provided in an embodiment of the present invention. This embodiment is optimized based on the above-mentioned technical solutions. In this embodiment, optionally, before using the Markov transition field to process the one-dimensional current data to obtain the two-dimensional current image, it also includes: performing cutoff processing on the one-dimensional current data according to a preset current cutoff frequency; and updating the one-dimensional current data according to the obtained cutoff processing result. Among them, the explanations of the terms that are the same or corresponding to the above-mentioned embodiments are not repeated here.
[0052] See also Figure 4 The method of this embodiment may specifically include the following steps:
[0053] S210 , obtaining an arc fault detection model, and collecting current from a collection circuit within a collection time period to obtain one-dimensional current data.
[0054] S220 , performing cutoff processing on the one-dimensional current data according to a preset current cutoff frequency.
[0055] The current cutoff frequency is a preset cutoff frequency for performing cutoff processing on one-dimensional current data.
[0056] In the embodiment of the present invention, for example, a cut-off current signal less than the current cut-off frequency in the one-dimensional current data may be cut-off processed; for another example, a cut-off current signal greater than or equal to the current cut-off frequency in the one-dimensional current data may be cut-off processed; etc. In the embodiment of the present invention, the manner of performing the cut-off processing on the one-dimensional current data according to the preset current cut-off frequency is not specifically limited.
[0057] S230 , updating the one-dimensional current data according to the obtained cutoff processing result.
[0058] The cutoff processing result is a result obtained by performing cutoff processing on the one-dimensional current data.
[0059] In the embodiment of the present invention, for example, the cutoff processing result can be updated to one-dimensional current data. In the embodiment of the present invention, there is no specific limitation on the way of updating the one-dimensional current data according to the obtained cutoff processing result.
[0060] S240, using a Markov transition field to perform image conversion on the one-dimensional current data to obtain a two-dimensional current image.
[0061] S250: Determine arc fault detection results of the acquisition circuit within an acquisition time period according to the two-dimensional current image and the arc fault detection model.
[0062] The technical solution of the embodiment of the present invention performs cutoff processing on one-dimensional current data according to a preset current cutoff frequency to update the one-dimensional current data, thereby retaining more important information in the one-dimensional current data, thereby improving the accuracy of arc fault detection.
[0063] An optional technical solution performs cutoff processing on one-dimensional current data according to a preset current cutoff frequency, including: obtaining the preset current cutoff frequency; using a high-pass filter to cutoff current signals in the one-dimensional current data that are less than the current cutoff frequency.
[0064] The cut-off current signal is a current signal smaller than the current cut-off frequency in the one-dimensional current data.
[0065] It is understandable that in practical applications, the high-frequency data in the one-dimensional current data contains more important information, such as the rapid change characteristics of the current, energy distribution, and arc characteristics. Therefore, analyzing the signals in the one-dimensional current data is beneficial to the arc fault detection of the acquisition circuit.
[0066] Exemplarily, a suitable sensor can be selected for collecting one-dimensional current data, for example, a current transformer or a Hall sensor can be selected. Here, a CIT100 current transformer can be used, which can collect current signals with a maximum frequency of 600MHz. A suitable signal conditioning circuit can be designed or selected to amplify, filter and adapt the sensor output so as to effectively process the high-frequency signal later. Figure 5 , a high-pass filter can be used here; a preset current cutoff frequency of 300MHz can be obtained; a high-pass filter is used to block or weaken the cutoff current signal less than 300MHz in the one-dimensional current data, and retain the current signal greater than or equal to 300MHz.
[0067] In the embodiment of the present invention, the current cut-off frequency can be determined by calculation. For example, the calculation formula of the current cut-off frequency can be: The gain calculation formula can be When at the current cut-off frequency, A=0.707.
[0068] In the embodiment of the present invention, there is no specific limitation on the method of using a high-pass filter to perform a cut-off process on a cut-off current signal that is less than a current cut-off frequency in the one-dimensional current data.
