An arc fault detection method and an arc fault detection model training method

Abstract

The application relates to an arc fault detection method and an arc fault detection model training method, and particularly relates to the technical field of electric signal processing. The method comprises the following steps: obtaining current data generated by a target device, and obtaining frequency domain data according to the current data; processing the frequency domain data through a feature extraction module of an arc fault detection model to obtain arc features; splicing a device number corresponding to the target device and the arc features to obtain to-be-identified features; processing the to-be-identified features through an identification module in the arc fault detection model to determine a fault result of the target device. In the above scheme, when identifying whether an arc fault exists, the arc fault detection model considers the distinguishing characteristics between different devices, so that the fault detection model has high identification accuracy when identifying current data of different devices.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric signal processing, and particularly relates to an arc fault detection method and an arc fault detection model training method. BACKGROUND

[0002] There are many single-chip microcomputer algorithms for direct current arc detection. In recent years, artificial intelligence algorithms have been widely used in arc detection due to high accuracy, so artificial intelligence algorithms are used for arc fault detection.

[0003] In the existing single-chip microcomputer algorithm process, first, it is judged whether the sampling time t of the data sent into the single-chip microcomputer reaches the pre-set time window T. After reaching, the one-dimensional current data is processed, the interference signal of the corresponding state is removed according to the current working state of the inverter, the representative characteristics of the current are extracted, and then the characteristic data is sent into the arc fault detection model for detection after normalization processing. The detection model uses an arc detection model based on a two-dimensional convolutional neural network, and other algorithm models can also be used. A large amount of arc data is used to train the model in advance to ensure accuracy before use.

[0004] In the above scheme, the arc fault detection model has poor detection effect on current data generated by different types of equipment. SUMMARY

[0005] The present application provides an arc fault detection method and an arc fault detection model training method, which can improve the speed detection accuracy. The technical scheme is as follows.

[0006] In one aspect, an arc fault detection method is provided, which comprises:

[0007] Obtaining current data generated by a target device, and obtaining frequency domain data according to the current data;

[0008] Processing the frequency domain data through a feature extraction module of an arc fault detection model to obtain arc features;

[0009] Splicing the device number corresponding to the target device and the arc features to obtain to-be-identified features;

[0010] Processing the to-be-identified features through an identification module in the arc fault detection model to determine the fault result of the target device.

[0011] In another aspect, an arc fault detection model training method is provided, which comprises:

[0012] Obtaining sample current data in a sample data set, and obtaining sample frequency domain data according to the sample current data; the sample data set includes current data generated by each sample device;

[0013] processing the sample frequency domain data through a feature extraction module of the arc fault detection model to obtain sample arc features;

[0014] splicing a device number corresponding to a sample device generating the sample current data and the sample arc features to obtain sample to-be-identified features;

[0015] processing the sample to-be-identified features through an identification module in the arc fault detection model to determine a sample identification result of the sample device generating the sample current data;

[0016] based on the sample identification result and a fault result of the sample device generating the sample current data, performing parameter updating on the arc fault detection model to obtain a trained arc fault detection model.

[0017] In another aspect, an arc fault detection device is provided, and the device comprises:

[0018] a frequency domain data acquisition module configured to acquire current data generated by a target device and obtain frequency domain data according to the current data;

[0019] an arc feature acquisition module configured to process the frequency domain data through a feature extraction module of an arc fault detection model to obtain arc features;

[0020] a feature splicing module configured to splice a device number corresponding to the target device and the arc features to obtain to-be-identified features;

[0021] a fault determination module configured to process the to-be-identified features through an identification module in the arc fault detection model to determine a fault result of the target device.

[0022] In a possible implementation, the identification module in the fault detection module is a fully connected layer;

[0023] the fault determination module is configured to,

[0024] process the to-be-identified features through the fully connected layer to obtain an arc fault probability corresponding to the target device;

[0025] when the arc fault probability is greater than or equal to a fault threshold, determine that the target device has an arc fault;

[0026] when the arc fault probability is less than the fault threshold, determine that the target device does not have an arc fault.

[0027] the fault determination module is further configured to,

[0028] generating a corresponding weight parameter according to the arc fault probability when the arc fault probability is less than the fault threshold;

[0029] accumulating the weight parameter with a historical weight parameter to obtain an updated historical weight parameter; the historical weight parameter is generated by the arc fault detection model based on historical current data of the target device within a specified time;

[0030] determining that the target device has an arc fault when the updated historical weight parameter is greater than a weight threshold.

[0031] In a possible implementation, the frequency domain data determination module is configured to,

[0032] obtain current data generated by the target device, and sample and normalize the current data to obtain a current data sequence;

[0033] performing fast Fourier transform on the current data sequence to obtain a frequency domain sequence;

[0034] equally dividing the frequency domain sequence according to a specified length to obtain each frequency domain sub-sequence;

[0035] splicing the each frequency domain sub-sequence into a matrix to obtain frequency domain data of the current data.

