Method, device and electronic equipment for generating photovoltaic support fault type diagnosis model

Abstract

The application provides a photovoltaic support fault type diagnosis model generation method and device and electronic equipment, the method comprises the following steps: adopting a convolution layer to extract features of target data to obtain a spatial feature map; introducing a self-attention mechanism on the spatial feature map, calculating the weight of each time step in the spatial feature map to obtain a weighted feature map; adopting a variational autoencoder to process the spatial feature map and the weighted feature map to obtain a low-dimensional hidden variable; inputting the low-dimensional hidden variable into a bidirectional gated recurrent unit to output the time sequence characteristics of the target data, and inputting the time sequence characteristics into a fully connected layer and a softmax layer to obtain a photovoltaic support fault type diagnosis model. Based on the above method, the photovoltaic system fault type can be effectively identified and diagnosed, and the accuracy of fault detection and diagnosis is improved.

Description

Technical Field

[0001] This application mainly relates to the field of photovoltaic technology, and in particular to the method, apparatus and electronic equipment for generating photovoltaic support failure type diagnosis models. Background Technology

[0002] Photovoltaic power generation, as one of the most cost-effective renewable clean energy sources, is developing at an increasingly rapid pace. People use photovoltaic modules to convert solar energy into electricity, which is then further converted into household electricity using inverters and energy storage. However, photovoltaic systems are complex in structure, with numerous components, making them susceptible to various factors that can lead to decreased power generation efficiency or even serious malfunctions such as fires. Traditional model-based fault detection methods are computationally intensive and inefficient, unable to effectively detect the existence and specific types of faults in photovoltaic systems. This hinders timely fault detection by staff, resulting in reduced power plant efficiency. Summary of the Invention

[0003] The purpose of this application is to provide a method, apparatus and electronic device for diagnosing photovoltaic bracket fault types, so as to solve the problem that existing photovoltaic fault detection methods cannot effectively detect fault types.

[0004] Firstly, this application provides a method for generating a photovoltaic support fault type diagnostic model, the method comprising:

[0005] Convolutional layers are used to extract features from the target data to obtain a spatial feature map. The target data is the preprocessed data of the dataset generated by the photovoltaic model and the monitored photovoltaic power station data.

[0006] A self-attention mechanism is introduced into the spatial feature map to calculate the weight of each time step in the spatial feature map, thereby obtaining a weighted feature map;

[0007] A variational autoencoder is used to process the spatial feature map and the weighted feature map to obtain low-dimensional latent variables;

[0008] The low-dimensional latent variables are input into a bidirectional gated recurrent unit, which outputs the temporal features of the target data. The temporal features are then input into a fully connected layer and a softmax layer to obtain a photovoltaic bracket fault type diagnosis model.

[0009] By employing a variational autoencoder combined with convolutional layers, a self-attention mechanism, and a bidirectional gate control loop unit for fault diagnosis, the output characteristics of the photovoltaic system can be extracted more comprehensively, the specific fault type of the photovoltaic system can be diagnosed, and the accuracy of fault diagnosis can be improved.

[0010] Optionally, before using a convolutional neural network to extract features from the target data and obtain a spatial feature map, the method further includes: constructing the photovoltaic model to simulate the failure conditions of the photovoltaic array under different light and temperature conditions, and obtaining a dataset, wherein the dataset includes: output current, output voltage, output power, solar irradiance, and temperature; and performing denoising and normalization processing on the dataset and the monitored photovoltaic power station data to complete the preprocessing, thereby obtaining the target data.

[0011] Optionally, the step of using a convolutional layer to extract features from the target data to obtain a spatial feature map includes: using the convolutional layer to perform a convolution operation on the target data with a set of convolutional kernels, followed by a pooling operation to obtain the spatial feature map, wherein the formula for the convolutional layer is as follows:

[0012] h l =ReLU(Conv(h) l-1 W l ,b l ))

[0013] Where ReLU is the ReLU (Rectified Linear Unit) activation function, Conv is the convolutional layer, and h l W represents the output feature map of the l-th layer. l and b l These are the weights and biases of the l-th convolutional kernel, respectively.