[0069] In an embodiment of the present invention, by adopting a high-pass filter to cut off the cutoff current signal in the one-dimensional current data that is less than the current cutoff frequency, the high-frequency band data in the one-dimensional current data that can reflect more important information can be retained, thereby further improving the accuracy of arc fault detection.
[0070] Figure 6 It is a flowchart of another arc fault detection method provided in an embodiment of the present invention. This embodiment is optimized based on the above-mentioned technical solutions. In this embodiment, optionally, the arc fault detection model is pre-trained by the following steps: obtaining at least two groups of training samples, wherein each group of training samples includes sample current data of the sample circuit in the sample time period and the arc fault label of the sample current data; based on at least two groups of training samples, the original detection model is trained to obtain the arc fault detection model. Among them, the explanations of the terms that are the same or corresponding to the above-mentioned embodiments are not repeated here.
[0071] See also Figure 6 The method of this embodiment may specifically include the following steps:
[0072] S310, obtaining at least two groups of training samples, wherein each group of training samples includes sample current data of a sample circuit within a sample time period and an arc fault label of the sample current data.
[0073] Among them, the training samples are samples used to train the arc fault detection model. The sample circuit is the circuit targeted when collecting sample current data. The sample current data is the data obtained by current collection and processing for the sample circuit; the sample current data can be image data obtained by converting the data obtained by current collection for the sample circuit using a Markov transition field, so as to improve the learning ability of the model and extract effective feature information, so that the judgment ability of the arc fault detection model obtained by training is enhanced, so that the arc fault detection model can be better used in field applications. The arc fault label is a label that can reflect whether there is an arc fault in the sample circuit.
[0074] S320: Based on at least two groups of training samples, train the original detection model to obtain an arc fault detection model.
[0075] Among them, the original detection model is an untrained arc fault detection model.
[0076] In the embodiment of the present invention, for example, based on at least two groups of training samples, the output can be calculated by forward propagation, the loss function value can be calculated, and the loss function can be used to evaluate the difference between the model output and the arc fault label. In the design of the loss function, not only the cross-entropy loss function is used to measure the gap between the original image and the recognized image, and then the loss function is adjusted to achieve the purpose of minimizing the loss function, but also the structural similarity of the image information is paid attention to. For this purpose, the Structural Similarity Index is introduced. The purpose of using SSIM as part of the loss function is to ensure that the original image is highly consistent with the recognized image. The SSIM indicator focuses more on evaluating the visual effect of the image, so the final generated image information is more complete in details and edges, and closer to the human eye's perception of the actual scene; back propagation is used to calculate the gradient, and the network parameters are adjusted through gradient descent to minimize the loss function, and the original detection model is trained to obtain the arc fault detection model. During the training process, the performance indicators of the model, such as accuracy and loss value, can be monitored. According to the actual situation, the internal parameters of the model are adjusted to minimize the difference between the predicted ultra-high frequency signal and the actual ultra-high frequency signal; the trained arc fault detection model can be verified using a validation data set to observe whether the arc fault detection model can effectively extract the characteristic information of the fault arc current, and use performance indicators such as accuracy (ACC), precision (P), recall (R), etc. to evaluate the accuracy and recognition quality of the arc fault detection model. Among them, SSIM is used to measure the structural similarity between the two signals, and its calculation formula can be μ I , μ K is the average value of signals I and K, is the variance of signals I and K, σ IK represents the covariance of signals I and K, c1=(K1L) 2 , c2=(K2L) 2 are two constants that maintain stability. The larger the SSIM, the higher the similarity between the two signals. The calculation formulas of ACC, P, and R are respectively as well as TP indicates that both the prediction and the actual are positive samples, TN indicates that both the prediction and the actual are negative samples, FP indicates that the predicted positive sample is actually a negative sample, and FN indicates that the predicted negative sample is actually a positive sample; the parameters, network architecture or training process of the arc fault detection model can be adjusted according to the results of the arc fault detection model verification and actual application to further improve the performance of the arc fault detection model. In the embodiment of the present invention, the method of training the original detection model based on at least two groups of training samples to obtain the arc fault detection model is not specifically limited.
