Fault detection methods and systems applied to photovoltaic energy storage equipment

Abstract

A fault detection method and system for photovoltaic (PV) energy storage devices are disclosed. First, the time series of operating parameters of each PV energy storage device are arranged to obtain a set of time-series input matrices of PV energy storage device operating parameters. Next, feature extraction is performed on the time-series input matrices of operating parameters of each PV energy storage device to obtain a set of time-series associated feature vectors of PV energy storage device operating parameters. Then, the semantic metric coefficient of the operating mode offset is calculated between the time-series associated feature vector of the analyzed PV energy storage device and the set of time-series associated feature vectors of all other PV energy storage devices in the set of time-series associated feature vectors of the analyzed PV energy storage device. Finally, based on the semantic metric coefficient of the operating mode offset, it is determined whether a fault exists in the analyzed PV energy storage device. This can improve the safety and reliability of the equipment and reduce maintenance costs and risks.

Description

Technical Field

[0001] This application relates to the field of photovoltaic energy storage, and more specifically, to a fault detection method and system for photovoltaic energy storage equipment. Background Technology

[0002] Photovoltaic energy storage equipment is a device that uses solar energy to generate electricity and stores excess power. It can provide a stable power supply at night or on cloudy days when power generation is impossible, offering advantages such as cleanliness, high efficiency, and renewability. However, due to long-term operation and the influence of the external environment, photovoltaic energy storage equipment may be prone to malfunctions, such as overcharging, over-discharging, short circuits, and overheating. These malfunctions can affect the performance and lifespan of the equipment, and may even lead to serious accidents such as fires and explosions. Therefore, fault detection for photovoltaic energy storage equipment is essential.

[0003] However, since the operational faults of photovoltaic energy storage equipment can be reflected from various aspects of different operating parameters, and the operating environment of photovoltaic energy storage equipment is usually quite complex, traditional fault detection methods based on thresholds or rules are difficult to adapt to such complex operating environments, and it is also difficult to use the interrelationships between different operating parameters to detect equipment faults, which can easily lead to false alarms or missed alarms.

[0004] Therefore, an optimized fault detection solution for photovoltaic energy storage equipment is desired. Summary of the Invention

[0005] To address the aforementioned technical problems, this application is proposed. This application provides a fault detection method and system for photovoltaic energy storage equipment. It can monitor different operating parameters of each photovoltaic energy storage device in a photovoltaic energy storage cluster, and utilize the correlation between multiple motion parameters of each device in the cluster to comprehensively detect faults in the analyzed photovoltaic energy storage device, thereby improving the accuracy of fault detection and judgment, and providing accurate fault diagnosis results.

[0006] According to one aspect of this application, a fault detection method for photovoltaic energy storage devices is provided, comprising:

[0007] The time series of operating parameters of each photovoltaic energy storage device in the photovoltaic energy storage device cluster is obtained, wherein the operating parameters include voltage value, current value, temperature value and power value;

[0008] The time series of operating parameters of each photovoltaic energy storage device in the photovoltaic energy storage device cluster are arranged according to the time dimension and the operating parameter sample dimension to form a photovoltaic energy storage device operating parameter time series input matrix, so as to obtain a set of photovoltaic energy storage device operating parameter time series input matrices;

[0009] The operating mode feature extractor based on a deep neural network model extracts features from each of the time-series input matrices of the photovoltaic energy storage device operating parameters in the set of time-series input matrices to obtain a set of time-series associated feature vectors of the photovoltaic energy storage device operating parameters.

[0010] Extract the time-series correlation feature vector of the photovoltaic energy storage device's operating parameters from the set of time-series correlation feature vectors of the photovoltaic energy storage device being analyzed as the query feature vector; calculate the semantic metric coefficient of the operating mode offset between the query feature vector and the set of time-series correlation feature vectors of all other photovoltaic energy storage devices' operating parameters in the set of time-series correlation feature vectors of the photovoltaic energy storage device; and

[0011] Based on the semantic metric coefficient of the operating mode offset, it is determined whether the analyzed photovoltaic energy storage device has a fault.

[0012] According to another aspect of this application, a fault detection system for photovoltaic energy storage devices is provided, comprising:

[0013] The data acquisition module is used to acquire the time series of operating parameters of each photovoltaic energy storage device in the photovoltaic energy storage device cluster, wherein the operating parameters include voltage value, current value, temperature value and power value;

[0014] The matrix module is used to arrange the time series of operating parameters of each photovoltaic energy storage device in the photovoltaic energy storage device cluster into a photovoltaic energy storage device operating parameter time series input matrix according to the time dimension and the operating parameter sample dimension, respectively, so as to obtain a set of photovoltaic energy storage device operating parameter time series input matrices;

[0015] The feature extraction module is used to extract features from each of the photovoltaic energy storage device operating parameter time-series input matrices in the set of photovoltaic energy storage device operating parameter time-series input matrices using a deep neural network model-based operating mode feature extractor to obtain a set of photovoltaic energy storage device operating parameter time-series associated feature vectors.

[0016] The coefficient calculation module is used to extract the time-series correlation feature vector of the photovoltaic energy storage device's operating parameters from the set of time-series correlation feature vectors of the photovoltaic energy storage device's operating parameters as a query feature vector, and to calculate the semantic metric coefficient of the operating mode offset between the query feature vector and the set of time-series correlation feature vectors of all other photovoltaic energy storage devices' operating parameters in the set of the photovoltaic energy storage device's operating parameters; and

[0017] The fault analysis module is used to determine whether the analyzed photovoltaic energy storage device has a fault based on the semantic metric coefficient of the operating mode offset.

