Arc fault detection method, device and equipment and storage medium

Abstract

The embodiment of the invention discloses an arc fault detection method and device, equipment and a storage medium. According to the technical scheme provided by the embodiment of the invention, the bus waveform is obtained by collecting the bus current value of photovoltaic power generation; carrying out segmentation processing on the bus waveform according to a fixed step length to obtain segmented waveforms; inputting the segmented waveforms into a preset auto-encoder for encoding and decoding to obtain a plurality of predicted values corresponding to each segmented waveform; performing error calculation processing on the predicted value and the corresponding bus current value to obtain an error value; the arc fault state is determined according to the comparison result of the error value and the preset threshold value, the problem that the accuracy of arc fault detection is low can be solved, and the accuracy of arc fault detection is improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of arc fault detection, and in particular, to an arc fault detection method, device, equipment and storage medium. Background Art

[0002] As people's environmental awareness continues to increase, renewable energy has become the first choice of energy. Solar energy has gradually become one of the most promising renewable energy sources due to its abundant reserves, pollution-free cleanliness and convenient operability. Therefore, photovoltaic power generation based on solar energy has also been deployed on a large scale in residential electricity and industrial fields.

[0003] Photovoltaic power generation systems are prone to fires. When a fire occurs in a photovoltaic power generation system, it not only damages the property and income of the corresponding equipment, but also causes damage to buildings and personal injuries in serious cases, and even spreads to the surrounding environment, leading to a series of secondary disasters. In photovoltaic power generation systems, arc faults are one of the main causes of fires. Therefore, accurately detecting arc faults is the key to ensuring the normal operation of photovoltaic power generation systems.

[0004] The existing detection of arc faults in photovoltaic power generation systems usually pre-sets a threshold (current threshold, voltage threshold or frequency domain threshold). When the electrical signal in the circuit exceeds the threshold, it is determined that an arc fault has occurred. In photovoltaic power generation systems, due to the presence of a large number of electrical switches and different types of loads, a large amount of noise will be generated; and sunlight will be affected by weather and time, and the power generation corresponding to different weather and time is also different. Therefore, the detection results obtained by performing arc fault detection based on a fixed threshold are not accurate. Therefore, the existing detection method for arc faults in photovoltaic power generation systems has low detection accuracy. Summary of the invention

[0005] The embodiments of the present application provide an arc fault detection method, apparatus, device and storage medium, which can solve the problem of low accuracy of arc fault detection and improve the accuracy of arc fault detection.

[0006] In a first aspect, an embodiment of the present application provides an arc fault detection method, comprising:

[0007] Collect the bus current value of photovoltaic power generation to obtain the bus waveform;

[0008] The bus waveform is segmented according to a fixed step size to obtain a segmented waveform;

[0009] The segmented waveform is input into a preset autoencoder for encoding and decoding to obtain multiple prediction values ​​corresponding to each segmented waveform;

[0010] Perform error calculation on the predicted value and the corresponding bus current value to obtain an error value;

[0011] The arc fault state is determined based on the comparison result between the error value and the preset threshold value.

[0012] In a second aspect, an embodiment of the present application provides an arc fault detection device, comprising:

[0013] A data acquisition unit is used to collect bus current values ​​of photovoltaic power generation to obtain bus waveforms;

[0014] A waveform segmentation unit is used to segment the bus waveform according to a fixed step size to obtain a segmented waveform;

[0015] A coding and decoding unit, used for inputting the segmented waveform into a preset autoencoder for coding and decoding, and obtaining a plurality of prediction values ​​corresponding to each segmented waveform;

[0016] An error calculation unit is used to perform error calculation processing on the predicted value and the corresponding bus current value to obtain an error value;

[0017] The arc fault judgment unit is used to determine the arc fault state according to the comparison result between the error value and the preset threshold value.

[0018] In a third aspect, an embodiment of the present application provides an arc fault detection device, including:

[0019] memory and one or more processors;

[0020] A memory for storing one or more programs;

[0021] When one or more programs are executed by one or more processors, the one or more processors implement the arc fault detection method as described in the first aspect.

[0022] In a fourth aspect, an embodiment of the present application provides a storage medium storing computer executable instructions, which, when executed by a computer processor, are used to execute the arc fault detection method as described in the first aspect.