[0077] In an embodiment of the present invention, the original detection model may be a ResNet model. In the training of the ResNet model, a variety of techniques are often used to improve model performance and training stability. Among them, an important technology is the random dropout technology, which is used to prevent overfitting by randomly shutting down a part of neurons in each iteration, forcing the network to learn more robust features, preventing excessive dependence between neurons, and improving the generalization ability of the model.
[0078] In the embodiment of the present invention, at least two groups of training samples may include a single disturbance and a compound disturbance, so as to increase the complexity of the at least two groups of training samples, thereby improving the generalization ability of the arc fault detection model.
[0079] Exemplarily, 8000 groups of samples are obtained; the 8000 groups of samples are divided into 5000 groups of training samples, 1000 groups of verification samples and 2000 groups of test samples, and there is no intersection between the three types of samples; based on the 5000 groups of training samples, the original detection model is trained to obtain an arc fault detection model; the optimal network weights in the training process are selected and saved using the 1000 groups of verification samples; and the optimal network weights are evaluated using the 2000 groups of test samples.
[0080] For example, see Figure 7, determine at least one group of normal current signals and at least one group of abnormal current signals corresponding to at least one sample load type, and preprocess at least one group of normal current signals and at least one group of abnormal current signals for each sample load type in the at least one sample load type; use Markov transition field to convert the preprocessed at least one group of normal current signals and at least one group of abnormal current signals, respectively, to obtain at least one normal current data and at least one abnormal current data represented in the form of images; determine at least two groups of training samples according to at least one normal current data and at least one abnormal current data corresponding to at least one sample load type; train the ResNet model based on the at least two groups of training samples to generate normal current characteristic signals and fault arc current characteristic signals corresponding to at least one sample load type, so as to obtain the arc fault detection model; use the arc fault detection model to perform arc fault detection; and adjust the parameters of the arc fault detection model and shorten the training time according to the obtained detection results.
[0081] S330, obtaining an arc fault detection model, and collecting current on a collection circuit within a collection time period to obtain one-dimensional current data.
[0082] S340, using a Markov transition field to perform image conversion on the one-dimensional current data to obtain a two-dimensional current image.
[0083] S350: Determine arc fault detection results of the acquisition circuit within an acquisition time period according to the two-dimensional current image and the arc fault detection model.
[0084] In the embodiment of the present invention, the original detection model can be trained based on at least two groups of training samples to obtain an arc fault detection model, thereby improving the accuracy of arc fault detection by the arc fault detection model.
[0085] An optional technical solution determines the arc fault detection result of the acquisition circuit within the acquisition time period based on a two-dimensional current image and an arc fault detection model, including: inputting the two-dimensional current image into the arc fault detection model; determining the arc fault detection result of the acquisition circuit within the acquisition time period based on the target load type and target arc fault type output by the arc fault detection model.
[0086] The target load type is the type of load of the acquisition circuit. The target arc fault type is the type of arc fault of the acquisition circuit; the target arc fault type may be, for example, one of a series fault arc, a parallel fault arc, a ground fault arc, and a fault-free arc. The target arc fault type may be, for example, one of a direct grounding arc, an indirect grounding arc, an arc in a winding, and a fault-free arc. In the embodiment of the present invention, the classification method of the target arc fault type is not specifically limited.
[0087] In the embodiment of the present invention, for example, the target load type and the target arc fault type outputted by the arc fault detection model may be used as the arc fault detection result of the acquisition circuit; it is understandable that under certain target load types, certain target arc fault types may not cause the acquisition circuit to have a fault or may have a fault that has a greater impact on the acquisition circuit. In this case, the arc fault detection result may be considered as the absence of an arc estimation. Therefore, for another example, the fault condition may be determined according to the target load type and the target arc fault type. The fault condition may characterize whether the acquisition circuit has a fault, and the arc fault detection result of the acquisition circuit may be determined according to the fault condition. In the embodiment of the present invention, the method of determining the arc fault detection result of the acquisition circuit according to the target load type and the target arc fault type outputted by the arc fault detection model is not specifically limited.
