Fault diagnosis method and system for photovoltaic tracking support

Abstract

A fault diagnosis method (10) and system (20) for a photovoltaic tracking support. The method comprises the following steps: in the operation process of a photovoltaic tracking support, tracking control units (21) that correspond to the photovoltaic tracking support continuously acquiring one-dimensional time-domain operating signals of the photovoltaic tracking support; converting the one-dimensional time-domain operating signals into two-dimensional time-frequency-domain operating signals; performing feature extraction on the two-dimensional time-frequency-domain operating signals to obtain an operating feature map; and determining whether the operating feature map is a fault feature map, so as to diagnose whether the photovoltaic tracking support has a fault. The fault diagnosis method (10) and system (20) for a photovoltaic tracking support can improve the stability, accuracy and intelligence level of fault diagnosis for photovoltaic tracking supports.

Description

Photovoltaic tracking bracket fault diagnosis method and system Technical Field

[0001] The present application mainly relates to the field of photovoltaic tracking brackets, and in particular to a fault diagnosis method and system for photovoltaic tracking brackets. Background Art

[0002] Photovoltaic tracking bracket fault diagnosis is an important technology in the field of photovoltaic cells. It can promptly detect and diagnose problems in the tracking system to ensure the normal operation of the system and maximize power generation efficiency. Because photovoltaic equipment is in the external environment for a long time, the bracket system is easily affected by dust, rain, etc., resulting in the photovoltaic tracking bracket being unable to accurately adjust its posture angle. Once a fault occurs, it will affect the power generation of the solar photovoltaic system and reduce power generation efficiency. Existing operation monitoring of photovoltaic tracking brackets is mostly based on sensor measurements and analysis algorithms based on sensor data. However, due to factors such as the conditions limiting the operating conditions of the photovoltaic system, existing fault diagnosis solutions for photovoltaic tracking brackets still have many problems such as low intelligence, low accuracy, and delayed reminders. Summary of the Invention

[0003] The technical problem to be solved by the present application is to provide a fault diagnosis method and system for a photovoltaic tracking bracket, which can improve the stability, accuracy and intelligence of the fault diagnosis of the photovoltaic tracking bracket.

[0004] In order to solve the above technical problems, the present application provides a fault diagnosis method for a photovoltaic tracking bracket, comprising the following steps: during the operation of the photovoltaic tracking bracket, continuously collecting the one-dimensional time domain operation signal of the photovoltaic tracking bracket through a tracking control unit corresponding to the photovoltaic tracking bracket; converting the one-dimensional time domain operation signal into a two-dimensional time-frequency domain operation signal; performing feature extraction on the two-dimensional time-frequency domain operation signal to obtain an operation characteristic diagram; and determining whether the operation characteristic diagram belongs to a fault characteristic diagram, thereby diagnosing whether the photovoltaic tracking bracket has a fault.

[0005] In one embodiment of the present application, it also includes extracting the information of the one-dimensional time domain operation signal on the time scale by wavelet transform to capture the change characteristics of the one-dimensional time domain operation signal on different time scales, thereby obtaining the two-dimensional time-frequency domain operation signal.

[0006] In one embodiment of the present application, the wavelet variation coefficient is obtained by the following formula:

[0007] in: For my mother Xiaobo, is the complex conjugate mother wavelet, x(t) is the input signal, t is the time, a is the scaling factor, and b is the translation factor.

[0008] In one embodiment of the present application, it also includes: pre-acquiring multiple operating characteristic diagrams for different fault types, and obtaining a fault type identification model through training based on a convolutional neural network; and during the operation of the photovoltaic tracking bracket, inputting the operating characteristic diagram into the fault type identification model in real time, and determining whether the operating characteristic diagram belongs to a fault characteristic diagram based on the output of the fault type identification model.

[0009] In one embodiment of the present application, the training process further includes: differentiating each or multiple of the multiple running feature maps into multiple image regions; generating a The first feature map is generated using maximum pooling a second feature map; connecting the first feature map and the second feature map into a two-pass feature map; and generating a spatial attention map through convolution to obtain the fault type recognition model.

[0010] In one embodiment of the present application, the one-dimensional time-domain operation signal includes a current signal of a driving motor of the photovoltaic tracking bracket.

