Battery pack fault diagnosis method and device, electronic equipment and storage medium

Through the attention convolution neural network, the voltage signal of the battery pack is preprocessed and feature extracted, combined with the fault classification detection model, the fault type of the battery pack is identified, and the problem of noise signal influence and inaccurate fault characteristics in the existing technology is solved, and the problem of inaccurate fault characteristics is realized. More accurate and adaptive fault diagnosis.

CN119087231APending Publication Date: 2024-12-06GAC AION NEW ENERGY AUTOMOBILE CO LTD
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411278175.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-12
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

The existing battery management system ignores noise signals in fault diagnosis, resulting in failure to effectively identify and locate system failures, and the fault feature extraction is not accurate enough, resulting in false alarms or missed alarms, affecting system performance and reliability.

Method used

The attention convolution neural network is used to preprocess and extract data of the original voltage signal of the battery pack. The fault type is identified through the attention convolution neural network and fault classification detection model, and the features with high correlation are extracted to avoid manually selecting features. Achieve more comprehensive and adaptive fault feature capture.

Benefits of technology

It improves the efficiency of fault feature capture and feature extraction effect, enhances the accuracy of the fault detection model, reduces the bottleneck of traditional manual feature extraction, and outputs more accurate fault type identification results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119087231A_ABST
    Figure CN119087231A_ABST
Patent Text Reader

Abstract

The invention provides a battery pack fault diagnosis method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining an original voltage signal of a battery pack, carrying out the data preprocessing of the original voltage signal of the battery pack, and obtaining a preprocessing signal, and taking the preprocessed signal as an input of an attention convolutional neural network, so that the attention convolutional neural network outputs a feature extraction result based on the preprocessed signal, and inputting the feature extraction result to a fault classification detection model, and identifying the fault type of the battery pack based on the fault classification detection model and the feature extraction result. According to the invention, the battery pack fault can be identified. Meanwhile, the fault feature capturing efficiency can be improved, the features can be improved, and the feature extraction effect can be improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of battery fault diagnosis, and in particular, to a battery pack fault diagnosis method, device, electronic device and storage medium. Background Art

[0002] For electric vehicle battery fault detection, the existing technologies mainly include the following:

[0003] The first type discloses a method, system and device for early warning of battery health status, which relates to the field of battery safety, including obtaining historical battery data; training a model based on the historical battery data and obtaining predicted battery data of the target time output by the model; adjusting the parameters in the model based on the historical battery data and the predicted data of the target time until the error between the historical battery data and the predicted battery data is less than the preset error; after training the model with the historical battery data, the error between the model and the actual historical battery data is less than the preset error, at which time the data output by the model is determined to be reliable, and then determining whether the battery data in the future time period after the current time output by the model exceeds the threshold value. Avoid the hysteresis of determining whether the battery data is faulty after determining the actual battery data.

[0004] The second type discloses a fault diagnosis system and a method for a lithium battery pack, which extracts the voltage signal and SPC signal components of each cell structure in the lithium battery pack in the time dimension through a context-based encoder model and a convolutional neural network to generate a first feature map, and also mines the voltage signal and SPC signal waveform of the lithium battery pack as a whole for local high-dimensional feature distribution to generate a second feature map, and further reuses the third convolutional neural network as a feature descriptor to make up for the difference in feature semantic expression between the first feature map as a shallow feature and the second feature map as a deep feature, so that iterative mutual weighting coefficients can be introduced during feature fusion to improve classification accuracy. In this way, the fault of the lithium battery pack can be detected to ensure the safety and service life of the battery pack.

[0005] The third method discloses a method for predicting the energy state of a series battery system, which is as follows: a battery system model is established by combining the battery cell model parameters with the series circuit characteristics, and a battery space state equation is constructed; then the battery system state of charge SOC is estimated by combining the battery system voltage online detection value U and the current online detection value I; SOC and U are used as inputs, and an energy state prediction base value SOE-b is obtained through an energy state base value prediction module; battery cell offline data is used as input, and the battery cell normal data is obtained by passing through a fault discriminator and a data filter in sequence, and then the voltage deviation value of each battery cell and the battery system current online detection value I are combined, and a BP neural network is used to construct an SOE corrector to output an energy state compensation value ΔSOE-b; the energy state prediction base value SOE-b is superimposed with the energy state compensation value ΔSOE-b to obtain the energy state prediction value SOE-r of the series battery system.