[0023] The embodiment of the present application obtains a bus waveform by collecting the bus current value of photovoltaic power generation, divides the bus waveform according to the step size to obtain a segmented waveform, inputs the segmented waveform into a preset autoencoder for encoding and decoding, obtains multiple prediction values ​​corresponding to each segmented waveform, performs error calculation processing on the prediction value and the corresponding bus current value to obtain an error value, and determines the arc fault state according to the comparison result between the error value and the preset threshold. By adopting the above technical means, the segmented waveform can be encoded and decoded by the autoencoder to obtain the prediction value, and the arc fault state can be confirmed by comparing the error value between the prediction value and the corresponding bus current value with the preset threshold, thereby avoiding the problem of low accuracy of arc fault detection, and improving the accuracy of arc fault detection compared to the existing method of directly confirming the arc fault state according to the comparison result between the bus current value and the preset threshold. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 is a flow chart of an arc fault detection method provided by an embodiment of the present application;

[0025] Figure 2 This is a schematic diagram of an arc fault bus current provided by an embodiment of the present application;

[0026] Figure 3 is a schematic diagram of a convolutional layer in an autoencoder provided in an embodiment of the present application;

[0027] Figure 4 This is a schematic diagram of an arc generation circuit provided in an embodiment of the present application;

[0028] Figure 5 is a first schematic diagram of a comparison result provided in an embodiment of the present application;

[0029] Figure 6 is a second schematic diagram of a comparison result provided in an embodiment of the present application;

[0030] Figure 7 is a third schematic diagram of a comparison result provided in an embodiment of the present application;

[0031] Figure 8 is a structural schematic diagram of an arc fault detection device provided in an embodiment of the present application;

[0032] Fig. 9 It is a structural schematic diagram of an arc fault detection device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0033] In order to make the purpose, technical scheme and advantages of the present application clearer, the specific embodiments of the present application are further described in detail below in conjunction with the accompanying drawings. It is understood that the specific embodiments described herein are only used to explain the present application, rather than to limit the present application. It should also be noted that, for the convenience of description, only the part related to the present application but not all the contents are shown in the accompanying drawings. Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow chart describes each operation (or step) as a sequential process, many of the operations therein can be implemented in parallel, concurrently or simultaneously. In addition, the order of each operation can be rearranged. The process can be terminated when its operation is completed, but it can also have additional steps not included in the accompanying drawings. The process can correspond to a method, a function, a procedure, a subroutine, a subprogram, etc.

[0034] The arc fault detection method, device, equipment and storage medium provided by the present application are intended to obtain a predicted value by encoding and decoding the segmented waveform through an autoencoder during arc fault detection, and confirm the arc fault state by comparing the error value between the predicted value and the corresponding bus current value and the preset threshold, so as to avoid the problem of low accuracy of arc fault detection and improve the accuracy of arc fault detection. Compared with the traditional arc fault detection method, it usually pre-sets a threshold (current threshold, voltage threshold or frequency domain threshold), and when the electrical signal in the circuit exceeds the threshold, it is judged that an arc fault occurs. In the photovoltaic power generation system, a large amount of noise will be generated based on the existence of a large number of electrical switches and different types of loads; and the sunlight will be affected by the weather time, and the power generation corresponding to different weather times is also different. Therefore, the detection result obtained by performing arc fault detection based on a fixed threshold is not accurate. Therefore, the existing detection method for arc faults in photovoltaic power generation systems has low detection accuracy. Based on this, an arc fault detection method of an embodiment of the present application is provided to solve the problem of low accuracy of existing arc fault detection.

[0035] Figure 1 A flow chart of an arc fault detection method provided in an embodiment of the present application is given. The arc fault detection method provided in this embodiment can be executed by an arc fault detection device, which can be implemented by software and / or hardware. The arc fault detection device can be composed of two or more physical entities, or can be composed of one physical entity. Generally speaking, the arc fault detection device can be a terminal device, such as a computer device.

[0036] The following description is made by taking a computer device as an example of the subject of the arc fault detection method. Figure 1 , the arc fault detection method specifically includes:

[0037] S101, collecting bus current values ​​of photovoltaic power generation to obtain bus waveforms.

[0038] Photovoltaic power generation is the process of converting light energy into electrical energy by using the photovoltaic effect of semiconductor interfaces. After being converted into electrical energy, the corresponding electrical energy can be transmitted to the corresponding load or storage device through the bus. The bus current value in the bus of photovoltaic power generation is collected to obtain the bus waveform.

[0039] Figure 2 This is a schematic diagram of an arc fault bus current provided by the embodiment of the present application, referring to Figure 2 When an arc fault occurs, due to the appearance of electric sparks, the bus current value will appear as follows Figure 2 As shown in the current oscillation, after the current oscillation occurs, the bus waveform will quickly return to normal. In the photovoltaic power generation system, due to the presence of a large number of electrical switches and different types of loads, a large amount of noise will be generated; and the sunlight will be affected by the weather time, and the corresponding power generation at different weather times is also different. Therefore, the corresponding bus current thresholds when arc faults occur at different times will be different. Since the bus current waveform when an arc fault occurs is different from the bus current waveform during normal operation, the bus current value of photovoltaic power generation can be collected to obtain the bus waveform, so that the collected bus waveform can be analyzed later to determine the arc fault status.

[0040] S102, segmenting the bus waveform according to a fixed step size to obtain a segmented waveform.

[0041] Since the bus current waveform is different from that in normal operation when an arc fault occurs, the bus current waveform can be segmented according to a fixed step size to obtain a segmented waveform. By segmenting the bus waveform and performing fault detection on the segmented waveforms, it is easy to determine the arc fault state in each time period, which helps to locate the arc fault time.