[0011] In one embodiment of the present application, the following cross entropy loss function is used to determine whether the operation characteristic graph belongs to a fault characteristic graph:

[0012] Where: N is the number of categories, w i is the weight vector of the i-th category in the fully connected layer, q i is the label smoothing factor, a is the scaling factor, T represents the transpose, and f is the integration function.

[0013] In one embodiment of the present application, the label smoothing factor q is calculated using the following formula: i :

[0014] Where: ε is the label smoothing parameter, which is used to reduce the weight of the true label when calculating the loss function to prevent the model from overfitting the training data, thereby improving its generalization ability, and y is the one-hot label.

[0015] Another aspect of the present application also proposes a fault diagnosis system for a photovoltaic tracking bracket, comprising: multiple tracking control units, suitable for continuously collecting one-dimensional time domain operation signals of the photovoltaic tracking bracket during the operation of the photovoltaic tracking bracket; a data acquisition and supervisory control system, suitable for diagnosing whether the photovoltaic tracking bracket has a fault according to the fault diagnosis method of any of the aforementioned embodiments.

[0016] Another aspect of the present application further provides a computer-readable medium storing computer program code, which implements the method of any of the aforementioned embodiments when executed by a processor.

[0017] Compared with the existing technology, the present application has the following advantages: the present application collects signals such as the motor current of the tracking bracket, converts them into time-frequency graphs through time-frequency analysis and wavelet transform, and uses an image classification algorithm based on deep learning to extract the features of the time-frequency graphs of each fault type for learning and modeling. The constructed model has the ability to identify the features of the time-frequency graph and establish connections with the corresponding fault types, and can be used for fault diagnosis of photovoltaic tracking brackets with higher stability, accuracy and intelligence.

[0018] Summary of the Figures

[0019] The accompanying drawings are included to provide a further understanding of the present application. They are incorporated into and constitute a part of the present application. The accompanying drawings illustrate embodiments of the present application and, together with the description, serve to explain the principles of the present application.

[0020] In the attached figure:

[0021] FIG1 is a schematic flow chart of a fault diagnosis method for a photovoltaic tracking bracket according to an embodiment of the present application;

[0022] FIG2 is a schematic diagram of monitoring one-dimensional time-domain operating signals of multiple photovoltaic tracking brackets in a fault diagnosis method for a photovoltaic tracking bracket according to an embodiment of the present application;

[0023] 3 and 4 are schematic diagrams showing the logical principles of a fault diagnosis method for a photovoltaic tracking bracket according to an embodiment of the present application;

[0024] 5 to 7 are characteristic diagrams of different types of faults of a photovoltaic tracking bracket using a fault diagnosis method of a photovoltaic tracking bracket according to an embodiment of the present application; and

[0025] FIG8 is a system block diagram of a fault diagnosis system for a photovoltaic tracking bracket according to an embodiment of the present application.

[0026] Preferred embodiments of the present invention

[0027] To more clearly illustrate the technical solutions of the embodiments of this application, the following is a brief introduction to the drawings required for describing the embodiments. Obviously, the drawings described below are merely examples or embodiments of this application. Those skilled in the art can apply this application to other similar scenarios based on these drawings without inventive effort. Unless otherwise apparent from the context or otherwise noted, the same reference numerals in the figures represent the same structure or operation.

[0028] As used in this application and the claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not intended to refer to the singular but may include the plural. Generally speaking, the terms "comprises" and "include" only indicate the inclusion of the steps and elements specifically identified, and these steps and elements do not constitute an exclusive list. A method or apparatus may also include other steps or elements.

[0029] Unless otherwise specifically stated, the relative arrangement of the parts and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present application. At the same time, it should be understood that, for ease of description, the sizes of the various parts shown in the drawings are not drawn according to actual proportional relationships. The techniques, methods and equipment known to those of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, the techniques, methods and equipment should be considered as part of the authorization specification. In all examples shown and discussed here, any specific values ​​should be interpreted as being merely exemplary and not as limitations. Therefore, other examples of the exemplary embodiments may have different values. It should be noted that similar numbers and letters represent similar items in the following figures, and therefore, once an item is defined in one figure, it does not need to be further discussed in subsequent figures.