[0018] Compared with existing technologies, the fault detection method and system for photovoltaic energy storage devices provided in this application first arranges the time series of operating parameters of each photovoltaic energy storage device to obtain a set of time series input matrices of photovoltaic energy storage device operating parameters. Next, features are extracted from the time series input matrices of operating parameters of each photovoltaic energy storage device to obtain a set of time series correlation feature vectors of photovoltaic energy storage device operating parameters. Then, the semantic metric coefficient of the operating mode offset is calculated between the time series correlation feature vector of the photovoltaic energy storage device being analyzed and the set of time series correlation feature vectors of all other photovoltaic energy storage devices in the set of time series correlation feature vectors of the photovoltaic energy storage device being analyzed. Finally, based on the semantic metric coefficient of the operating mode offset, it is determined whether a fault exists in the photovoltaic energy storage device being analyzed. This improves the safety and reliability of the equipment and reduces maintenance costs and risks. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. The following drawings are not deliberately drawn to scale according to the actual size, but are intended to show the main idea of ​​this application.

[0020] Figure 1 This is a flowchart of a fault detection method applied to photovoltaic energy storage equipment according to an embodiment of this application.

[0021] Figure 2 This is a schematic diagram of the architecture of a fault detection method applied to photovoltaic energy storage equipment according to an embodiment of this application.

[0022] Figure 3 This is a block diagram of a fault detection system applied to photovoltaic energy storage equipment according to an embodiment of this application.

[0023] Figure 4 This is an application scenario diagram of the fault detection method for photovoltaic energy storage equipment according to an embodiment of this application. Detailed Implementation

[0024] The technical solutions in 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, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are also within the scope of protection of this application.

[0025] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not specifically singular and may include plural forms. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0026] While this application makes various references to certain modules of the systems according to embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The modules described are merely illustrative, and different aspects of the systems and methods may use different modules.

[0027] Flowcharts are used in this application to illustrate the operations performed by the system according to embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

[0028] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0029] To address the aforementioned technical problems, this application proposes a fault detection method for photovoltaic (PV) energy storage devices. This method monitors different operating parameters of each PV energy storage device within a cluster, such as voltage, current, temperature, and power. By utilizing the correlation between multiple motion parameters of each device in the cluster, it comprehensively detects faults in the analyzed PV energy storage devices, thereby improving the accuracy of fault detection and diagnosis and providing accurate fault diagnosis results. This enables real-time monitoring and intelligent diagnosis of PV energy storage devices, facilitating timely maintenance, improving equipment safety and reliability, and reducing maintenance costs and risks.

[0030] Figure 1 This is a flowchart of a fault detection method applied to photovoltaic energy storage equipment according to an embodiment of this application. Figure 2 This is a schematic diagram of the architecture of a fault detection method applied to photovoltaic energy storage equipment according to an embodiment of this application. Figure 1 and Figure 2As shown, the fault detection method for photovoltaic energy storage devices according to an embodiment of this application includes the following steps: S110, obtaining the time series of operating parameters of each photovoltaic energy storage device in the photovoltaic energy storage device cluster, wherein the operating parameters include voltage, current, temperature and power values; S120, arranging the time series of operating parameters of each photovoltaic energy storage device in the photovoltaic energy storage device cluster according to the time dimension and the operating parameter sample dimension to form a photovoltaic energy storage device operating parameter time series input matrix to obtain a set of photovoltaic energy storage device operating parameter time series input matrices; S130, using an operating mode feature extractor based on a deep neural network model to process the set of photovoltaic energy storage device operating parameter time series input matrices. S140: Extract features from the time-series input matrices of the operating parameters of each photovoltaic energy storage device in the set to obtain a set of time-series associated feature vectors of photovoltaic energy storage device operating parameters; S150: Extract the time-series associated feature vectors of the operating parameters of the photovoltaic energy storage device being analyzed from the set of time-series associated feature vectors of the photovoltaic energy storage device operating parameters as query feature vectors, and calculate the semantic metric coefficient of the operating mode offset between the query feature vector and the set of time-series associated feature vectors of all other photovoltaic energy storage devices in the set of time-series associated feature vectors of the photovoltaic energy storage device operating parameters; and S160: Determine whether the photovoltaic energy storage device being analyzed has a fault based on the semantic metric coefficient of the operating mode offset.

[0031] Specifically, in the technical solution of this application, firstly, the time series of operating parameters of each photovoltaic energy storage device in the photovoltaic energy storage device cluster are obtained, wherein the operating parameters include voltage, current, temperature, and power values. Next, considering that the operating parameters of each photovoltaic energy storage device in the photovoltaic energy storage device cluster include voltage, current, temperature, and power values, these operating parameter items not only have a dynamic change pattern in the time dimension, but also have a temporal synergistic relationship among them. This is of great significance for monitoring the operating status of photovoltaic energy storage devices and analyzing the overall operating mode of the photovoltaic energy storage device cluster. Based on this, in the technical solution of this application, the time series of operating parameters of each photovoltaic energy storage device in the photovoltaic energy storage device cluster need to be arranged into a photovoltaic energy storage device operating parameter time series input matrix according to the time dimension and the operating parameter sample dimension, respectively, to obtain a set of photovoltaic energy storage device operating parameter time series input matrices.