[0006] However, there are some common defects in battery management system fault detection. One of the major problems is that existing technologies often ignore the potential impact of noise signals on fault signals. This neglect may lead to the failure to effectively identify and locate system faults in actual operation. Secondly, the challenge of fault diagnosis is mainly reflected in finding features that can accurately reflect fault information. Determining these features is crucial to building a reliable fault detection system, because relying solely on imperfect or inaccurate features may lead to false positives or false negatives, affecting the performance and reliability of the battery management system. Therefore, current technology still needs more in-depth research and improvement in fault diagnosis to improve the system's robustness to noise signals and find more accurate and effective fault features. Summary of the invention

[0007] The purpose of the embodiments of the present application is to provide a battery pack fault diagnosis method, device, electronic device and storage medium for identifying battery pack faults. At the same time, the embodiments of the present application can improve the efficiency of fault feature capture and improve the features and feature extraction effect.

[0008] In a first aspect, the present invention provides a battery pack fault diagnosis method, the method comprising:

[0009] Obtain the original voltage signal of the battery pack;

[0010] Performing data preprocessing on the original voltage signal of the battery pack to obtain a preprocessed signal;

[0011] Using the preprocessed signal as an input of an attention convolutional neural network, so that the attention convolutional neural network outputs a feature extraction result based on the preprocessed signal;

[0012] The feature extraction result is input into a fault classification detection model to identify the fault type of the battery pack based on the feature extraction result based on the fault classification detection model.

[0013] The method of the first aspect of the present application obtains the original voltage signal of the battery pack, and then can perform data preprocessing on the original voltage signal of the battery pack and obtain a preprocessed signal, and then can use the preprocessed signal as the input of the attention convolutional neural network, so that the attention convolutional neural network outputs a feature extraction result based on the preprocessed signal, and then can input the feature extraction result into the fault classification detection model, so as to identify the fault type of the battery pack based on the fault classification detection model based on the feature extraction result, and finally realize the identification of the fault type of the battery pack based on the voltage signal of the battery pack. At the same time, compared with the prior art, the present application can extract features with a high correlation with the fault type based on the attention convolutional neural network, thereby avoiding manual selection of such features, thereby eliminating the bottleneck of traditional manual feature extraction, and realizing more comprehensive and adaptive fault feature capture, and through the features with a high correlation, the fault detection model can output a more accurate fault type identification result.

[0014] In an optional implementation manner, performing data preprocessing on the original voltage signal of the battery pack to obtain a preprocessed signal includes:

[0015] Performing data filtering on the original voltage signal of the battery pack and obtaining a filtering result;

[0016] Performing wavelet transform on the filtering result to obtain a reconstructed voltage signal;

[0017] The preprocessed signal is generated based on the reconstructed voltage signal.

[0018] This optional implementation can perform data filtering on the original voltage signal of the battery pack and obtain a filtering result, thereby removing the noise signal in the original voltage signal. On the other hand, since the value of the original voltage signal is small, it is difficult for the model to distinguish, so the value of the original voltage signal can be amplified by wavelet transformation, so that the model can identify features based on larger values.

[0019] In an optional implementation, during the training process, generating the preprocessed signal based on the reconstructed voltage signal includes:

[0020] Performing data expansion based on the reconstructed voltage signal to obtain an expanded sample;

[0021] The expanded samples and the reconstructed voltage signal are used as the preprocessed signal.

[0022] This optional implementation method can obtain expanded samples by reconstructing the voltage signal for data expansion. This can increase the number of samples, thereby solving the problem of small samples and enabling the model to be trained based on more abundant samples and have higher fault type identification accuracy.

[0023] In an optional implementation manner, performing data filtering on the original voltage signal of the battery pack and obtaining a filtering result includes:

[0024] Decomposing the original voltage signal of the battery pack into a plurality of intrinsic mode function frequency characteristic components based on an empirical mode decomposition algorithm;

[0025] Calculating the correlation coefficient between the frequency characteristic component of the intrinsic mode function and the original voltage signal, wherein the correlation coefficient is a Pearson correlation coefficient;

[0026] Based on the correlation coefficient between the intrinsic mode function frequency characteristic component and the original voltage signal, a component that meets a first preset condition is screened out from a plurality of the intrinsic mode function frequency characteristic components, and the component that meets the first preset condition is used as the filtering result, wherein the component that meets the first preset condition is a signal component that carries a fault signal.

[0027] This optional implementation can decompose the original voltage signal based on the empirical mode decomposition algorithm, filter out the signal component carrying the fault signal based on the Pearson correlation coefficient, and use the signal component carrying the fault signal as the filtering result.