[0030] Furthermore, it should be noted that the use of terms such as "first" and "second" to define components is solely 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. Furthermore, while the terms used in this application are selected from commonly known and commonly used terms, some terms mentioned in this specification may have been selected by the applicant at his or her discretion, and their detailed meanings are explained in the relevant sections of this description. Furthermore, this application should be understood not only by the actual terms used, but also by the meaning implied by each term.

[0031] It should be understood that when a component is referred to as being “on another component,” “connected to another component,” “coupled to another component,” or “contacting another component,” it can be directly on, connected to, coupled to, or contacting the other component, or intervening components may be present. In contrast, when a component is referred to as being “directly on another component,” “directly connected to,” “directly coupled to,” or “directly contacting” another component, there are no intervening components. Similarly, when a first component is referred to as being “electrically in contact with” or “electrically coupled to” a second component, an electrical path exists between the first and second components that allows current to flow. This electrical path may include capacitors, coupled inductors, and / or other components that allow current to flow, even without direct contact between the conductive components.

[0032] The present application proposes a fault diagnosis method 10 for a photovoltaic tracking bracket (hereinafter referred to as "fault diagnosis method 10") with reference to FIG1. ​​Fault diagnosis method 10 can improve the stability, accuracy and intelligence of fault diagnosis of photovoltaic tracking brackets. FIG1 in the present application uses a flowchart to illustrate the operations performed by the system according to the embodiment of the present application. It should be understood that the previous or following operations are not necessarily performed in exact sequence. On the contrary, various steps can be processed in reverse order or simultaneously. At the same time, other operations may be added to these processes, or one or more steps may be removed from these processes.

[0033] 1 , a fault diagnosis method 10 includes the following steps.

[0034] Step 11 involves continuously collecting the one-dimensional time-domain operating signal of the photovoltaic tracking bracket via the tracking control unit corresponding to the photovoltaic tracking bracket during operation. Exemplarily, the one-dimensional time-domain operating signal in step 11 includes the current signal of the photovoltaic tracking bracket's drive motor. FIG2 illustrates the current signals of tracking control units T17, T18, T19, and T20 corresponding to some photovoltaic tracking brackets collected using the fault diagnosis method 10 in one embodiment of the present application. It can be seen that different tracking control units have different one-dimensional time-domain operating signal variation characteristics. Subsequent steps will further process the one-dimensional time-domain operating signal of each tracking control unit.

[0035] Continuing with reference to FIG1 , step 12 is to convert the one-dimensional time-domain operation signal into a two-dimensional time-frequency domain operation signal. Further preferably, based on the fault diagnosis method 10 of FIG1 , a preferred embodiment of the present application further proposes extracting information on the time scale of the one-dimensional time-domain operation signal through wavelet transform, so as to capture the variation characteristics of the one-dimensional time-domain operation signal at different time scales, thereby obtaining a two-dimensional time-frequency domain operation signal.

[0036] For example, the wavelet variation coefficient can be obtained by the following formula:

[0037] in: For my mother Xiaobo, is the complex conjugate mother wavelet, x(t) is the input signal, t is time, a is the scaling factor, and b is the translation factor. Wavelet time-frequency plots can display signal information such as time, frequency, and energy in a two-dimensional image, carrying richer information than traditional one-dimensional signals. Therefore, using two-dimensional time-frequency domain operating signals processed by wavelet transforms for diagnosis can achieve greater diagnostic stability and accuracy.

[0038] Continuing with Figure 1, step 13 is to extract features from the two-dimensional time-frequency domain operating signal to obtain an operating characteristic graph. Step 14 is to determine whether the operating characteristic graph is a fault characteristic graph, thereby diagnosing whether the photovoltaic tracking bracket has a fault.

[0039] It can be seen that steps 13 and 14 of the fault diagnosis method 10 shown in Figure 1 rely on image processing methods based on two-dimensional time-frequency domain operating signals. In a preferred embodiment of the fault diagnosis method 10 shown in Figure 1 of the present application, a fault type recognition model can be created in the following manner, and the fault diagnosis detection of the operating characteristic diagram in step 14 can be completed based on the trained fault type recognition model.