[0032] Then, the time-series input matrices of the operating parameters of each photovoltaic energy storage device in the set of time-series input matrices of the photovoltaic energy storage device are subjected to feature mining through an operating mode feature extractor based on a convolutional neural network model, so as to extract the time-series collaborative correlation feature information between multiple operating parameter items of each photovoltaic energy storage device in the photovoltaic energy storage device cluster, thereby reflecting the operating mode of each photovoltaic energy storage device in the cluster, and thus obtaining a set of time-series correlation feature vectors of photovoltaic energy storage device operating parameters.

[0033] Accordingly, in step S130, the deep neural network model is a convolutional neural network model; in other words, the operating mode feature extractor based on the deep neural network model is an operating mode feature extractor based on the convolutional neural network model.

[0034] Furthermore, during the fault detection process of the analyzed photovoltaic energy storage device, it is considered that if the analyzed photovoltaic energy storage device is operating normally, its operating mode is consistent with that of other photovoltaic energy storage devices in the cluster; otherwise, the semantic differences in the operating mode are relatively high. Based on this, in order to more accurately detect the fault of the analyzed photovoltaic energy storage device, it is necessary to compare and analyze the operating mode of the analyzed photovoltaic energy storage device with the overall operating mode of each photovoltaic energy storage device in the cluster, thereby determining its semantic deviation from the operating mode of other devices. Specifically, in the technical solution of this application, the time-series correlation feature vector of the photovoltaic energy storage device's operating parameters is further extracted from the set of time-series correlation feature vectors of the photovoltaic energy storage device's operating parameters as a query feature vector. The semantic metric coefficient of the operating mode deviation between the query feature vector and the set of time-series correlation feature vectors of all other photovoltaic energy storage devices in the set of time-series correlation feature vectors of the photovoltaic energy storage device's operating parameters is calculated. By calculating the semantic metric coefficient of the operating mode offset, the overall difference in operating modes between the analyzed photovoltaic energy storage device and other photovoltaic energy storage devices in the cluster can be measured. If the semantic metric coefficient of the operating mode offset is large, it indicates that there are significant differences in the operating modes between the devices, which means that the analyzed device is faulty. Conversely, if the semantic metric coefficient of the operating mode offset is small, it indicates that the operating modes between the devices are relatively similar, which means that the analyzed device is operating normally.

[0035] Accordingly, in step S140, the photovoltaic energy storage device's operating parameter time-series correlation feature vector is extracted from the set of photovoltaic energy storage device operating parameter time-series correlation feature vectors as a query feature vector. The semantic metric coefficient of the operating mode offset between the query feature vector and the set of all other photovoltaic energy storage device operating parameter time-series correlation feature vectors in the set of photovoltaic energy storage device operating parameter time-series correlation feature vectors is calculated, including: calculating the semantic metric coefficient of the operating mode offset between the query feature vector and the set of all other photovoltaic energy storage device operating parameter time-series correlation feature vectors in the set of photovoltaic energy storage device operating parameter time-series correlation feature vectors using the following coefficient calculation formula; wherein, the coefficient calculation formula is:

[0036]

[0037] Among them, v i It is the query feature vector, v j It refers to all the time-series correlation feature vectors of photovoltaic energy storage device operating parameters in the set of photovoltaic energy storage device operating parameter time-series correlation feature vectors other than the query feature vector, ||·||1 represents the 1-norm of the feature vector, M is the number of photovoltaic energy storage device operating parameter time-series correlation feature vectors in the set of photovoltaic energy storage device operating parameter time-series correlation feature vectors - 1, V k D is the set of time-series correlation feature vectors of the operating parameters of the photovoltaic energy storage device. i This represents the semantic metric coefficient for runtime mode offset.

[0038] Furthermore, in the technical solution of this application, each photovoltaic energy storage device operating parameter time-series correlation feature vector in the set of photovoltaic energy storage device operating parameter time-series correlation feature vectors expresses the local high-order semantic correlation features of the time series of operating parameters of each photovoltaic energy storage device in the photovoltaic energy storage device cluster within the spatial domain woven by the time dimension and the cross dimension of operating parameter samples. Therefore, when extracting the photovoltaic energy storage device operating parameter time-series correlation feature vector of the analyzed photovoltaic energy storage device from the set of photovoltaic energy storage device operating parameter time-series correlation feature vectors as a query feature vector, and calculating the semantic metric coefficient of the operating mode offset between the query feature vector and the set of all other photovoltaic energy storage device operating parameter time-series correlation feature vectors in the set of photovoltaic energy storage device operating parameter time-series correlation feature vectors, the vector set composed of all other photovoltaic energy storage device operating parameter time-series correlation feature vectors will have a significantly different correlation feature distribution information due to the differences in the correlation feature distribution of the source data subsets of each photovoltaic energy storage device operating parameter time-series correlation feature vector, thus affecting the accuracy of the calculation of the semantic metric coefficient of the operating mode offset.

[0039] Based on this, in another preferred embodiment of this application, the photovoltaic energy storage device's operating parameter time-series correlation feature vector is extracted from the set of photovoltaic energy storage device operating parameter time-series correlation feature vectors as a query feature vector. The semantic metric coefficient of the operating mode offset between the query feature vector and the set of all other photovoltaic energy storage device operating parameter time-series correlation feature vectors in the set of photovoltaic energy storage device operating parameter time-series correlation feature vectors is calculated, including:

[0040] The set of time-series correlated feature vectors of the photovoltaic energy storage device's operating parameters is arranged into a feature matrix to obtain the initial key matrix;

[0041] The query feature vector and the initial key matrix are input into a cross-domain attention gate query encoder based on local association enhancement of the key matrix to obtain the self-associative cross-domain query matching feature vector of the device to be analyzed.

