Battery short-circuit fault diagnosis method and device, computer equipment and storage medium

Abstract

The invention relates to a battery short circuit fault diagnosis method and device, computer equipment and a storage medium. The method comprises the steps of obtaining multiple pieces of operation state information and health state information of a target battery, and performing feature extraction on the multiple pieces of operation state information and health state information to obtain multiple pieces of operation state feature information and health state feature information; performing feature fusion on the multiple pieces of operation state feature information to obtain operation state fusion feature information; performing data reconstruction processing on the running state fusion feature information to obtain a first reconstruction error, and performing data reconstruction processing on the health state feature information to obtain a second reconstruction error; and generating a short-circuit fault diagnosis result of the target battery based on the first reconstruction error and the second reconstruction error. Through the method, the short-circuit fault diagnosis of the battery can be accurately realized.

Description

Technical Field

[0001] The present application relates to the technical field of fault detection, and particularly to a method, device, computer device, and storage medium for diagnosing battery short - circuit faults. Background Art

[0002] Batteries play a crucial role in many fields. In recent years, battery - related accidents have occurred frequently, causing serious economic losses and social impacts. Therefore, safety issues have gradually become the primary problem in battery applications. However, taking the example of a battery with an internal short - circuit fault, it is difficult to identify the battery in the initial stage of the internal short - circuit because there are almost no obvious electrical and thermal characteristics. Once the internal short - circuit fault develops to the middle and late stages, the battery will generate a large amount of heat in a short time, and effective countermeasures cannot be used to prevent battery thermal runaway at this stage.

[0003] Regarding the thermal runaway problem that may be caused by battery short - circuit faults, generally, it is necessary to obtain capacity increment analysis or data - driven analysis results, such as isolation forest, neural network, etc., through capacity increment analysis or data - driven methods in the initial stage of the battery short - circuit fault, and detect the battery short - circuit fault based on the capacity increment analysis or data - driven analysis results.

[0004] However, the current battery short - circuit fault diagnosis methods still have the problem of insufficient accuracy. Summary of the Invention

[0005] Based on this, in view of the above - mentioned technical problems, it is necessary to provide an accurate method, device, computer device, computer - readable storage medium, and computer program product for diagnosing battery short - circuit faults.

[0006] In a first aspect, the present application provides a method for diagnosing battery short - circuit faults, including:

[0007] Obtain multiple operating - state information and health - state information of a target battery, and respectively perform feature extraction on the multiple operating - state information and health - state information to obtain multiple operating - state feature information and health - state feature information;

[0008] Perform feature fusion on the multiple operating - state feature information to obtain operating - state fusion feature information;

[0009] Perform data reconstruction processing on the operating - state fusion feature information to obtain a first reconstruction error, and perform data reconstruction processing on the health - state feature information to obtain a second reconstruction error;

[0010] Generate a short - circuit fault diagnosis result of the target battery based on the first reconstruction error and the second reconstruction error.

[0011] In one embodiment, data reconstruction processing is performed on the operation state fusion feature information to obtain a first reconstruction error, including:

[0012] Map the operation state fusion feature information into a data space of a preset dimension to obtain operation state latent feature information;

[0013] Perform data recovery processing on the operation state latent feature information in the data space of the preset dimension to obtain target operation state feature information;

[0014] Detect the first reconstruction error between the operation state latent feature information and the target operation state feature information;

[0015] Perform data reconstruction processing on the health state feature information to obtain a second reconstruction error, including:

[0016] Map the health state feature information into a data space of a preset dimension to obtain health state latent feature information;

[0017] Perform data recovery processing on the health state latent feature information in the data space of the preset dimension to obtain target health state feature information;

[0018] Detect the second reconstruction error between the health state latent feature information and the target health state feature information.

[0019] In one embodiment, based on the first reconstruction error and the second reconstruction error, a short - circuit fault diagnosis result of the target battery is generated, including:

[0020] Perform weighted processing on the first reconstruction error and the second reconstruction error to generate a target reconstruction error of the target battery;

[0021] Obtain a target reconstruction error threshold corresponding to the target reconstruction error, and generate a short - circuit fault diagnosis result of the target battery based on the target reconstruction error and the target reconstruction error threshold.

[0022] In one embodiment, obtaining a target reconstruction error threshold corresponding to the target reconstruction error includes:

[0023] Obtain standard reconstruction error distribution information when the target battery has no fault;

[0024] Generate a target reconstruction error threshold corresponding to the target reconstruction error based on the standard reconstruction error distribution information.

[0025] In one embodiment, feature fusion is performed on multiple operation state feature information to obtain operation state fusion feature information, including:

[0026] Perform a linear transformation on multiple operating state characteristic information to obtain multiple intermediate vectors, where the intermediate vectors include a value vector, a query vector, and a key vector;

[0027] Use multiple attention heads to perform scaled dot-product attention operations on the value vector, query vector, and key vector in each intermediate vector respectively to obtain multiple intermediate operating characteristic data;

[0028] Combine multiple intermediate operating characteristic data to obtain operating state fusion characteristic information.

[0029] In one embodiment, obtaining multiple operating state information of a target battery includes:

[0030] Obtain multiple initial operating state information of the target battery;

[0031] Perform a timing alignment process on multiple initial operating state information to obtain an initial operating state timing sequence;

[0032] Perform a normalization process on multiple timing-aligned initial operating state information in the initial operating state timing sequence to obtain an operating state timing sequence, where the operating state timing sequence is used to represent multiple operating state information of the target battery in the form of a time series.