[0036] In another aspect, an arc fault detection model training device is provided, and the device comprises:

[0037] a sample data acquisition module configured to obtain sample current data in a sample data set, and obtain sample frequency domain data based on the sample current data; the sample data set comprises current data generated by each sample device;

[0038] a sample feature acquisition module configured to process the sample frequency domain data through a feature extraction module of an arc fault detection model to obtain sample arc features;

[0039] a sample splicing module configured to splice a device number corresponding to a sample device generating the sample current data with the sample arc features to obtain sample to-be-recognized features;

[0040] a sample identification result generation module configured to process the sample to-be-recognized features through an identification module in the arc fault detection model to determine a sample identification result of a sample device generating the sample current data;

[0041] The model training module is configured to update parameters of the arc fault detection model based on the sample identification result and a fault result of a sample device generating the sample current data, to obtain a trained arc fault detection model.

[0042] In a possible implementation, the model training module is further configured to,

[0043] extract each parameter in the trained arc fault detection model;

[0044] determine a first difference between a maximum value and a minimum value of the each parameter;

[0045] obtain a second difference between a maximum value and a minimum value of a target integer range;

[0046] determine a quantization interval of a target integer number based on a ratio of the first difference and the second difference;

[0047] quantize the each parameter to the target integer range based on the quantization interval, to obtain a quantized arc fault detection model.

[0048] In another aspect, a computer device is provided, which includes a processor and a memory. The memory stores at least one instruction, which is loaded and executed by the processor to implement the arc fault detection method or the arc fault detection model training method.

[0049] In another aspect, a computer readable storage medium is provided, which stores at least one instruction. The at least one instruction is loaded and executed by a processor to implement the arc fault detection method or the arc fault detection model training method.

[0050] In yet another aspect, a computer program product or a computer program is provided, which includes computer instructions stored in a computer readable storage medium. A processor of a computer device reads the computer instructions from the computer readable storage medium. The processor executes the computer instructions, so that the computer device performs the arc fault detection method or the arc fault detection model training method.

[0051] The technical solutions provided in the present application can have the following beneficial effects:

[0052] In order to realize arc fault detection, the current data of the target device is obtained first, and the frequency domain data is obtained according to the current data, and the arc feature is obtained by processing the frequency domain data through the feature extraction model of the arc fault detection model. At this time, the arc feature is spliced with the device number corresponding to the target device to obtain the to-be-recognized feature. When the recognition model processes the to-be-recognized feature, the current characteristics of the current data are considered, and the characteristics of the device generating the current data are also considered. At this time, when the arc fault detection model identifies whether there is an arc fault, the distinguishing characteristics between different devices are considered, so that the fault detection model has high recognition accuracy when identifying the current data of different devices. BRIEF DESCRIPTION OF DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the specific embodiments or prior art of the present application, the drawings needed in the specific embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0054] Figure 1 It is a structural schematic diagram of an arc fault recognition system according to an example embodiment.

[0055] Figure 2 It is a method flowchart of an arc fault detection method according to an example embodiment.

[0056] Figure 3 It is a method flowchart of an arc fault detection model training method according to an example embodiment.

[0057] Figure 4 It is a method flowchart of an arc fault detection method according to an example embodiment.

[0058] Figure 5 An implementation logic diagram of a neural network model related to an embodiment of the present application is shown.

[0059] Figure 6 An arc detection flowchart related to an embodiment of the present application is shown.

[0060] Figure 7 A framework diagram of an arc fault detection device related to an embodiment of the present application is shown Figure 8 A framework diagram of an arc fault detection model training device related to an embodiment of the present application is shown.

[0061] Figure 9 It is a schematic diagram of a computer device provided according to an example embodiment of the present application. DETAILED DESCRIPTION

[0062] The technical solutions of the present application will be described clearly and completely below with reference to the drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0063] It should be understood that the "indication" mentioned in the embodiments of the present application can be direct indication, indirect indication, or can be an indication of a correlation relationship. For example, A indicates B, which can mean that B can be obtained by A; or A indirectly indicates B, for example, A indicates C, and B can be obtained by C; or A and B have a correlation relationship.

[0064] In the description of the embodiments of the present application, the term "corresponding" can mean a direct or indirect corresponding relationship between the two, or a correlation relationship between the two, or an indication and being indicated, configuration and being configured, etc.

[0065] In the embodiments of the present application, "predefined" can be realized by pre-saving corresponding codes, tables or other means for indicating related information in devices (such as terminal devices and network devices), and the specific implementation manner is not limited in the present application.

[0066] Figure 1 is a structural schematic diagram of an arc fault identification system according to an exemplary embodiment. The arc fault identification system includes a data processing device 110 and a current collection device 120.