[0042] The self-associative cross-domain query matching feature vector of the device to be analyzed is input into the decoder to obtain the semantic metric coefficient of the operating mode offset.

[0043] Specifically, in this preferred embodiment, the query feature vector and the initial key matrix are input into a cross-domain attention gate query encoder based on key matrix local association enhancement to obtain the self-associative cross-domain query matching feature vector of the device to be analyzed, including:

[0044] Calculate the feature distribution difference energy coefficient between the query feature vector and each row vector in the initial key matrix to obtain a set of feature distribution difference energy coefficients;

[0045] Each feature distribution difference energy coefficient in the set of feature distribution difference energy coefficients is compared with a predetermined energy threshold. In response to a feature distribution difference energy coefficient being less than the predetermined energy threshold, the mean vector between the row vector corresponding to the feature distribution difference energy coefficient and the query feature vector is used as the updated row vector to obtain the update key matrix.

[0046] The query feature vector is processed using a value embedding matrix to obtain a value vector;

[0047] The query feature vector is used as the query vector, each row vector in the update key matrix is ​​used as the key vector, and the value vector is input into the cross-domain attention gate query module based on the Transformer structure to obtain a sequence of cross-domain query attention vectors;

[0048] Calculate the positional mean vector of the sequence of cross-domain query attention vectors to obtain the self-associative cross-domain query matching feature vector of the device to be analyzed.

[0049] This process can be expressed by the following formula:

[0050] K = {k1,k2,...,k} n}

[0051] M o ={k1;k2;...;k n}

[0052]

[0053] Wherein, K represents the set of time-series correlation feature vectors of the operating parameters of the photovoltaic energy storage device, k1, k2, k... n Let k represent the first, second, and nth time-series correlation feature vectors of the photovoltaic energy storage device's operating parameters in the set of time-series correlation feature vectors, respectively, where n represents the total number of vectors in the set of time-series correlation feature vectors of the photovoltaic energy storage device's operating parameters, and k represents the total number of vectors in the set of time-series correlation feature vectors of the photovoltaic energy storage device's operating parameters. i Let {·} represent the i-th time-series correlation feature vector of the photovoltaic energy storage device's operating parameters in the set of time-series correlation feature vectors, and M represent the set of vectors. o Let v represent the initial bond matrix. q This represents the query feature vector. The circle (⊙) represents subtraction by position, and the dot product (⊙) represents dot product by position. This indicates addition by position, [·,·,·] indicates vector concatenation, conv1D indicates one-dimensional convolution, MaxPool indicates max pooling, and f int R represents the coefficient of eigenvalues ​​between vectors. i Let μ and σ be the feature correlation coefficients corresponding to the time-series correlation feature vector of the operating parameters of the i-th photovoltaic energy storage device. 2 Let λ represent the mean and variance of the features corresponding to the time-series correlation feature vector of the operating parameters of the i-th photovoltaic energy storage device, λ represent the predetermined hyperparameter, max(·) represent the calculated maximum value, and e(R) represent the maximum value. i ) represents the i-th characteristic distribution difference energy coefficient in the set of characteristic distribution difference energy coefficients, θ represents the predetermined energy threshold, and k i ' represents the updated i-th row vector, M i W represents the update of the key matrix. q The value embedding matrix, b q Represents the value bias vector, v v Let T represent the value vector corresponding to the query feature vector, d represent the transpose of the vector, d represent the length of each row vector in the update key matrix, softmax represent the normalization exponential function, and v represent the value vector corresponding to the query feature vector. pi v represents the i-th cross-domain query attention vector in the sequence of cross-domain query attention vectors. pThis represents the self-associative cross-domain query matching feature vector of the device to be analyzed.

[0054] Specifically, the cross-domain attention gate query encoder based on key matrix local association enhancement first calculates the feature distribution difference between the query feature vector and the row vectors in the initial key matrix to obtain a set of difference energy coefficients. That is, the query feature vector is used as a soft anchoring reference point, and the feature distribution energy difference coefficients are used to measure the feature distribution difference between the query feature vector and each key vector in the initial key matrix. It should be understood that, considering the query feature vector is a shared soft anchoring reference point, the set of feature distribution energy difference coefficients also contains local association information between the key feature vectors. Next, the feature distribution difference energy coefficients are compared with a predetermined energy threshold. For values ​​less than the threshold, the row vectors in the key matrix are updated using the mean vector of the query feature vector and the corresponding row vector. This update operation enhances the time-series association feature vectors of photovoltaic energy storage device operating parameters that are more similar to the query vector. Through the local association enhancement mechanism, the specificity and matching quality of the key matrix are improved.

[0055] Furthermore, query vectors and value vectors are constructed based on the query feature vectors, and each row vector in the update key matrix is ​​used as the key vector. These are then input into a Transformer-based cross-domain attention gate query module to obtain a sequence of cross-domain query attention vectors. In other words, the Transformer-based cross-domain attention gate query module combines the query vector, key vector, and value vector to perform self-attention calculation of features, achieving cross-domain information integration. The effect of the cross-domain attention mechanism is that it allows the model to flow and integrate information between different feature domains, improving the richness of feature representation and the accuracy of matching. Finally, by calculating the positional mean of the cross-domain query attention vector sequence, the final self-associative cross-domain query matching feature vector of the device to be analyzed is obtained. The purpose is to comprehensively consider the information of all cross-domain query attention vectors in the sequence, generating a comprehensive feature representation, which helps improve the model's overall understanding and response capability to complex queries.

