A method, apparatus, equipment and medium for fault detection of photovoltaic equipment

Abstract

This application provides a method, apparatus, device, and medium for photovoltaic (PV) equipment fault detection, relating to the field of PV equipment maintenance. The method includes: acquiring the output voltage and current values ​​of the PV equipment under different temperatures and light intensities; determining a first training sample set; augmenting the training samples to obtain a second training sample set; training the second training sample set using a local Mahalanobis distance algorithm; determining the anchor point set and fault region boundary for detecting PV equipment faults; and finally, performing fault detection on the target PV equipment based on this set. This approach reduces the number of samples required for training, improving work efficiency. Furthermore, the augmented training samples cover various operating conditions and models of PV equipment, greatly enhancing the versatility of the fault detection method. Moreover, the local Mahalanobis distance algorithm can identify subtle deviations, resulting in higher sensitivity of the fault detection results, achieving accurate and efficient detection of PV equipment faults.

Description

Technical Field

[0001] This application relates to the field of photovoltaic equipment maintenance, and in particular to a method, apparatus, equipment and medium for detecting photovoltaic equipment faults. Background Technology

[0002] In the development of new energy power generation, photovoltaic (PV) power generation equipment has experienced rapid growth. However, PV equipment often operates in complex outdoor environments, making it prone to various malfunctions. These malfunctions not only reduce power generation efficiency but can also endanger the power grid and personnel safety. Therefore, rapid and accurate fault detection is crucial for ensuring the safety and reliability of new energy power generation systems.

[0003] Fault detection technology for photovoltaic (PV) equipment typically relies on the physical characteristics of the equipment. It identifies anomalies by mathematically modeling the power generation process. Common PV equipment fault detection algorithms include autoencoders, principal component analysis, and mathematical morphology methods. However, differences in PV equipment models, operating conditions, and system configurations often lead to the failure of these methods, resulting in insufficient accuracy. Furthermore, an insufficient number of fault samples can cause class imbalance in the dataset, leading to poor prediction results. Collecting and organizing a large number of fault samples also increases the workload.

[0004] Therefore, an accurate and efficient method for detecting photovoltaic equipment faults is urgently needed. Summary of the Invention

[0005] Based on the above-mentioned technical problems, this application provides a photovoltaic equipment fault detection method, device, equipment and medium, aiming to achieve high-precision, high-efficiency and robust photovoltaic equipment fault detection.

[0006] The first aspect of this application provides a method for detecting faults in photovoltaic equipment, the method comprising:

[0007] The output voltage and current values ​​of the photovoltaic device are obtained under different temperatures and different light intensities to determine a first training sample set, wherein each training sample in the first training sample set includes: open circuit voltage, short circuit current, maximum power, maximum power point voltage and maximum power point current;

[0008] An offset is introduced into the first training sample set to perform data augmentation, and the second training sample set after data augmentation is determined.

[0009] The second training sample set is trained using the local Mahalanobis distance algorithm to generate an anchor point set and determine the fault area boundary of the photovoltaic equipment.

[0010] Fault detection of the target photovoltaic equipment is performed based on the set of anchor points and the boundary of the fault area.

[0011] Optionally, an offset is introduced into the first training sample set for data augmentation to determine the augmented second training sample set, including:

[0012] Data augmentation is performed by introducing offsets into each training sample in the first training sample set through various dimensions, wherein each dimension includes at least: open-circuit voltage, short-circuit current, maximum power, maximum power point voltage, and maximum power point current.

[0013] The training samples after introducing the offset, along with the training samples in the first training sample set, are determined as the second training sample set.

[0014] Optionally, data augmentation is performed by introducing offsets into the first training sample set across various dimensions, including:

[0015] Simultaneously, a first offset is introduced into each dimension of each training sample in the first training sample set to determine the first offset training sample set;

[0016] A second offset is introduced into one dimension of each training sample in the first training sample set to determine the second offset training sample set. The second offset should be greater than the first offset.

[0017] The first training sample set, the first offset training sample set, and the second offset training sample set are determined as the second training sample set.

[0018] Optionally, the step of determining the data dimensions included in each training sample in the first training sample set, through various dimensions, includes:

[0019] Under the same temperature and light intensity, the voltage and current values ​​output by normal photovoltaic equipment and each faulty photovoltaic equipment are collected as test results. The faulty photovoltaic equipment includes at least: partial open circuit, partial short circuit, partial shading, and partial aging.

[0020] The test results of each faulty photovoltaic device are matched with the test results of the normal photovoltaic devices to determine the abnormal indicator dimensions in the test results of the faulty photovoltaic devices.

[0021] The dimension of the anomaly indicator is determined as the data dimension included in the training samples of the first training sample set.

[0022] Optionally, the second training sample set is trained using a local Mahalanobis distance algorithm to generate an anchor point set and determine the fault region boundary of the photovoltaic equipment, including:

[0023] Based on the second training sample set, determine the minimum number of local samples and the local radius;

[0024] The anchor set of the second training sample set is determined based on the minimum number of local samples and the local radius.

[0025] Optionally, training the second training sample set using a local Mahalanobis distance algorithm to generate an anchor point set and determine the fault region boundary of the photovoltaic equipment further includes:

[0026] Set the confidence level value;

[0027] Based on the confidence level value and the second training sample set, the fault region boundary of the second training sample set is determined.

[0028] Optionally, fault detection of the target photovoltaic equipment is performed based on the set of anchor points and the boundary of the fault area, including:

[0029] Collect the voltage and current values ​​output by the target photovoltaic device, and calculate the open-circuit voltage, short-circuit current, maximum power, maximum power point voltage, and maximum power point current of the target device;

[0030] Calculate the target local Mahalanobis distance of the target photovoltaic device based on the set of anchor points;

[0031] The target local Mahalanobis distance is compared with the boundary of the fault area. When the target local Mahalanobis distance is greater than the boundary of the fault area, the target photovoltaic device is determined to be in a fault state.

[0032] Once the target photovoltaic device is determined to be in a fault state, a fault alarm is issued for the fault state.

