Lithium battery cell self-discharge anomaly detection method, device, equipment and medium

Abstract

The invention discloses a lithium battery cell anomaly detection method, device and equipment and a medium, and the method comprises the steps: obtaining the charging and discharging data of each to-be-detected cell in a capacity grading stage, extracting the corresponding dynamic characteristics according to the charging and discharging data, so as to generate the characteristic data of each to-be-detected cell in the capacity grading stage, and obtaining the characteristic data of each to-be-detected cell in the capacity grading stage; according to the feature data, self-discharge anomaly detection is conducted on the to-be-detected battery cells through an anomaly detection model, and the anomaly detection model is obtained through training of a self-attention-enhanced generative adversarial auto-encoder model. According to the method, the self-discharge abnormal condition of the to-be-tested battery cell can be predicted only according to the charge and discharge data of the to-be-tested battery cell in the capacity grading stage, the time for testing the self-discharge rate of the lithium battery cell in a subsequent standing manner is greatly shortened, meanwhile, the production cycle of the lithium battery is shortened, and energy waste is avoided.

Description

Technical Field

[0001] The present invention belongs to the technical field of lithium battery cell self-discharge anomaly detection, and in particular relates to a lithium battery cell self-discharge anomaly detection method, device, equipment and medium. Background Art

[0002] Lithium-ion batteries have the advantages of high energy density, long cycle life, low cost, and environmental protection and pollution-free. They are widely used in civilian fields such as electric vehicles and energy storage.

[0003] In the manufacturing process of lithium batteries, the self-discharge rate is an important quality evaluation index, which is usually obtained by static experiment on lithium-ion batteries after the capacity division process. The static experiment usually takes 15-30 days to measure the change rate of the battery open circuit voltage during this period and screen out the cells with abnormal self-discharge rate. In this process, a lot of time and energy are consumed, so it is of great significance to quickly and accurately screen out the lithium battery cells with abnormal self-discharge rate. Summary of the invention

[0004] The purpose of the present invention is to provide a method for detecting abnormal self-discharge of lithium battery cells. By utilizing an abnormal detection model to perform abnormal detection on the charge and discharge data of the cell to be tested in the capacity division stage, the technical problems of long time and low efficiency in the prior art for detecting abnormal self-discharge of lithium battery cells are solved.

[0005] In order to solve the above technical problems, the present invention is achieved through the following technical solutions:

[0006] The present invention provides a method for detecting abnormal self-discharge of a lithium battery cell, which comprises:

[0007] Obtain the charge and discharge data of each battery cell to be tested during the capacity division stage;

[0008] Extracting corresponding dynamic features according to each of the charge and discharge data to generate feature data of each of the cells to be tested in the capacity division stage;

[0009] According to each of the feature data, an anomaly detection model is used to perform self-discharge anomaly detection on each of the battery cells to be tested, wherein the anomaly detection model is obtained by training a self-attention enhanced generative adversarial autoencoder model.

[0010] In one embodiment of the present invention, the anomaly detection model is obtained in the following manner:

[0011] A preset data set of charging and discharging of lithium battery cells in the capacity division stage is obtained, and the anomaly detection model is obtained in the following way:

[0012] Establishing the self-attention enhanced generative adversarial autoencoder model;

[0013] Acquire a standard data set of charging and discharging of lithium battery cells in a capacity division stage, wherein the standard data set includes data of lithium battery cells with normal self-discharge and data of lithium battery cells with abnormal self-discharge;

[0014] Dividing the standard data set to generate a training set, a validation set, and a test set respectively;

[0015] According to the training set, the validation set and the test set, the self-attention enhanced generative adversarial autoencoder model is trained, validated and tested to obtain the anomaly detection model.