[0056] In this way, the accuracy of calculating the semantic metric coefficient of the operating mode offset between the analyzed photovoltaic energy storage device's operating parameter time-series correlation feature vector and all other photovoltaic energy storage devices' operating parameter time-series correlation feature vectors is improved.

[0057] Subsequently, based on a comparison between the semantic metric coefficient of the operating mode offset and a predetermined threshold, it is determined whether the analyzed photovoltaic energy storage device is faulty. Specifically, in a specific example of this application, a fault is determined in the analyzed photovoltaic energy storage device in response to the operating mode offset semantic metric coefficient being greater than the predetermined threshold. In this way, the difference between the operating mode of the overall photovoltaic energy storage device cluster and the operating mode of the analyzed photovoltaic energy storage device can be used to detect faults in the analyzed photovoltaic energy storage device, thereby improving the accuracy of fault detection and judgment of photovoltaic energy storage devices and providing accurate fault diagnosis results.

[0058] Accordingly, in step S150, determining whether the analyzed photovoltaic energy storage device has a fault based on the operating mode offset semantic metric coefficient includes: determining whether the analyzed photovoltaic energy storage device has a fault based on the comparison between the operating mode offset semantic metric coefficient and a predetermined threshold.

[0059] Furthermore, in the technical solution of this application, the fault detection method applied to photovoltaic energy storage equipment further includes a training step: training a feature extractor for operating modes based on a convolutional neural network model.

[0060] In one example, the training step includes: acquiring training data, which includes time series of training operation parameters for each photovoltaic energy storage device in the photovoltaic energy storage device cluster, and the true value of the semantic metric coefficient of the operation mode offset, wherein the training operation parameters include training voltage, training current, training temperature, and training power values; arranging the time series of training operation parameters for each photovoltaic energy storage device in the photovoltaic energy storage device cluster according to the time dimension and the operation parameter sample dimension to form a set of training photovoltaic energy storage device operation parameter time series input matrices; and using the operation mode feature extractor based on the convolutional neural network model to extract features from each training photovoltaic energy storage device operation parameter time series input matrix in the set of training photovoltaic energy storage device operation parameter time series input matrices to obtain the training photovoltaic energy storage device operation parameters. The system extracts a set of time-series associated feature vectors; extracts the time-series associated feature vectors of the training photovoltaic energy storage device's operating parameters from the set of time-series associated feature vectors of the training photovoltaic energy storage device's operating parameters as a training query feature vector; calculates the training operation mode offset semantic metric coefficient between the training query feature vector and the set of all other training photovoltaic energy storage device operating parameter time-series associated feature vectors in the set of time-series associated feature vectors of the training photovoltaic energy storage device's operating parameters; calculates the cross-entropy function value between the training operation mode offset semantic metric coefficient and the true value of the operation mode offset semantic metric coefficient; and trains the operation mode feature extractor based on the convolutional neural network model using the cross-entropy function value as the loss function value, wherein, in each iteration of the training, the set of time-series associated feature vectors of the training photovoltaic energy storage device's operating parameters is optimized.

[0061] In the technical solution of this application, each training photovoltaic energy storage device operating parameter time-series correlation feature vector in the set of training photovoltaic energy storage device operating parameter time-series correlation feature vectors expresses the local high-order semantic correlation features of the time series of training operating parameters of each photovoltaic energy storage device in the photovoltaic energy storage device cluster within the spatial domain woven by the time dimension and the cross dimension of operating parameter samples. Therefore, when extracting the training photovoltaic energy storage device's operating parameter time-series correlation feature vector as the training query feature vector from the set of training photovoltaic energy storage device operating parameter time-series correlation feature vectors, and calculating the training operation mode offset semantic metric coefficient of the training query feature vector and all other sets of training photovoltaic energy storage device operating parameter time-series correlation feature vectors in the set of training photovoltaic energy storage device operating parameter time-series correlation feature vectors, the vector set composed of all other training photovoltaic energy storage device operating parameter time-series correlation feature vectors will have saliency in the correlation feature distribution information of each training photovoltaic energy storage device operating parameter time-series correlation feature vector due to the difference in the correlation feature distribution of each training photovoltaic energy storage device operating parameter time-series correlation feature vector's source data subset. This makes it difficult for the multiple training photovoltaic energy storage device operating parameter time-series correlation feature vectors as a set to stably focus on the significant local distribution of features during the training process, thereby affecting the accuracy of the calculation of the training operation mode offset semantic metric coefficient.

[0062] Based on this, the applicant of this application optimizes the set of time-series correlation feature vectors of the photovoltaic energy storage device operating parameters during each model iteration training, for example, when backpropagating the loss function based on the difference between the inferred training operating mode offset semantic metric coefficient and the actual operating mode offset semantic metric coefficient through the set of training photovoltaic energy storage device operating parameter time-series correlation feature vectors.