[0088] In an embodiment of the present invention, the arc fault detection model can first determine the target load type based on the input two-dimensional current image. Since the method of determining the target arc fault type may be different under different target load types, the arc fault detection model can determine the target arc fault type based on the target load type and the two-dimensional current image, and output the target load type and the target arc fault type. That is, a part of the arc fault detection model will be trained to process the data corresponding to the normal current in the two-dimensional current image to achieve the judgment of the target load type, and the other part will be used to learn the data corresponding to the abnormal current in the two-dimensional current image to judge whether a fault arc is generated and determine the target arc fault type.
[0089] In an embodiment of the present invention, the arc fault detection model can output a target load type and a target arc fault type, and then determine the arc fault detection result of the acquisition circuit according to the target load type and the target arc fault type, thereby improving the accuracy of the arc fault detection result.
[0090] Based on the above scheme, another optional technical scheme is that at least two groups of training samples are pre-constructed through the following steps: determining at least one sample load type of the sample circuit; for each sample load type in the at least one sample load type, determining at least one normal current data and at least one abnormal current data of the sample circuit under the sample load type; for each normal current data in the at least one normal current data, taking the normal current data as a sample current data, and taking the sample load type and the fault-free arc as arc fault labels of the sample current data, and taking the sample current data and the arc fault label as a group of training samples; for each abnormal current data in the at least one abnormal current data, determining the sample arc fault type of the abnormal current data, taking the abnormal current data as a sample current data, taking the sample load type and the sample arc fault type as arc fault labels of the sample current data, and taking the sample current data and the arc fault label as a group of training samples.
[0091] Among them, the sample load type is the type of load of the sample circuit. Normal current data is current data obtained by current collection for the sample circuit when there is no arc fault in the sample circuit. Abnormal current data is current data obtained by current collection for the sample circuit when there is an arc fault in the sample circuit. The sample arc fault type is the type of arc fault of the sample circuit corresponding to the abnormal current data; the sample arc fault type can be, for example, one of a series fault arc, a parallel fault arc, a grounding fault arc, and a fault-free arc. The sample arc fault type can be, for example, one of a direct grounding arc, an indirect grounding arc, an arc in a winding, and a fault-free arc. In the embodiment of the present invention, the classification method of the sample arc fault type is not specifically limited.
[0092] In an embodiment of the present invention, the time lengths corresponding to at least one normal current data and at least one abnormal current data may be the same as or different from the time lengths corresponding to the one-dimensional current data; the data volumes corresponding to at least one normal current data and at least one abnormal current data may be the same as or different from the data volumes corresponding to the one-dimensional current data.
[0093] In an embodiment of the present invention, after determining at least one normal current data and at least one abnormal current data of the sample circuit under the sample load type, the at least one normal current data and the at least one abnormal current data can be pre-processed such as denoising, normalization and / or data format conversion, and the at least one normal current data and the at least one abnormal current data can be updated according to the pre-processing result. The pre-processing of denoising the at least one normal current data and the at least one abnormal current data can use wavelet denoising to decompose the signal in the data into sub-bands of different frequencies, and then apply threshold processing to each sub-band to remove noise, effectively removing noise while retaining details; the pre-processing of normalizing the at least one normal current data and the at least one abnormal current data can use substantial The Z-Score normalization method is performed, wherein μ represents the mean of the original data (normal current data or abnormal current data), σ represents the standard deviation of the original data, x represents the original data, and x′ represents the normalized data.
[0094] In the embodiment of the present invention, each load type that may appear in the sample circuit can be used as at least one sample load type. In the embodiment of the present invention, the method for determining at least one sample load type of the sample circuit is not specifically limited.