[0014] Optionally, the convolutional layer is a Conv1D layer, and the convolution formula of the Conv1D layer is:

[0015]

[0016] Where x(t) is the sequence data of the input target data, t+i-1 represents the value of the input sequence data corresponding to the i-th element of the convolution kernel when the convolution kernel slides to the position at time t, W(i) is the convolution kernel, k is the size of the convolution kernel, and b l It is the bias of the l-th convolutional kernel.

[0017] Optionally, the step of introducing a self-attention mechanism on the spatial feature map and calculating the weight of each time step in the spatial feature map to obtain a weighted feature map includes: introducing the self-attention mechanism on the spatial feature map, obtaining the attention weight of each input feature by calculating the weighted sum between the query vector, key vector and value vector of each input feature; and obtaining the weighted feature map based on the attention weight of each input feature.

[0018] Optionally, the formula for the self-attention mechanism is:

[0019]

[0020] Where Q = W Q X is the query vector, W Q This is the weight matrix for the query, X is the input feature, and K = W. K X is the key vector, W K It is the weight matrix of the keys, V = W V X is a value vector, W V It is the weight matrix of values, d k is the dimension of the key vector, and softmax is the normalized exponential function used to calculate the attention weight at each position.

[0021] Optionally, the formula for calculating the weighted feature map is:

[0022]

[0023] Where, α i V is the attention weight of the i-th element. i It is the value vector of the i-th element, and Attention(X) represents the weighted summation of the attention weights for each input feature.

[0024] Optionally, the step of processing the spatial feature map and the weighted feature map using a variational autoencoder to obtain low-dimensional latent variables includes: mapping the spatial feature map and the weighted feature map to a low-dimensional space through the encoder of the variational autoencoder to generate latent variables; and processing the latent variables through a gradient descent training model and the decoder of the autoencoder to obtain the low-dimensional latent variables.

[0025] Optionally, the encoding process of the variational autoencoder is as follows:

[0026] μ, log(σ) 2 )=f φ (x Conv1D+attention )

[0027] z=μ+σ·∈

[0028] Where μ represents the mean parameter, σ represents the variance parameter, σ ranges from the natural number 0 to 1, z is the latent variable sampled from the latent space, and ∈ represents noise randomly sampled from the standard normal distribution N(0,1).

[0029] Secondly, this application provides a method for diagnosing photovoltaic support fault types, the method comprising:

[0030] Acquire data to be diagnosed, wherein the data to be diagnosed is the data after preprocessing the acquired photovoltaic power station data, and the preprocessing is noise reduction and normalization processing;

[0031] The data to be diagnosed is input into the photovoltaic support fault type diagnosis model to obtain the photovoltaic support fault type corresponding to the data to be diagnosed.

[0032] Thirdly, this application provides a photovoltaic support fault type diagnosis device, applied to the photovoltaic support fault type diagnosis method described in the second aspect, the device comprising:

[0033] The acquisition module is used to acquire data to be diagnosed, wherein the data to be diagnosed is the acquired photovoltaic power station data after preprocessing, and the preprocessing is noise reduction and normalization processing;

[0034] The diagnostic module is used to input the data to be diagnosed into the photovoltaic support fault type diagnostic model to obtain the photovoltaic support fault type corresponding to the data to be diagnosed.

[0035] Fourthly, this application provides a photovoltaic support system, including: a host computer, the host computer being configured to execute a photovoltaic support fault type diagnosis model generated by the photovoltaic support fault type diagnosis model generation method described in any one of the first aspects.

[0036] Fifthly, this application provides an electronic device, the electronic device comprising:

[0037] Memory, used to store computer programs;

[0038] When a processor executes a computer program stored in the memory, it implements the method steps provided in the first or second aspect described above.