[0033] A second aspect of this application provides a photovoltaic equipment fault detection device, the device comprising:

[0034] The first training sample set acquisition module is used to acquire the output voltage and current values ​​of the photovoltaic device under different temperatures and different light intensities, and to determine the first training sample set. Each training sample in the first training sample set includes: open circuit voltage, short circuit current, maximum power, maximum power point voltage, and maximum power point current.

[0035] The second training sample set determination module is used to introduce an offset into the first training sample set for data augmentation and determine the second training sample set after data augmentation.

[0036] The local Mahalanobis distance algorithm training module is used to train the local Mahalanobis distance algorithm on the second training sample set, generate an anchor point set, and determine the fault area boundary of the photovoltaic equipment.

[0037] The fault detection module is used to detect faults in the target photovoltaic equipment based on the set of anchor points and the boundary of the fault area.

[0038] Optionally, the second training sample set determination module includes:

[0039] The offset introduction submodule is used to introduce offsets into each training sample in the first training sample set for data augmentation through various dimensions, wherein each dimension includes at least: open circuit voltage, short circuit current, maximum power, maximum power point voltage and maximum power point current.

[0040] The second training sample set augmentation submodule is used to determine the second training sample set by taking the training samples after introducing the offset and the training samples in the first training sample set.

[0041] Optionally, the offset is introduced into a submodule, including:

[0042] The first offset introduction unit is used to simultaneously introduce a first offset into each dimension of each training sample in the first training sample set to determine the first offset training sample set.

[0043] The second offset introduction unit is used to introduce a second offset into one dimension of each training sample in the first training sample set to determine the second offset training sample set. The second offset should be greater than the first offset.

[0044] The second training sample set determination unit is used to determine the first training sample set, the first offset training sample set, and the second offset training sample set as the second training sample set.

[0045] Optionally, the first training sample set acquisition module includes:

[0046] The test result acquisition submodule is used to collect the voltage and current values ​​output by normal photovoltaic equipment and each faulty photovoltaic equipment under the same temperature and light intensity as test results. The faulty photovoltaic equipment includes at least: partial open circuit, partial short circuit, partial shading, and partial aging.

[0047] The abnormal indicator dimension determination submodule is used to match the test results of each faulty photovoltaic device with the test results of the normal photovoltaic device to determine the abnormal indicator dimension in the test results of the faulty photovoltaic device.

[0048] The data dimension determination submodule is used to determine the anomaly indicator dimension as the data dimension included in the training samples of the first training sample set.

[0049] Optionally, the local Mahalanobis distance algorithm training module includes:

[0050] The anchor point generation condition determination submodule is used to determine the minimum number of local samples and the local radius based on the second training sample set;

[0051] An anchor point generation submodule is used to determine the anchor point set of the second training sample set based on the minimum number of local samples and the local radius.

[0052] Optionally, the local Mahalanobis distance algorithm training module further includes:

[0053] The confidence level setting submodule is used to set the confidence level value;

[0054] The fault region boundary determination submodule is used to determine the fault region boundary of the second training sample set based on the confidence level value and the second training sample set.

[0055] Optionally, the fault detection module includes:

[0056] The output signal acquisition submodule is used to acquire the voltage and current values ​​output by the target photovoltaic device, and to calculate the open-circuit voltage, short-circuit current, maximum power, maximum power point voltage, and maximum power point current of the target device.

[0057] The target local Mahalanobis distance calculation submodule is used to calculate the target local Mahalanobis distance of the target photovoltaic device based on the set of anchor points.

[0058] The target local Mahalanobis distance comparison submodule is used to compare the target local Mahalanobis distance with the fault region boundary, and determine that the target photovoltaic device is in a fault state when the target local Mahalanobis distance is greater than the fault region boundary.

[0059] The fault alarm submodule is used to issue a fault alarm for the target photovoltaic equipment after determining that the equipment is in a fault state.

[0060] A third aspect of this application provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the photovoltaic equipment fault detection method of the first aspect of this application.

[0061] A fourth aspect of this application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the photovoltaic equipment fault detection method of the first aspect of this application.

[0062] The fifth aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements the photovoltaic equipment fault detection method of the first aspect of this application.

[0063] The photovoltaic equipment fault detection method of this application obtains the output voltage and current values ​​of the photovoltaic equipment under different temperatures and light intensities to determine a first training sample set. Each training sample in the first training sample set includes: open circuit voltage, short circuit current, maximum power, maximum power point voltage, and maximum power point current. Then, by adding offsets to multiple dimensions most relevant to photovoltaic equipment faults, the training samples are augmented to obtain a second training sample set for training. The local Mahalanobis distance algorithm is then used to train the second training sample set. The training results are obtained through training, which are the anchor point set and fault region boundary used to detect photovoltaic equipment faults. Finally, the target photovoltaic equipment is fault detected based on the anchor point set and fault region boundary.

[0064] Unlike previous photovoltaic (PV) equipment fault detection methods, this application does not require collecting a large amount of PV equipment fault data. Instead, it expands the training samples by introducing offsets based on normal data under different operating conditions, thus determining the anchor point set and the fault region boundary. This reduces the number of samples required for training and improves efficiency. By forward training on the non-faulty sample set under normal conditions, the range of the output results of non-faulty PV equipment, i.e., the fault region boundary, is determined. Then, by calculating the local Mahalanobis distance between the target PV equipment and the anchor point set and comparing the result with the fault region boundary, the deviation between the target PV equipment's signal and normal conditions is effectively evaluated. The calculation of the local Mahalanobis distance can identify subtle deviations, resulting in higher sensitivity and enabling rapid and accurate determination of whether the target PV equipment is faulty. Furthermore, the introduction of multi-dimensional offsets greatly improves the reliability of the detection method, enabling it to handle PV equipment fault detection under various environmental conditions and is not limited to a specific model or operating condition of PV equipment, significantly enhancing the robustness of the proposed detection method. Attached Figure Description

[0065] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0066] Figure 1 This is a flowchart of a photovoltaic equipment fault detection method proposed in one embodiment of this application;

[0067] Figure 2 This is a schematic diagram of a photovoltaic equipment simulation system proposed in one embodiment of this application;