[0095] In the embodiment of the present invention, for example, for each sample load type in at least one sample load type, current can be collected in a laboratory for a sample circuit under the sample load type when an arc fault exists, and current can be collected for a sample circuit under the sample load type when no arc fault exists, to obtain at least one normal current data and at least one abnormal current data. In the embodiment of the present invention, the method for determining at least one normal current data and at least one abnormal current data of the sample circuit under the sample load type is not specifically limited.
[0096] In the embodiment of the present invention, the method for determining the type of the arc fault of the sample of abnormal current data is not specifically limited.
[0097] In an embodiment of the present invention, at least two groups of training samples are pre-constructed in the above manner, and based on the at least two groups of training samples, cross entropy loss and mean square error loss are used as loss functions of the original detection model, and the network parameters are optimized by gradient descent to train the original detection model. The information of the fault arc current is learned through multiple iterations of the model to obtain an arc fault detection model. After a large amount of training of at least two groups of training samples, the obtained arc fault detection model can extract features from the input two-dimensional current image representing the normal current and the fault arc current, and generate normal current and fault arc characteristic signals of different load types, thereby improving the accuracy of arc fault detection by the trained arc fault detection model, wherein the formula of the cross entropy loss function can be: y represents the real data, which is a C-dimensional vector, in which only one element is 1 and the other elements are 0; x represents the input data; x i Represents the i-th element of the real data; y i It represents the probability that the model predicts that x belongs to the i-th category.
[0098] In an embodiment of the present invention, at least two groups of training samples can be constructed by determining at least one sample load type of the sample circuit, and determining at least one normal current data and at least one abnormal current data corresponding to the sample circuit under each sample load type, thereby ensuring the diversity of the training samples, covering a variety of training samples, and improving the generalization ability of the trained arc fault detection model.
[0099] In order to better understand the technical solution of the above embodiment of the present invention, an optional example is provided here. Figure 8 , taking the one-dimensional current data from the arc generating device as an example, the arc fault detection model is obtained, and within the acquisition time period, the current signal is collected for the acquisition circuit to obtain the one-dimensional current data; the one-dimensional current data is high-pass filtered by a high-pass filter, and only the high-frequency components in the one-dimensional current data are retained and updated to the one-dimensional current data; the one-dimensional current data is image converted by MTF to extract features and obtain a two-dimensional current image; the two-dimensional current image is input into the arc fault detection model, and the two-dimensional current image is processed by the attention mechanism, convolution layer, residual module, fully connected layer and output layer in the arc fault detection model to obtain the output result of the arc fault detection model, wherein the arc fault detection model is a residual network based on the attention mechanism and deep learning; according to the output result, the arc fault detection result of the acquisition circuit within the acquisition time period is determined.
[0100] Fig. 9This is a structural block diagram of an arc fault detection device provided in an embodiment of the present invention, and the device is used to execute the arc fault detection method provided in any of the above embodiments. The device and the arc fault detection method of the above embodiments belong to the same inventive concept. For details not described in detail in the embodiment of the arc fault detection device, refer to the embodiment of the above arc fault detection method. Fig. 9 The device may specifically include: a one-dimensional current data obtaining module 410, a two-dimensional current image obtaining module 420 and an arc fault detection result determining module 430.
[0101] Among them, the one-dimensional current data acquisition module 410 is used to obtain the arc fault detection model, and to perform current acquisition on the acquisition circuit within the acquisition time period to obtain the one-dimensional current data; the two-dimensional current image acquisition module 420 is used to use the Markov transition field to perform image conversion on the one-dimensional current data to obtain a two-dimensional current image; the arc fault detection result determination module 430 is used to determine the arc fault detection result of the acquisition circuit within the acquisition time period according to the two-dimensional current image and the arc fault detection model.
[0102] Optionally, the device may also include: a cutoff processing module, which is used to perform cutoff processing on the one-dimensional current data according to a preset current cutoff frequency before using a Markov transition field to process the one-dimensional current data to obtain a two-dimensional current image; and a one-dimensional current data updating module, which is used to update the one-dimensional current data according to the obtained cutoff processing result.