[0039] In a sixth aspect, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method steps provided in the first or second aspect described above.

[0040] For the various aspects of the second to sixth aspects mentioned above, and the technical effects that each aspect may achieve, please refer to the technical effects that can be achieved for the first aspect or the various possible solutions in the first aspect mentioned above, which will not be repeated here. Attached Figure Description

[0041] The accompanying drawings are included to provide a further understanding of this application; they are incorporated into and constitute a part of this application. The drawings illustrate embodiments of this application and, together with this specification, serve to explain the principles of this application. In the drawings:

[0042] Figure 1 This is a flowchart of a method for generating a photovoltaic support fault type diagnosis model provided in this application;

[0043] Figure 2 This is a flowchart of a method for obtaining target data provided in this application;

[0044] Figure 3 This is a curve showing the accuracy of a photovoltaic bracket fault type diagnosis model during its training process, as provided in this application.

[0045] Figure 4 This is a curve showing the change in loss rate during the training process of a photovoltaic bracket fault type diagnosis model provided in this application;

[0046] Figure 5 This is a flowchart of a photovoltaic support fault type diagnosis method provided in this application;

[0047] Figure 6 This is a schematic diagram of a photovoltaic bracket fault type diagnosis device provided in this application;

[0048] Figure 7 This is a schematic diagram of the structure of an electronic device provided in this application. Detailed Implementation

[0049] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are merely some examples or embodiments of this application. For those skilled in the art, these drawings can be applied to other similar scenarios without creative effort. Unless obvious from the context or otherwise specified, the same reference numerals in the drawings represent the same structures or operations.

[0050] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not specifically singular and may include plural forms. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0051] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of this application. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale. Techniques, methods, and devices known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and devices should be considered part of the specification. In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters in the following drawings denote similar items; therefore, once an item is defined in one drawing, it need not be further discussed in subsequent drawings.

[0052] Furthermore, it should be noted that the use of terms such as "first" and "second" to define components is merely for the purpose of distinguishing the corresponding components. Unless otherwise stated, these terms have no special meaning and therefore should not be construed as limiting the scope of protection of this application. In addition, although the terminology used in this application is selected from commonly known and used terms, some terms mentioned in this application's specification may have been chosen by the applicant according to his or her judgment, and their detailed meanings are explained in the relevant sections of this description. Moreover, this application should be understood not only through the actual terms used, but also through the meaning implied by each term.

[0053] Flowcharts are used in this application to illustrate the operations performed by the system according to embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, various steps can be processed in reverse order or simultaneously. Furthermore, other operations may be added to these processes, or one or more steps may be removed from these processes.

[0054] See Figure 1 The diagram shown is a flowchart of a method for generating a photovoltaic support fault type diagnostic model according to an embodiment of this application. The specific process is as follows:

[0055] Step 101: Use convolutional layers to extract features from the target data to obtain a spatial feature map.

[0056] First, obtain the target data, such as... Figure 2 The diagram shown is a flowchart of a method for obtaining target data according to an embodiment of this application. The specific process is as follows:

[0057] Step 201: Construct a photovoltaic model to simulate the failure conditions of the photovoltaic array under different light and temperature conditions, and obtain a dataset.

[0058] In one embodiment, the PVsyst model is selected as the framework for the photovoltaic system model. This model can accurately simulate the performance of the photovoltaic array under different illumination and temperature conditions. A photovoltaic model is constructed based on the PVsyst model to simulate different fault conditions, including normal operation of the photovoltaic system, short-circuit fault conditions, open-circuit fault conditions, and shading fault conditions, to create a dataset. The dataset includes: output current, output voltage, output power, solar irradiance, and temperature.

[0059] By constructing a photovoltaic model, a large amount of data is generated to simulate the photovoltaic system under different light and temperature conditions, thereby improving the accuracy of subsequent training of photovoltaic support fault type diagnosis models.

[0060] Step 202: Denoise and normalize the dataset and monitored photovoltaic power station data to complete the preprocessing, thereby obtaining the target data.