[0040] Specifically, in this preferred embodiment, it is necessary to obtain multiple operating characteristic diagrams for different fault types in advance, and obtain a fault type recognition model through training based on a convolutional neural network. Furthermore, during the operation of the photovoltaic tracking bracket, the operating characteristic diagram is input into the fault type recognition model in real time, and whether the operating characteristic diagram belongs to a fault characteristic diagram is determined based on the output of the fault type recognition model. Exemplarily, Figures 3 and 4 respectively show the logical principle diagrams for creating a fault type recognition model and applying the fault type recognition model for detection. According to Figure 3, a fault experiment is carried out, the operating current data of the motor corresponding to the photovoltaic tracking bracket is collected and analyzed in the time-frequency domain, and after the time-frequency diagram is output, image classification is performed for different fault types based on the time-frequency diagram; the image features of the fault time-frequency diagram are further extracted based on the neural network, and finally the fault type recognition model is output.

[0041] Further according to FIG4 , after obtaining the fault type identification model, step 11 can be performed in real time during the operation of the photovoltaic tracking bracket, and the one-dimensional time domain operation signal of the photovoltaic tracking bracket can be continuously collected through the tracking control unit TCU. It can be understood that, as mentioned above, when training the fault type identification model, the current data is mainly trained, but the present application is not limited to this. In some other embodiments of the present application, when the one-dimensional time domain operation signal to be monitored includes other inputs besides current, the fault type identification model can be trained in a targeted manner according to the actual application scenario, even for example, structural faults of the bracket, etc. On the basis of the current input, other types of fault defects are introduced as training inputs, thereby enriching the types of faults that can be detected by the trained fault type identification model. Therefore, by inputting the time-frequency graph collected and processed during the actual operation of the photovoltaic tracking bracket into the fault type identification model, the diagnosis result of the fault type can be directly obtained.

[0042] The diagnostic model proposed in this application, which is based on machine learning and big data solutions, has good detection effects and a high degree of intelligence. In this way, the drawbacks of the existing technology that mainly relies on regular manual data collection by sensors are avoided, and all-weather intelligent collection and diagnosis can be achieved. In addition, the same training model can be used to detect multiple different categories, and the data input for diagnosis is diversified. The overall accuracy and reliability of the detection are significantly improved. For example, Figures 5 to 7 show several different types of fault types that can be identified by using the above-mentioned fault type identification model. Among them, Figure 5 is a characteristic diagram under a short circuit defect, Figure 6 is a characteristic diagram of a larger current intensity, and Figure 7 is a characteristic diagram of a higher frequency transient current change. Of course, Figures 5 to 7 only provide some schematic enumerations, and this application is not limited to this.

[0043] Further preferably, in the above training process, the method further includes dividing each or more of the plurality of running feature maps into a plurality of image regions and generating the image regions by using average pooling. The first feature map is generated using maximum pooling The second feature map connects the first feature map and the second feature map into a two-pass feature map. Finally, a spatial attention map is generated by convolution to obtain a further optimized fault type recognition model. In this preferred embodiment, it is proposed to use a channel attention mechanism to improve the classification accuracy of the model. The use of global maximum pooling and global average pooling can avoid information loss in the pooling operation. After that, a multi-layer perceptron is used to splice the fully connected layer, and the Sigmoid activation function is used to output the weight matrix of the channel dimension of the feature map, thereby realizing the extraction of important information in the time-frequency map. On this basis, considering that different areas of the feature map have different effects on feature recognition, the use of spatial attention can help the model learn the relationship between feature maps and find important parts for processing, thereby improving the accuracy and reliability of detection.

[0044] In any embodiment of the present application, preferably, it further includes using the following cross entropy loss function to determine whether the operation characteristic graph belongs to a fault characteristic graph:

[0045] Where: N is the number of categories, w i is the weight vector of the i-th category in the fully connected layer, q i is the label smoothing factor, a is the scaling factor, and f is the integration function. Specifically, when the true category of the sample belongs to the fault in the fault catalog, the integration function f takes 1, otherwise the integration function f takes 0.

[0046] The integration function selection process primarily includes detection, isolation, and identification steps. The detection step involves determining whether a system fault has occurred. In this application, when a sample case is determined to be a fault, it belongs to the true category. The identification step involves distinguishing the fault type of the sample case in the true category. In some embodiments, fault types are distinguished using a trained model with a predefined fault catalog. During model training, fault categories and corresponding fault characteristics are defined to determine the fault catalog.