[0016] In one embodiment of the present invention, dividing the standard data set to generate a training set, a validation set, and a test set respectively includes:

[0017] Randomly divide the lithium battery cell data with normal self-discharge in proportion to generate the training set, positive validation set and positive test set;

[0018] The lithium battery cell data with abnormal self-discharge is randomly divided in proportion to generate an abnormal verification set and an abnormal test set;

[0019] Combining the positive validation set and the abnormal validation set to generate the validation set;

[0020] The positive test set and the abnormal test set are combined to generate the test set.

[0021] In one embodiment of the present invention, extracting corresponding dynamic features according to each of the charge and discharge data to generate feature data of each of the cells to be tested in the capacity division stage includes:

[0022] According to the process of the lithium battery cell in the capacity division stage, each of the charging and discharging data is divided into constant current and constant voltage charging stage data, static stage data and constant current discharging stage data;

[0023] The dynamic characteristics of the constant-current and constant-voltage charging stage data, the static stage data, and the constant-current discharging stage data are extracted respectively to generate characteristic data of each of the cells to be tested in the capacity division stage.

[0024] In one embodiment of the present invention, the charge and discharge data includes voltage data, current data, capacity data and temperature data.

[0025] In one embodiment of the present invention, performing self-discharge anomaly detection on each of the tested cells using an anomaly detection model according to each of the characteristic data includes:

[0026] Inputting each of the characteristic data into the abnormality detection model to obtain the abnormality index of each of the battery cells to be tested;

[0027] According to the abnormal index and the preset threshold, the battery cell to be tested with abnormal self-discharge is detected.

[0028] In one embodiment of the present invention, inputting each of the characteristic data into the abnormality detection model to obtain the abnormality index of each of the battery cells to be tested includes:

[0029] Using the anomaly detection model to process each of the feature data to obtain reconstructed data corresponding to each of the feature data;

[0030] A reconstruction difference is calculated according to each of the characteristic data and the corresponding reconstruction data, and an abnormality indicator of each of the lithium battery cells is obtained according to the reconstruction difference.

[0031] Based on the same inventive concept, another embodiment of the present invention further provides a lithium battery cell self-discharge abnormality detection device, the device comprising:

[0032] A data acquisition module is used to obtain the charge and discharge data of each battery cell to be tested in the capacity division stage;

[0033] A feature extraction module, used to extract corresponding dynamic features according to each of the charge and discharge data, so as to generate feature data of each of the cells to be tested in the capacity division stage;

[0034] An anomaly detection module is used to perform self-discharge anomaly detection on each of the battery cells to be tested according to each of the feature data using an anomaly detection model, wherein the anomaly detection model is obtained by training a self-attention enhanced generative adversarial autoencoder model.

[0035] Based on the same inventive concept, another embodiment of the present invention further provides an electronic device, the electronic device comprising:

[0036] one or more processors;

[0037] A storage device is used to store one or more programs. When the one or more programs are executed by the one or more processors, the electronic device implements any of the above-mentioned lithium battery cell self-discharge abnormality detection methods.

[0038] Based on the same inventive concept, another embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor of a computer, the computer executes any of the above-mentioned methods for detecting abnormal self-discharge of lithium battery cells.

[0039] As described above, the present invention provides a method for detecting abnormal self-discharge of lithium battery cells, which has the following beneficial effects: by acquiring the charge and discharge data of each cell to be tested in the capacity division stage, respectively extracting corresponding dynamic features according to each of the charge and discharge data to generate feature data of each cell to be tested in the capacity division stage, and using an anomaly detection model to perform self-discharge anomaly detection on each cell to be tested according to each of the feature data, wherein the anomaly detection model is obtained by training a self-attention enhanced generative adversarial autoencoder model. The method for detecting abnormal self-discharge of lithium battery cells can predict the abnormal self-discharge of the cell to be tested based on the charge and discharge data of the cell to be tested in the capacity division stage, which greatly reduces the time for subsequent static self-discharge rate testing of the lithium battery cells, while shortening the production cycle of lithium batteries and avoiding energy waste. Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0041] Figure 1 The present invention is a flowchart of a method for detecting abnormal self-discharge of a lithium battery cell provided by an exemplary embodiment of the present application.