[0033] In a second aspect, the present application also provides a battery short-circuit fault diagnosis device, including:

[0034] A data acquisition module, configured to acquire multiple operating state information and health state information of a target battery, and respectively perform feature extraction on the multiple operating state information and health state information to obtain multiple operating state characteristic information and health state characteristic information;

[0035] A feature fusion module, configured to perform feature fusion on multiple operating state characteristic information to obtain operating state fusion characteristic information;

[0036] A data reconstruction processing module, configured to perform data reconstruction processing on the operating state fusion characteristic information to obtain a first reconstruction error, and perform data reconstruction processing on the health state characteristic information to obtain a second reconstruction error;

[0037] A short-circuit fault diagnosis module, configured to generate a short-circuit fault diagnosis result of the target battery based on the first reconstruction error and the second reconstruction error.

[0038] In a third aspect, the present application also provides a computer device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0039] Obtain multiple operating state information and health state information of the target battery, and respectively perform feature extraction on the multiple operating state information and health state information to obtain multiple operating state feature information and health state feature information;

[0040] Perform feature fusion on the multiple operating state feature information to obtain operating state fusion feature information;

[0041] Perform data reconstruction processing on the operating state fusion feature information to obtain a first reconstruction error, and perform data reconstruction processing on the health state feature information to obtain a second reconstruction error;

[0042] Generate a short - circuit fault diagnosis result of the target battery based on the first reconstruction error and the second reconstruction error.

[0043] In a fourth aspect, the present application also provides a computer - readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0044] Obtain multiple operating state information and health state information of the target battery, and respectively perform feature extraction on the multiple operating state information and health state information to obtain multiple operating state feature information and health state feature information;

[0045] Perform feature fusion on the multiple operating state feature information to obtain operating state fusion feature information;

[0046] Perform data reconstruction processing on the operating state fusion feature information to obtain a first reconstruction error, and perform data reconstruction processing on the health state feature information to obtain a second reconstruction error;

[0047] Generate a short - circuit fault diagnosis result of the target battery based on the first reconstruction error and the second reconstruction error.

[0048] In a fifth aspect, the present application also provides a computer program product, including a computer program. When the computer program is executed by a processor, the following steps are implemented:

[0049] Obtain multiple operating state information and health state information of the target battery, and respectively perform feature extraction on the multiple operating state information and health state information to obtain multiple operating state feature information and health state feature information;

[0050] Perform feature fusion on the multiple operating state feature information to obtain operating state fusion feature information;

[0051] Perform data reconstruction processing on the operating state fusion feature information to obtain a first reconstruction error, and perform data reconstruction processing on the health state feature information to obtain a second reconstruction error;

[0052] Generate a short - circuit fault diagnosis result for the target battery based on the first reconstruction error and the second reconstruction error.

[0053] In the above - mentioned battery short - circuit fault diagnosis method, device, computer device, computer - readable storage medium, and computer program product, throughout the entire process, first, a joint diagnosis of the short - circuit fault of the target battery is carried out by combining multiple operating - state information and health - state information of the target battery. This joint diagnosis can more comprehensively detect the short - circuit fault of the battery in the case of a short - circuit fault, reduce misdiagnosis and missed diagnosis, and can provide a more accurate short - circuit fault diagnosis. Further, by performing feature fusion on multiple operating - state information, the feature information of multiple operating - state information can be effectively fused. Compared with common fault feature extraction, the input data for short - circuit fault diagnosis in this application is more accurate, improving the accuracy of short - circuit fault diagnosis. Finally, data reconstruction processing is respectively performed on the fused feature information of the operating state and the feature information of the health state to obtain the first reconstruction error and the second reconstruction error. Compared with existing technologies such as capacity - increment diagnosis analysis or data - driven diagnosis methods, the method of diagnosing the short - circuit fault of the target battery by whether the reconstruction error is abnormal is more accurate and can generate an accurate short - circuit fault diagnosis result for the target battery. Brief Description of the Drawings

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments of the present application or related technologies. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0055] Figure 1 It is an application environment diagram of the battery short - circuit fault diagnosis method in an embodiment;

[0056] Figure 2 It is a flow schematic diagram of the battery short - circuit fault diagnosis method in an embodiment;

[0057] Figure 3 It is a flow schematic diagram of the battery short - circuit fault diagnosis method in another embodiment;

[0058] Figure 4 It is a structural schematic diagram of the Transformer model in an embodiment;

[0059] Figure 5 It is a network structure schematic diagram of the variational auto - encoder in an embodiment;

[0060] Figure 6 It is a structural block diagram of the battery short - circuit fault diagnosis device in an embodiment;

[0061] Figure 7 It is the internal structure diagram of a computer device in an embodiment. Detailed implementation manners

[0062] In order to make the objectives, technical solutions and advantages of the present application clearer and more understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are used to explain the present application and are not used to limit the present application.

[0063] The battery short-circuit fault diagnosis method provided by the embodiments of the present application can be applied to, for example Figure 1 the application environment shown in the figure. Among them, the terminal 102 communicates with the information collection device 104 through a network.

[0064] When the user triggers the fault detection control on the battery short-circuit fault detection interface on the terminal 102, the terminal 102 responds to the trigger message of the short-circuit fault detection control of the user, controls the information collection device 104 to collect parameter information related to multiple operating state information and health state information of the target battery, and controls the information collection device 104 to feed back the collected parameter information to the terminal 102. The terminal 102 generates multiple operating state information and health state information of the target battery based on the collected parameter information, respectively extracts features from the multiple operating state information and health state information to obtain multiple operating state feature information and health state feature information; fuses the multiple operating state feature information to obtain operating state fusion feature information; performs data reconstruction processing on the operating state fusion feature information to obtain a first reconstruction error, and performs data reconstruction processing on the health state feature information to obtain a second reconstruction error; generates a short-circuit fault diagnosis result of the target battery based on the first reconstruction error and the second reconstruction error. Further, the terminal 102 can also display the fault diagnosis result of the target battery to the user.