[0068] Figure 3 This is a schematic diagram of the simulation results of a photovoltaic device proposed in one embodiment of this application;

[0069] Figure 4 This is a schematic diagram of the normal photovoltaic equipment test results according to an embodiment of this application;

[0070] Figure 5 This is a schematic diagram of the abnormal photovoltaic equipment detection results proposed in an embodiment of this application;

[0071] Figure 6 This is a statistical chart showing the distribution of photovoltaic equipment test results according to an embodiment of this application;

[0072] Figure 7 This is a structural block diagram of a photovoltaic equipment fault detection device proposed in an embodiment of this application;

[0073] Figure 8 This is a schematic diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0074] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0075] In the accompanying drawings, the size of constituent elements, the thickness of layers, or areas may sometimes be exaggerated for clarity. Therefore, any implementation of this disclosure is not necessarily limited to the dimensions shown in the drawings, and the shapes and sizes of the components in the drawings do not reflect true proportions. Furthermore, the drawings schematically illustrate ideal examples, and any implementation of this disclosure is not limited to the shapes or values ​​shown in the drawings.

[0076] In related technologies, fault detection methods for photovoltaic (PV) equipment typically involve mathematically modeling the PV power generation process, forming a fault detection model based on the fault conditions, and then using this model to detect faults in the PV equipment. Commonly used PV equipment detection methods in related technologies include: autoencoder method, which analyzes the output characteristics of PV equipment and uses an autoencoder to extract and classify fault features, thereby achieving fault diagnosis; principal component analysis method, which analyzes the output characteristic curves of different faults, extracts feature quantities reflecting different fault characteristics, uses principal component analysis (PCA) to reduce the dimensionality of fault data and perform preliminary detection, uses the CatBoost algorithm to classify fault data, and establishes a fault diagnosis model based on the PCA-CatBoost algorithm; and mathematical morphology method, which treats power curtailment anomaly data as noise signals in the original data through mathematical morphology, and performs denoising processing through basic operations such as expansion and erosion in mathematical morphology to achieve adaptive identification of anomaly data.

[0077] However, while the method of building a fault detection model based on fault data can also achieve the purpose of fault detection for related photovoltaic equipment, its application is limited and can only meet the fault detection needs of photovoltaic equipment related to the collection of fault data. Furthermore, building a fault detection model requires a large amount of fault data as support. This process not only consumes a lot of the testing personnel's energy, but may also lead to class imbalance in the dataset due to insufficient sample size, thereby affecting the fault detection results.

[0078] Therefore, in order to achieve accurate and efficient fault detection of photovoltaic equipment, this application proposes a method for photovoltaic equipment fault detection, the details of which can be found in the following reference. Figure 1 , Figure 1 This is a flowchart of a photovoltaic equipment fault detection method according to this application. Figure 1 As shown, the method may include steps S101 to S104:

[0079] Step S101: Obtain the output voltage and current values ​​of the photovoltaic device under different temperatures and different light intensities, and determine the first training sample set, wherein each training sample in the first training sample set includes: open circuit voltage, short circuit current, maximum power, maximum power point voltage and maximum power point current.

[0080] Step S102: Introduce an offset into the first training sample set to perform data augmentation, and determine the second training sample set after data augmentation;

[0081] Step S103: Train the second training sample set using the local Mahalanobis distance algorithm to generate an anchor point set and determine the fault area boundary of the photovoltaic equipment;

[0082] Step S104: Perform fault detection on the target photovoltaic equipment based on the set of anchor points and the boundary of the fault area.

[0083] This application's fault detection for photovoltaic equipment mainly consists of two stages. The first stage involves training a local Mahalanobis distance algorithm on a sample set. First, voltage and current values ​​output by normally operating photovoltaic equipment under different temperatures and light intensities are collected to calculate training samples for various conditions, including open-circuit voltage, short-circuit current, maximum power, maximum power point voltage, and maximum power point current. These data under normal operating conditions form the first training sample set, which is the foundation of the training samples. Then, to further enrich the sample space of the training sample set and cover more possible situations, different dimensional offsets are introduced into the first training sample set for data augmentation, resulting in a second training sample set with richer sample conditions. Next, the local Mahalanobis distance algorithm is trained on the second training sample set. Multiple anchor points covering the samples in the second training sample set are generated using an anchor point generation algorithm, and then region boundary optimization is performed to determine the fault region boundary.

[0084] The second step involves using the anchor point set and fault area boundary calculated in the previous steps to detect the target photovoltaic equipment that needs fault detection. Specifically, by collecting the voltage and current values ​​output by the target photovoltaic equipment, the open-circuit voltage, short-circuit current, maximum power, maximum power point voltage, and maximum power point current of the target photovoltaic equipment are determined. The local Mahalanobis distance of the target photovoltaic equipment is then calculated using these data. By comparing the calculation results with the fault area boundary obtained in the first step, it is verified whether the local Mahalanobis distance of the target photovoltaic equipment is within the fault area boundary, thereby determining whether the target photovoltaic equipment has a fault.

[0085] Unlike related technologies that require a large amount of fault data as samples, the photovoltaic equipment fault detection method proposed in this application is data-driven. It does not require research on the entire charging, discharging, and power generation processes of the photovoltaic equipment; it only needs to collect data from normal operation under different working conditions to train the local Mahalanobis distance algorithm. Furthermore, because this application collects data under different working conditions and further augments it based on multiple dimensions, it contains a sufficient number of samples to cover various types of photovoltaic equipment and data under various operating conditions. Compared to related technologies, the method proposed in this application is more versatile. Moreover, the advantage of using the local Mahalanobis distance algorithm lies in its ability to identify subtle deviations, resulting in higher sensitivity and more accurate detection results for the photovoltaic equipment fault detection method.

[0086] Step S101: Obtain the output voltage and current values ​​of the photovoltaic device under different temperatures and light intensities, and determine the first training sample set, wherein each training sample in the first training sample set includes: open circuit voltage, short circuit current, maximum power, maximum power point voltage, and maximum power point current.