[0042] It should be noted that the fixed step size can be set according to actual conditions and is not limited in this embodiment.

[0043] S103, inputting the segmented waveform into a preset autoencoder for encoding and decoding, and obtaining a plurality of prediction values ​​corresponding to each segmented waveform.

[0044] Before arc fault detection, an autoencoder (model) is pre-built, two one-dimensional convolutional layers are used as encoders, and two one-dimensional transposed convolutional layers are used as decoders. The corresponding nonlinear transformation processing and the corresponding random inactivation processing are performed through the activation function to complete the encoding and decoding processing of the data. The autoencoder is trained, and the bus current value of the normal operation of the bus of the photovoltaic power generation to be detected is collected to obtain the bus waveform. The collected normal operation bus waveform is segmented according to a fixed step size to obtain a segmented waveform. The segmented waveform is input into the initial autoencoder for training processing, and a set of predicted values ​​corresponding to each segmented waveform is obtained through the preset encoding and decoding processing in the autoencoder. It should be noted that the waveform of this group of predicted values ​​has a trend regularity with the corresponding segmented waveform, but the values ​​are not the same. An error threshold is obtained by comparing a large number of predicted values ​​with the true value (that is, the corresponding bus current value). In the subsequent detection process, the error value between the predicted value and the true value within the error threshold range is considered to be normal operation.

[0045] Four convolutional layers are defined in the autoencoder, namely the first convolutional layer, the second convolutional layer, the third convolutional layer and the fourth convolutional layer, wherein the first convolutional layer and the second convolutional layer are encoding processes, and the third convolutional layer and the fourth convolutional layer are decoding processes. When training the autoencoder, the segmented waveform is input into the autoencoder, and in the autoencoder, the segmented waveform is input into the first convolutional layer, and in the first convolutional layer, the segmented waveform is subjected to the first convolutional process to obtain the first sub-data. The first sub-data is subjected to nonlinear change processing through an activation function to obtain the second sub-data. The neurons corresponding to the second sub-data are subjected to random inactivation processing to obtain the first data. The random inactivation process can reduce the dependency between neurons and prevent overfitting. The first data is output through the first branch channel. The first data is input into the second convolutional layer, and in the second convolutional layer, the first data is subjected to the second convolutional process to obtain the third sub-data; the third sub-data is subjected to nonlinear transformation processing through an activation function to obtain the second data; the second data is output through the second branch channel, and the number of the second branch channels is half of the number of the first branch channels. Input the second data into the third convolution layer; in the third convolution layer, perform the first convolution transposition process on the second data to obtain the fourth sub-data; perform nonlinear transformation process on the fourth sub-data through the activation function to obtain the fourth sub-data; perform random inactivation process on the neurons corresponding to the fourth sub-data to obtain the third data; output the third data through the third branch channel, and the number of the third branch channels is twice the number of the second branch channels. Input the third data into the fourth convolution layer to perform the second convolution transposition process to obtain the predicted value, and output the corresponding predicted value through the fourth branch channel, and the number of the fourth branch channels is one. Through the above-mentioned convolution process and convolution transposition process, a set of predicted values ​​corresponding to each segmented waveform can be obtained.

[0046] In one embodiment, the activation function is a ReLU activation function.

[0047] The autoencoder is trained using the normal operating bus current waveform. After the training is completed, the training of the autoencoder is optimized using an adaptive optimizer (such as an adam optimizer). The learning rate is adjusted based on historical gradient information. A higher learning rate can be used in the early stages of training to achieve a fast convergence model. A smaller learning rate can be used in the later stages of training to more accurately find the minimum value of the loss function.

[0048] When performing arc fault detection, after obtaining the bus current waveform and the corresponding segmented waveform, the segmented waveform is input into the trained autoencoder, and the segmented waveform is encoded and decoded in the autoencoder to obtain the corresponding preset value, and each predicted value is output through the corresponding branch channel. The corresponding predicted value is obtained through the encoding and decoding processing of the autoencoder, so that the waveform of the predicted value corresponding to each segmented waveform has the same trend law as the segmented waveform itself. Based on the error value between the predicted value and the corresponding bus current value, the error of the corresponding segmented waveform can be obtained. When the error is large, it is judged that there is an arc fault, and when the error is small, it is judged to be normal operation.