[0028] In an optional embodiment, the method of decomposing the original voltage signal of the battery pack into a plurality of intrinsic mode function frequency characteristic components based on an empirical mode decomposition algorithm includes:

[0029] Determine all local maxima of the original voltage signal and all local minima of the original voltage signal, and determine the upper and lower envelopes of the original voltage signal based on the cubic spline interpolation algorithm and all local maxima of the original voltage signal and all local minima of the original voltage signal, wherein the upper envelope of the original voltage signal is represented by E 1 Indicates that the lower envelope of the original voltage signal is represented by E 2 express;

[0030] The signal average value is calculated based on the upper envelope of the original voltage signal and the upper and lower envelopes of the original voltage signal, wherein the calculation formula is:

[0031] m 1 (t) represents the signal average value;

[0032] Calculate the target component h based on the signal average value and the original signal n (t), where:

[0033] h n (t) = x(t) - m 1 (t), x(t) represents the original signal;

[0034] Determine whether the target component satisfies a second preset condition, wherein the second preset condition includes a first condition and a second condition, the first condition refers to that the difference between the number of extreme values ​​of the target component and the number of zero crossings of the target component is less than or equal to 1, and the second condition refers to that the average value of the upper and lower envelopes of the target component is 0;

[0035] When the target component satisfies the second preset condition, the target component is used as the intrinsic mode function frequency characteristic component of the original voltage signal;

[0036] When the target component does not satisfy the preset condition, the target component is taken as the original signal for iterative signal decomposition to obtain the intrinsic mode function frequency characteristic component of the original voltage signal.

[0037] This optional implementation can obtain the intrinsic mode function frequency characteristic component of the original voltage signal based on a cubic spline interpolation algorithm.

[0038] In an optional implementation, the calculation formula for calculating the correlation coefficient between the frequency characteristic component of the intrinsic mode function and the original voltage signal is:

[0039]

[0040] Among them, X i represents the frequency characteristic component of the intrinsic mode function, Y i represents the original voltage signal, P(x i ,y i ) is X i and Y i At the same time, take the value x i and i The probability, P(x i ) and P(y i ) are X i and Y i The marginal probability, I(X i ,Y i ) represents the correlation coefficient between the frequency characteristic component of the intrinsic mode function and the original voltage signal.

[0041] In an optional implementation manner, performing wavelet transform on the filtering result to obtain a reconstructed voltage signal includes:

[0042] The filtering result is subjected to wavelet transformation to obtain wavelet coefficients at different frequency scales, where:

[0043]

[0044] W(a,b) represents the wavelet coefficients, x * (t) represents the filtering result, Ψ a,b (t) represents the wavelet function, a and b represent the scale parameter and translation parameter respectively;

[0045] The wavelet coefficients are subjected to threshold processing and a threshold processing result is obtained, and the calculation formula is:

[0046]

[0047] W thresh (a, b) represents the threshold processing result, T represents the target threshold used in the threshold processing, the target threshold is determined based on the normal distribution, and σ represents the standard deviation of the normal distribution;

[0048] Based on the threshold processing result, an inverse wavelet transform is performed to obtain the reconstructed voltage signal, and the calculation formula is:

[0049] x recon (t)=∑ a ∑ b W thresh (a,b)Ψ a,b (t);

[0050] x recon (t) represents the reconstructed voltage signal.

[0051] In a second aspect, the present invention provides a battery pack fault diagnosis device, the device comprising:

[0052] A data acquisition module, used to acquire the original voltage signal of the battery pack;

[0053] A data preprocessing module, used for performing data preprocessing on the original voltage signal of the battery pack and obtaining a preprocessed signal;

[0054] A feature extraction module, used for taking the preprocessed signal as an input of an attention convolutional neural network, so that the attention convolutional neural network outputs a feature extraction result based on the preprocessed signal;

[0055] A fault diagnosis module is used to input the feature extraction result into a fault classification detection model to identify the fault type of the battery pack based on the feature extraction result based on the fault classification detection model.

[0056] The device of the second aspect of the present application obtains the original voltage signal of the battery pack, and then can perform data preprocessing on the original voltage signal of the battery pack and obtain a preprocessed signal, and then can use the preprocessed signal as the input of the attention convolutional neural network, so that the attention convolutional neural network outputs a feature extraction result based on the preprocessed signal, and then can input the feature extraction result into the fault classification detection model, so as to identify the fault type of the battery pack based on the fault classification detection model based on the feature extraction result, and finally realize the identification of the fault type of the battery pack based on the voltage signal of the battery pack. At the same time, compared with the prior art, the present application can extract features with a high correlation with the fault type based on the attention convolutional neural network, thereby avoiding manual selection of such features, thereby eliminating the bottleneck of traditional manual feature extraction, and realizing more comprehensive and adaptive fault feature capture, and through the features with a high correlation, the fault detection model can output a more accurate fault type identification result.