[0061] In one embodiment, the monitored photovoltaic power station data is the actual monitoring data of the photovoltaic power station. This monitoring data can be the performance data of the photovoltaic power station over a year, including: output current, output voltage, output power, solar irradiance, and temperature. Noise and outlier removal are performed on the created dataset and the actual monitored data to ensure data quality. Then, the data is normalized so that its values ​​are within the range of [0, 1]. After these preprocessing steps, the target data is obtained.

[0062] After obtaining the target data, convolutional layers are used to extract features from the target data, resulting in a spatial feature map. In one embodiment, convolutional layers are introduced into the variational autoencoder, meaning that the encoder and decoder parts of the variational autoencoder consist of multiple convolutional layers to extract local spatial features of the target data. Specifically, the target data is convolved with a set of convolutional kernels using convolutional layers, followed by pooling to extract the spatial features of the target data, resulting in a spatial feature map. The formula for the convolutional layer is as follows:

[0063] h l =ReLU(Conv(h) l-1 W l ,b l ))

[0064] Where ReLU is the ReLU (Rectified Linear Unit) activation function, Conv is the convolutional layer, and h l W represents the output feature map of the l-th layer. l and b l These are the weights and biases of the l-th convolutional kernel, respectively.

[0065] In one embodiment, the convolutional layer used is a Conv1D layer. The Conv1D layer can identify spatial patterns and structures in the data. The convolution formula for the Conv1D layer operation is as follows:

[0066]

[0067] Where x(t) is the sequence data of the input target data, t+i-1 is a specific position of the input sequence data, representing the value of the input sequence data corresponding to the i-th element of the convolution kernel when the convolution kernel slides to the position at time t, W(i) is the convolution kernel, k is the size of the convolution kernel, and b l It is the bias of the l-th convolutional kernel.

[0068] By using convolutional layers to extract spatial features from photovoltaic system data and then performing pooling operations, the number of model parameters is reduced, thereby increasing the speed of subsequent model training.

[0069] Step 102: Introduce a self-attention mechanism on the spatial feature map, calculate the weight of each time step in the spatial feature map, and obtain a weighted feature map.

[0070] In one embodiment, a self-attention mechanism is introduced onto the spatial feature map extracted by convolution. The weights at each time step in the spatial feature map are calculated, i.e., the relationships between different time steps in the spatial feature map are calculated to obtain global spatial dependencies. Finally, a weighted feature map with adjusted weights is output. The self-attention mechanism is used to capture global dependencies between different parts of the input data. It is particularly effective in handling long sequences or data with complex spatial relationships. The attention weights for each input feature are obtained by calculating the weighted sum of the query vector, key vector, and value vector. For each input feature X, the formula for the self-attention mechanism is:

[0071]

[0072] Where Q = W Q X is the query vector, W Q This is the weight matrix for the query, X is the input feature, and K = W. K X is the key vector, W K It is the weight matrix of the keys, V = W V X is a value vector, W V It is the weight matrix of values, d k Here, is the dimension of the key vector, and softmax is the normalized exponential function used to calculate the attention weight at each position. Then, the attention weights of each input feature are weighted and summed to obtain the weighted feature map. The final formula for calculating the weighted feature map is:

[0073]

[0074] Where, α i V is the attention weight of the i-th element. i It is the value vector of the i-th element, and Attention(X) represents the weighted summation of the attention weights for each input feature.

[0075] By introducing convolutional layers and self-attention mechanisms into the variational autoencoder, the encoder can not only extract local spatial features of the target data through convolutional layers, but also capture global dependencies through self-attention mechanisms. This can effectively handle the local spatial features of the photovoltaic system, such as local fluctuations in solar irradiance, and also capture long-term trends, enabling a more comprehensive extraction of the output features of the photovoltaic system.

[0076] Step 103: Use a variational autoencoder to process the spatial feature map and the weighted feature map to obtain low-dimensional latent variables.