[0067] Optionally, the current collection device 120 includes a data storage. When the current collection device collects the target current to obtain target current data, the current data can be saved in the data storage. For example, the current collection device can be a current sensor, a current collection instrument, and a current collection current.

[0068] Optionally, in the present application, the current collection device is used to collect photovoltaic current data generated by a photovoltaic device.

[0069] Optionally, the data processing device 110 can be a computer device with high computing power, and the data processing device is used to analyze the collected target current data to obtain the characteristics of the target current data.

[0070] Optionally, the data processing device 110 can be a terminal device installed with current analysis software. When the terminal device receives an instruction for current data analysis, the terminal device can read corresponding current data from the data storage in the current collection device 120 and analyze the current data to obtain the characteristics of the target current data.

[0071] Optionally, the terminal device can process the target current data through an arc detection model to obtain the characteristics of the target current data.

[0072] Optionally, the data processing device 110 can also be a server installed with current analysis software. The current collection device can be a terminal device. When the terminal device collects target current data, the terminal device can transmit the target current data to the server to complete the analysis of the characteristics of the target current data.

[0073] Optionally, the data processing device 110 and the current collection device 120 can be connected through a wired or wireless network.

[0074] Optionally, the server can be a server cluster composed of multiple physical servers or a distributed system. The server can also be a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDNs, and big data and artificial intelligence platforms.

[0075] Optionally, the system can further include a management device for managing the system (such as managing the connection state between each module and the server). The management device and the server are connected through a communication network. Optionally, the communication network is a wired network or a wireless network.

[0076] Optionally, the wireless network or wired network uses standard communication techniques and / or protocols. The network is usually the Internet, but can also be any other network, including but not limited to a local area network, a metropolitan area network, a wide area network, a mobile, limited or wireless network, a private network or any combination of virtual private networks. In some embodiments, technologies and / or formats including hypertext markup language, extensible markup language, etc. are used to represent data exchanged through the network. In addition, all or some links can be encrypted using conventional encryption techniques such as secure sockets layer, transport layer security, virtual private network, Internet Protocol Security, etc. In other embodiments, custom and / or dedicated data communication techniques can be used instead of or in addition to the above data communication techniques.

[0077] Figure 2 is a method flowchart of an arc fault detection method according to an exemplary embodiment. The method is performed by a system as shown in Figure 1The computer device in the arc fault detection system shown performs. Figure 2 The arc fault detection method shown can include the following steps:

[0078] Step 201, obtaining current data generated by a target device, and obtaining frequency domain data according to the current data.

[0079] In the embodiments of the present application, the current data generated by the target device can be current data in a specified time period. Taking the target device as a photovoltaic current device, for example, when the photovoltaic current device generates electricity, the current acquisition device can obtain the photovoltaic current data in a specified time period (for example, five minutes), and determine it as the current data of the target device.

[0080] After the computer device obtains the current data, it needs to perform analog-to-digital conversion on the current data to convert the continuous signal into discrete digital signal format current data. When the computer device obtains the discrete digital signal format current data, it can convert the discrete data signal format current data into frequency domain data through Fourier transform, and at this time the frequency domain data is also in discrete digital signal format.

[0081] Step 202, processing the frequency domain data through a feature extraction module of an arc fault detection model to obtain arc features.

[0082] When the computer device obtains the discrete digital signal format frequency domain data, it can directly process the frequency domain data through the trained arc fault detection model.

[0083] Optionally, the arc fault detection model includes a feature extraction module, which can be composed of various convolutional layers. Each convolutional layer sequentially extracts features from the frequency domain data to obtain arc features.

[0084] Step 203, concatenating the device number corresponding to the target device with the arc features to obtain the to-be-recognized features.

[0085] After the feature extraction module in the arc fault detection model extracts the arc features from the frequency domain data, in order to ensure that the arc fault detection model can consider the device features when processing the data, the number of the target device can be used as artificial features, which are concatenated with the extracted arc features to generate the to-be-recognized features that need to be recognized. At this time, the to-be-recognized features contain both the features extracted from the current data and the device features of the target device that generated the current data, so the to-be-recognized features obtained after fusion are more suitable for the actual production environment compared to the original arc features.

[0086] Step 204: The feature to be identified is processed by the identification module in the arc fault detection model to determine the fault result of the target device.

[0087] At this point, the computer equipment processes the features to be identified through the recognition module in the arc fault detection model to determine the fault result of the target device, i.e., whether the target device has experienced an arc fault. Since this fault result is obtained based on the features to be identified, it comprehensively considers the characteristics of the current and the device characteristics of the target device that generates the current, resulting in higher accuracy in arc fault detection. Furthermore, this arc fault detection model can perform arc fault detection on different devices. When performing arc fault detection on different devices, due to the different added artificial features (i.e., device numbers), it also has excellent recognition capabilities when detecting arc faults on different devices.