[0049] In one embodiment, Figure 3 is a schematic diagram of a convolutional layer in an autoencoder provided in an embodiment of the present application, referring to Figure 3 , the segmented waveform is input into the autoencoder. In the autoencoder, the segmented waveform is input into the first convolution layer. In the first convolution layer, the number of input channels is 1, the number of output channels is 32, the convolution kernel size is 7, and the step size is 2. In the first convolution layer, the segmented waveform is subjected to the first convolution process to obtain the first sub-data. The first sub-data is subjected to nonlinear change processing through the ReLU activation function to obtain the second sub-data. For example, according to Output=max(0,W T X+B) performs nonlinear change processing and outputs the second sub-data, where X represents the input vector, W represents the encoder weight matrix, and W TRepresents the transpose of the encoder weight matrix, B represents the bias vector, max represents the maximum value, and Output represents the output value (i.e., the second sub-data). The neurons corresponding to the second sub-data are randomly deactivated by 20% to obtain the first data. It should be noted that in the training process and the actual detection process, the output of each neuron has a 20% probability of being set to zero. The random deactivation process can reduce the dependency between neurons and prevent overfitting. The first data is output through 32 output channels. The 32-channel first data is input into the second convolution layer. In the second convolution layer, the number of input channels is set to 32 channels, the number of output channel data is set to 16 channels, the convolution kernel size is 7, and the step size is 2. In the second convolution layer, the first data is subjected to the second convolution process to obtain the third sub-data. The third sub-data is subjected to nonlinear transformation processing by the ReLU activation function to obtain the second data. The second data is output through 16 output channels. The 16-channel second data is input into the third convolution layer. In the third convolution layer, the number of input channels is set to 16 channels, the number of output channels is set to 32 channels, the convolution kernel size is 7, and the step size is 2. In the third convolution layer, the second data is subjected to the first convolution transposition process to obtain the fourth sub-data. The fourth sub-data is subjected to nonlinear transformation processing through the ReLU activation function to obtain the fourth sub-data. The neurons corresponding to the fourth sub-data are subjected to 20% random inactivation processing to obtain the third data. The third data is output through 32 output channels. The 32-channel third data are input into the fourth convolution layer. In the fourth convolution layer, the number of input channels is set to 32, the number of output channels is set to 1, the convolution kernel size is 7, and the step size is 2. In the fourth convolution layer, the third data is subjected to the second convolution transposition process to obtain the predicted value, and the corresponding predicted value is output through 1 output channel. Through the encoding and decoding processing of the above-mentioned autoencoder, a set of predicted values ​​corresponding to each segmented waveform can be obtained.

[0050] As mentioned above, by encoding and decoding the segmented waveform through the autoencoder, a set of prediction values ​​corresponding to each segmented waveform can be obtained, which is convenient for subsequent judgment of the arc fault state according to the error between the predicted value and the true value, and the arc fault can be detected by the waveform change error. Compared with the existing method of directly detecting arc faults based on the comparison between the bus electrical signal and the electrical signal threshold (current threshold, voltage threshold or frequency domain threshold), this embodiment has higher accuracy and finer precision.

[0051] S104: Perform error calculation on the predicted value and the corresponding bus current value to obtain an error value.

[0052] Through the encoding and decoding processing of the above-mentioned autoencoder, a group of predicted values ​​corresponding to each segmented waveform can be obtained. According to the error calculation processing between the predicted value and the corresponding bus current value, the error value between the predicted value and the true value (i.e., the corresponding bus current value) can be obtained. Based on the training of the autoencoder, a large number of segmented waveforms corresponding to the bus current values ​​of normal operation are used for training processing to obtain a large number of predicted values; according to the comparison results of a large number of predicted values ​​of normal operation and the true value (i.e., the corresponding bus current value), a preset threshold (i.e., the error threshold) is obtained by statistics. During the detection process, the error value between the predicted value and the true value within the preset threshold range is regarded as normal operation. When an arc fault occurs, the error value between the predicted value and the true value (i.e., the corresponding bus current value) in the corresponding segmented waveform will exceed the corresponding preset error threshold. Therefore, the arc fault state can be confirmed by the error value obtained by performing error calculation processing on each predicted value. Compared with the existing method of directly confirming the arc fault state based on the comparison result of the bus electrical signal and the electrical signal threshold (current threshold, voltage threshold or frequency domain threshold), this embodiment can reduce the influence of electrical signal fluctuations caused by noise or different power generation on the judgment threshold of arc fault detection, and improves the accuracy of arc fault detection by using the error value between the predicted value and the true value to judge the corresponding arc fault state.

[0053] In one embodiment, according to the first formula The error calculation process is performed on the predicted value and the corresponding bus current value to obtain the error value, where MSE represents the error value, y pred Represents the predicted value, y true represents the bus current value, and n represents the number of samples.

[0054] S105 . Determine an arc fault state according to a comparison result between the error value and a preset threshold value.

[0055] According to the comparison result between the error value and the preset threshold value (i.e., the preset error threshold value), the fault state of the arc is determined. If the error value is greater than the preset threshold value, it is judged that an arc fault has occurred. If the error value is less than or equal to the preset threshold value, it is judged that the operation is normal. By using the error value between the predicted value and the true value to judge the corresponding arc fault state, the accuracy of arc fault detection is improved.