[0077] In one embodiment, the spatial feature maps extracted by the convolutional layer and the weighted feature maps extracted by the self-attention mechanism are mapped to a latent low-dimensional space through the encoder part of the variational autoencoder to generate latent variables. The model is then trained using gradient descent, and the decoder of the variational autoencoder is used to restore the original pattern of the latent variables, resulting in low-dimensional latent variables and reducing overfitting. The complete encoding process of the variational autoencoder is as follows:

[0078] Based on the spatial feature map extracted by the convolutional layer and the weighted feature map extracted by the self-attention mechanism, the parameters μ and σ of the latent distribution are generated:

[0079] μ, log(σ) 2 )=f φ (x Conv1D+attention )

[0080] Where μ represents the mean parameter and σ represents the variance parameter, and σ ranges from the natural number 0 to 1.

[0081] Then, sampling is performed in the latent space to obtain the latent vector z:

[0082] z=μ+σ·∈

[0083] Where z is a latent variable sampled from the latent space, and ∈ is noise randomly sampled from the standard normal distribution N(0,1).

[0084] Finally, the latent vector z is decoded using the decoder of the variational autoencoder to generate reconstructed data, i.e., low-dimensional latent variables.

[0085] The core objective of a variational autoencoder (VAE) is to compress, encode, and reconstruct input data through an encoder and decoder, while introducing regularization into the latent space. In one embodiment, the loss function of a VAE consists of a reconstruction loss and KL divergence regularization. The expression for the loss function of a VAE is as follows:

[0086]

[0087] Among them, E qφ(z|x) [logp θ [x|z)] is the reconstruction loss, which is the logarithmic reconstruction probability between the decoder output and the original data x. It is usually calculated using the mean squared error (MSE); D KL (q φ (z|x)||p(z)) is the regularization term, which measures the distribution q of the latent variables generated by the encoder. φ The difference between (z|x) and the standard normal distribution p(z) is calculated using the KL divergence.

[0088] When the loss function of the variational autoencoder meets the preset requirements, such as when the loss function value is less than a preset threshold, the variational autoencoder outputs the reconstructed low-dimensional latent variables.

[0089] Step 104: Input the low-dimensional latent variables into the bidirectional gated recurrent unit, output the time series features of the target data, and input the time series features into the fully connected layer and the softmax layer to obtain the photovoltaic bracket fault type diagnosis model.

[0090] In one embodiment, the low-dimensional latent variables output by the variational autoencoder are used as inputs to a bidirectional gated recurrent unit. Based on these inputs, the bidirectional gated recurrent unit can capture the temporal dependencies of the data, namely the time correlation between the output voltage, output current and the maximum power point, to obtain the temporal features of the target data. The temporal features are then input into a fully connected layer and a softmax layer for multi-class classification to obtain a photovoltaic bracket fault type diagnosis model.

[0091] Furthermore, in one embodiment, a grid search method is used to optimize the hyperparameters of the photovoltaic support fault type diagnosis model, thereby improving the performance of the photovoltaic support fault type diagnosis model.

[0092] Furthermore, in one embodiment, the performance of the photovoltaic support fault type diagnosis model is evaluated using a test set and a validation set. The training set is used to train the model's performance, and the validation set is used to verify whether the model's performance meets expectations. The performance of the photovoltaic support fault type diagnosis model is evaluated using metrics such as training accuracy, validation accuracy, and test accuracy. Figure 3 and Figure 4 As shown, Figure 3This is a graph showing the accuracy curve of a photovoltaic bracket fault type diagnosis model provided in this application, as the number of training iterations increases. Figure 4 This diagram illustrates the change in the loss rate of a photovoltaic support fault type diagnosis model provided in this application, as the number of training iterations increases. With the increase in the number of training iterations, the accuracy of the photovoltaic support fault type diagnosis model improves while the loss rate decreases. During training, the model effectively learns potential patterns and demonstrates high accuracy in determining whether the photovoltaic system is in a fault state. The low loss value reflects the model's robustness in accurate photovoltaic system fault classification and its strong ability to learn complex patterns and the features necessary for accurate classification.