[0088] In summary, to achieve arc fault detection, the current data of the target device is first acquired, and frequency domain data is obtained from the current data. The frequency domain data is then processed by the feature extraction model of the arc fault detection model to obtain arc features. At this point, the arc features are concatenated with the device number corresponding to the target device to obtain the features to be identified. When the identification model processes the features to be identified, it takes into account both the current characteristics of the current data and the characteristics of the device that generated the current data. Thus, when the fault detection model identifies whether an arc fault exists, it takes into account the differences between different devices, resulting in high identification accuracy when identifying the current data of different devices.

[0089] Figure 3 This is a flowchart illustrating a method for training an arc fault detection model according to an exemplary embodiment. The method comprises, as follows: Figure 1 The computer equipment in the arc fault detection system shown executes the commands. For example... Figure 3 As shown, the training method for the arc fault detection model may include the following steps:

[0090] Step 301: Obtain sample current data from the sample dataset and obtain sample frequency domain data based on the sample current data; the sample dataset includes current data generated by each sample device.

[0091] In this embodiment of the application, each sample device will generate its own current data, that is, sample current data. At this time, during the training process of a model, the computer device can select any sample current data in the sample dataset and determine the sample device corresponding to the sample current data.

[0092] The computer equipment then performs sampling and time-domain-frequency-domain conversion on the sample current data to obtain sample frequency-domain data.

[0093] Step 302, the sample frequency domain data is processed by a feature extraction module of the arc fault detection model to obtain sample arc features.

[0094] Step 303, the device number corresponding to the sample device generating the sample current data is spliced with the sample arc features to obtain sample to-be-recognized features.

[0095] Step 304, the sample to-be-recognized features are processed by a recognition module in the arc fault detection model to determine a sample recognition result of the sample device generating the sample current data.

[0096] The above steps 302 to 304 are similar to steps 202 to 204 in the embodiment shown in Figure 2 The above steps 302 to 304 are similar to steps 202 to 204 in the embodiment shown in

[0097] Step 305, based on the sample recognition result and a fault result of the sample device generating the sample current data, the arc fault detection model is updated in parameters to obtain a trained arc fault detection model.

[0098] When the arc fault detection model generates a sample recognition result of a sample device, the sample recognition result of the sample device can be compared with an actual fault result saved in the sample data set, and the actual fault result is used to indicate whether the sample device generates the sample current data with an arc fault.

[0099] The arc fault detection model can update each parameter in the arc fault detection model in parameters according to an error between the fault result and the sample recognition result through a back propagation algorithm, so as to complete a training process of the arc fault detection model.

[0100] When the arc fault detection model is trained for several times, the prediction accuracy of the arc fault detection model can be verified through a verification data set, and after verification, the arc fault detection model can be used to detect an arc fault of actually collected current data.

[0101] In a possible implementation, each parameter in the trained arc fault detection model is extracted;

[0102] A first difference between a maximum value and a minimum value of the each parameter is determined;

[0103] A second difference between a maximum value and a minimum value of the target integer range is obtained;

[0104] Based on a ratio of the first difference and the second difference, a quantization interval of the target integer is determined;

[0105] Based on the quantization interval, the parameters are quantized to a target integer range to obtain a quantized arc fault detection model.

[0106] In the actual test process, the neural model may not achieve the original detection effect under the limited resources of the hardware device. The application also supports a model optimization method of a quantized model, which can change the parameters in the neural network model from floating-point numbers to integer numbers. The specific method is as follows: first, find the maximum value Q max and the minimum value Q min of all parameters, determine the value range [I min , I max ] of the integer number to be converted, then calculate T=(Q max -Q min ) / (I max -I min ), S=I max -Q max / T, for any floating-point number Q in the parameter, it can be converted to the desired integer number by the formula I=Q / T+S. By this way, all parameters in the trained model are converted to integer numbers to obtain a quantized model. Since the integer number occupies less hardware memory than the floating-point number, the inference operation speed is faster, so this method can be used on hardware (which can support integer number operation) with limited memory and inference time.

[0107] In summary, in order to realize arc fault detection, the current data of the target device is first obtained, and the frequency domain data is obtained according to the current data, and the arc feature is obtained by processing the frequency domain data through the feature extraction model of the arc fault detection model. At this time, the arc feature and the device number corresponding to the target device are spliced to obtain the to-be-recognized feature. When the recognition model processes the to-be-recognized feature, it considers the current characteristics of the current data and the characteristics of the device generating the current data. At this time, when the fault detection model identifies whether there is an arc fault, it considers the different characteristics between different devices, so that the fault detection model has high recognition accuracy when identifying the current data of different devices.

[0108] Figure 4 is a method flowchart of an arc fault detection method according to an example embodiment. The method is executed by a computer device in the arc fault detection system as shown in Figure 1 . As shown in Figure 4 , the arc fault detection method can include the following steps:

[0109] Step 401, obtaining current data generated by a target device, and obtaining frequency domain data according to the current data.