[0056] In one embodiment, Figure 4 This is a schematic diagram of an arc generating circuit provided in an embodiment of the present application, referring to Figure 4, the arc generating circuit includes a photovoltaic DC power supply 10, an arc generator 20, a Hall sampling subcircuit 30, a control panel 40, a sampling power supply 50, an inverter 60, a battery power supply 70 and a power grid 80. The photovoltaic DC power supply 10 is connected to the first end of the arc generator 20 and the first end of the inverter 60, and the second end of the arc generator 20 is connected to the first end of the Hall sampling subcircuit 30. The second end of the Hall sampling subcircuit 30 is connected to the first end of the control panel 40, the third end of the Hall sampling subcircuit 30 is connected to the sampling power supply 50, and the fourth end of the Hall sampling subcircuit 30 is connected to the second end of the inverter 60. The second end of the control panel 40 is connected to the sampling power supply 50. The sampling power supply 50 is used to provide a first voltage (for example, 3.3V) for the control panel 40, and is used to provide a second voltage (for example, 12V) for the Hall sampling subcircuit 30. The third end of the inverter 60 is connected to the battery power supply 70, and the fourth end of the inverter 60 is connected to the power grid 80, which is used to provide electric energy to the user. The arc generator 20 is used to simulate the DC arc in the photovoltaic bus, and the inverter 60, the battery power supply 70 and the power grid 80 are used as interference items to verify the stability of the arc fault detection. The control panel 40 controls the Hall sampling subcircuit 30 to perform data sampling, collect the bus current value, and after the collection is completed, the collected bus current value is transmitted to the control panel 40. The control panel 40 performs the corresponding arc fault detection processing through the above arc fault detection method and determines the arc fault state.

[0057] When performing arc fault detection through the arc generating circuit, when the arc generator 20 is set to the closed state, no arc is generated, and the circuit is in a normal operating state. When the circuit is in a normal operating state, the current range is 5-10A, the maximum power voltage is 300V, the open circuit voltage of the battery power supply 70 is 349.5V, the short circuit current is 5.28-10.5A, and the power of the photovoltaic DC power supply 10 is 15000W-30000W. The data (i.e., the bus current value) is collected through the Hall sampling subcircuit 30, and the collected data is transmitted to the control panel 40, and the control panel 40 stores the recorded data through the register. Based on the register recording data has certain limitations, the maximum number of current points that can be recorded at one time is 6250, and the sampling frequency is 2kHz, so the sampling time can be obtained as 3.12s.

[0058] The autoencoder is trained using a normal operating data set in a photovoltaic generating circuit. It should be noted that the data set needs to be preprocessed during training. For example, the training data set contains 25,000 normal operating points and 8,476 normal segment data before the arc occurs. These data are initialized and the obtained mean and variance are saved for data standardization. After standardization, the original data in the data set needs to be processed in time series, and each time step data is used as a sample. It should be noted that the value of the specific time step can be determined according to the actual data situation. For example, the number of time steps in this embodiment is set to 288.

[0059] In order to detect an arc fault, it is necessary to simulate the generation of an arc. Wait for the photovoltaic bus current to rise to the specified current value. After the bus current stabilizes, according to the preset standard regulations, start to pull the two electrodes of the arc generator 20 apart at a speed of 2.5mm / s until the gap value is 0.8mm. The arc luminescence phenomenon is observed accompanied by the sound of the current. During this process, the bus current data during the arc fault is collected through the Hall sampling subcircuit 30 and saved in the register of the control panel 40. The saved bus current data during the arc fault is processed and segmented by the control panel 40, and the segmented waveform obtained by the segmentation process is input into the autoencoder for encoding and decoding to obtain the corresponding predicted value. The error calculation process is performed based on the obtained predicted value and the corresponding true value (that is, the corresponding bus current value) to obtain the corresponding error value. Refer to Figure 5-7 It can be seen that when an arc fault occurs, the corresponding error value is significantly different from the true value, which is much larger than the preset threshold.

[0060] As mentioned above, by inferring the arc fault state through the normal operation data of the circuit, the collection of circuit waveform types can be reduced. Normal operation data is easier to collect than arc fault data, thereby reducing the complexity of data collection, thereby improving the efficiency of data collection and model training. In addition, the arc fault state is judged by the deep learning algorithm of the autoencoder to ensure that most of the characteristics of the bus current are used in the judgment, thereby improving the accuracy of arc fault detection. In addition, by changing the data content of the training set, the normal working state of different systems can be detected, adapting to different working environments, and improving the compatibility and adaptability of arc fault detection.

[0061] As described above, the training set can be continuously updated, and continuous training and model optimization can be performed through more normal status data in actual operation to continuously improve the accuracy of detection and be iterative.