[0065] Among them, the terminal 102 can be but is not limited to various personal computers, laptop computers, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart vehicle-mounted devices, projection devices, etc. The portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The head-mounted device can be a virtual reality (VR) device, an augmented reality (AR) device, smart glasses, etc.

[0066] In an exemplary embodiment, as Figure 2 shown, a battery short-circuit fault diagnosis method is provided. Taking the method applied to Figure 1 the terminal 102 in the figure as an example for description. Among them:

[0067] S100. Obtain multiple operating status information and health status information of the target battery, and respectively perform feature extraction on the multiple operating status information and health status information to obtain multiple operating status feature information and health status feature information.

[0068] Among them, the status information of the target battery includes multiple operating status information and health status information. The operating status information includes, but is not limited to, information such as voltage, current, temperature, charge and discharge time, and capacity increment curve. The health status information is also known as SOH (state of health), which reflects the health life status of the battery and is the embodiment of the battery's power, energy, charge and discharge power and other states. Accurately evaluating the health status can fully understand the current state of the battery, make a maintenance plan according to the underlying premise, correct various parameter performance indicators, reduce or avoid the risk factor, or maintain and replace single cells whose performance cannot meet the usage requirements, reducing the usage cost.

[0069] Specifically, the user triggers the fault detection control on the battery fault detection interface of the terminal. The terminal responds to the user's fault detection control trigger message, controls the information collection device to collect parameter information related to the status information of the target battery, and controls the information collection device to feedback the collected parameter information of the target battery to the terminal. The terminal generates multiple operating status information and health status information of the target battery based on the collected parameter information, and respectively performs feature extraction on the multiple operating status information and health status information to obtain multiple operating status feature information and health status feature information.

[0070] In an exemplary embodiment, the information collection device does not necessarily need to collect all the multiple operating status information and health status information of the target battery. If at least one of the multiple operating status information and health status information of the target battery is collected, the battery short-circuit fault diagnosis operation can also be carried out.

[0071] That is to say, in the case where the information collection device collects multiple operating status information and does not collect health status information, for example, when information such as voltage, current, temperature, charge and discharge time, and capacity increment curve is collected, the terminal controls the information collection device to feedback the collected information of the target battery to the terminal. The terminal uses information such as voltage, current, temperature, charge and discharge time, and capacity increment curve as the multiple operating status information of the target battery, and performs feature extraction on the multiple operating status information to obtain multiple operating status feature information.

[0072] When the collected status information includes health status information and does not include multiple operating status information, for example, when information such as battery power, energy, charge and discharge power is collected, the control information acquisition device feeds back the information of the target battery collected to the terminal. The terminal combines information such as battery power, energy, charge and discharge power to obtain the health status information of the target battery, and performs feature extraction on the health status information of the operating status of the target battery to obtain the health status feature information of the target battery. Further, in the process of the terminal combining information such as battery power, energy, charge and discharge power to obtain the health status information of the target battery, the terminal is obtained by real-time estimation based on a specific SOH prediction model.

[0073] S200. Perform feature fusion on multiple operating status feature information to obtain operating status fusion feature information.

[0074] Specifically, since the operating data of the battery usually contains a variety of different physical signals, such as operating status information such as voltage, current, temperature, charge and discharge time, capacity increment curve, etc., therefore, the multi-modal operating status information of the target battery can be obtained first, and then feature extraction is performed on the operating status information to obtain operating status feature information. Finally, fusing multiple operating status feature information can obtain operating status fusion feature information.

[0075] Further, the fusion of multiple operating status feature information is realized by using a deep learning model. For example, a Transformer model can be used to naturally process these multi-modal operating status feature information and effectively fuse their features. Compared with common fault feature extraction, it can capture more comprehensively various semantic associations latent in multiple operating status feature information, and the generated operating status fusion feature information will be more accurate. That is to say, this model improves the accuracy of fault diagnosis. Among them, the Transformer model is a deep learning architecture based on the self-attention mechanism and is widely used in the field of natural language processing. It consists of an encoder and a decoder, and each part contains multiple identical layers. The core advantage of this model is that it can process sequence data in parallel and effectively capture long-distance dependencies.

[0076] S300. Perform data reconstruction processing on the operating status fusion feature information to obtain a first reconstruction error, and perform data reconstruction processing on the health status feature information to obtain a second reconstruction error.

[0077] Specifically, a short - circuit fault in the battery will cause changes in the internal state of the battery, such as a voltage drop and a temperature rise. Under normal conditions, the model can reconstruct the input data well with a small reconstruction error. However, a short - circuit fault will cause the battery behavior to deviate from the normal range, making the model unable to accurately reconstruct the input data and resulting in a significant increase in the reconstruction error. Therefore, by separately performing data reconstruction processing on the fused feature information of the operating state and the feature information of the healthy state, the first reconstruction error and the second reconstruction error can be obtained to accurately diagnose whether the battery has a short - circuit fault.

[0078] Furthermore, separately performing data reconstruction processing on the fused feature information of the operating state and the feature information of the healthy state can be achieved through a variational auto - encoder in a time - series manner. The variational auto - encoder learns the probability distribution of normal - state data and can use the reconstruction error to determine whether there is a fault when detecting anomalies. For a short - circuit fault, the variational auto - encoder can capture a significant increase in the reconstruction error when the fault signal appears, thus accurately identifying the short - circuit fault.

[0079] S400, generate a short - circuit fault diagnosis result for the target battery based on the first reconstruction error and the second reconstruction error.