[0110] In a possible implementation, current data generated by the target device is acquired, and the current data is sampled and normalized to obtain a current data sequence;

[0111] The current data sequence is subjected to fast Fourier transform to obtain a frequency domain sequence;

[0112] The frequency domain sequence is equally divided according to a specified length to obtain each frequency domain subsequence;

[0113] The frequency domain subsequence is spliced into a matrix to obtain frequency domain data of the current data.

[0114] That is, after the current collection system collects current data, the current data can be sampled. At this time, the form of the current data is a sequence (that is, a current data sequence), and assuming that the frequency of the current is A hz, T * A current data points are obtained by sampling the current for T seconds, which form a sequence. The sequence is the original one-dimensional current data. The detection model detects whether the current data with T seconds as a period has failed.

[0115] The current data of a time period is subjected to normalization processing first, so that all data sizes are limited to 0-1; then the time-frequency domain features are extracted. There are many methods for extracting time-frequency domain features, and the present application adopts the fast Fourier transform method to obtain the frequency domain information of one-dimensional current data.

[0116] The frequency domain signal obtained after short-time Fourier transform is still a sequence, and before the sequence is input into the neural network, it is equally divided into M parts, each part having N data, and then they are combined into a M*N matrix. The sequence is changed into a matrix to input the neural network, which can reduce the size of the network model and more effectively extract the features of the current data.

[0117] In step 402, the frequency domain data is processed by a feature extraction module of the arc fault detection model to obtain arc features.

[0118] Optionally, in the embodiment of the present application, the feature extraction module of the arc fault detection model can be a deep convolution module constructed based on a CNN network, which can extract features in the frequency domain data layer by layer to obtain important arc features.

[0119] In step 403, the device number corresponding to the target device is spliced with the arc features to obtain a to-be-identified feature.

[0120] In step 404, the to-be-identified feature is processed through a full connection layer to obtain an arc fault probability corresponding to the target device.

[0121] Please refer to Figure 5Figure 1 shows an implementation logic diagram of a neural network model according to an embodiment of the present application. As shown in Figure 1, in the embodiment of the present application, the original current signal input into the neural network model (i.e., the arc fault detection model) includes two parts, one part is an artificial feature, i.e., a machine model number; the other part is a current feature, which can be converted into a frequency domain feature after processing. Figure 5

[0122] The frequency domain feature can be processed by a deep convolution module to obtain an arc feature, which is then spliced with the machine model number, and the spliced feature (i.e., the to-be-identified feature) is input into a fully connected layer for processing. When the to-be-identified feature is processed by the fully connected layer, the fully connected layer outputs a value P, which is the probability of the neural network initially judging the arc fault.

[0123] Further, after processing the to-be-identified feature by the fully connected layer, the to-be-identified feature can also be processed by a softmax function to output the arc fault probability and the normal probability corresponding to the target device.

[0124] Step 405: When the arc fault probability is greater than or equal to the fault threshold, it is determined that the target device has an arc fault; when the arc fault probability is less than the fault threshold, it is determined that the target device does not have an arc fault.

[0125] Optionally, in the embodiment of the present application, the trained arc fault detection model can process the input current data and generate the corresponding fault probability. At this time, a fault threshold can be set to determine whether the target device has an arc fault. The fault threshold can be determined manually, for example, the fault threshold is determined to be 80%. After processing the to-be-identified feature by the fully connected layer and the softmax function, the arc fault probability corresponding to the target device can be obtained. When the arc fault probability is greater than 80%, it can be considered that the target device has an arc fault; if the arc fault probability is less than 80%, it is determined that the target device does not have an arc fault.

[0126] In one possible implementation, when the arc fault probability is less than the fault threshold, a corresponding weight parameter is generated according to the arc fault probability;

[0127] The weight parameter is accumulated with a historical weight parameter to obtain an updated historical weight parameter; the historical weight parameter is generated by the arc fault detection model for historical current data of the target device within a specified time;

[0128] When the updated historical weight parameter is greater than a weight threshold, it is determined that the target device has an arc fault.

[0129] Please refer to Figure 6 ​Fig. 1 shows an arc detection flowchart to which embodiments of the present application relate. As shown in Fig. 1, in embodiments of the present application, a computer device first collects current data for a time period T, at which time the current data is discrete digital signals obtained by sampling, and then the computer device extracts frequency domain features by FFT (Fast Fourier Transform) after normalizing the discrete digital signals, thereby obtaining a frequency domain signal. Figure 6

[0130] The computer device then segments and concatenates the frequency domain signal into a matrix form and inputs it into a deep learning neural network model (i.e., an arc fault detection model), thereby obtaining a probability P of the neural network initially judging an arc fault. arc In order to improve the accuracy of the entire process, an additional weight determination module is added as an insurance before fault determination: h weights are set according to the size of P, when P is within [P1, P2], a weight W1 is assigned to it, when P is within [P2, P3], a weight W2 is assigned to it, and so on, when the weights of several time periods are accumulated to a certain limit, it is output as an arc fault.