[0087] In this embodiment, the first step is to obtain original training samples, which need to include normal operating data under various working conditions. Specifically, due to the operation of photovoltaic equipment, light intensity and operating temperature significantly affect the output results of the photovoltaic equipment. Therefore, it is necessary to collect output data of photovoltaic equipment under different temperatures and light intensities. Then, based on the output current and voltage values, the open-circuit voltage, short-circuit current, maximum power, maximum power point voltage, and maximum power point current of each training sample are determined. In an optional embodiment, the operating temperature and light intensity can be set to different levels, and different operating conditions can be selected through permutation and combination. For example, the temperature can be set from 15°C to 45°C, with each level in 5°C increments, and the light intensity can be set from 400W / m² to 1000W / m², with each level in 100m² increments, thus obtaining data under 49 different operating conditions. Generally speaking, more training samples can enrich the possible situations included in the final fault detection method and make the fault detection results more accurate.

[0088] Step S102: Introduce an offset into the first training sample set to perform data augmentation, and determine the second training sample set after data augmentation.

[0089] In this embodiment of the application, although a large amount of data under different working conditions has been selected as the first training sample set in step S101, the actual working environment of photovoltaic equipment is far more complex. Therefore, in order to further improve the adaptability of the local Mahalanobis distance algorithm to the complex working environment of photovoltaic equipment, it is also necessary to augment the collected first training sample set to further expand the number of training samples so that it can cover more possibilities.

[0090] Step S103: Train the second training sample set using the local Mahalanobis distance algorithm to generate an anchor point set and determine the fault area boundary of the photovoltaic equipment.

[0091] In this embodiment, after determining the second training sample set for the training algorithm, it is necessary to train it using the Local Mahalanobis Distance (LMD) algorithm. LMD originates from the optimal decision boundary problem in single-class learning, aiming to fit an optimal decision boundary based on single-class data. Then, based on the decision boundary, it determines whether the test sample belongs to this decision region covering normal data. The decision region consists of multiple sub-regions with different centers that can overlap but have the same radius. The center of each sub-region is called an anchor point, and the radius of the sub-region is set to a fixed value, i.e., the fault region boundary. Using this method, when a fault occurs, the test sample will gradually deviate from the decision region. When the sample exceeds the decision region, it can be determined as a faulty behavior.

[0092] Step S104: Perform fault detection on the target photovoltaic equipment based on the set of anchor points and the boundary of the fault area.

[0093] In this embodiment of the application, after training with the local Mahalanobis distance algorithm, the fault status of the target photovoltaic equipment can be detected. Specifically, by collecting the output signal of the target photovoltaic equipment, calculating the local Mahalanobis distance between the target photovoltaic equipment and each anchor point, and comparing it with the boundary of the fault area, it can be determined whether it is within the normal range, thereby determining whether the target photovoltaic equipment has a fault.

[0094] In conjunction with the above embodiments, in one implementation, this application also provides a photovoltaic equipment fault detection method, which introduces an offset into the first training sample set for data augmentation, and determines a second training sample set after data augmentation, specifically including the following:

[0095] First, offsets are introduced into each training sample in the first training sample set to augment the data through various dimensions, including at least: open-circuit voltage, short-circuit current, maximum power, maximum power point voltage, and maximum power point current.

[0096] Then, the training samples after introducing the offset and the training samples in the first training sample set are determined as the second training sample set.

[0097] In this embodiment, since the first training sample set only contains a portion of training samples obtained from detections under different light intensities and temperatures, augmentation of the first training sample set is necessary to improve the adaptability of the fault detection method to complex environments and different operating conditions. Each training sample includes data in five dimensions: open-circuit voltage, short-circuit current, maximum power, maximum power point voltage, and maximum power point current. Therefore, during data augmentation, at least these five dimensions can be used to introduce offsets to simulate more data on the normal operation of photovoltaic devices under different conditions. After data augmentation, the training samples with introduced offsets, along with the original training samples in the first training sample set, are collectively determined as the second training sample set for subsequent training using the local Mahalanobis distance algorithm.

[0098] In conjunction with the above embodiments, in one implementation, this application also provides a photovoltaic equipment fault detection method, which introduces offsets into the first training sample set through various dimensions to augment the data, specifically including the following:

[0099] First, a first offset is introduced into each dimension of each training sample in the first training sample set to determine the first offset training sample set.

[0100] In this application embodiment, offsets can be introduced into the first training sample set through various methods. This application proposes a method for introducing offsets by simultaneously introducing the same offset into the data of each dimension in the training samples, to simulate the situation where the data of each dimension is simultaneously affected under different environments. In an optional embodiment, based on the variance μ0 of each training sample in the first training sample set, white noise with a variance of 0.01μ0 is simultaneously introduced into the data of each dimension, thereby obtaining the first offset training sample set after introducing noise.

[0101] Then, a second offset is introduced into one dimension of each training sample in the first training sample set to determine the second offset training sample set, wherein the second offset should be greater than the first offset.

[0102] In this embodiment, a second offset, larger than the first offset, can be introduced into one dimension of the training samples. The sample set after introducing the second offset serves as the second offset training sample set, used to simulate the differences between the training samples in the first training sample set caused by variations in photovoltaic equipment models, operating conditions, etc. Since the first offset in this application simulates data under normal conditions with noise influence, the data offset caused by differences in photovoltaic equipment should generally be greater than the offset caused by noise. Therefore, the second offset used to determine the second offset training sample set should be greater than the first offset. In an optional embodiment, as described above, the second offset training sample set can be obtained by introducing an offset of ±0.1μ0 into one of the five dimensions, while keeping the data in the other dimensions unchanged.

[0103] The first training sample set, the first offset training sample set, and the second offset training sample set are determined as the second training sample set.

[0104] In this embodiment, the first offset training sample set and the second offset training sample set with introduced offsets are combined with the original first training sample set to form the second training sample set. During the preparation of training samples, if only actual test samples are collected, a large number of test samples need to be collected, which is inefficient. By introducing offsets into the actual measured first training sample set, different offset sizes are selected to simulate various working scenarios that photovoltaic equipment may actually experience, increasing the number of samples in the training sample set and making the trained results more stable.