[0080] Specifically, combining the first reconstruction error and the second reconstruction error is often achieved by weighted summation, and other methods such as error fusion can also be used for combination. By combining the first reconstruction error and the second reconstruction error, the operating state information and the healthy state information can be jointly used to generate an accurate short - circuit fault diagnosis result for the target battery.

[0081] Furthermore, the fault diagnosis result of the target battery can also be pushed to the staff's terminal for display to the user through the staff's terminal.

[0082] In the above - mentioned battery fault diagnosis method, throughout the process, first, a joint diagnosis of the short - circuit fault of the target battery is carried out by combining multiple operating state information and healthy state information of the target battery. This joint diagnosis can more comprehensively detect the short - circuit fault of the battery in the case of a short - circuit fault, reducing misdiagnosis and missed diagnosis and providing a more accurate short - circuit fault diagnosis. Further, by performing feature fusion on multiple operating state information, the feature information of multiple operating state information can be effectively fused. Compared with common fault feature extraction, the input data for short - circuit fault diagnosis in this application is more accurate, improving the accuracy of short - circuit fault diagnosis. Finally, separately performing data reconstruction processing on the fused feature information of the operating state and the feature information of the healthy state to obtain the first reconstruction error and the second reconstruction error. Compared with existing technologies such as capacity increment diagnosis analysis or data - driven diagnosis methods, the method of diagnosing the short - circuit fault of the target battery by whether the reconstruction error is abnormal is more accurate and can generate an accurate short - circuit fault diagnosis result for the target battery.

[0083] In an exemplary embodiment, as Figure 3 shown, S300 includes:

[0084] S310, mapping the operation state fusion feature information into a data space of a preset dimension to obtain operation state latent feature information.

[0085] S320, performing data recovery processing on the operation state latent feature information in the data space of the preset dimension to obtain target operation state feature information.

[0086] S330, detecting a first reconstruction error between the operation state latent feature information and the target operation state feature information.

[0087] S340, mapping the health state feature information into a data space of a preset dimension to obtain health state latent feature information.

[0088] S350, performing data recovery processing on the health state latent feature information in the data space of the preset dimension to obtain target health state feature information.

[0089] S360, detecting a second reconstruction error between the health state latent feature information and the target health state feature information.

[0090] Specifically, the present application also designs a variational autoencoder. The variational autoencoder processes the operation state fusion feature information to generate a first reconstruction error corresponding to the operation state fusion feature information, and processes the health state feature information through the variational autoencoder to generate a second reconstruction error corresponding to the health state feature information.

[0091] Among them, the variational autoencoder includes an encoder and a decoder. The encoder is generally a multi-layer convolutional network. Taking the data reconstruction processing of the operation state fusion feature information as an example, in the process of fusing to obtain the operation state fusion feature information, it may cause the operation state fusion feature information to be mapped to a higher dimension. That is to say, the obtained operation state fusion feature information is a high-dimensional fusion feature information. Therefore, it is necessary to compress the high-dimensional operation state fusion feature information into a latent space of a preset dimension through the encoder to obtain operation state latent feature information. In practical applications, the preset dimension is generally a low dimension; the decoder can use a symmetric decoding network to perform data recovery processing on the operation state latent feature information in the data space of the preset dimension, and restore the operation state latent feature information to the target operation state feature information corresponding to the operation state fusion feature information, and the dimension of the target operation state feature information is the same as that of the operation state latent feature information. Furthermore, the first reconstruction error between the operation state latent feature information and the target operation state feature information can be obtained.

[0092] Similarly, in the process of data reconstruction processing for health status characteristic information, it is also necessary to compress the health status characteristic information into a latent space of a preset dimension through an encoder to obtain health status latent characteristic information; and perform data recovery processing on the health status latent characteristic information in a data space of a preset dimension through a decoder to obtain target health status characteristic information, and the target health status characteristic information has the same dimension as the health status latent characteristic information. Furthermore, the second reconstruction error between the health status latent characteristic information and the target health status characteristic information is detected.

[0093] In the above embodiment, the reconstruction errors corresponding to multiple operating status information and health status information are output through the variational autoencoder, so as to use the reconstruction errors to judge whether there is a fault when detecting a battery short-circuit fault, and the battery short-circuit fault diagnosis process is more accurate.

[0094] In an exemplary embodiment, based on the first reconstruction error and the second reconstruction error, a short-circuit fault diagnosis result of the target battery is generated, including:

[0095] Perform weighted processing on the first reconstruction error and the second reconstruction error to generate a target reconstruction error of the target battery; obtain a target reconstruction error threshold corresponding to the target reconstruction error, and generate a short-circuit fault diagnosis result of the target battery based on the target reconstruction error and the target reconstruction error threshold.

[0096] Specifically, perform weighted summation on the first reconstruction error and the second reconstruction error to generate a target reconstruction error of the target battery. At this time, it is necessary to obtain a target reconstruction error threshold corresponding to the target reconstruction error, and generate a short-circuit fault diagnosis result of the target battery based on the target reconstruction error and the target reconstruction error threshold. That is to say, when the target reconstruction error is greater than the target reconstruction error threshold, the short-circuit fault diagnosis result of the target battery indicates that the target battery is in a short-circuit fault state; when the target reconstruction error is less than or equal to the target reconstruction error threshold, the short-circuit fault diagnosis result of the target battery indicates that the target battery is not in a short-circuit fault state.

[0097] In an exemplary embodiment, in the process of performing weighted processing on the first reconstruction error and the second reconstruction error, the corresponding de-weighting coefficient can be dynamically adjusted through experiments to balance the relationship between the health status information and the operating status information.