[0063] Accordingly, in one example, in each iteration of the training, the set of time-series correlation feature vectors of the operating parameters of the trained photovoltaic energy storage device is optimized using the following optimization formula to obtain the optimized set of time-series correlation feature vectors of the operating parameters of the trained photovoltaic energy storage device; wherein, the optimization formula is:

[0064]

[0065] Wherein, V1 is a cascaded feature vector obtained by cascading the set of time-series correlation feature vectors of the operating parameters of the trained photovoltaic energy storage device, v 1i It is the eigenvalue at the i-th position of the cascaded eigenvector. and These are the squares of the 1-norm and 2-norm of the cascaded feature vector, respectively; L is the length of the cascaded feature vector V1; ω is the weight hyperparameter; log[·] represents the logarithmic function to the base 2; v 1i ' is the feature value at the i-th position of the optimized cascaded feature vector obtained by cascading the set of time-series correlation feature vectors of the optimized training photovoltaic energy storage device operating parameters.

[0066] Next, by further expanding the optimized cascaded feature vectors, we can obtain the set of optimized training photovoltaic energy storage device operating parameter time-series correlation feature vectors.

[0067] Specifically, by geometrically registering the high-dimensional feature manifold shape based on the scale and structural parameters of the cascaded feature vector V1, we can focus on features with rich semantic information in the feature set composed of the feature values ​​of the cascaded feature vector V1. That is, during model training, we can identify distinguishable and stable interest features based on local context information to represent dissimilarity. This enables the saliency labeling of the feature information of the cascaded feature vector V1 during model training, improving the accuracy of calculating the semantic metric coefficient of the operating mode offset between the analyzed photovoltaic energy storage device's operating parameter time-series correlation feature vector and all other photovoltaic energy storage devices' operating parameter time-series correlation feature vectors. In this way, we can utilize the difference between the overall operating mode of the photovoltaic energy storage device cluster and the operating mode of the analyzed photovoltaic energy storage device to perform fault detection of the analyzed photovoltaic energy storage device, thereby improving the accuracy of fault detection and judgment of photovoltaic energy storage devices. In this way, we can realize real-time monitoring and intelligent diagnosis of photovoltaic energy storage devices, so as to take timely maintenance measures, improve the safety and reliability of the equipment, and reduce maintenance costs and risks.

[0068] In summary, the fault detection method for photovoltaic energy storage equipment based on the embodiments of this application has been clarified, which can improve the safety and reliability of the equipment and reduce maintenance costs and risks.

[0069] Figure 3 This is a block diagram of a fault detection system 100 applied to a photovoltaic energy storage device according to an embodiment of this application. Figure 3As shown, the fault detection system 100 applied to photovoltaic energy storage equipment according to an embodiment of this application includes: a data acquisition module 110, used to acquire the time series of operating parameters of each photovoltaic energy storage device in the photovoltaic energy storage equipment cluster, wherein the operating parameters include voltage, current, temperature and power values; a matrixing module 120, used to arrange the time series of operating parameters of each photovoltaic energy storage device in the photovoltaic energy storage equipment cluster according to the time dimension and the operating parameter sample dimension to form a photovoltaic energy storage equipment operating parameter time series input matrix to obtain a set of photovoltaic energy storage equipment operating parameter time series input matrices; and a feature extraction module 130, used to extract the photovoltaic energy storage equipment operating parameter time series input matrices respectively through an operating mode feature extractor based on a deep neural network model. The set of matrices extracts features from the time-series input matrices of the operating parameters of each photovoltaic energy storage device to obtain a set of time-series associated feature vectors of the operating parameters of the photovoltaic energy storage devices; the coefficient calculation module 140 is used to extract the time-series associated feature vectors of the operating parameters of the photovoltaic energy storage devices under analysis from the set of time-series associated feature vectors of the operating parameters of the photovoltaic energy storage devices as query feature vectors, and calculates the semantic measurement coefficient of the operating mode offset between the query feature vector and the set of time-series associated feature vectors of the operating parameters of all other photovoltaic energy storage devices in the set of time-series associated feature vectors of the operating parameters of the photovoltaic energy storage devices; and the fault analysis module 150 is used to determine whether there is a fault in the photovoltaic energy storage device under analysis based on the semantic measurement coefficient of the operating mode offset.

[0070] In one example, in the fault detection system 100 applied to photovoltaic energy storage equipment described above, the deep neural network model is a convolutional neural network model.

[0071] In one example, in the fault detection system 100 applied to photovoltaic energy storage equipment described above, the coefficient calculation module 140 is used to: arrange the set of time-series correlation feature vectors of the photovoltaic energy storage equipment operating parameters into a feature matrix to obtain an initial key matrix; input the query feature vector and the initial key matrix into a cross-domain attention gate query encoder based on key matrix local correlation enhancement to obtain the self-correlation cross-domain query matching feature vector of the device to be analyzed; and input the self-correlation cross-domain query matching feature vector of the device to be analyzed into a decoder to obtain the semantic metric coefficient of the operating mode offset.

[0072] In one example, in the fault detection system 100 applied to photovoltaic energy storage equipment described above, the fault analysis module 150 is used to: determine whether the analyzed photovoltaic energy storage equipment has a fault based on the comparison between the semantic metric coefficient of the operating mode offset and a predetermined threshold.

[0073] Here, those skilled in the art will understand that the specific functions and operations of each module in the fault detection system 100 applied to photovoltaic energy storage equipment have been referenced above. Figures 1 to 2 The method for fault detection in photovoltaic energy storage equipment has been described in detail, and therefore, its repeated description will be omitted.

[0074] As described above, the fault detection system 100 for photovoltaic energy storage devices according to the embodiments of this application can be implemented in various wireless terminals, such as servers with fault detection algorithms for photovoltaic energy storage devices. In one example, the fault detection system 100 for photovoltaic energy storage devices according to the embodiments of this application can be integrated into a wireless terminal as a software module and / or hardware module. For example, the fault detection system 100 for photovoltaic energy storage devices can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the fault detection system 100 for photovoltaic energy storage devices can also be one of many hardware modules of the wireless terminal.