[0047] More specifically, the label smoothing factor q is calculated using the following formula: i :

[0048] Wherein: ε is the label smoothing parameter, which is used to reduce the weight of the true label when calculating the loss function, prevent the model from overfitting the training data, and thus improve its generalization ability, and y is the one-hot label. Specifically, y represents the original one-hot label, which is a special label encoding method in which only one element of each category has a value of 1, and the values ​​of the remaining elements are 0. Specifically, if there is a classification problem in which the total number of categories is C, for any sample, its one-hot label y is a vector of length C, in which only one element is 1, representing the category to which the sample belongs, and the remaining elements are 0. On the other hand, the present application also proposes a fault diagnosis system 20 for a photovoltaic tracking bracket with reference to FIG8, comprising:

[0049] A plurality of tracking control units 21, adapted to continuously collect one-dimensional time domain operation signals of the photovoltaic tracking bracket during operation of the photovoltaic tracking bracket;

[0050] The data acquisition and supervisory control system 22 is adapted to diagnose whether a photovoltaic tracker has experienced a fault according to the fault diagnosis method of any embodiment of the present application. For example, the data acquisition and supervisory control system 22 may directly utilize the SCADA system in an existing photovoltaic tracker operating system, i.e., directly deploy the fault diagnosis method proposed earlier in this application in the SCADA system, thereby conveniently improving the diagnostic method for the photovoltaic tracker without introducing additional hardware costs.

[0051] Another aspect of the present application further provides a computer-readable medium storing computer program code, which implements the method of any embodiment of the present application when executed by a processor.

[0052] A photovoltaic tracking system fault diagnosis algorithm based on wavelet transform and attention mechanism is proposed. Wavelet transform is used to map the one-dimensional time-domain signal of the motor current collected from the TCU to the two-dimensional time-frequency domain. A convolutional neural network is used to extract features from the transformed two-dimensional time-frequency map. This feature map is then fed into a residual network for fault pattern recognition. The model distinguishes different fault types and is saved as a model for subsequent fault diagnosis.

[0053] The basic concepts have been described above. It will be apparent to those skilled in the art that the above disclosures are merely examples and do not limit the present application. Although not explicitly stated herein, those skilled in the art may make various modifications, improvements, and revisions to the present application. Such modifications, improvements, and revisions are suggested in the present application and remain within the spirit and scope of the exemplary embodiments of the present application.

[0054] At the same time, this application uses specific terms to describe the embodiments of this application. For example, "one embodiment," "an embodiment," and / or "some embodiments" refer to a certain feature, structure, or characteristic related to at least one embodiment of this application. Therefore, it should be emphasized and noted that "one embodiment," "an embodiment," or "an alternative embodiment" mentioned twice or multiple times in different locations in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures, or characteristics in one or more embodiments of this application may be appropriately combined.

[0055] Some aspects of the present application can be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software can be referred to as "data blocks", "modules", "engines", "units", "components" or "systems". The processor can be one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DAPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors or combinations thereof. In addition, various aspects of the present application may be expressed as computer products located in one or more computer-readable media, which include computer-readable program code. For example, computer-readable media may include, but are not limited to, magnetic storage devices (e.g., hard disks, floppy disks, tapes...), optical disks (e.g., compact disks CDs, digital versatile disks DVDs...), smart cards, and flash memory devices (e.g., cards, sticks, key drives...).

[0056] A computer-readable medium may include a propagated data signal embodying computer program code, for example, in baseband or as part of a carrier wave. The propagated signal may be in a variety of forms, including electromagnetic, optical, etc., or a suitable combination thereof. A computer-readable medium may be any computer-readable medium other than a computer-readable storage medium that can be connected to an instruction execution system, apparatus, or device to communicate, propagate, or transmit the program for use. The program code on the computer-readable medium may be transmitted via any suitable medium, including radio, cable, fiber optic cable, radio frequency signal, or similar medium, or any combination of the above.

[0057] Similarly, it should be noted that, in order to simplify the description of this application and thus facilitate understanding of one or more embodiments of the application, the foregoing description of the embodiments of this application sometimes combines multiple features into a single embodiment, figure, or description thereof. However, this disclosure method does not mean that the subject matter of this application requires more features than those recited in the claims. In fact, the features of an embodiment may be fewer than all the features of the individual embodiments disclosed above.