[0131] In summary, in order to realize arc fault detection, current data of a target device is first obtained, and frequency domain data is obtained according to the current data, the arc feature is obtained by processing the frequency domain data through the feature extraction model of the arc fault detection model, at this time the arc feature and the device number corresponding to the target device are spliced to obtain the to-be-recognized feature, at this time the recognition model considers the current characteristics of the current data and the characteristics of the device generating the current data when processing the to-be-recognized feature, at this time the fault detection model considers the different characteristics between different devices when identifying whether there is an arc fault, so that the fault detection model has high recognition accuracy when identifying current data of different devices.

[0132] Please refer to Figure 7 Fig. 2 shows a framework diagram of an arc fault detection device to which embodiments of the present application relate, the device comprising:

[0133] A frequency domain data acquisition module 701 is configured to acquire current data generated by a target device, and obtain frequency domain data according to the current data;

[0134] An arc feature acquisition module 702 is configured to process the frequency domain data through a feature extraction module of an arc fault detection model, and obtain arc features;

[0135] A feature splicing module 703 is configured to splice a device number corresponding to the target device with the arc features, and obtain to-be-recognized features;

[0136] ​The fault determination module 704 is configured to determine the fault result of the target device by processing the to-be-identified feature through an identification module in the arc fault detection model.

[0137] In a possible implementation, the identification module in the fault detection module is a fully connected layer.

[0138] The fault determination module is configured to,

[0139] process the to-be-identified feature through the fully connected layer to obtain an arc fault probability corresponding to the target device.

[0140] When the arc fault probability is greater than or equal to a fault threshold, it is determined that the target device has an arc fault.

[0141] When the arc fault probability is less than the fault threshold, it is determined that the target device does not have an arc fault.

[0142] The fault determination module is further configured to,

[0143] When the arc fault probability is less than the fault threshold, generate a corresponding weight parameter according to the arc fault probability.

[0144] accumulate the weight parameter and a historical weight parameter to obtain an updated historical weight parameter; the historical weight parameter is generated by the arc fault detection model based on historical current data of the target device within a specified time;

[0145] When the updated historical weight parameter is greater than a weight threshold, it is determined that the target device has an arc fault.

[0146] In a possible implementation, the frequency domain data determination module is configured to,

[0147] obtain current data generated by the target device, sample and normalize the current data to obtain a current data sequence;

[0148] perform fast Fourier transform on the current data sequence to obtain a frequency domain sequence;

[0149] equally divide the frequency domain sequence according to a specified length to obtain each frequency domain subsequence;

[0150] splice the frequency domain subsequences into a matrix to obtain frequency domain data of the current data.

[0151] In summary, in order to realize arc fault detection, the current data of the target device is first obtained, and the frequency domain data is obtained according to the current data. The arc feature is obtained by processing the frequency domain data through the feature extraction model of the arc fault detection model. At this time, the arc feature and the device number corresponding to the target device are spliced to obtain the to-be-recognized feature. When the recognition model processes the to-be-recognized feature, the current characteristics of the current data are considered, and the characteristics of the device generating the current data are also considered. When the fault detection model identifies whether there is an arc fault, the distinguishing characteristics between different devices are considered, so that the fault detection model has high recognition accuracy when identifying the current data of different devices.

[0152] Please refer to Figure 8 which shows a framework diagram of an arc fault detection model training device related to an embodiment of the present application. The device comprises:

[0153] The sample data acquisition module 801 is configured to acquire sample current data in a sample data set, and obtain sample frequency domain data according to the sample current data. The sample data set includes current data generated by each sample device.

[0154] The sample feature acquisition module 802 is configured to process the sample frequency domain data through a feature extraction module of an arc fault detection model to obtain sample arc features.

[0155] The sample splicing module 803 is configured to splice the device number corresponding to the sample device generating the sample current data with the sample arc features to obtain sample to-be-recognized features.

[0156] The sample recognition result generation module 804 is configured to process the sample to-be-recognized features through a recognition module in the arc fault detection model to determine a sample recognition result of the sample device generating the sample current data.

[0157] The model training module 805 is configured to update parameters of the arc fault detection model based on the sample recognition result and a fault result of the sample device generating the sample current data, to obtain a trained arc fault detection model.

[0158] In a possible implementation, the model training module is further configured to,

[0159] extract each parameter in the trained arc fault detection model;

[0160] determine a first difference between the maximum value and the minimum value of the each parameter;

[0161] obtain a second difference between the maximum value and the minimum value of the target integer range;

[0162] determine a quantization interval of a target integer number based on a ratio of the first difference value and the second difference value;

[0163] quantize the respective parameters to a target integer range based on the quantization interval to obtain a quantized arc fault detection model.