[0075] Alternatively, in another example, the fault detection system 100 applied to the photovoltaic energy storage device and the wireless terminal can also be separate devices, and the fault detection system 100 applied to the photovoltaic energy storage device can be connected to the wireless terminal via wired and / or wireless networks, and transmit interactive information in accordance with the agreed data format.

[0076] Figure 4 This is an application scenario diagram of the fault detection method for photovoltaic energy storage equipment according to an embodiment of this application. For example... Figure 4 As shown, in this application scenario, firstly, the time series of operating parameters of each photovoltaic energy storage device in the photovoltaic energy storage device cluster is obtained (e.g., ...). Figure 4 As shown in D), the operating parameters include voltage, current, temperature, and power values. Then, the time series of the operating parameters of each photovoltaic energy storage device in the photovoltaic energy storage device cluster is input to a server deployed with a fault detection algorithm for photovoltaic energy storage devices (e.g., ...). Figure 4 In the S shown, the server can use the fault detection algorithm applied to photovoltaic energy storage devices to process the time series of operating parameters of each photovoltaic energy storage device in the photovoltaic energy storage device cluster to obtain the semantic metric coefficient of operating mode offset. Then, based on the comparison between the semantic metric coefficient of operating mode offset and a predetermined threshold, it determines whether the analyzed photovoltaic energy storage device has a fault.

[0077] According to another aspect of this application, a non-volatile computer-readable storage medium is also provided, on which computer-readable instructions are stored, which, when executed by a computer, can perform the methods described above.

[0078] The program portion of a technology can be considered a "product" or "artifact" existing in the form of executable code and / or related data, and is involved in or implemented through a computer-readable medium. Tangible, permanent storage media can include memory or storage used by any computer, processor, or similar device or related module. For example, various semiconductor memories, tape drives, disk drives, or any similar device capable of providing storage functionality for software.

[0079] Furthermore, those skilled in the art will understand that aspects of this application can be described and illustrated through several patentable types or situations, including any new and useful combination of processes, machines, products, or substances, or any new and useful improvements thereof. Accordingly, aspects of this application can be implemented entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. All of the above hardware or software may be referred to as a “data block,” “module,” “engine,” “unit,” “component,” or “system.” Furthermore, aspects of this application may manifest as a computer product located on one or more computer-readable media, the product including computer-readable program code.

[0080] Unless otherwise defined, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in a common dictionary shall be interpreted as having a meaning consistent with their meaning in the context of the relevant art, and not as having an idealized or highly formalized meaning, unless expressly defined herein.

[0081] The foregoing description is a illustrative description of the present application and should not be construed as limiting it. Although several exemplary embodiments of the present application have been described, those skilled in the art will readily understand that many modifications can be made to the exemplary embodiments without departing from the novel teachings and advantages of the present application. Therefore, all such modifications are intended to be included within the scope of the present application as defined by the claims. It should be understood that the foregoing description is a illustrative description of the present application and should not be construed as limiting it to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present application is defined by the claims and their equivalents.

Claims

1. A fault detection method for photovoltaic energy storage equipment, characterized in that, include: The time series of operating parameters of each photovoltaic energy storage device in the photovoltaic energy storage device cluster is obtained, wherein the operating parameters include voltage value, current value, temperature value and power value; The time series of operating parameters of each photovoltaic energy storage device in the photovoltaic energy storage device cluster are arranged according to the time dimension and the operating parameter sample dimension to form a photovoltaic energy storage device operating parameter time series input matrix, so as to obtain a set of photovoltaic energy storage device operating parameter time series input matrices; The operating mode feature extractor based on a deep neural network model extracts features from each of the time-series input matrices of the photovoltaic energy storage device operating parameters in the set of time-series input matrices to obtain a set of time-series associated feature vectors of the photovoltaic energy storage device operating parameters. Extracting the time-series correlation feature vectors of the photovoltaic energy storage device's operating parameters from the set of time-series correlation feature vectors of the photovoltaic energy storage device's operating parameters as query feature vectors, and calculating the semantic metric coefficient of the operating mode offset between the query feature vector and the set of all other sets of time-series correlation feature vectors of the photovoltaic energy storage device's operating parameters, including: arranging the set of time-series correlation feature vectors of the photovoltaic energy storage device's operating parameters into a feature matrix to obtain an initial key matrix; calculating the feature distribution difference energy coefficients between the query feature vector and each row vector in the initial key matrix to obtain a set of feature distribution difference energy coefficients; and comparing each feature distribution difference energy coefficient in the set of feature distribution difference energy coefficients with a predetermined energy threshold. In response to the feature distribution difference energy coefficient being less than a predetermined energy threshold, the mean vector between the row vector corresponding to the feature distribution difference energy coefficient and the query feature vector is used as the updated row vector to obtain an update key matrix; the query feature vector is processed using a value embedding matrix to obtain a value vector; the query feature vector is used as the query vector, each row vector in the update key matrix is ​​used as the key vector, and the value vector is input into a cross-domain attention gate query module based on a Transformer structure to obtain a sequence of cross-domain query attention vectors; the position-wise mean vector of the sequence of cross-domain query attention vectors is calculated to obtain the self-associative cross-domain query matching feature vector of the device to be analyzed; the self-associative cross-domain query matching feature vector of the device to be analyzed is input into a decoder to obtain the semantic metric coefficient of the running mode offset. Based on the semantic metric coefficient of the operating mode offset, it is determined whether the analyzed photovoltaic energy storage device has a fault.