[0058] In some embodiments, numbers are used to describe the quantity of components and attributes. It should be understood that such numbers used in the description of the embodiments are modified by the modifiers "about", "approximately" or "substantially" in some examples. Unless otherwise stated, "about", "approximately" or "substantially" indicate that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the description and claims are approximate values, which may change according to the required features of individual embodiments. In some embodiments, the numerical parameters should take into account the specified significant digits and adopt the general method of retaining digits. Although the numerical domains and parameters used to confirm the breadth of their range in some embodiments of the present application are approximate values, in specific embodiments, the settings of such numerical values ​​are as accurate as possible within the feasible range.

[0059] Although the present application has been described with reference to the current specific embodiments, ordinary technicians in this technical field should recognize that the above embodiments are only used to illustrate the present application, and various equivalent changes or substitutions can be made without departing from the spirit of the present application. Therefore, as long as the changes and modifications to the above embodiments are within the scope of the essential spirit of the present application, they will fall within the scope of the claims of the present application.

Claims

1. A fault diagnosis method for a photovoltaic tracking bracket, characterized in that, It includes the following steps: During the operation of the photovoltaic tracking bracket, the one-dimensional time-domain operation signal of the photovoltaic tracking bracket is continuously collected through the tracking control unit corresponding to the photovoltaic tracking bracket; Convert the one-dimensional time-domain operation signal into a two-dimensional time-frequency domain operation signal; Extract features from the two-dimensional time-frequency domain operation signal to obtain an operation feature map; And Determine whether the operation feature map belongs to a fault feature map, so as to diagnose whether the photovoltaic tracking bracket has a fault.

2. The fault diagnosis method according to claim 1, characterized in that, It also includes extracting the information of the one-dimensional time-domain operation signal on the time scale through wavelet transform to capture the change characteristics of the one-dimensional time-domain operation signal on different time scales, so as to obtain the two-dimensional time-frequency domain operation signal.

3. The fault diagnosis method according to claim 2, wherein, It also includes obtaining wavelet transform coefficients through the following formula Wherein: is the mother wavelet, Ψ is the complex conjugate mother wavelet, x(t) is the input signal, t is the time, a is the scaling factor, and b is the translation factor.

4. The fault diagnosis method according to claim 1, wherein It also includes: Pre-acquire multiple operation feature maps for different fault types, and obtain a fault type recognition model through training based on a convolutional neural network; And During the operation of the photovoltaic tracking bracket, input the operation feature map into the fault type recognition model in real time, and determine whether the operation feature map belongs to a fault feature map according to the output of the fault type recognition model.

5. The fault diagnosis method according to claim 4, wherein, During the training process, it also includes: Divide each or multiple of the multiple operation feature maps into multiple image regions; Generated by average pooling The first feature map and generate it using max pooling Second feature map, where: R 1×H×W represents an image with 1 channel, height H, and width W. F is the feature map input to the network, and its superscript S is used to distinguish the feature map after the pooling operation; Connect the first feature map and the second feature map into a two-way feature map; and Generate a spatial attention map through convolution to obtain the fault type recognition model.

6. The fault diagnosis method according to claim 1, wherein The one-dimensional time-domain operation signal includes the current signal of the drive motor of the photovoltaic tracking bracket.

7. The fault diagnosis method according to any one of claims 1 to 6, characterized in that It also includes determining whether the running feature map belongs to a fault feature map by using the following cross-entropy loss function: where: N is the number of categories, w i is the weight vector of the i-th category in the fully connected layer, q i is the label smoothing factor, a is the scaling factor, T represents transpose, and f is the integration function.

8. The fault diagnosis method according to claim 7, wherein It also includes calculating the label smoothing factor q using the following formula i : Where: ε is the label smoothing parameter, which is used to reduce the weight of the true label when calculating the loss function, prevent the model from overfitting the training data, and thus improve its generalization ability, and y is the one-hot label.

9. A fault diagnosis system for a photovoltaic tracking bracket, including: Multiple tracking control units, adapted to continuously collect the one-dimensional time-domain operation signal of the photovoltaic tracking bracket during the operation of the photovoltaic tracking bracket; A data acquisition and supervision control system, adapted to diagnose whether the photovoltaic tracking bracket has a fault according to the fault diagnosis method according to any one of claims 1 to 8.

10. A computer-readable medium storing computer program code, where the computer program code implements the method according to any one of claims 1-8 when executed by a processor.

Images