[0098] In the above embodiment, the joint optimization of the health status information and the operating status information can achieve adaptive learning of the battery state, enabling the model to continuously adapt to the real-time state and health changes of the battery pack, continuously optimize the fault diagnosis and health status prediction, and by comparing the target reconstruction error with the target reconstruction error threshold, the short-circuit fault diagnosis result is more accurate.

[0099] In an exemplary embodiment, obtaining a target reconstruction error threshold corresponding to a target reconstruction error includes:

[0100] Obtaining standard reconstruction error distribution information when the target battery has no faults; generating a target reconstruction error threshold corresponding to the target reconstruction error based on the standard reconstruction error distribution information.

[0101] Specifically, obtaining the standard reconstruction error distribution information when the target battery has no faults to accurately detect whether the target battery has faults by combining the reconstruction error with the standard reconstruction error distribution information. More specifically, accurately detecting whether the target battery has faults by combining the reconstruction error with the standard reconstruction error distribution information is to generate a target reconstruction error threshold corresponding to the target reconstruction error based on the standard reconstruction error distribution information, and generate a short - circuit fault diagnosis result of the target battery by combining the target reconstruction error threshold with the target reconstruction error.

[0102] Furthermore, the standard reconstruction error distribution information can be a distribution curve formed by standard reconstruction errors. Through the standard reconstruction error distribution curve, a target reconstruction error threshold corresponding to the standard reconstruction error distribution curve is generated, and the target reconstruction error is compared with the target reconstruction error threshold to generate a short - circuit fault diagnosis result of the target battery.

[0103] More specifically, generating a target reconstruction error threshold corresponding to the standard reconstruction error distribution curve through the standard reconstruction error distribution curve includes: processing the standard reconstruction error distribution curve based on the three - standard - deviation principle to generate a target reconstruction error threshold corresponding to the standard reconstruction error distribution curve. That is, obtaining the standard reconstruction error mean and the standard reconstruction error standard deviation of all standard reconstruction errors in the standard reconstruction error distribution curve, and generating a target reconstruction error threshold through the standard reconstruction error mean and the standard reconstruction error standard deviation. The expression of the target reconstruction error threshold δ can be:

[0104]

[0105] where u is the standard reconstruction error mean, σ is the standard reconstruction error standard deviation, and δ is the target reconstruction error threshold.

[0106] In an exemplary embodiment, the variational auto - encoder can continuously adjust the target reconstruction error threshold through an online learning and updating mechanism to adapt to the real - time state changes of the battery pack. Even after the battery pack has been running for a long time, the model can still timely detect new fault types or change trends.

[0107] In the above embodiments, by generating a target reconstruction error threshold corresponding to the standard reconstruction error distribution information based on the standard reconstruction error distribution information, and then accurately determining whether the standard reconstruction error is abnormal based on the target reconstruction error threshold, a short - circuit fault diagnosis result of the battery is accurately generated.

[0108] In an exemplary embodiment, feature fusion is performed on multiple operation - state feature information to obtain operation - state fusion feature information, including:

[0109] Perform a linear transformation on multiple operation - state feature information to obtain multiple intermediate vectors, where the intermediate vectors include value vectors, query vectors, and key vectors; use multiple attention heads to perform scaled dot - product attention operations on the value vectors, query vectors, and key vectors in each intermediate vector respectively to obtain multiple intermediate operation feature data; combine the multiple intermediate operation feature data to obtain operation - state fusion feature information.

[0110] Specifically, in this application, the fusion of multiple operation - state feature information is realized by using the multi - head attention mechanism in the Transformer model. Among them, the multi - head attention mechanism is an extended form of the attention mechanism widely adopted in the Transformer model. It obtains the attention distribution of different sub - spaces of the input sequence by running multiple independent attention mechanisms in parallel, so as to more comprehensively capture various potential semantic associations in the sequence. That is to say, it can be used to capture the global dependencies between time steps in multi - modal data and calculate the correlation between features. In practical applications, parallel processing of multiple attention patterns means that each attention head uses a different linear transformation, which means that they can learn different feature associations from different sub - spaces of the input sequence. In this way, the model can simultaneously focus on different aspects of the input sequence through multiple attention heads, such as syntactic structure, semantic role, topic shift, etc.

[0111] In the multi - head attention mechanism, the input sequence first passes through three different linear transformation layers to obtain three intermediate vectors, namely the query vector Query, the key vector Key, and the value vector Value respectively. Then, these transformed vectors are divided into several "heads", and each head has its own independent Query, Key, and Value matrices. For each head, a scaled dot - product attention operation is performed once. Finally, the outputs of all heads are concatenated together and then fused through a linear layer to obtain the final attention output vector.

[0112] The expression of the scaled dot - product attention operation includes:

[0113]

[0114] Among them, Q, K, and V are the input query vector, key vector, and value vector respectively. is the intermediate running feature data output by each head, and d k is the dimension of the key, the Softmax function is the activation function, and K T represents the transposed matrix of the key vector.

[0115] The expressions for concatenating the outputs of all heads include:

[0116] MHSA(Q, K, V) = Concat(head 1 , head 2 ,..., head h )W 0

[0117] Among them, h is the number of attention heads, head 1 , head 2 ,..., head h are the intermediate running feature data output by the 1st, 2nd,..., h-th attention heads respectively, concat represents concatenating multiple intermediate running feature data, and W 0 is the linear transformation matrix of the output, MHSA(Q, K, V) represents the running state feature information, and MHSA represents the multi-head self-attention network.