2. The fault detection method for photovoltaic energy storage equipment according to claim 1, characterized in that, The deep neural network model is a convolutional neural network model.

3. The fault detection method for photovoltaic energy storage equipment according to claim 2, characterized in that, Based on the semantic metric coefficient of the operating mode offset, determine whether the analyzed photovoltaic energy storage device has a fault, including: Based on the comparison between the semantic metric coefficient of the operating mode offset and a predetermined threshold, it is determined whether the analyzed photovoltaic energy storage device has a fault.

4. The fault detection method for photovoltaic energy storage equipment according to claim 3, characterized in that, It also includes a training step: used to train the operating pattern feature extractor based on the convolutional neural network model.

5. The fault detection method for photovoltaic energy storage equipment according to claim 4, characterized in that, The training steps include: Acquire training data, which includes the time series of training operation parameters of each photovoltaic energy storage device in the photovoltaic energy storage device cluster, and the true value of the semantic metric coefficient of the operation mode offset, wherein the training operation parameters include training voltage value, training current value, training temperature value and training power value; The time series of training operation parameters of each photovoltaic energy storage device in the photovoltaic energy storage device cluster are arranged according to the time dimension and the operation parameter sample dimension to form a time series input matrix of training photovoltaic energy storage device operation parameters, so as to obtain a set of time series input matrices of training photovoltaic energy storage device operation parameters; The operation mode feature extractor based on the convolutional neural network model extracts features from each of the time-series input matrices of the training photovoltaic energy storage device operation parameters in the set of training photovoltaic energy storage device operation parameter time-series input matrices to obtain a set of time-series associated feature vectors of the training photovoltaic energy storage device operation parameters; Extract the training photovoltaic energy storage device's operating parameter time-series correlation feature vector from the set of training photovoltaic energy storage device operating parameter time-series correlation feature vectors as the training query feature vector, and calculate the training operation mode offset semantic metric coefficient between the training query feature vector and the set of all other training photovoltaic energy storage device operating parameter time-series correlation feature vectors in the set of training photovoltaic energy storage device operating parameter time-series correlation feature vectors. Calculate the cross-entropy function value between the training run mode offset semantic metric coefficient and the true value of the run mode offset semantic metric coefficient; and The cross-entropy function value is used as the loss function value to train the operation mode feature extractor based on the convolutional neural network model. In each iteration of the training, the set of time-series correlation feature vectors of the photovoltaic energy storage device operation parameters is optimized.

6. A fault detection system for photovoltaic energy storage equipment, characterized in that, include: The data acquisition module is used to acquire the time series of operating parameters of each photovoltaic energy storage device in the photovoltaic energy storage device cluster, wherein the operating parameters include voltage value, current value, temperature value and power value; The matrix module is used to arrange the time series of operating parameters of each photovoltaic energy storage device in the photovoltaic energy storage device cluster into a photovoltaic energy storage device operating parameter time series input matrix according to the time dimension and the operating parameter sample dimension, respectively, so as to obtain a set of photovoltaic energy storage device operating parameter time series input matrices; The feature extraction module is used to extract features from each of the photovoltaic energy storage device operating parameter time-series input matrices in the set of photovoltaic energy storage device operating parameter time-series input matrices using a deep neural network model-based operating mode feature extractor to obtain a set of photovoltaic energy storage device operating parameter time-series associated feature vectors. The coefficient calculation module is used to extract the time-series correlation feature vector of the photovoltaic energy storage device's operating parameters from the set of time-series correlation feature vectors of the photovoltaic energy storage device's operating parameters as a query feature vector, and to calculate the semantic metric coefficient of the operating mode offset between the query feature vector and the set of all other sets of time-series correlation feature vectors of the photovoltaic energy storage device's operating parameters in the set of time-series correlation feature vectors of the photovoltaic energy storage device's operating parameters. This includes: arranging the set of time-series correlation feature vectors of the photovoltaic energy storage device's operating parameters into a feature matrix to obtain an initial key matrix; calculating the feature distribution difference energy coefficient between the query feature vector and each row vector in the initial key matrix to obtain a set of feature distribution difference energy coefficients; and comparing each feature distribution difference energy coefficient in the set of feature distribution difference energy coefficients with a predetermined energy threshold. A comparison is made, and in response to the feature distribution difference energy coefficient being less than a predetermined energy threshold, the mean vector between the row vector corresponding to the feature distribution difference energy coefficient and the query feature vector is used as the updated row vector to obtain an update key matrix; the query feature vector is processed using a value embedding matrix to obtain a value vector; the query feature vector is used as the query vector, each row vector in the update key matrix is ​​used as the key vector, and the value vector is input into a cross-domain attention gate query module based on a Transformer structure to obtain a sequence of cross-domain query attention vectors; the position-wise mean vector of the sequence of cross-domain query attention vectors is calculated to obtain the self-associative cross-domain query matching feature vector of the device to be analyzed; the self-associative cross-domain query matching feature vector of the device to be analyzed is input into a decoder to obtain the semantic metric coefficient of the running mode offset; The fault analysis module is used to determine whether the analyzed photovoltaic energy storage device has a fault based on the semantic metric coefficient of the operating mode offset.

7. The fault detection system for photovoltaic energy storage equipment according to claim 6, characterized in that, The deep neural network model is a convolutional neural network model.

Images