[0103] Optionally, based on the above device, the cutoff processing module includes: a current cutoff frequency acquisition submodule, used to obtain a preset current cutoff frequency; a cutoff processing submodule, used to use a high-pass filter to cut off the cutoff current signal in the one-dimensional current data that is less than the current cutoff frequency.
[0104] Optionally, the device may also include the following modules for pre-training an arc fault detection model: a training sample acquisition module for acquiring at least two groups of training samples, wherein each group of training samples includes sample current data of the sample circuit within a sample time period and an arc fault label of the sample current data; an arc fault detection model acquisition module for training the original detection model based on at least two groups of training samples to obtain an arc fault detection model.
[0105] Optionally, based on the above-mentioned device, the arc fault detection result determination module 430 includes: a two-dimensional current image input submodule, used to input the two-dimensional current image into the arc fault detection model; an arc fault detection result determination submodule, used to determine the arc fault detection result of the acquisition circuit within the acquisition time period according to the target load type and the target arc fault type output by the arc fault detection model.
[0106] Optionally, based on the above-mentioned device, the device may also include the following modules to pre-construct at least two groups of training samples: a sample load type determination module, used to determine at least one sample load type of the sample circuit; an abnormal current data determination module, used to determine, for each sample load type in at least one sample load type, at least one normal current data and at least one abnormal current data of the sample circuit under the sample load type; a first training sample is used as a module, used to, for each normal current data in at least one normal current data, take the normal current data as a sample current data, and take the sample load type and the fault-free arc as arc fault labels of the sample current data, and take the sample current data and the arc fault label as a group of training samples; a second training sample is used as a module, used to determine, for each abnormal current data in at least one abnormal current data, the sample arc fault type of the abnormal current data, take the abnormal current data as a sample current data, take the sample load type and the sample arc fault type as arc fault labels of the sample current data, and take the sample current data and the arc fault label as a group of training samples.
[0107] Optionally, the one-dimensional current data obtaining module 410 includes: a minimum resistance determination submodule, used to determine the minimum resistance with the smallest resistance value in the acquisition circuit; a one-dimensional current data obtaining submodule, used to perform voltage acquisition for the minimum resistance within the acquisition time period, obtain the acquisition voltage, and use a voltage division algorithm to obtain one-dimensional current data based on the acquisition voltage and the resistance value of the minimum resistance.
[0108] The arc fault detection device provided by the embodiment of the present invention obtains an arc fault detection model through a one-dimensional current data acquisition module, and performs current acquisition on an acquisition circuit within an acquisition time period to obtain one-dimensional current data, so as to perform arc fault detection through the one-dimensional current data; through a two-dimensional current image acquisition module, a Markov transformation field is used to perform image conversion on the one-dimensional current data to obtain a two-dimensional current image, so as to obtain a two-dimensional current image that can better characterize the regularity and information of the one-dimensional current data through the Markov transformation field, and can be subsequently processed by an arc fault detection model that can perform image feature extraction; through an arc fault detection result determination module, the arc fault detection result of the acquisition circuit within the acquisition time period is determined according to the two-dimensional current image and the arc fault detection model, so as to determine arc fault detection with higher accuracy through the two-dimensional current image and the arc fault detection model. The above-mentioned device converts one-dimensional current data into a two-dimensional current image by using a Markov transition field, so that an arc fault detection model that can perform image feature extraction can process the two-dimensional current image that better characterizes the rules and information of the one-dimensional current data, determine arc fault detection with higher accuracy, and thus improve the accuracy of arc fault detection.
[0109] The arc fault detection device provided in the embodiment of the present invention can execute the arc fault detection method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.
[0110] It is worth noting that in the above-mentioned embodiment of the arc fault detection device, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.
[0111] Fig.10 A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.
[0112] like Fig.10As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.
[0113] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.