[0093] The above method uses a variational autoencoder combined with convolutional layers, self-attention mechanisms, and bidirectional gated recurrent units for specific fault diagnosis. This method can more comprehensively extract the output characteristics of the photovoltaic system, effectively identify and diagnose fault types, improve the accuracy of fault detection and diagnosis, and thus improve the reliability and safety of the photovoltaic system.

[0094] Based on the same inventive concept, this application also provides a method for diagnosing photovoltaic support fault types, see [link to relevant documentation]. Figure 5 As shown, this method is applied to Figure 1 The method for generating the photovoltaic support fault type diagnosis model is shown below. The specific implementation process of the photovoltaic support fault type diagnosis method is as follows:

[0095] Step 501: Obtain the data to be diagnosed.

[0096] In this embodiment of the application, the data to be diagnosed is the data obtained from the photovoltaic power station after preprocessing. The preprocessing specifically includes noise reduction and normalization.

[0097] Step 502: Input the data to be diagnosed into the photovoltaic support fault type diagnosis model to obtain the photovoltaic support fault type corresponding to the data to be diagnosed.

[0098] The photovoltaic fault type diagnostic model can effectively identify and diagnose the photovoltaic support fault type corresponding to the data to be diagnosed, thereby improving the accuracy of fault detection and diagnosis.

[0099] Based on the same inventive concept, this application also provides a photovoltaic support fault type diagnosis device, see [link to relevant documentation]. Figure 6 As shown, this device is applied to Figure 5 The photovoltaic support fault type diagnosis method shown includes the following device:

[0100] The acquisition module 601 is used to acquire data to be diagnosed, wherein the data to be diagnosed is data after preprocessing the acquired photovoltaic power station data, and the preprocessing is noise reduction and normalization processing;

[0101] The diagnostic module 602 is used to input the data to be diagnosed into the photovoltaic support fault type diagnostic model to obtain the photovoltaic support fault type corresponding to the data to be diagnosed.

[0102] Based on the same inventive concept, this application also provides a photovoltaic bracket control system, including a host computer, which is configured to execute... Figure 1 The photovoltaic support fault type diagnostic model generated by any of the methods shown in the figure.

[0103] Based on the same inventive concept, this application also provides an electronic device that can realize the function of the aforementioned photovoltaic bracket fault type diagnosis device. (Refer to...) Figure 7 The electronic device includes:

[0104] At least one processor 701 and a memory 702 connected to at least one processor 701. In this embodiment, the specific connection medium between the processor 701 and the memory 702 is not limited. Figure 7 The example shown is the connection between processor 701 and memory 702 via bus 700. Bus 700 is... Figure 7 The connections between other components are indicated by thick lines and are for illustrative purposes only, not as limiting information. The 700 bus can be divided into address bus, data bus, control bus, etc., for ease of representation. Figure 7 The term is represented by a single thick line, but this does not imply that there is only one bus or one type of bus. Alternatively, the processor 701 can also be called a controller; there is no restriction on the name.

[0105] In this embodiment, memory 702 stores instructions executable by at least one processor 701. By executing the instructions stored in memory 702, at least one processor 701 can execute the production method for the photovoltaic support fault type diagnosis model and / or the photovoltaic support fault type diagnosis method discussed above. Processor 701 can implement... Figure 6 The functions of each module in the device shown.

[0106] The processor 701 is the control center of the device. It can connect to various parts of the control device through various interfaces and lines. By running or executing instructions stored in memory 702 and calling data stored in memory 702, the processor can perform various functions and process data, thereby monitoring the device as a whole.

[0107] In one possible design, processor 701 may include one or more processing units. Processor 701 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may also not be integrated into processor 701. In some embodiments, processor 701 and memory 702 may be implemented on the same chip; in some embodiments, they may also be implemented on separate chips.