[0062] In the above, the bus waveform is obtained by collecting the bus current value of photovoltaic power generation, the bus waveform is segmented according to the step size to obtain a segmented waveform, the segmented waveform is input into the preset autoencoder for encoding and decoding, and multiple prediction values ​​corresponding to each segmented waveform are obtained. The prediction value and the corresponding bus current value are subjected to error calculation processing to obtain the error value, and the arc fault state is determined according to the comparison result between the error value and the preset threshold. By adopting the above technical means, the segmented waveform can be encoded and decoded by the autoencoder to obtain the prediction value, and the arc fault state can be confirmed by comparing the error value between the prediction value and the corresponding bus current value with the preset threshold, thereby avoiding the problem of low accuracy of arc fault detection. Compared with the existing method of directly confirming the arc fault state according to the comparison result between the bus current value and the preset threshold, this embodiment can reduce the influence of the electric signal fluctuation caused by noise or different power generation on the judgment threshold of arc fault detection, and the corresponding arc fault state is judged by using the error value between the predicted value and the true value, thereby improving the accuracy of arc fault detection.

[0063] Based on the above embodiments, Figure 8 This is a schematic diagram of the structure of an arc fault detection device provided in an embodiment of the present application. Figure 8 The arc fault detection device provided in this embodiment specifically includes: a data acquisition unit 21, a waveform segmentation unit 22, a coding and decoding unit 23, an error calculation unit 24 and an arc fault judgment unit 25.

[0064] The data acquisition unit 21 is used to collect the bus current value of photovoltaic power generation to obtain the bus waveform;

[0065] The waveform segmentation unit 22 is used to segment the bus waveform according to a fixed step size to obtain a segmented waveform;

[0066] The encoding and decoding unit 23 is used to input the segmented waveform into a preset autoencoder for encoding and decoding processing to obtain multiple prediction values ​​corresponding to each segmented waveform;

[0067] The error calculation unit 24 is used to perform error calculation processing on the predicted value and the corresponding bus current value to obtain an error value;

[0068] The arc fault judgment unit 25 is used to determine the arc fault state according to the comparison result between the error value and the preset threshold value.

[0069] In one embodiment, the encoding and decoding unit 23 is also used to input the segmented waveform into a preset autoencoder, perform convolution transformation on the segmented waveform in the preset autoencoder to obtain corresponding prediction values, and output each prediction value through a corresponding branch channel.

[0070] In one embodiment, the encoding and decoding unit 23 includes an input module, a first convolution module, a second convolution module, a third convolution module and a fourth convolution module;

[0071] An input module, used for inputting the segmented waveform into a preset autoencoder;

[0072] A first convolution module is used to input the segmented waveform into a first convolution layer in a preset autoencoder to perform a first convolution transformation process to obtain first data, and output the first data through a first branch channel;

[0073] A second convolution module, used for inputting the first data into the second convolution layer to perform a second convolution transformation process to obtain second data, and outputting the second data through a second branch channel, where the number of the second branch channels is half the number of the first branch channels;

[0074] a third convolution module, configured to input the second data into a third convolution layer to perform the first convolution transposition process to obtain third data, and output the third data through a third branch channel, wherein the number of the third branch channels is twice the number of the second branch channels;

[0075] The fourth convolution module is used to input the third data into the fourth convolution layer to perform the second convolution transposition process to obtain a prediction value, and output the corresponding prediction value through a fourth branch channel, and the number of the fourth branch channel is one.

[0076] In one embodiment, the first convolution module includes a first input submodule, a first convolution submodule, a first activation submodule, a first random deactivation submodule, and a first output submodule;

[0077] A first input submodule, used for inputting the segmented waveform into a first convolutional layer in a preset autoencoder;

[0078] A first convolution submodule, used for performing a first convolution process on the segmented waveform in a first convolution layer to obtain first sub-data;

[0079] A first activation submodule, used for performing nonlinear change processing on the first sub-data through an activation function to obtain second sub-data;

[0080] A first random deactivation submodule, used for performing random deactivation processing on neurons corresponding to the second sub-data to obtain first data;

[0081] The first output submodule is used to output first data through a first branch channel.

[0082] In one embodiment, the second convolution module includes a second input submodule, a second convolution submodule, a second activation submodule, and a second output submodule;

[0083] A second input submodule, used for inputting the first data into the second convolutional layer;

[0084] A second convolution submodule, used for performing a second convolution process on the first data in a second convolution layer to obtain third sub-data;

[0085] A second activation submodule, used for performing nonlinear transformation processing on the third sub-data through an activation function to obtain second data;

[0086] The second output submodule is used to output second data through a second branch channel.

[0087] In one embodiment, the third convolution module includes a third input submodule, a third convolution submodule, a third activation submodule, a second random deactivation submodule and a third output submodule;

[0088] A third input submodule, used for inputting the second data into the third convolutional layer;

[0089] A third convolution submodule, used for performing a first convolution transposition process on the second data in a third convolution layer to obtain fourth sub-data;

[0090] A third activation submodule, used for performing nonlinear transformation processing on the fourth sub-data through an activation function to obtain the fourth sub-data;

[0091] A second random deactivation submodule, used for performing random deactivation processing on neurons corresponding to the fourth sub-data to obtain third data;

[0092] The third output submodule is used to output third data through a third branch channel.