[0057] In a third aspect, the present invention provides an electronic device, comprising:

[0058] Processor; and

[0059] A memory is configured to store machine-readable instructions, which, when executed by the processor, execute the battery pack fault diagnosis method as described in any of the aforementioned embodiments.

[0060] The electronic device of the third aspect of the present application can obtain the original voltage signal of the battery pack by executing the battery pack fault diagnosis method, and then can perform data preprocessing on the original voltage signal of the battery pack and obtain a preprocessed signal, and then can use the preprocessed signal as the input of the attention convolutional neural network, so that the attention convolutional neural network outputs a feature extraction result based on the preprocessed signal, and then can input the feature extraction result into the fault classification detection model, so as to identify the fault type of the battery pack based on the feature extraction result based on the fault classification detection model, and finally realize the identification of the fault type of the battery pack based on the voltage signal of the battery pack. At the same time, compared with the prior art, the present application can extract features with a high correlation with the fault type based on the attention convolutional neural network, thereby avoiding manual selection of such features, thereby eliminating the bottleneck of traditional manual feature extraction, and realizing more comprehensive and adaptive fault feature capture, and the fault detection model can output more accurate fault type identification results through features with a high correlation.

[0061] In a fourth aspect, the present invention provides a storage medium storing a computer program, wherein the computer program is executed by a processor to perform a battery pack fault diagnosis method as described in any one of the aforementioned embodiments.

[0062] The storage medium of the third aspect of the present application can obtain the original voltage signal of the battery pack by executing the battery pack fault diagnosis method, and then can perform data preprocessing on the original voltage signal of the battery pack and obtain a preprocessed signal, and then can use the preprocessed signal as the input of the attention convolutional neural network, so that the attention convolutional neural network outputs a feature extraction result based on the preprocessed signal, and then can input the feature extraction result into the fault classification detection model, so as to identify the fault type of the battery pack based on the feature extraction result based on the fault classification detection model, and finally realize the identification of the fault type of the battery pack based on the voltage signal of the battery pack. At the same time, compared with the prior art, the present application can extract features with a high correlation with the fault type based on the attention convolutional neural network, thereby avoiding manual selection of such features, thereby eliminating the bottleneck of traditional manual feature extraction, and realizing more comprehensive and adaptive fault feature capture, and the fault detection model can output more accurate fault type identification results through features with a high correlation. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0064] Figure 1 This is a flow chart of a battery pack fault diagnosis method disclosed in an embodiment of the present application.

[0065] Figure 2 It is a schematic diagram of the structure of an attention convolutional neural network disclosed in an embodiment of the present application;

[0066] Figure 3 It is a schematic diagram of sample expansion disclosed in the embodiment of the present application;

[0067] Figure 4 It is a structural schematic diagram of a battery pack fault diagnosis device disclosed in an embodiment of the present application;

[0068] Figure 5 It is a structural schematic diagram of an electronic device disclosed in an embodiment of the present application. DETAILED DESCRIPTION

[0069] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0070] Embodiment 1

[0071] See also Figure 1 , Figure 1 is a flowchart of a battery pack fault diagnosis method disclosed in an embodiment of the present application, such as Figure 1 As shown, the method of the embodiment of the present application includes the following steps:

[0072] 101. Obtain an original voltage signal of the battery pack;

[0073] 102. Perform data preprocessing on the original voltage signal of the battery pack and obtain a preprocessing signal;

[0074] 103. Using the preprocessed signal as an input of the attention convolutional neural network, so that the attention convolutional neural network outputs a feature extraction result based on the preprocessed signal;

[0075] 104. Input the feature extraction result into a fault classification detection model to identify the fault type of the battery pack based on the fault classification detection model and the feature extraction result.

[0076] The method of the embodiment of the present application obtains the original voltage signal of the battery pack, and then can perform data preprocessing on the original voltage signal of the battery pack and obtain a preprocessed signal, and then can use the preprocessed signal as the input of the attention convolutional neural network, so that the attention convolutional neural network outputs a feature extraction result based on the preprocessed signal, and then can input the feature extraction result into the fault classification detection model, so as to identify the fault type of the battery pack based on the fault classification detection model and the feature extraction result, and finally realize the identification of the fault type of the battery pack based on the voltage signal of the battery pack. At the same time, compared with the prior art, the present application can extract features with a high correlation with the fault type based on the attention convolutional neural network, thereby avoiding manual selection of such features, thereby eliminating the bottleneck of traditional manual feature extraction, and realizing more comprehensive and adaptive fault feature capture, and through the features with a high correlation, the fault detection model can output a more accurate fault type identification result.