[0164] In summary, in order to realize arc fault detection, the current data of the target device is first obtained, and the frequency domain data is obtained according to the current data. The arc feature is obtained by processing the frequency domain data through the feature extraction model of the arc fault detection model. At this time, the arc feature and the device number corresponding to the target device are spliced to obtain the to-be-recognized feature. When the recognition model processes the to-be-recognized feature, the current characteristics of the current data and the characteristics of the device generating the current data are considered. When the fault detection model identifies whether there is an arc fault, the distinguishing characteristics between different devices are considered, so that the fault detection model has high recognition accuracy when identifying the current data of different devices.

[0165] Referring to Figure 9 According to an exemplary embodiment of the present application, a computer device schematic diagram is provided, which comprises a memory and a processor. The memory is used to store a computer program. When the computer program is executed by the processor, the above method is implemented.

[0166] The processor can be a central processing unit (CPU). The processor can also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, or combinations thereof.

[0167] The memory is a non-transitory computer readable storage medium, which can be used to store non-transitory software programs, non-transitory computer executable programs and modules, such as program instructions / modules corresponding to the method in the embodiments of the present application. The processor executes various functions and data processing of the processor by running the non-transitory software programs, instructions and modules stored in the memory, that is, implements the method in the above method embodiments.

[0168] The memory can include a program storage area and a data storage area. The program storage area can store an operating system and applications required by at least one function. The data storage area can store data created by the processor and the like. In addition, the memory can include a high-speed random access memory and can further include a non-transitory memory such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid state memory device. In some embodiments, the memory can optionally include a memory that is remotely located with respect to the processor and can be connected to the processor through a network. Examples of the network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0169] In an exemplary embodiment, a computer readable storage medium storing at least one computer program is also provided. The at least one computer program is loaded and executed by a processor to implement all or part of the steps of the above method. For example, the computer readable storage medium can be a Read-Only Memory (ROM), a Random Access Memory (RAM), a Compact Disc Read-Only Memory (CD-ROM), a magnetic tape, a floppy disk, and an optical data storage device, etc.

[0170] In an exemplary embodiment, a computer program product or computer program is also provided. The computer program product or computer program includes computer instructions stored in a computer readable storage medium. A processor of a computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions to cause the computer device to perform all or part of the above method. Figures 2 to 4 all or part of the steps of the method shown in any of the embodiments.

[0171] Other embodiments of the application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. It is intended that the application encompass any and all variations or modifications of the application which fall within the general scope of the application. This application is to be limited only by the claims sufficiently supported by this specification. The specification and examples given are to be considered exemplary of the application, with the true scope and spirit of the application indicated by the following claims.

[0172] It is to be understood that the application is not limited to the precise construction described and as shown in the attached figures, and that various modifications and changes can be made by those skilled in the art without departing from the scope of the application. The scope of the application is to be limited only by the claims appended hereto.

Claims

1. A method for detecting electric arc faults, characterized in that, The method includes: Acquire the current data generated by the target device, and obtain frequency domain data based on the current data; The frequency domain data is processed by the feature extraction module of the arc fault detection model to obtain arc features; The device number corresponding to the target device is concatenated with the arc feature to obtain the feature to be identified; The identification module in the arc fault detection model processes the features to be identified to determine the fault result of the target equipment. The identification module in the arc fault detection model is a fully connected layer; the process of processing the features to be identified by the identification module in the arc fault detection model to determine the fault result of the target equipment includes: The fully connected layer processes the features to be identified to obtain the arc fault probability corresponding to the target device. When the arc fault probability is greater than or equal to the fault threshold, it is determined that the target equipment has experienced an arc fault. When the arc fault probability is less than the fault threshold, it is determined that the target device has not experienced an arc fault; a corresponding weight parameter is generated based on the arc fault probability; the weight parameter is accumulated with the historical weight parameter to obtain the updated historical weight parameter; the historical weight parameter is generated by the arc fault detection model using the historical current data of the target device within a specified time period; when the updated historical weight parameter is greater than the weight threshold, it is determined that the target device has experienced an arc fault.

2. The method according to claim 1, characterized in that, The step of acquiring the current data generated by the target device and obtaining frequency domain data based on the current data includes: Acquire the current data generated by the target device, and sample and normalize the current data to obtain a current data sequence; Perform a Fast Fourier Transform on the current data sequence to obtain a frequency domain sequence; The frequency domain sequence is divided into equal parts of a specified length to obtain each frequency domain subsequence; The various frequency domain subsequences are concatenated into a matrix to obtain the frequency domain data of the current data.