[0108] The processor 701 can be a general-purpose processor, such as a central processing unit (CPU), digital signal processor, application-specific integrated circuit, field-programmable gate array or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the seat collision avoidance method disclosed in the embodiments of this application can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0109] Memory 702, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules. Memory 702 may include at least one type of storage medium, such as flash memory, hard disk, multimedia card, card-type memory, random access memory (RAM), static random access memory (SRAM), programmable read-only memory (PROM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), magnetic storage, magnetic disk, optical disk, etc. Memory 702 can be any other medium capable of carrying or storing desired program code in the form of instructions or data structures that can be accessed by a computer, but is not limited thereto. In the embodiments of this application, memory 702 can also be a circuit or any other device capable of implementing storage functions for storing program instructions and / or data.

[0110] By designing and programming the processor 701, the production method of the photovoltaic bracket fault type diagnosis model and / or the corresponding code of the photovoltaic bracket fault type diagnosis method described in the foregoing embodiments can be embedded into the chip, thereby enabling the chip to execute the code during operation. Figure 1 The method for producing a photovoltaic bracket fault type diagnosis model as illustrated in the embodiment and / or Figure 5 The steps of the photovoltaic bracket fault type diagnosis method in the illustrated embodiment are as follows. How to design and program the processor 701 is a technique well-known to those skilled in the art and will not be described further here.

[0111] Based on the same inventive concept, embodiments of this application also provide a storage medium storing computer instructions that, when executed on a computer, cause the computer to perform the production method and / or the photovoltaic bracket fault type diagnosis method described above.

[0112] In some possible implementations, the method for producing a photovoltaic bracket fault type diagnosis model and / or various aspects of the photovoltaic bracket fault type diagnosis method provided in this application can also be implemented in the form of a program product, which includes program code. When the program product is run on a device, the program code is used to cause the control device to perform the steps in the method for producing a photovoltaic bracket fault type diagnosis model and / or the photovoltaic bracket fault type diagnosis method according to various exemplary embodiments of this application described above.

[0113] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0114] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0115] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0116] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0117] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for generating a photovoltaic support fault type diagnostic model, characterized in that, The method includes: Convolutional layers are used to extract features from the target data to obtain a spatial feature map. The target data is the preprocessed data of the dataset generated by the photovoltaic model and the monitored photovoltaic power station data. A self-attention mechanism is introduced into the spatial feature map to calculate the weight of each time step in the spatial feature map, thereby obtaining a weighted feature map; A variational autoencoder is used to process the spatial feature map and the weighted feature map to obtain low-dimensional latent variables; The low-dimensional latent variables are input into a bidirectional gated recurrent unit, which outputs the temporal features of the target data. The temporal features are then input into a fully connected layer and a softmax layer to obtain a photovoltaic bracket fault type diagnosis model.

2. The method as described in claim 1, characterized in that, Before using a convolutional neural network to extract features from the target data and obtain a spatial feature map, the method further includes: The photovoltaic model is constructed to simulate the failure conditions of the photovoltaic array under different light and temperature conditions, and a dataset is obtained. The dataset includes: output current, output voltage, output power, solar irradiance, and temperature. The dataset and the monitored photovoltaic power station data are subjected to denoising and normalization to complete the preprocessing, thereby obtaining the target data.

3. The method as described in claim 1, characterized in that, The step of using convolutional layers to extract features from the target data to obtain a spatial feature map includes: The target data is convolved with a set of convolutional kernels using the convolutional layer, followed by pooling, to obtain the spatial feature map. The formula for the convolutional layer is as follows: h l =ReLU(Conv(h l-1 ,W l ,b l )) Where ReLU is the ReLU (Rectified Linear Unit) activation function, Conv is the convolutional layer, and h l W represents the output feature map of the l-th layer. l and b l These are the weights and biases of the l-th convolutional kernel, respectively.