[0077] For steps 101-104, steps 101 and 104 can also be used as a training process to train the attention convolutional neural network and the fault classification detection model. When the fault classification detection model and the attention convolutional neural network meet the training conditions through iterative training, the original voltage signal of the target battery pack can be processed based on steps 101-104 to identify the fault type of the target battery pack.

[0078] In an embodiment of the present application, the original voltage signal of the battery pack may refer to a voltage signal of the battery pack composed of multiple battery cells, wherein the multiple battery cells are connected in series, and thus the voltage signal of the battery pack may be calculated based on the voltage signals of the multiple battery cells.

[0079] In the embodiment of the present application, the attention convolutional neural network refers to a neural network with an attention head, wherein Figure 2 , Figure 2 Schematic diagram of the structure of an attention convolutional neural network disclosed in the embodiment of the present application. Figure 2 As shown in the figure, the neural network consists of convolutional layers, attention mechanisms, pooling layers, and fully connected layers. All convolutional layers use 2×2 convolutional filters with a stride of 1 and an activation function of tanh. All maximum pooling layers use a 2×2 window for pooling in 2 steps. The attention mechanisms use channel attention and spatial attention to ensure effective feature extraction.

[0080] Furthermore, the channel attention mechanism is used to enhance the correlation between different channels in the feature map, and its formula is as follows:

[0081]

[0082] in, is the input feature map after the first convolutional layer, W 1 and W 2 is the learned weight matrix, σ is the sigmoid activation function, δ represents the global average pooling operation, and F CA It is a feature map enhanced by the channel attention mechanism.

[0083] Furthermore, the spatial attention mechanism is used to enhance the correlation between different spatial positions in the feature map, and its formula is as follows:

[0084]

[0085] in, is the feature corresponding to the cth channel of the input feature map, C is the total number of data channels, σ is the sigmoid activation function, and F CA It is a feature map enhanced by the channel attention mechanism.

[0086] Furthermore, the convolution of the first layer uses the tanh activation function, the second layer uses the ReLU activation function, and the third and fourth layers use the ELU activation function. The subsequent fully connected output layer uses softamx to solve the multi-classification problem.

[0087] It should be noted that the attention convolutional neural network can be integrated into the fault classification detection model.

[0088] In an optional implementation, performing data preprocessing on the original voltage signal of the battery pack to obtain a preprocessed signal includes the following steps:

[0089] Performing data filtering on the original voltage signal of the battery pack and obtaining a filtering result;

[0090] Perform wavelet transform on the filtering result and obtain the reconstructed voltage signal;

[0091] A preprocessed signal is generated based on the reconstructed voltage signal.

[0092] This optional implementation can perform data filtering on the original voltage signal of the battery pack and obtain a filtering result, thereby removing the noise signal in the original voltage signal. Therefore, the original voltage signal can be retained through wavelet transformation, so that the model can learn effective features.

[0093] In an optional embodiment, during the training process, generating a preprocessing signal based on the reconstructed voltage signal includes the following steps:

[0094] Performing data expansion based on the reconstructed voltage signal to obtain an expanded sample;

[0095] The expanded samples and the reconstructed voltage signal are used as preprocessing signals.

[0096] This optional implementation method can obtain expanded samples by reconstructing the voltage signal for data expansion. This can increase the number of samples, thereby solving the problem of small samples and enabling the model to be trained based on more abundant samples and have higher fault type identification accuracy.

[0097] For the above optional implementation, the data expansion of the reconstructed voltage signal adopts a sliding window, wherein the length of the sliding window can be a preset value. Figure 3 , Figure 3 is a schematic diagram of sample expansion disclosed in the embodiment of this application. Figure 3 As shown, by sliding the sliding window on the reconstructed voltage signal, n sub-data can be selected from the reconstructed voltage signal, and the n sub-data constitute a new sample.

[0098] In an optional implementation, the step of performing data filtering on the original voltage signal of the battery pack and obtaining a filtering result comprises the following steps:

[0099] Based on the empirical mode decomposition algorithm, the original voltage signal of the battery pack is decomposed into several intrinsic mode function frequency characteristic components;

[0100] Calculate the correlation coefficient between the frequency characteristic component of the intrinsic mode function and the original voltage signal, where the correlation coefficient is the Pearson correlation coefficient;

[0101] Based on the correlation coefficient between the intrinsic mode function frequency characteristic component and the original voltage signal, a component that meets the first preset condition is screened out from a number of intrinsic mode function frequency characteristic components, and the component that meets the first preset condition is used as the filtering result, wherein the component that meets the first preset condition is a signal component that carries a fault signal.

[0102] This optional implementation can decompose the original voltage signal based on the empirical mode decomposition algorithm, filter out the signal component carrying the fault signal based on the Pearson correlation coefficient, and use the signal component carrying the fault signal as the filtering result.