[0105] In conjunction with the above embodiments, in one implementation, this application also provides a photovoltaic equipment fault detection method. The step of determining the data dimensions included in each training sample in the first training sample set specifically includes the following:

[0106] First, under the same temperature and light intensity, the voltage and current values ​​output by normal photovoltaic equipment and each faulty photovoltaic equipment are collected as test results. The faulty photovoltaic equipment includes at least: partial open circuit, partial short circuit, partial shading, and partial aging.

[0107] In this embodiment, the local Mahalanobis distance algorithm is typically used to process time-series signals. However, in fault detection of photovoltaic (PV) equipment, the focus is usually on the current-voltage characteristic curve of the PV equipment output. To utilize the local Mahalanobis distance algorithm for PV equipment fault detection, this application needs to determine several data dimensions for processing using the algorithm. To determine which specific parameters are more closely related to the PV equipment's fault condition, this application uses PV equipment operating under simulated temperature and light intensity conditions, along with PV equipment exhibiting common faults such as partial open circuits, partial short circuits, partial shading, and partial aging. By comparing the differences between them, the data dimensions are determined.

[0108] Specifically, in one optional embodiment, this application has built a 3×3 photovoltaic equipment simulation system based on the Simulink platform, such as... Figure 2 As shown, Figure 2 This is a schematic diagram of a photovoltaic equipment simulation system proposed in an embodiment of this application. The system simulates a partial open circuit fault by disconnecting one branch while keeping the other branches normal; a partial short circuit fault by short-circuiting both sides of a photovoltaic panel in one branch while keeping the other branches normal; a partial shading fault by setting the light intensity of one photovoltaic panel in one branch to half the normal light intensity of the other photovoltaic panels while keeping the other branches normal; and a partial aging fault by connecting a resistor in series in one branch while keeping the other branches normal.

[0109] Then, the test results of each faulty photovoltaic device are matched with the test results of the normal photovoltaic devices to determine the abnormal indicator dimensions in the test results of the faulty photovoltaic devices.

[0110] In this embodiment, the test results of photovoltaic equipment under various fault conditions are compared with the test results of normal photovoltaic equipment to perform data matching. The differences between the test results and those of normal photovoltaic equipment are used to determine the abnormal indicator dimensions. Specifically, this application analyzes the current-voltage characteristic curves and power-voltage characteristic curves of photovoltaic equipment under normal and fault conditions to compare the abnormal indicator dimensions.

[0111] In an optional embodiment, the photovoltaic equipment simulation system described above is used for testing to obtain the output characteristic curves of the photovoltaic equipment under different conditions, such as... Figure 3 As shown, Figure 3This is a schematic diagram of the simulation results of a photovoltaic device according to an embodiment of this application. The temperature is set to 25°C and the light intensity is 1000W / ㎡. The solid line represents the simulation result curve of the photovoltaic device output under normal conditions. By comparing it with the simulation result curves of the photovoltaic device output under other fault conditions, it can be found that:

[0112] The multi-point line represents the simulation result curve when there is a partial open circuit fault. Comparing it with the simulation result under normal conditions, the main difference is that the short-circuit current decreases and the current at the maximum power point decreases.

[0113] The long dashed line represents the simulation result curve when there is a partial short circuit fault. Comparing it with the simulation result under normal conditions, the main difference is that the open circuit voltage is significantly reduced and the maximum power point voltage is reduced.

[0114] The dotted line represents the simulation result curve when there is a partial shading fault. Comparing it with the simulation result under normal conditions, the main difference is that the output characteristic curve of the photovoltaic equipment will show a "multiple knees" phenomenon.

[0115] The dashed line represents the simulation result curve when there is a local aging fault. Comparing it with the simulation result under normal conditions, the main difference is that the open circuit voltage and short circuit current do not change, but the inflection point of the current-voltage curve will shift downward and the maximum power point current will also decrease.

[0116] Finally, the dimension of the anomaly indicator is determined to be the data dimension included in the training samples of the first training sample set.

[0117] In this embodiment, based on the comparison of test results of faulty photovoltaic equipment and normal photovoltaic equipment as described above, it is evident that the data changes in five dimensions—open-circuit voltage, short-circuit current, maximum power, maximum power point voltage, and maximum power point current—are particularly significant. Therefore, these dimensions are identified as key areas for fault identification and detection and are determined as the data dimensions included in the training samples of the first training sample set. Similarly, since these five data dimensions are significantly affected by photovoltaic equipment faults, they are prioritized for augmentation of the training samples during training data augmentation.

[0118] In conjunction with the above embodiments, in one implementation, this application also provides a photovoltaic equipment fault detection method, which trains the second training sample set using a local Mahalanobis distance algorithm to generate an anchor point set and determine the fault region boundary of the photovoltaic equipment, specifically including the following:

[0119] First, based on the second training sample set, determine the minimum number of local samples and the local radius.

[0120] In this embodiment of the application, after the training sample set is determined, it is necessary to train it with the local Mahalanobis distance algorithm. Specifically, the training of the local Mahalanobis distance algorithm is an unsupervised process. Its purpose is to use the training samples under normal conditions to determine a set of anchor points and determine a region radius to mark the spatial range to which the normal samples belong. Therefore, the training of the local Mahalanobis distance algorithm is mainly divided into two steps. The first step is to use the anchor point generation algorithm to determine the anchor point set.

[0121] Specifically, the anchor point generation algorithm uses the positions of several spatially adjacent normal samples to determine the center point that can cover its spatial range as an anchor point. By setting a minimum local sample size η, the minimum number of normal samples covered around each selected anchor point is determined. By setting a reasonable value, outliers that may exist in the training samples are filtered out, ensuring the rationality of the training results. It is also necessary to determine the local radius γ, which is used to determine the size of the local region around each anchor point. The local radius can affect the accuracy of the final generated decision region. Specifically, choosing a larger local radius requires fewer anchor points to cover all normal samples, but the number of anchor points available for calculating the local Mahalanobis distance is also reduced, resulting in lower accuracy of the final generated decision region. However, the computational load generated by fewer anchor points is also reduced, improving overall efficiency. Conversely, choosing a smaller local radius increases the number of anchor points, which can improve the accuracy of the decision region, but also increases the computational load. Therefore, choosing an appropriate minimum sample size and an appropriate local radius helps to improve the accuracy and efficiency of the fault detection method.