3. A method for training an arc fault detection model, characterized in that, The method includes: Obtain sample current data from the sample dataset, and obtain sample frequency domain data based on the sample current data; the sample dataset includes current data generated by each sample device; The sample frequency domain data is processed by the feature extraction module of the arc fault detection model to obtain the sample arc features; The device number corresponding to the sample device that generated the sample current data is concatenated with the sample arc feature to obtain the sample feature to be identified. The identification module in the arc fault detection model processes the features to be identified in the sample to determine the sample identification result of the sample device that generated the sample current data. Based on the sample identification results and the fault results of the sample device that generated the sample current data, the parameters of the arc fault detection model are updated to obtain the trained arc fault detection model. The step of processing the features to be identified in the sample using the identification module in the arc fault detection model to determine the sample identification result of the sample device that generated the sample current data includes: The sample features to be identified are processed to obtain the arc fault probability corresponding to the sample device. When the arc fault probability is greater than or equal to the fault threshold, it is determined that the sample device has an arc fault. When the arc fault probability is less than the fault threshold, it is determined that the sample device has not experienced an arc fault; a corresponding weight parameter is generated based on the arc fault probability; the weight parameter is accumulated with the historical weight parameter to obtain the updated historical weight parameter; the historical weight parameter is generated by the arc fault detection model using the historical current data of the sample device within a specified time period; when the updated historical weight parameter is greater than the weight threshold, it is determined that the sample device has experienced an arc fault.

4. The arc fault detection model training method according to claim 3, characterized in that, The method further includes: Extract the parameters from the trained arc fault detection model; Determine the first difference between the maximum and minimum values ​​of each parameter; Obtain the second difference between the maximum and minimum values ​​within the target integer range; The quantization interval of the target integer is determined based on the ratio of the first difference to the second difference. Based on the quantization interval, each parameter is quantized to the target integer range to obtain the quantized arc fault detection model.

5. An arc fault detection device, characterized in that, The device includes: The frequency domain data acquisition module is used to acquire the current data generated by the target device and obtain frequency domain data based on the current data. The arc feature acquisition module is used to process the frequency domain data through the feature extraction module of the arc fault detection model to obtain arc features; The feature splicing module is used to splice the device number corresponding to the target device with the arc feature to obtain the feature to be identified; The fault determination module is used to process the features to be identified through the identification module in the arc fault detection model to determine the fault result of the target equipment; The identification module in the arc fault detection model is a fully connected layer; the fault determination module is further used to process the features to be identified through the fully connected layer to obtain the arc fault probability corresponding to the target device; when the arc fault probability is greater than or equal to a fault threshold, it is determined that the target device has experienced an arc fault; when the arc fault probability is less than the fault threshold, it is determined that the target device has not experienced an arc fault; corresponding weight parameters are generated based on the arc fault probability; the weight parameters are accumulated with historical weight parameters to obtain updated historical weight parameters; the historical weight parameters are generated by the arc fault detection model using historical current data of the target device within a specified time period; when the updated historical weight parameters are greater than the weight threshold, it is determined that the target device has experienced an arc fault.

6. A training device for an arc fault detection model, characterized in that, The device includes: The sample data acquisition module is used to acquire sample current data from the sample dataset and obtain sample frequency domain data based on the sample current data; the sample dataset includes current data generated by each sample device. The sample feature acquisition module is used to process the sample frequency domain data through the feature extraction module of the arc fault detection model to obtain the sample arc features. The sample stitching module is used to stitch together the device number corresponding to the sample device that generates the sample current data with the sample arc feature to obtain the sample feature to be identified. The sample identification result generation module is used to process the features to be identified in the sample through the identification module in the arc fault detection model, and determine the sample identification result of the sample device that generates the sample current data. The model training module is used to update the parameters of the arc fault detection model based on the sample identification results and the fault results of the sample device that generated the sample current data, so as to obtain the trained arc fault detection model. The sample identification result generation module is further configured to process the features to be identified in the sample to obtain the arc fault probability corresponding to the sample device; when the arc fault probability is greater than or equal to a fault threshold, it is determined that the sample device has experienced an arc fault; when the arc fault probability is less than the fault threshold, it is determined that the sample device has not experienced an arc fault; generate corresponding weight parameters based on the arc fault probability; accumulate the weight parameters with historical weight parameters to obtain updated historical weight parameters; the historical weight parameters are generated by the arc fault detection model using historical current data of the sample device within a specified time period; when the updated historical weight parameters are greater than the weight threshold, it is determined that the sample device has experienced an arc fault.

7. A computer device, characterized in that, The computer device includes a processor and a memory, the memory storing at least one instruction, the at least one instruction being loaded and executed by the processor to implement the arc fault detection method as described in any one of claims 1 to 2; or the at least one instruction being loaded and executed by the processor to implement the arc fault detection model training method as described in claim 3 or 4.

8. A computer-readable storage medium, characterized in that, The storage medium stores at least one instruction, which is loaded and executed by a processor to implement the arc fault detection method as described in any one of claims 1 to 2; or the at least one instruction is loaded and executed by a processor to implement the arc fault detection model training method as described in claim 3 or 4.

Images