[0103] In an optional embodiment, decomposing the original voltage signal of the battery pack into a plurality of intrinsic mode function frequency characteristic components based on an empirical mode decomposition algorithm includes the following steps:

[0104] Determine all local maxima and all local minima of the original voltage signal, and determine the upper and lower envelopes of the original voltage signal based on the cubic spline interpolation algorithm and all local maxima and all local minima of the original voltage signal, where the upper envelope of the original voltage signal is represented by E 1 Indicates that the lower envelope of the original voltage signal is E v express;

[0105] The signal average is calculated based on the upper envelope of the original voltage signal and the upper and lower envelopes of the original voltage signal, where the calculation formula is:

[0106] m 1 (t) represents the signal average value;

[0107] Calculate the target component h based on the signal average and the original signal n (t), where:

[0108] h n (t) = x(t) - m 1 (t), x(t) represents the original signal;

[0109] Determine whether the target component satisfies a second preset condition, wherein the second preset condition includes a first condition and a second condition, the first condition refers to that the difference between the number of extreme values ​​of the target component and the number of zero crossings of the target component is less than or equal to 1, and the second condition refers to that the average value of the upper and lower envelopes of the target component is 0;

[0110] When the target component satisfies the second preset condition, the target component is used as the intrinsic mode function frequency characteristic component of the original voltage signal;

[0111] When the target component does not meet the preset conditions, the target component is taken as the original signal for iterative signal decomposition to obtain the intrinsic mode function frequency characteristic component of the original voltage signal.

[0112] This optional implementation can obtain the intrinsic mode function frequency characteristic component of the original voltage signal based on the cubic spline interpolation algorithm.

[0113] In an optional implementation, the calculation formula for calculating the correlation coefficient between the frequency characteristic component of the intrinsic mode function and the original voltage signal is:

[0114]

[0115] Among them, X i represents the frequency characteristic component of the intrinsic mode function, Y i Represents the original voltage signal, P(x i ,y i ) is X i and Y i At the same time, take the value x i and i The probability, P(x i ) and P(y i ) are X i and Y i The marginal probability, I(X i ,Y i ) represents the correlation coefficient between the frequency characteristic component of the intrinsic mode function and the original voltage signal.

[0116] In an optional implementation, performing wavelet transform on the filtering result and obtaining a reconstructed voltage signal includes:

[0117] Perform wavelet transform on the filtering result and obtain wavelet coefficients at different frequency scales, where:

[0118]

[0119] W(a,b) represents the wavelet coefficients, x * (t) represents the filtering result, Ψ a,b (t) represents the wavelet function, a and b represent the scale parameter and translation parameter respectively;

[0120] Perform threshold processing on the wavelet coefficients and obtain the threshold processing result. The calculation formula is:

[0121]

[0122] W thresh (a, b) represents the result of threshold processing, T represents the target threshold used in threshold processing, the target threshold is determined based on normal distribution, and σ represents the standard deviation of normal distribution;

[0123] Based on the threshold processing result, an inverse wavelet transform is performed to obtain the reconstructed voltage signal. The calculation formula is:

[0124] x recon (t)=∑ a ∑ b W thresh (a,b)Ψ a,b(t);

[0125] x recon (t) represents the reconstructed voltage signal.

[0126] Embodiment 2

[0127] See also Figure 4 , Figure 4 is a schematic diagram of the structure of a battery pack fault diagnosis device disclosed in an embodiment of the present application, such as Figure 4 As shown, the device of the embodiment of the present application includes the following functional modules:

[0128] The data acquisition module 201 is used to acquire the original voltage signal of the battery pack;

[0129] The data preprocessing module 202 is used to perform data preprocessing on the original voltage signal of the battery pack and obtain a preprocessed signal;

[0130] A feature extraction module 203 is used to use the preprocessed signal as an input of the attention convolutional neural network, so that the attention convolutional neural network outputs a feature extraction result based on the preprocessed signal;

[0131] The fault diagnosis module 204 is used to input the feature extraction result into the fault classification detection model to identify the fault type of the battery pack based on the fault classification detection model and the feature extraction result.

[0132] The device of the embodiment of the present application obtains the original voltage signal of the battery pack, and then can perform data preprocessing on the original voltage signal of the battery pack and obtain a preprocessed signal, and then can use the preprocessed signal as the input of the attention convolutional neural network, so that the attention convolutional neural network outputs a feature extraction result based on the preprocessed signal, and then can input the feature extraction result into the fault classification detection model, so as to identify the fault type of the battery pack based on the fault classification detection model and the feature extraction result, and finally realize the identification of the fault type of the battery pack based on the voltage signal of the battery pack. At the same time, compared with the prior art, the present application can extract features with a high correlation with the fault type based on the attention convolutional neural network, thereby avoiding manual selection of such features, thereby eliminating the bottleneck of traditional manual feature extraction, and realizing more comprehensive and adaptive fault feature capture, and through the features with a high correlation, the fault detection model can output a more accurate fault type identification result.