[0093] In one embodiment, the error calculation unit 24 is further configured to calculate the error according to the first formula The error calculation process is performed on the predicted value and the corresponding bus current value to obtain the error value, where MSE represents the error value, y pred Represents the predicted value, y true represents the bus current value, and n represents the number of samples.

[0094] In the above, the bus waveform is obtained by collecting the bus current value of photovoltaic power generation, the bus waveform is segmented according to the step size to obtain a segmented waveform, the segmented waveform is input into the preset autoencoder for encoding and decoding, and multiple prediction values ​​corresponding to each segmented waveform are obtained. The prediction value and the corresponding bus current value are subjected to error calculation processing to obtain the error value, and the arc fault state is determined according to the comparison result between the error value and the preset threshold. By adopting the above technical means, the segmented waveform can be encoded and decoded by the autoencoder to obtain the prediction value, and the arc fault state can be confirmed by comparing the error value between the prediction value and the corresponding bus current value with the preset threshold, thereby avoiding the problem of low accuracy of arc fault detection. Compared with the existing method of directly confirming the arc fault state according to the comparison result between the bus current value and the preset threshold, this embodiment can reduce the influence of the electric signal fluctuation caused by noise or different power generation on the judgment threshold of arc fault detection, and the corresponding arc fault state is judged by using the error value between the predicted value and the true value, thereby improving the accuracy of arc fault detection.

[0095] The arc fault detection device provided in the embodiment of the present application can be used to execute the arc fault detection method provided in the above embodiment, and has corresponding functions and beneficial effects.

[0096] The present application embodiment provides an arc fault detection device, referring to Fig. 9 The arc fault detection device includes: a processor 31, a memory 32, a communication module 33, an input device 34 and an output device 35. The number of processors in the arc fault detection device can be one or more, and the number of memories in the arc fault detection device can be one or more. The processor, memory, communication module, input device and output device of the arc fault detection device can be connected through a bus or other methods.

[0097] The memory 32, as a computer-readable storage medium, can be used to store software programs, computer executable programs and modules, such as program instructions / modules corresponding to the arc fault detection method of any embodiment of the present application (for example, the data acquisition unit, waveform segmentation unit, encoding and decoding unit, error calculation unit and arc fault judgment unit in the arc fault detection device). The memory may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and an application required for at least one function; the data storage area may store data created according to the use of the device, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device or other non-volatile solid-state storage device. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network and a combination thereof.

[0098] The communication module 33 is used for data transmission.

[0099] The processor 31 executes various functional applications and data processing of the device by running software programs, instructions and modules stored in the memory, that is, implements the above-mentioned arc fault detection method.

[0100] The input device 34 may be used to receive input digital or character information and generate key signal input related to user settings and function control of the device. The output device 35 may include a display device such as a display screen.

[0101] The arc fault detection device provided above can be used to execute the arc fault detection method provided in the above embodiment, and has corresponding functions and beneficial effects.

[0102] The embodiment of the present application also provides a storage medium storing computer executable instructions. When the computer executable instructions are executed by a computer processor, they are used to execute an arc fault detection method. The arc fault detection method includes: collecting the bus current value of photovoltaic power generation to obtain a bus waveform; segmenting the bus waveform according to a fixed step size to obtain a segmented waveform; inputting the segmented waveform into a preset autoencoder for encoding and decoding to obtain multiple predicted values ​​corresponding to each segmented waveform; performing error calculation on the predicted value and the corresponding bus current value to obtain an error value; and determining the arc fault state according to the comparison result of the error value and a preset threshold.

[0103] Storage medium - any of various types of memory devices or storage devices. The term "storage medium" is intended to include: installation media, such as CD-ROM, floppy disk or tape device; computer system memory or random access memory, such as DRAM, DDR RAM, SRAM, EDO RAM, Rambus RAM, etc.; non-volatile memory, such as flash memory, magnetic media (such as hard disk or optical storage); registers or other similar types of memory elements, etc. Storage media may also include other types of memory or combinations thereof. In addition, the storage medium may be located in the first computer system in which the program is executed, or may be located in a different second computer system, which is connected to the first computer system via a network (such as the Internet). The second computer system can provide program instructions to the first computer for execution. The term "storage medium" may include two or more storage media residing in different locations (for example, in different computer systems connected by a network). The storage medium may store program instructions (for example, embodied as a computer program) that can be executed by one or more processors.

[0104] Of course, the computer executable instructions of a storage medium storing computer executable instructions provided in an embodiment of the present application are not limited to the above arc fault detection method, and can also execute related operations in the arc fault detection method provided in any embodiment of the present application.

[0105] The arc fault detection apparatus, storage medium and arc fault detection device provided in the above embodiments can execute the arc fault detection method provided in any embodiment of the present application. For technical details not described in detail in the above embodiments, please refer to the arc fault detection method provided in any embodiment of the present application.