[0118] In an exemplary embodiment, the structural diagram of the Transformer model in the present application is as Figure 4 shown. The structure of the Transformer model includes an encoder and a decoder, each part consisting of multiple layers. The core advantage of this model is its ability to process sequence data in parallel and effectively capture long-range dependencies. The encoder is responsible for processing the input sequence, while the decoder generates the target sequence based on the output of the encoder. Each encoder layer contains a multi-head self-attention mechanism and a feed-forward neural network, and each decoder layer contains a masked multi-head self-attention mechanism, an encoder-decoder attention mechanism, and a feed-forward neural network.

[0119] Among them, FFN (Feed-Forward Neural Network, feed-forward neural network): The FFN layer is actually a linear transformation layer used to complete the dimensionality transformation of the input data to the output data. The FFN layer is a sequential structure: including the first fully connected layer + relu activation layer + the second fully connected layer. The FFN layer enhances the representation ability by performing a non-linear transformation on the feature information. The expression of the FFN layer includes:

[0120] FFN(x) = ReLU(xW 1 + b 1 )W2 +b 2

[0121] Among them, FFN(x) is the output of the feed-forward neural network, and xW 1 +b 1 is the calculation formula of the first fully connected layer, and ReLU(xW 1 +b 1 ) is the calculation formula of the relu activation layer. FFN(x) = ReLU(xW 1 +b 1 )W 2 +b 2 is the calculation formula of the second fully connected layer. W 1 , b 1 are the model parameters of the first fully connected layer, and W 2 , b 2 are the model parameters of the second fully connected layer, and x is the input data of the feed-forward neural network.

[0122] In an exemplary embodiment, in order to prevent the vanishing gradient and ensure the stability of training, a residual connection and a normalization layer can also be set in the Transformer model. By further enhancing the depth of feature extraction through multiple layers of Transformer, the final output feature representation is X ∈ R Txdmodel .

[0123] Among them, X is the running state fusion feature information output by the Transformer model, d model is the hidden layer dimension of the model, T is the time step, and R is the running state feature information.

[0124] Furthermore, the model expression of the residual connection and the normalization layer can be: x out = LayerNorm(x + MHSA(x)), x out = LayerNorm(x + FFN(x)), where MHSA represents the multi-head self-attention network and FFN represents the feed-forward network.

[0125] In an exemplary embodiment, a fully connected layer can also be set in the Transformer model to map the running state fusion feature information to an autoencoder model, and the autoencoder model is used to perform data reconstruction processing on the input data to obtain a reconstruction error.

[0126] In the above embodiment, by using the multi-level network structure in the Transformer model to perform attention mechanism-based feature fusion on multiple running state feature information, the features of multiple running state feature information can be effectively fused to accurately obtain the running state fusion feature information.

[0127] In an exemplary embodiment, a plurality of operating state information of a target battery is obtained, including:

[0128] Obtain a plurality of initial operating state information of the target battery; perform time series alignment processing on the plurality of initial operating state information to obtain an initial operating state time series; perform normalization processing on the plurality of time series-aligned initial operating state information in the initial operating state time series to obtain an operating state time series, and the operating state time series is used to represent the plurality of operating state information of the target battery in the form of a time series.

[0129] Specifically, first, obtain multi-modal initial operating state information of the target battery, such as voltage, current, temperature, charge and discharge time, and capacity increment curve and other information. And since there will be a certain time error in each collected multi-modal initial operating state information when collecting the multi-modal initial operating state information of the target battery, it is necessary to perform time series alignment processing on the plurality of initial operating state information to obtain an initial operating state time series. Generally, the methods of time series alignment processing include but are not limited to achieving time series alignment through a sliding window to gather the plurality of initial operating state information corresponding to each time stamp, and based on the plurality of initial operating state information corresponding to all time stamps, obtain an initial operating state time series.

[0130] Secondly, in order to improve the convergence and accuracy of model processing in fault detection, normalization processing can also be performed on each item of the initial operating state information in the initial operating state time series, that is, linearly transform each item of the initial operating state feature information in the initial operating state feature time series into a specified range, usually between [0, 1], to obtain the final operating state time series, and the operating state time series includes a plurality of operating state information of the target battery.

[0131] In the above embodiment, by performing time series alignment processing on the plurality of initial operating state information, the time error between the plurality of initial operating state information can be reduced, making the acquisition of the operating state information more accurate. Further, by performing normalization processing on the initial operating state time series to obtain the operating state time series, the data scales can also be made consistent to eliminate the influence of the dimensions of different features.

[0132] In an exemplary embodiment, the present application proposes a battery short circuit fault diagnosis method for multi-modal feature fusion, jointly extracts the operating state feature information of multiple physical quantities through a Transformer model, and realizes high-precision anomaly detection through a variational autoencoder model. At the same time, by comprehensively considering the plurality of operating state information and health state information of the target battery, dynamically adjusts the diagnosis weight through a joint optimization strategy, so as to achieve more accurate fault diagnosis of the battery pack. The method includes:

[0133] 1. Data processing: Obtain the battery multi-modal data, that is, multiple initial operating state information of the target battery, including but not limited to parameters such as voltage, current, temperature, charge and discharge time, capacity increment curve, etc., and perform time series alignment and min-max normalization processing on the data to make the data scales consistent, so as to eliminate the dimensionality effects of different features and obtain the operating state time series.

[0134] 2. The Transformer model realizes feature fusion in the operating state time series:

[0135] Represent the processed multi-modal signal as a time series matrix X 0 ∈R TxF , where R is the operating state feature information, T is the time step, and F is the feature dimension. Use a linear embedding layer to map the input features to a high-dimensional space, and through the attention mechanism, perform a linear transformation on multiple operating state feature information in the operating state time series to obtain multiple intermediate vectors. The intermediate vectors include value vectors, query vectors, and key vectors; use multiple attention heads to perform scaled dot-product attention operations on the value vectors, query vectors, and key vectors in each intermediate vector respectively to obtain multiple intermediate operating feature data; combine multiple intermediate operating feature data to obtain the operating state fusion feature information.