[0133] Embodiment 3

[0134] See also Figure 5 , Figure 5 is a schematic diagram of the structure of an electronic device disclosed in an embodiment of the present application, such as Figure 5 As shown, the electronic device of the embodiment of the present application includes:

[0135] Processor 301; and

[0136] The memory 302 is configured to store machine-readable instructions, and when the instructions are executed by the processor 301, the battery pack fault diagnosis method as described in any one of the aforementioned implementations is executed.

[0137] The electronic device of the embodiment of the present application can obtain the original voltage signal of the battery pack by executing the battery pack fault diagnosis method, and then can perform data preprocessing on the original voltage signal of the battery pack and obtain a preprocessed signal, and then can use the preprocessed signal as the input of the attention convolutional neural network, so that the attention convolutional neural network outputs a feature extraction result based on the preprocessed signal, and then can input the feature extraction result into the fault classification detection model, so as to identify the fault type of the battery pack based on the fault classification detection model and the feature extraction result, and finally realize the identification of the fault type of the battery pack based on the voltage signal of the battery pack. At the same time, compared with the prior art, the present application can extract features with a high correlation with the fault type based on the attention convolutional neural network, thereby avoiding manual selection of such features, thereby eliminating the bottleneck of traditional manual feature extraction, and realizing more comprehensive and adaptive fault feature capture, and the fault detection model can output more accurate fault type identification results through features with a high correlation.

[0138] Embodiment 4

[0139] An embodiment of the present application provides a storage medium, wherein the storage medium stores a computer program, and the computer program is executed by a processor as a battery pack fault diagnosis method according to any one of the aforementioned implementations.

[0140] The storage medium of the embodiment of the present application can obtain the original voltage signal of the battery pack by executing the battery pack fault diagnosis method, and then can perform data preprocessing on the original voltage signal of the battery pack and obtain the preprocessed signal, and then can use the preprocessed signal as the input of the attention convolutional neural network, so that the attention convolutional neural network outputs a feature extraction result based on the preprocessed signal, and then can input the feature extraction result into the fault classification detection model, so as to identify the fault type of the battery pack based on the fault classification detection model and the feature extraction result, and finally realize the identification of the fault type of the battery pack based on the voltage signal of the battery pack. At the same time, compared with the prior art, the present application can extract features with a high correlation with the fault type based on the attention convolutional neural network, thereby avoiding manual selection of such features, thereby eliminating the bottleneck of traditional manual feature extraction, and realizing more comprehensive and adaptive fault feature capture, and the fault detection model can output more accurate fault type identification results through features with a high correlation.

[0141] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.

[0142] In addition, the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0143] Furthermore, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0144] It should be noted that if the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can essentially be embodied in the form of a software product, or in other words, the part that contributes to the prior art or the part of the technical solution. The computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM) random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0145] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0146] The above are only embodiments of the present application and are not intended to limit the scope of protection of the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the scope of protection of the present application.

Claims

1. A battery pack fault diagnosis method, characterized in that: The method comprises: Obtain the original voltage signal of the battery pack; Performing data preprocessing on the original voltage signal of the battery pack to obtain a preprocessed signal; Using the preprocessed signal as an input of an attention convolutional neural network, so that the attention convolutional neural network outputs a feature extraction result based on the preprocessed signal; The feature extraction result is input into a fault classification detection model to identify the fault type of the battery pack based on the feature extraction result based on the fault classification detection model.

2. The method according to claim 1, characterized in that The performing data preprocessing on the original voltage signal of the battery pack to obtain a preprocessed signal includes: Performing data filtering on the original voltage signal of the battery pack and obtaining a filtering result; Performing wavelet transform on the filtering result to obtain a reconstructed voltage signal; The preprocessed signal is generated based on the reconstructed voltage signal.

3. The method according to claim 2, characterized in that During the training process, generating the preprocessing signal based on the reconstructed voltage signal includes: Performing data expansion based on the reconstructed voltage signal to obtain an expanded sample; The expanded samples and the reconstructed voltage signal are used as the preprocessed signal.