[0106] The above are only preferred embodiments of the present application and the technical principles used. The present application is not limited to the specific embodiments herein, and various obvious changes, readjustments and substitutions that can be made by those skilled in the art will not deviate from the protection scope of the present application. Therefore, although the present application is described in more detail through the above embodiments, the present application is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present application, and the scope of the present application is determined by the scope of the claims.

Claims

1. A method for detecting an arc fault, characterized in that: include: Collect the bus current value of photovoltaic power generation to obtain the bus waveform; The bus waveform is segmented according to a fixed step size to obtain a segmented waveform; Inputting the segmented waveform into a preset autoencoder for encoding and decoding to obtain a plurality of prediction values ​​corresponding to each segmented waveform; Perform error calculation processing on the predicted value and the corresponding bus current value to obtain an error value; The arc fault state is determined according to a comparison result between the error value and a preset threshold value.

2. The method according to claim 1, characterized in that: The step of inputting the segmented waveform into a preset autoencoder for encoding and decoding to obtain a plurality of prediction values ​​corresponding to each segmented waveform includes: The segmented waveform is input into a preset autoencoder, the segmented waveform is subjected to convolution transformation in the preset autoencoder to obtain corresponding prediction values, and each prediction value is output through a corresponding branch channel.

3. The method according to claim 2, characterized in that The step of inputting the segmented waveform into a preset autoencoder, performing convolution transformation on the segmented waveform in the preset autoencoder to obtain corresponding prediction values, and outputting each prediction value through a corresponding branch channel includes: Inputting the segmented waveform into a preset autoencoder; In the preset autoencoder, the segmented waveform is input into a first convolution layer to perform a first convolution transformation process to obtain first data, and the first data is output through a first branch channel; Inputting the first data into a second convolution layer to perform a second convolution transformation process to obtain second data, and outputting the second data through a second branch channel, wherein the number of the second branch channels is half the number of the first branch channels; Input the second data into the third convolution layer to perform the first convolution transposition process to obtain third data, and output the third data through a third branch channel, where the number of the third branch channels is twice the number of the second branch channels; The third data is input into the fourth convolution layer for performing a second convolution transposition process to obtain the prediction value, and the corresponding prediction value is output through a fourth branch channel, where the number of the fourth branch channel is one.

4. The method according to claim 3, characterized in that In the preset autoencoder, the segmented waveform is input into the first convolution layer to perform a first convolution transformation process to obtain first data, and the first data is output through the first branch channel, including: In the preset autoencoder, inputting the segmented waveform into a first convolutional layer; In the first convolution layer, a first convolution process is performed on the segmented waveform to obtain first sub-data; Performing nonlinear change processing on the first sub-data through an activation function to obtain second sub-data; Performing random inactivation processing on neurons corresponding to the second sub-data to obtain first data; The first data is output through a first branch channel.

5. The method according to claim 3, characterized in that: The step of inputting the first data into a second convolution layer for performing a second convolution transformation process to obtain second data, and outputting the second data through a second branch channel includes: Inputting the first data into a second convolutional layer; In the second convolution layer, performing a second convolution process on the first data to obtain third sub-data; Performing nonlinear transformation processing on the third sub-data through an activation function to obtain second data; The second data is output through a second branch channel.

6. The method according to claim 3, characterized in that The step of inputting the second data into the third convolution layer to perform the first convolution transposition process to obtain third data, and outputting the third data through the third branch channel includes: Inputting the second data into a third convolutional layer; In the third convolution layer, performing a first convolution transposition process on the second data to obtain fourth sub-data; Performing nonlinear transformation processing on the fourth sub-data through an activation function to obtain fourth sub-data; Performing random inactivation processing on neurons corresponding to the fourth sub-data to obtain third data; The third data is output through a third branch channel.

7. The method according to claim 1, characterized in that The performing error calculation processing on the predicted value and the corresponding bus current value to obtain the error value includes: According to the first formula The predicted value and the corresponding bus current value are subjected to error calculation processing to obtain an error value, where MSE represents the error value, and y pred represents the predicted value, y true represents the bus current value, and n represents the number of samples.

8. An arc fault detection device, characterized in that: include: A data acquisition unit is used to collect bus current values ​​of photovoltaic power generation to obtain bus waveforms; A waveform segmentation unit, used for segmenting the bus waveform according to a fixed step size to obtain a segmented waveform; A coding and decoding unit, used for inputting the segmented waveform into a preset autoencoder for coding and decoding, and obtaining a plurality of prediction values ​​corresponding to each segmented waveform; An error calculation unit, used for performing error calculation processing on the predicted value and the corresponding bus current value to obtain an error value; The arc fault judgment unit is used to determine the arc fault state according to the comparison result between the error value and a preset threshold value.

9. An arc fault detection device, characterized in that: include: memory and one or more processors; The memory is used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A storage medium storing computer executable instructions, characterized in that: The computer executable instructions are used to perform the method according to any one of claims 1 to 7 when executed by a processor.

Images