[0136] 3. Variational autoencoder: Achieve precise detection of short-circuit faults through the reconstruction error.

[0137] The network structure diagram of the variational autoencoder is as Figure 5 shown. At this time, there may be high-dimensional information in the operating state fusion feature information. Through the variational autoencoder, the state feature information x1, x2, x3,..., xn can be mapped to a low-dimensional latent space, that is, the z space respectively, to obtain the operating state latent feature information z1, z2, z3,..., zn, and perform data recovery processing on the operating state latent feature information in the latent space of the preset dimension to obtain the target operating state feature information x1’, x2’, x3’,..., xn’.

[0138] Furthermore, by detecting the first reconstruction error between the potential feature information of the operating state and the target operating state feature information, for the short-circuit fault data, its reconstruction error will increase significantly. Therefore, it is also possible to obtain the first standard reconstruction error distribution information when the target battery has no fault, and based on the first standard reconstruction error distribution information, generate the first reconstruction error threshold corresponding to the first standard reconstruction error distribution information. Among them, the first reconstruction error threshold is obtained through the mean and variance of the first standard reconstruction error distribution information, so as to combine the first reconstruction error threshold with the first reconstruction error for comparison. When the first reconstruction error is greater than the standard reconstruction error distribution information, it is considered that the battery is in a short-circuit fault state. When the second reconstruction error is less than or equal to the standard reconstruction error distribution information, it is considered that the battery is not in a short-circuit fault state.

[0139] Furthermore, through the online learning and updating mechanism, the variational autoencoder can continuously adjust the threshold of the reconstruction error to adapt to the real-time state changes of the battery pack. Even after the battery pack has been running for a long time, the model can still detect new fault types or change trends in a timely manner.

[0140] 4. Joint SOH optimization strategy: Combine SOH prediction with fault feature diagnosis to improve the diagnosis accuracy.

[0141] Specifically, combine the operating state information and health state information of the battery. After extracting the features of the health state information, input it into the variational autoencoder model to generate the potential feature information of the healthy operating state corresponding to the health state information and the target health state feature information. At this time, the second reconstruction error between the potential feature information of the healthy operating state and the target health state feature information can be detected, and the first reconstruction error and the second reconstruction error are weighted to generate the target reconstruction error of the target battery; obtain the target reconstruction error threshold corresponding to the target reconstruction error, and generate the fault diagnosis result of the target battery based on the target reconstruction error and the target reconstruction error threshold.

[0142] It should be understood that although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the arrows, these steps do not necessarily need to be executed in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages. These steps or stages do not necessarily need to be executed at the same time, but can be executed at different times. The execution order of these steps or stages does not necessarily need to be sequential, but can be executed alternately or alternately with at least a part of other steps or steps or stages in other steps.

[0143] Based on the same inventive concept, an embodiment of the present application further provides a battery short - circuit fault diagnosis device for implementing the battery short - circuit fault diagnosis method involved above. The solution provided by this device to solve the problem is similar to the solution described in the above - mentioned method. Therefore, the specific limitations in one or more embodiments of the battery short - circuit fault diagnosis device provided below can refer to the limitations on the battery short - circuit fault diagnosis method in the above text, and will not be elaborated here.

[0144] In an exemplary embodiment, as Figure 6 shown, a battery short - circuit fault diagnosis device is provided, including: a data acquisition module 100, a feature fusion module 200, a data reconstruction processing module 300, and a short - circuit fault diagnosis module 400, where:

[0145] The data acquisition module 100 is configured to acquire a plurality of operating state information and health state information of a target battery, and perform feature extraction on the plurality of operating state information and health state information respectively to obtain a plurality of operating state feature information and health state feature information;

[0146] The feature fusion module 200 is configured to perform feature fusion on the plurality of operating state feature information to obtain operating state fusion feature information;

[0147] The data reconstruction processing module 300 is configured to perform data reconstruction processing on the operating state fusion feature information to obtain a first reconstruction error, and perform data reconstruction processing on the health state feature information to obtain a second reconstruction error;

[0148] The short - circuit fault diagnosis module 400 is configured to generate a short - circuit fault diagnosis result of the target battery based on the first reconstruction error and the second reconstruction error.

[0149] In an exemplary embodiment, the data reconstruction processing module 300 is further configured to map the operating state fusion feature information into a data space of a preset dimension to obtain operating state latent feature information; perform data recovery processing on the operating state latent feature information in the data space of the preset dimension to obtain target operating state feature information; detect the first reconstruction error between the operating state latent feature information and the target operating state feature information; the data reconstruction processing module 300 is further configured to map the health state feature information into a data space of a preset dimension to obtain health state latent feature information; perform data recovery processing on the health state latent feature information in the data space of the preset dimension to obtain target health state feature information; detect the second reconstruction error between the health state latent feature information and the target health state feature information.

[0150] In an exemplary embodiment, the short - circuit fault diagnosis module 400 is further configured to perform a weighted processing on the first reconstruction error and the second reconstruction error to generate a target reconstruction error of the target battery; obtain a target reconstruction error threshold corresponding to the target reconstruction error, and generate a short - circuit fault diagnosis result of the target battery based on the target reconstruction error and the target reconstruction error threshold.

[0151] In an exemplary embodiment, the short - circuit fault diagnosis module 400 is further configured to obtain standard reconstruction error distribution information when the target battery has no fault; generate a target reconstruction error threshold corresponding to the target reconstruction error based on the standard reconstruction error distribution information.