4. The method as described in claim 1, characterized in that, The convolutional layer is a Conv1D layer, and the convolution formula for the Conv1D layer is: Where x(t) is the sequence data of the input target data, t+i-1 represents the value of the input sequence data corresponding to the i-th element of the convolution kernel when the convolution kernel slides to the position at time t, W(i) is the convolution kernel, k is the size of the convolution kernel, and b l It is the bias of the l-th convolutional kernel.

5. The method as described in claim 1, characterized in that, The step of introducing a self-attention mechanism on the spatial feature map and calculating the weight of each time step in the spatial feature map to obtain a weighted feature map includes: The self-attention mechanism is introduced onto the spatial feature map, and the attention weight of each input feature is obtained by calculating the weighted sum between the query vector, key vector and value vector of each input feature. The weighted feature map is obtained based on the attention weights of each input feature.

6. The method as described in claim 5, characterized in that, The formula for the self-attention mechanism is: Where Q = W Q X is the query vector, W Q This is the weight matrix of the query, X is the input feature, and K = W. K X is the key vector, W K It is the weight matrix of the keys, V = W V X is a value vector, W V It is the weight matrix of values, d k is the dimension of the key vector, and softmax is the normalized exponential function used to calculate the attention weight at each position.

7. The method as described in claim 6, characterized in that, The formula for calculating the weighted feature map is: Where, α i V is the attention weight of the i-th element. i It is the value vector of the i-th element, and Attention(X) represents the weighted summation of the attention weights for each input feature.

8. The method as described in claim 1, characterized in that, The process of using a variational autoencoder to process the spatial feature map and the weighted feature map to obtain low-dimensional latent variables includes: The spatial feature map and the weighted feature map are mapped to a low-dimensional space through the encoder of the variational autoencoder to generate latent variables; The latent variables are processed by the gradient descent training model and the decoder of the autoencoder to obtain the low-dimensional latent variables.

9. The method as described in claim 1, characterized in that, The encoding process of the variational autoencoder is as follows: μ, log(σ 2 )=f φ (x Conv1D+attention ) z=μ+σ·∈ Where μ represents the mean parameter, σ represents the variance parameter, σ ranges from the natural number 0 to 1, z is the latent variable sampled from the latent space, and ∈ represents noise randomly sampled from the standard normal distribution N(0,1).

10. A method for diagnosing photovoltaic support fault types, characterized in that, A method for generating a photovoltaic support fault type diagnostic model according to any one of claims 1-9, the method comprising: Acquire data to be diagnosed, wherein the data to be diagnosed is the data after preprocessing the acquired photovoltaic power station data, and the preprocessing is noise reduction and normalization processing; The data to be diagnosed is input into the photovoltaic support fault type diagnosis model to obtain the photovoltaic support fault type corresponding to the data to be diagnosed.

11. A photovoltaic support fault type diagnosis device, characterized in that, The device, applied to the photovoltaic support fault type diagnosis method of claim 10, comprises: The acquisition module is used to acquire data to be diagnosed, wherein the data to be diagnosed is the acquired photovoltaic power station data after preprocessing, and the preprocessing is noise reduction and normalization processing; The diagnostic module is used to input the data to be diagnosed into the photovoltaic support fault type diagnostic model to obtain the photovoltaic support fault type corresponding to the data to be diagnosed.

12. A photovoltaic support control system, characterized in that, Includes: a host computer, the host computer being configured to execute a photovoltaic support fault type diagnosis model generated by the generation method of the photovoltaic support fault type diagnosis model according to any one of claims 1-9.

13. An electronic device, characterized in that, include: Memory, used to store computer programs; The processor, when executing the computer program stored in the memory, implements the production method of the photovoltaic bracket fault type diagnosis model according to any one of claims 1-9 or the photovoltaic bracket fault type diagnosis method according to claim 10.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the method for producing the photovoltaic bracket fault type diagnosis model according to any one of claims 1-9 or the photovoltaic bracket fault type diagnosis method according to claim 10.

Images