[0042] Figure 2 A schematic diagram of a system flow of a method for detecting abnormal self-discharge of a lithium battery cell provided by an exemplary embodiment of the present application.

[0043] Figure 3 A flowchart of an anomaly detection model provided for an exemplary embodiment of the present application.

[0044] Figure 4 A schematic structural diagram of a lithium battery cell self-discharge abnormality detection device provided by another exemplary embodiment of the present application.

[0045] Figure 5 A schematic structural diagram of an electronic device provided for another exemplary embodiment of the present application. DETAILED DESCRIPTION

[0046] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0047] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and thus the drawings only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0048] In the following description, numerous details are discussed to provide a more thorough explanation of the embodiments of the present invention. However, it is obvious to those skilled in the art that the embodiments of the present invention can be implemented without these specific details. In other embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present invention difficult to understand.

[0049] In order to solve the technical problem of long time and low efficiency in the prior art of lithium battery cell self-discharge abnormality detection process, the present invention provides a lithium battery cell self-discharge abnormality detection method, which uses an abnormality detection model to perform abnormality detection on the charge and discharge data of the tested cell in the capacity division stage, and then predicts the self-discharge abnormality of the tested cell, thereby reducing the time for subsequent static self-discharge rate testing of the lithium battery cell, shortening the production cycle of lithium batteries and avoiding energy waste. Figure 1 As shown, the lithium battery cell self-discharge abnormality detection method comprises the following steps:

[0050] S100: Acquire charging and discharging data of each battery cell to be tested in the capacity division stage;

[0051] S200: extracting corresponding dynamic features according to each of the charge and discharge data to generate feature data of each of the cells to be tested in the capacity division stage;

[0052] S300: Performing self-discharge anomaly detection on each of the battery cells to be tested according to each of the feature data using an anomaly detection model, wherein the anomaly detection model is obtained by training a self-attention enhanced generative adversarial autoencoder model.

[0053] The steps of the above-mentioned lithium battery cell self-discharge abnormality detection method will be discussed in detail below.

[0054] First, before executing step S100, it is necessary to establish an abnormality detection model, which is used for subsequent detection of abnormal self-discharge conditions of the battery cell to be tested.

[0055] In an exemplary embodiment of the present application, the establishment of the anomaly detection model includes the following steps:

[0056] S010: Establish the self-attention enhanced generative adversarial autoencoder model;

[0057] S020: Acquire a standard data set of charging and discharging of lithium battery cells in the capacity division stage, wherein the standard data set includes data of lithium battery cells with normal self-discharge and data of lithium battery cells with abnormal self-discharge;

[0058] S030: Divide the standard data set to generate a training set, a validation set, and a test set respectively;

[0059] S040: According to the training set, the verification set and the test set, the self-attention enhanced generative adversarial autoencoder model is trained, verified and tested to obtain the anomaly detection model.

[0060] In an exemplary embodiment of the present application, in step S030, dividing the standard data set to generate a training set, a validation set, and a test set further includes the following steps:

[0061] S031: randomly dividing the lithium battery cell data with normal self-discharge according to proportion to generate the training set, the positive validation set and the positive test set;

[0062] S032: randomly divide the lithium battery cell data with abnormal self-discharge according to proportion to generate an abnormal verification set and an abnormal test set;

[0063] S033: combining the positive validation set and the abnormal validation set to generate the validation set;

[0064] S034: Combine the positive test set and the abnormal test set to generate the test set.