[0152] In an exemplary embodiment, the feature fusion module 200 is further configured to perform a linear transformation on multiple operation state feature information to obtain multiple intermediate vectors, where the intermediate vectors include a value vector, a query vector, and a key vector; use multiple attention heads to perform scaled dot - product attention operations on the value vector, the query vector, and the key vector in each intermediate vector respectively to obtain multiple intermediate operation feature data; combine the multiple intermediate operation feature data to obtain operation state fusion feature information.

[0153] In an exemplary embodiment, the data acquisition module 100 is further configured to obtain multiple initial operation state information of the target battery; perform a timing alignment process on the multiple initial operation state information to obtain an initial operation state timing sequence; perform a normalization process on the multiple timing - aligned initial operation state information in the initial operation state timing sequence to obtain an operation state timing sequence, and the operation state timing sequence is used to represent multiple operation state information of the target battery in the form of a time series.

[0154] Each module in the above battery short - circuit fault diagnosis device can be implemented in whole or in part by software, hardware, and their combination. The above - mentioned modules can be embedded in the processor of the computer device in hardware form or be independent of it, or be stored in the memory of the computer device in software form so that the processor can call and execute the operations corresponding to the above - mentioned modules.

[0155] In an exemplary embodiment, a computer device is provided. The computer device can be a terminal, and its internal structure diagram can be as Figure 7As shown in the figure. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface, the display unit, and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a method for diagnosing battery short-circuit faults. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0156] Those skilled in the art can understand that Figure 7 the structure shown in the figure is a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0157] In one embodiment, a computer device is further provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above method embodiments are implemented.

[0158] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by the processor, the steps in the above method embodiments are implemented.

[0159] In one embodiment, a computer program product is provided, including a computer program. When the computer program is executed by the processor, the steps in the above method embodiments are implemented.

[0160] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in this application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in this application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, data processing logics based on quantum computing, artificial intelligence (AI) processors, etc., without limitation.

[0161] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this application.

[0162] The above embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all fall within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. A battery short circuit fault diagnosis method, characterized in that: The method comprises: Acquire multiple operating state information and health state information of the target battery, and perform feature extraction on the multiple operating state information and health state information respectively to obtain multiple operating state feature information and health state feature information; Performing feature fusion on the plurality of running state feature information to obtain running state fusion feature information; Performing data reconstruction processing on the running state fusion feature information to obtain a first reconstruction error, and performing data reconstruction processing on the health state feature information to obtain a second reconstruction error; A short circuit fault diagnosis result of the target battery is generated based on the first reconstruction error and the second reconstruction error.

2. The method according to claim 1, characterized in that The performing data reconstruction processing on the running state fusion feature information to obtain a first reconstruction error includes: Mapping the running state fusion feature information to a data space of a preset dimension to obtain running state potential feature information; Performing data recovery processing on the potential characteristic information of the operating state in the data space of the preset dimension to obtain target operating state characteristic information; Detecting a first reconstruction error between the running state potential feature information and the target running state feature information; The performing data reconstruction processing on the health status characteristic information to obtain a second reconstruction error includes: Mapping the health status characteristic information to a data space of a preset dimension to obtain health status potential characteristic information; Performing data recovery processing on the health status potential characteristic information in the data space of the preset dimension to obtain target health status characteristic information; A second reconstruction error between the health status potential feature information and the target health status feature information is detected.

3. The method according to claim 1, characterized in that The generating a short circuit fault diagnosis result of the target battery based on the first reconstruction error and the second reconstruction error includes: Performing weighted processing on the first reconstruction error and the second reconstruction error to generate a target reconstruction error of the target battery; A target reconstruction error threshold corresponding to the target reconstruction error is obtained, and a short circuit fault diagnosis result of the target battery is generated based on the target reconstruction error and the target reconstruction error threshold.

4. The method according to claim 3, characterized in that The obtaining a target reconstruction error threshold corresponding to the target reconstruction error includes: Obtaining standard reconstruction error distribution information when the target battery is not faulty; A target reconstruction error threshold corresponding to the target reconstruction error is generated based on the standard reconstruction error distribution information.

5. The method according to claim 1, characterized in that The step of fusing the plurality of running state feature information to obtain the running state fused feature information includes: Performing linear transformation on the plurality of running state feature information to obtain a plurality of intermediate vectors, wherein the intermediate vectors include a value vector, a query vector, and a key vector; Using multiple attention heads, scaled dot product attention operations are performed on the value vector, query vector, and key vector in each intermediate vector to obtain multiple intermediate running feature data; The plurality of intermediate operation feature data are combined to obtain operation status fusion feature information.

6. The method according to claim 1, characterized in that The obtaining of multiple operating status information of the target battery includes: Acquire multiple initial operating status information of the target battery; Performing timing alignment processing on the multiple initial running state information to obtain an initial running state timing sequence; The initial operating state information after multiple time series alignment in the initial operating state time series is normalized to obtain an operating state time series, and the operating state time series is used to represent multiple operating state information of the target battery in the form of a time series.

7. A battery short circuit fault diagnosis device, characterized in that: The device comprises: A data acquisition module, used to acquire multiple operating status information and health status information of the target battery, and perform feature extraction on the multiple operating status information and health status information respectively to obtain multiple operating status feature information and health status feature information; A feature fusion module, used for performing feature fusion on the plurality of running state feature information to obtain running state fusion feature information; A data reconstruction processing module, used to perform data reconstruction processing on the running state fusion feature information to obtain a first reconstruction error, and to perform data reconstruction processing on the health state feature information to obtain a second reconstruction error; A short-circuit fault diagnosis module is used to generate a short-circuit fault diagnosis result of the target battery based on the first reconstruction error and the second reconstruction error.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images