[0122] In an alternative embodiment, the local radius can be determined by taking into account the region fitting error. Regardless of the number of local samples, the region fitting error is always a convex function of the local radius, and the optimal local radius value can be searched by minimizing the equation.

[0123] Then, based on the minimum number of local samples and the local radius, the anchor set of the second training sample set is determined.

[0124] In this embodiment of the application, after determining the minimum number of local samples and the local radius error, the anchor points that can cover the normal data can be calculated to form the anchor point set of the training sample set, which serves as the basis for determining the decision region of normal working conditions.

[0125] In conjunction with the above embodiments, in one implementation, this application also provides a photovoltaic equipment fault detection method, which trains the second training sample set using a local Mahalanobis distance algorithm to generate an anchor point set and determine the fault region boundary of the photovoltaic equipment, specifically including the following:

[0126] First, set the confidence level.

[0127] In this embodiment, after determining the set of anchor points, it is also necessary to determine the boundary of the fault region. Based on the properties of local Mahalanobis distance, this index can be considered a random variable under the null hypothesis. Once its probability distribution model is determined, the corresponding fault region boundary can be determined based on a given confidence level value α. The confidence level value is used to determine that the normal data covered by the anchor point set and the fault region boundary may not be 100%, in order to avoid the impact of outliers in the training samples on the overall accuracy range.

[0128] Then, based on the confidence level value and the second training sample set, the fault region boundary of the second training sample set is determined.

[0129] In this embodiment of the application, after setting the confidence level value, the boundary of the fault area can be determined based on the second training sample set and the confidence level value. Specifically, in an optional embodiment, a distribution model of the local Mahalanobis distance exponent under the null hypothesis can be established by using a generalized extreme value distribution model. The model parameters are obtained by optimizing the mean square error between the generalized extreme value distribution model and the empirical cumulative density. This parameter estimation process can be described by the convex optimization problem shown in the following formula.

[0130] minimize

[0131] Satisfy -β-τ(ρ-ξ) i )≤0,β>0

[0132] Where Φ(ξ) i ;ρ,β,τ) is the generalized extremum distribution function, This represents the accumulated empirical density. To solve convex optimization problems with multiple constraints, the Nelder-Mead method can be used for iterative search until the objective function (as shown in the formula above) converges. After estimating the model parameters, the fault region boundary ∈ can be determined by the confidence level value α, i.e.:

[0133]

[0134] Therefore, the boundary of the fault region can be determined based on the second training sample set and the confidence level value.

[0135] In conjunction with the above embodiments, in one implementation, this application also provides a photovoltaic equipment fault detection method, which performs fault detection on the target photovoltaic equipment based on the anchor point set and the fault area boundary, specifically including the following:

[0136] First, the voltage and current values ​​output by the target photovoltaic device are collected, and the open-circuit voltage, short-circuit current, maximum power, maximum power point voltage, and maximum power point current of the target device are calculated.

[0137] In this embodiment of the application, after training the training model using the local Mahalanobis distance algorithm, the target photovoltaic equipment can be fault detected based on the local Mahalanobis distance algorithm. Specifically, it is necessary to first collect the data output by the target photovoltaic equipment, and calculate the data of each dimension of the target equipment based on its output current and voltage values, namely: open circuit voltage, short circuit current, maximum power, maximum power point voltage, and maximum power point current.

[0138] Then, the target local Mahalanobis distance of the target photovoltaic device is calculated based on the set of anchor points.

[0139] In this embodiment of the application, once the open-circuit voltage, short-circuit current, maximum power, maximum power point voltage, and maximum power point current of the target photovoltaic device are determined, the Mahalanobis distance between it and each anchor point in the anchor point set can be calculated.

[0140] Specifically, when calculating the local Mahalanobis distance for two sample vectors X and Y, which come from the same distribution with mean μ and covariance ∑, the method for calculating the multivariate Mahalanobis distance for sample vector X is as follows:

[0141]

[0142] The formula for calculating the Mahalanobis distance between two sample vectors X and Y is as follows:

[0143]

[0144] Thus, the local Mahalanobis distance is defined as the shortest Mahalanobis distance from the sample to the decision region. In order to quantify the local Mahalanobis distance between the sample and the decision region, the decision region is represented as a finite number of points and called anchor points, and the fault region radius is set as another factor to limit the range of the decision region.

[0145] Let S denote the set of K anchor points, i.e., S = {C1, ..., C2}. k ,…,C K}, then the local Mahalanobis distance of the sample vector X is:

[0146]

[0147] In this way, the local Mahalanobis distance between the target photovoltaic device and the set of anchor points obtained from training can be calculated.

[0148] Next, the target local Mahalanobis distance is compared with the boundary of the fault area. When the target local Mahalanobis distance is greater than the boundary of the fault area, the target photovoltaic device is determined to be in a fault state.

[0149] In this embodiment, a larger local Mahalanobis distance indicates that the corresponding sample is further away from the healthy decision region. Therefore, it can be used to determine whether a fault state exists. Thus, by comparing the target local Mahalanobis distance of the target photovoltaic device with the boundary of the fault region obtained through training, it can be determined whether the target photovoltaic device is faulty. Specifically, when the target local Mahalanobis distance is greater than the boundary of the fault region, the target photovoltaic device is determined to be in a fault state; when the target local Mahalanobis distance is not greater than the boundary of the fault region, the target photovoltaic device is determined to be in a normal state.

[0150] Finally, after determining that the target photovoltaic equipment is in a fault state, a fault alarm is issued for the fault state.

[0151] In this embodiment of the application, after establishing the photovoltaic equipment fault detection method, when a fault condition is detected in the target photovoltaic equipment, the relevant personnel can be promptly notified of the fault condition through a fault alarm. Since the fault detection method proposed in this application is only a method of calculation based on data, the specific fault condition still needs to be judged by the relevant personnel. Therefore, timely fault alarms are issued to reduce the impact of faulty photovoltaic equipment.