[0114] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 performs the various methods and processes described above, such as an arc fault detection method.
[0115] In some embodiments, the arc fault detection method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the arc fault detection method described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to perform the arc fault detection method in any other appropriate manner (e.g., by means of firmware).
[0116] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.
[0117] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.
[0118] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
[0119] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).
[0120] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.
[0121] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.
[0122] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.
[0123] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for detecting an arc fault, characterized in that: include: Obtain an arc fault detection model, and collect current from the collection circuit within a collection time period to obtain one-dimensional current data; Using a Markov transition field, the one-dimensional current data is image-converted to obtain a two-dimensional current image; The arc fault detection result of the acquisition circuit within the acquisition time period is determined according to the two-dimensional current image and the arc fault detection model.
2. The method according to claim 1, characterized in that Before the Markov transition field is used to process the one-dimensional current data to obtain a two-dimensional current image, the method further includes: According to a preset current cutoff frequency, the one-dimensional current data is subjected to a cutoff process; The one-dimensional current data is updated according to the obtained cutoff processing result.
3. The method according to claim 2, characterized in that The step of performing cutoff processing on the one-dimensional current data according to a preset current cutoff frequency includes: Obtaining a preset current cutoff frequency; A high-pass filter is used to perform cut-off processing on a cut-off current signal in the one-dimensional current data that is less than the current cut-off frequency.
4. The method according to claim 1, characterized in that: The arc fault detection model is pre-trained by the following steps: Acquire at least two groups of training samples, wherein each group of the training samples includes sample current data of a sample circuit within a sample time period and arc fault labels of the sample current data; Based on the at least two groups of training samples, the original detection model is trained to obtain the arc fault detection model.
5. The method according to claim 4, characterized in that The step of determining the arc fault detection result of the acquisition circuit within the acquisition time period according to the two-dimensional current image and the arc fault detection model includes: Inputting the two-dimensional current image into the arc fault detection model; The arc fault detection result of the acquisition circuit within the acquisition time period is determined according to the target load type and the target arc fault type output by the arc fault detection model.
6. The method according to claim 5, characterized in that The at least two groups of training samples are pre-constructed by the following steps: determining at least one sample load type for the sample circuit; For each sample load type of the at least one sample load type, determining at least one normal current data and at least one abnormal current data of the sample circuit under the sample load type; For each normal current data in the at least one normal current data, the normal current data is used as a sample current data, the sample load type and the fault-free arc are used as arc fault labels of the sample current data, and the sample current data and the arc fault labels are used as a group of training samples; For each abnormal current data among the at least one abnormal current data, determine a sample arc fault type of the abnormal current data, take the abnormal current data as a sample current data, take the sample load type and the sample arc fault type as an arc fault label of the sample current data, and take the sample current data and the arc fault label as a group of training samples.
7. The method according to claim 1, characterized in that During the acquisition time period, current acquisition is performed on the acquisition circuit to obtain one-dimensional current data, including: Determine the minimum resistor with the lowest resistance in the acquisition circuit; During the collection time period, voltage collection is performed for the minimum resistor to obtain a collection voltage, and a voltage division algorithm is used to obtain one-dimensional current data according to the collection voltage and the resistance value of the minimum resistor.
8. An arc fault detection device, characterized in that: include: A one-dimensional current data acquisition module is used to acquire an arc fault detection model and collect current from a collection circuit within a collection period to obtain one-dimensional current data. A two-dimensional current image obtaining module is used to perform image conversion on the one-dimensional current data by using a Markov transformation field to obtain a two-dimensional current image; The arc fault detection result determination module is used to determine the arc fault detection result of the acquisition circuit within the acquisition time period according to the two-dimensional current image and the arc fault detection model.
9. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor performs the arc fault detection method according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the arc fault detection method according to any one of claims 1 to 7 when executed.
Citation Information
Cited By
Wind power system converter open-circuit fault robustness diagnosis method considering wind speed fluctuation
CN120068644A