4. The method according to claim 2, characterized in that The performing data filtering on the original voltage signal of the battery pack and obtaining a filtering result includes: Decomposing the original voltage signal of the battery pack into a plurality of intrinsic mode function frequency characteristic components based on an empirical mode decomposition algorithm; Calculating the correlation coefficient between the frequency characteristic component of the intrinsic mode function and the original voltage signal, wherein the correlation coefficient is a Pearson correlation coefficient; Based on the correlation coefficient between the intrinsic mode function frequency characteristic component and the original voltage signal, a component that meets a first preset condition is screened out from a plurality of the intrinsic mode function frequency characteristic components, and the component that meets the first preset condition is used as the filtering result, wherein the component that meets the first preset condition is a signal component that carries a fault signal.

5. The method according to claim 4, characterized in that The original voltage signal of the battery pack is decomposed into a plurality of intrinsic mode function frequency characteristic components based on the empirical mode decomposition algorithm, including: Determine all local maxima of the original voltage signal and all local minima of the original voltage signal, and determine the upper and lower envelopes of the original voltage signal based on a cubic spline interpolation algorithm and all local maxima of the original voltage signal and all local minima of the original voltage signal, wherein the upper envelope of the original voltage signal is represented by E1, and the lower envelope of the original voltage signal is represented by E2; The signal average value is calculated based on the upper envelope of the original voltage signal and the upper and lower envelopes of the original voltage signal, wherein the calculation formula is: m1(t) represents the signal average value; Calculate the target component h based on the signal average value and the original signal n (t), where: h n (t)=x(t)-m1(t), x(t) represents the original signal; Determine whether the target component satisfies a second preset condition, wherein the second preset condition includes a first condition and a second condition, the first condition refers to that the difference between the number of extreme values ​​of the target component and the number of zero crossings of the target component is less than or equal to 1, and the second condition refers to that the average value of the upper and lower envelopes of the target component is 0; When the target component satisfies the second preset condition, the target component is used as the intrinsic mode function frequency characteristic component of the original voltage signal; When the target component does not satisfy the preset condition, the target component is taken as the original signal for iterative signal decomposition to obtain the intrinsic mode function frequency characteristic component of the original voltage signal.

6. The method according to claim 5, characterized in that The calculation formula for calculating the correlation coefficient between the frequency characteristic component of the intrinsic mode function and the original voltage signal is: Among them, X i represents the frequency characteristic component of the intrinsic mode function, Y i represents the original voltage signal, P(x i ,y i ) is X i and Y i At the same time, take the value x i and i The probability, P(x i ) and P(y i ) are X i and Y i The marginal probability, I(X i , Y i ) represents the correlation coefficient between the frequency characteristic component of the intrinsic mode function and the original voltage signal.

7. The method according to claim 6, characterized in that The performing wavelet transform on the filtering result to obtain a reconstructed voltage signal includes: The filtering result is subjected to wavelet transformation to obtain wavelet coefficients at different frequency scales, where: W(a, b) represents the wavelet coefficients, x * (t) represents the filtering result, Ψ a,b (t) represents the wavelet function, a and b represent the scale parameter and translation parameter respectively; The wavelet coefficients are subjected to threshold processing and a threshold processing result is obtained, and the calculation formula is: W thresh (a, b) represents the threshold processing result, T represents the target threshold used in the threshold processing, the target threshold is determined based on the normal distribution, and σ represents the standard deviation of the normal distribution; Based on the threshold processing result, an inverse wavelet transform is performed to obtain the reconstructed voltage signal, and the calculation formula is: x recon (t)=∑ a ∑ b W thresh (a,b)Ψ a,b (t); x recon (t) represents the reconstructed voltage signal.

8. A battery pack fault diagnosis device, characterized in that: The device comprises: A data acquisition module, used to acquire the original voltage signal of the battery pack; A data preprocessing module, used for performing data preprocessing on the original voltage signal of the battery pack and obtaining a preprocessed signal; A feature extraction module, used for taking the preprocessed signal as an input of an attention convolutional neural network, so that the attention convolutional neural network outputs a feature extraction result based on the preprocessed signal; A fault diagnosis module is used to input the feature extraction result into a fault classification detection model to identify the fault type of the battery pack based on the feature extraction result based on the fault classification detection model.

9. An electronic device, characterized in that: include: processor; as well as A memory configured to store machine-readable instructions, which, when executed by the processor, execute the battery pack fault diagnosis method as described in any one of claims 1-7.

10. A storage medium, characterized in that: The storage medium stores a computer program, and the computer program is executed by a processor to implement the battery pack fault diagnosis method as described in any one of claims 1 to 7.

Citation Information

Cited By

  • Battery pack fault diagnosis method and system based on multi-domain characteristics

    CN119355537A

  • Fault identification method and device for battery cluster and nonvolatile storage medium

    CN121208691A