[0152] In an optional embodiment, this application also proposes a method for verifying the established photovoltaic equipment fault detection method. Specifically, after determining the anchor point set and fault area boundary based on the second training sample set, the output signal of the photovoltaic equipment is re-collected under various normal and abnormal operating states (normal operation, partial open circuit, partial short circuit, partial shading, and partial aging) at different temperatures and light intensities. This includes 500 sets of data samples with temperatures ranging from 15°C to 45°C and light intensities ranging from 400W / ㎡ to 1000W / ㎡. Fault detection is then performed on these samples according to the target photovoltaic equipment fault detection method in this application embodiment, and the detection results are as follows: Figures 4-6 As shown, where, Figure 4 This is a schematic diagram of the detection results of normal photovoltaic equipment according to an embodiment of this application, wherein the local Mahalanobis distance calculated for all normal photovoltaic equipment is less than the boundary of the fault area; Figure 5 This is a schematic diagram of the abnormal photovoltaic equipment detection results proposed in an embodiment of this application. In this diagram, the calculated local Mahalanobis distance for all abnormal photovoltaic equipment is greater than the fault area boundary, and fault alarms are also triggered. This proves that all fault conditions can be successfully detected, and there are no missed or false detections. Furthermore, according to... Figure 6 As shown, Figure 6This is a statistical distribution chart of photovoltaic equipment test results proposed in an embodiment of this application. The local Mahalanobis distance index of normal photovoltaic equipment is concentrated in the range of 1-1.5, which is much smaller than the fault area boundary of 5.15 detected in this case. On the other hand, the local Mahalanobis distance index of faulty photovoltaic equipment is concentrated in the range of 100-500, which is much larger than the fault area boundary of 5.15 detected in this case. Therefore, it can be found that the photovoltaic equipment fault detection method proposed in this application can effectively distinguish between normal photovoltaic equipment and faulty photovoltaic equipment.

[0153] Based on the same design concept, one embodiment of this application provides a photovoltaic equipment fault detection device. (Reference) Figure 7 , Figure 7 This is a structural block diagram of a photovoltaic equipment fault detection device provided in one embodiment of this application. Figure 7 As shown, the device includes:

[0154] The first training sample set acquisition module is used to acquire the output voltage and current values ​​of the photovoltaic device under different temperatures and different light intensities, and to determine the first training sample set. Each training sample in the first training sample set includes: open circuit voltage, short circuit current, maximum power, maximum power point voltage, and maximum power point current.

[0155] The second training sample set determination module is used to introduce an offset into the first training sample set for data augmentation and determine the second training sample set after data augmentation.

[0156] The local Mahalanobis distance algorithm training module is used to train the local Mahalanobis distance algorithm on the second training sample set, generate an anchor point set, and determine the fault area boundary of the photovoltaic equipment.

[0157] The fault detection module is used to detect faults in the target photovoltaic equipment based on the set of anchor points and the boundary of the fault area.

[0158] Optionally, the second training sample set determination module includes:

[0159] The offset introduction submodule is used to introduce offsets into each training sample in the first training sample set for data augmentation through various dimensions, wherein each dimension includes at least: open circuit voltage, short circuit current, maximum power, maximum power point voltage and maximum power point current.

[0160] The second training sample set augmentation submodule is used to determine the second training sample set by taking the training samples after introducing the offset and the training samples in the first training sample set.

[0161] Optionally, the offset is introduced into a submodule, including:

[0162] The first offset introduction unit is used to simultaneously introduce a first offset into each dimension of each training sample in the first training sample set to determine the first offset training sample set.

[0163] The second offset introduction unit is used to introduce a second offset into one dimension of each training sample in the first training sample set to determine the second offset training sample set. The second offset should be greater than the first offset.

[0164] The second training sample set determination unit is used to determine the first training sample set, the first offset training sample set, and the second offset training sample set as the second training sample set.

[0165] Optionally, the first training sample set acquisition module includes:

[0166] The test result acquisition submodule is used to collect the voltage and current values ​​output by normal photovoltaic equipment and each faulty photovoltaic equipment under the same temperature and light intensity as test results. The faulty photovoltaic equipment includes at least: partial open circuit, partial short circuit, partial shading, and partial aging.

[0167] The abnormal indicator dimension determination submodule is used to match the test results of each faulty photovoltaic device with the test results of the normal photovoltaic device to determine the abnormal indicator dimension in the test results of the faulty photovoltaic device.

[0168] The data dimension determination submodule is used to determine the anomaly indicator dimension as the data dimension included in the training samples of the first training sample set.

[0169] Optionally, the local Mahalanobis distance algorithm training module includes:

[0170] The anchor point generation condition determination submodule is used to determine the minimum number of local samples and the local radius based on the second training sample set;

[0171] An anchor point generation submodule is used to determine the anchor point set of the second training sample set based on the minimum number of local samples and the local radius.

[0172] Optionally, the local Mahalanobis distance algorithm training module further includes:

[0173] The confidence level setting submodule is used to set the confidence level value;

[0174] The fault region boundary determination submodule is used to determine the fault region boundary of the second training sample set based on the confidence level value and the second training sample set.

[0175] Optionally, the fault detection module includes:

[0176] The output signal acquisition submodule is used to acquire the voltage and current values ​​output by the target photovoltaic device, and to calculate the open-circuit voltage, short-circuit current, maximum power, maximum power point voltage, and maximum power point current of the target device.

[0177] The target local Mahalanobis distance calculation submodule is used to calculate the target local Mahalanobis distance of the target photovoltaic device based on the set of anchor points.

[0178] The target local Mahalanobis distance comparison submodule is used to compare the target local Mahalanobis distance with the fault region boundary, and determine that the target photovoltaic device is in a fault state when the target local Mahalanobis distance is greater than the fault region boundary.

[0179] The fault alarm submodule is used to issue a fault alarm for the target photovoltaic equipment after determining that the equipment is in a fault state.

[0180] A third aspect of this application provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the photovoltaic equipment fault detection method of the first aspect of this application.

[0181] A fourth aspect of this application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the photovoltaic equipment fault detection method of the first aspect of this application.

[0182] The fifth aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements the photovoltaic equipment fault detection method of the first aspect of this application.

[0183] Based on the same design concept, another embodiment of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the photovoltaic equipment fault method as described in any of the above embodiments of this application.

[0184] Based on the same design concept, another embodiment of this application provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps in the photovoltaic equipment fault method as described in any of the above embodiments of this application.

[0185] Based on the same design concept, another embodiment of this application provides an electronic device, such as... Figure 8 As shown. Figure 8This is a schematic diagram of an electronic device according to an embodiment of this application. The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When executed by the processor, the program implements the steps of the photovoltaic device fault diagnosis method described in any of the above embodiments of this application.

[0186] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0187] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0188] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus, or computer program products. Therefore, embodiments of this application can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of this application can take the form of computer program products implemented 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.

[0189] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of this application. It should 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 terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0190] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate 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.

[0191] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal 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.

[0192] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0193] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0194] The above provides a detailed description of a photovoltaic equipment fault diagnosis method, apparatus, device, and medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for fault detection in photovoltaic equipment, characterized in that, include: The output voltage and current values ​​of the photovoltaic device under different temperatures and light intensities are obtained to determine a first training sample set, wherein each training sample in the first training sample set includes: open circuit voltage, short circuit current, maximum power, maximum power point voltage and maximum power point current. An offset is introduced into the first training sample set to perform data augmentation, and a second training sample set after data augmentation is determined; the offset is used to simulate data of photovoltaic equipment operating normally under different conditions; The second training sample set is trained using the local Mahalanobis distance algorithm to generate an anchor point set and determine the fault area boundary of the photovoltaic equipment. Fault detection of the target photovoltaic equipment is performed based on the set of anchor points and the boundary of the fault area. The second training sample set is trained using the local Mahalanobis distance algorithm to generate an anchor point set and determine the fault region boundary of the photovoltaic equipment, including: Based on the second training sample set, determine the minimum number of local samples and the local radius; wherein the local radius is determined by the region fitting error. The anchor set of the second training sample set is determined based on the minimum number of local samples and the local radius.

2. The photovoltaic equipment fault detection method according to claim 1, characterized in that, Introduce an offset into the first training sample set to perform data augmentation, and determine the augmented second training sample set, including: Data augmentation is performed by introducing offsets into each training sample in the first training sample set through various dimensions, wherein each dimension includes at least: open-circuit voltage, short-circuit current, maximum power, maximum power point voltage, and maximum power point current. The training samples after introducing the offset, along with the training samples in the first training sample set, are determined as the second training sample set.

3. The photovoltaic equipment fault detection method according to claim 2, characterized in that, Data augmentation is performed by introducing offsets into the first training sample set across various dimensions, including: Simultaneously, a first offset is introduced into each dimension of each training sample in the first training sample set to determine the first offset training sample set; A second offset is introduced into one dimension of each training sample in the first training sample set to determine the second offset training sample set. The second offset should be greater than the first offset. The first training sample set, the first offset training sample set, and the second offset training sample set are determined as the second training sample set.

4. The photovoltaic equipment fault detection method according to claim 1, characterized in that, The steps for determining the data dimensions included in each training sample in the first training sample set include: Under the same temperature and light intensity, the voltage and current values ​​output by normal photovoltaic equipment and each faulty photovoltaic equipment are collected as test results. The faulty photovoltaic equipment includes at least: partial open circuit, partial short circuit, partial shading, and partial aging. The test results of each faulty photovoltaic device are matched with the test results of the normal photovoltaic devices to determine the abnormal indicator dimensions in the test results of the faulty photovoltaic devices. The dimension of the anomaly indicator is determined as the data dimension included in the training samples of the first training sample set.

5. The photovoltaic equipment fault detection method according to claim 1, characterized in that, The second training sample set is trained using the local Mahalanobis distance algorithm to generate an anchor point set and determine the fault region boundary of the photovoltaic equipment. The method also includes: Set the confidence level value; Based on the confidence level value and the second training sample set, the fault region boundary of the second training sample set is determined.

6. The photovoltaic equipment fault detection method according to claim 1, characterized in that, Fault detection of the target photovoltaic equipment is performed based on the set of anchor points and the boundary of the fault area, including: Collect the voltage and current values ​​output by the target photovoltaic device, and calculate the open-circuit voltage, short-circuit current, maximum power, maximum power point voltage, and maximum power point current of the target device; Calculate the target local Mahalanobis distance of the target photovoltaic device based on the set of anchor points; The target local Mahalanobis distance is compared with the boundary of the fault area. When the target local Mahalanobis distance is greater than the boundary of the fault area, the target photovoltaic device is determined to be in a fault state. Once the target photovoltaic device is determined to be in a fault state, a fault alarm is issued for the fault state.

7. A photovoltaic equipment fault detection device, characterized in that, The device includes: The first training sample set acquisition module is used to acquire the output voltage and current values ​​of the photovoltaic device under different temperatures and different light intensities, and to determine the first training sample set. Each training sample in the first training sample set includes: open circuit voltage, short circuit current, maximum power, maximum power point voltage, and maximum power point current. The second training sample set determination module is used to introduce an offset into the first training sample set for data augmentation and determine the second training sample set after data augmentation; the offset is used to simulate data of photovoltaic equipment working normally under different conditions. The local Mahalanobis distance algorithm training module is used to train the local Mahalanobis distance algorithm on the second training sample set, generate an anchor point set, and determine the fault area boundary of the photovoltaic equipment. The fault detection module is used to detect faults in the target photovoltaic equipment based on the anchor point set and the boundary of the fault area. The local Mahalanobis distance algorithm training module includes: The anchor point generation condition determination submodule is used to determine the minimum number of local samples and the local radius based on the second training sample set; wherein, the local radius is determined by the region fitting error. An anchor point generation submodule is used to determine the anchor point set of the second training sample set based on the minimum number of local samples and the local radius.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is executed by the processor, it implements the photovoltaic equipment fault detection method as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program thereon, characterized in that, When the computer program is executed by the processor, it implements the photovoltaic equipment fault detection method as described in any one of claims 1 to 6.

Images