[0065] It should be noted that, in this embodiment, the anomaly detection model is obtained by training, verifying and testing the self-attention enhanced generative adversarial autoencoder (SAE-GAAE) model. The training set is used to train the self-attention enhanced generative adversarial autoencoder model. The verification set is used to optimize the self-attention enhanced generative adversarial autoencoder model, and the hyperparameters in the self-attention enhanced generative adversarial autoencoder model are optimized by using the Bayesian optimization algorithm based on the TPE (Tree-structured Parzen Estimator, TPE) modeling strategy to improve the anomaly detection accuracy of the model. The test set is used to test the performance of the optimized model to obtain the final anomaly detection model. The anomaly detection model aims to improve the accuracy of self-discharge anomaly detection of lithium battery cells by utilizing the advantages of the self-attention mechanism for capturing feature relationships and key features, the reconstruction ability of the autoencoder for data, and the generation of the adversarial training mechanism of the generative adversarial network for the encoder encoding effect. The advantage of the anomaly detection model is that it is highly sensitive to features, can accurately capture the feature correlation and key features in the lithium battery cell data, and the model becomes more robust due to the constraints of the generative adversarial mechanism. In addition, given the significant impact of temperature changes on the self-discharge effect of lithium-ion batteries, temperature indicators are also extracted during the establishment of the anomaly detection model. By combining the extracted temperature indicators with the self-attention mechanism, the correlation between features can be established more accurately, thereby improving the anomaly detection performance of the model.

[0066] After the abnormality detection model is established, step S100 is executed to obtain the charge and discharge data of each battery cell to be tested in the capacity division stage.

[0067] It should be noted that the capacity sorting stage, also known as the capacity sorting or capacity testing stage, is an important link in the lithium battery production process, which is mainly used to evaluate the battery's performance parameters such as capacity and internal resistance, and to grade the battery according to performance. The process of the capacity sorting stage includes constant current discharge, standing, constant current charging, standing again and constant voltage charging. It is worth noting that according to actual needs, the above charging and discharging process can be repeated multiple times to further evaluate the cycle performance and stability of the lithium battery cell. It is worth noting that different manufacturers have certain differences in the process steps and process combinations of the capacity sorting stage.

[0068] In an exemplary embodiment of the present application, the charge and discharge data includes voltage data, current data, capacity data and temperature data. It should be noted that, in this embodiment, the charge and discharge data can be measured by a high-precision battery charge and discharge tester.

[0069] Next, step S200 is executed to extract corresponding dynamic features according to each of the charge and discharge data, so as to generate feature data of each of the battery cells to be tested in the capacity division stage.

[0070] In an exemplary embodiment of the present application, in step S200, extracting corresponding dynamic features according to each of the charge and discharge data to generate feature data of each of the cells to be tested in the capacity division stage further includes the following steps:

[0071] S201: dividing the charging and discharging data into constant current and constant voltage charging stage data, static stage data and constant current discharging stage data according to the process of the lithium battery cell in the capacity division stage;

[0072] S202: extracting dynamic features of the constant-current and constant-voltage charging stage data, the static stage data, and the constant-current discharging stage data respectively, to generate feature data of each of the cells to be tested in the capacity division stage.

[0073] Specifically, the processes of lithium battery cells in the capacity division stage include constant current discharge, static, constant current charging, static again and constant voltage charging, which can be divided into constant current and constant voltage charging stage, static stage and constant current discharge stage. The complete charging and discharging data of each cell to be tested are divided into constant current and constant voltage charging stage data, static stage data and constant current discharge stage data according to the processes of lithium battery cells in the capacity division stage. Then, the dynamic characteristics of the constant current and constant voltage charging stage data, the static stage data and the constant current discharge stage data of each cell to be tested are extracted respectively, so as to generate the characteristic data of each cell to be tested in the capacity division stage. Among them, the core characteristics of the constant current and constant voltage charging stage include but are not limited to isobaric rise time, polarization voltage, isovolumetric voltage and charging voltage difference, the core characteristics of the static stage include but are not limited to static voltage difference, and the core characteristics of the constant current discharge stage include but are not limited to isobaric drop time, polarization voltage, discharge voltage difference and capacity voltage difference. It should be noted that, in view of the impact of temperature changes on the self-discharge effect of lithium-ion batteries, temperature indicators are extracted for each battery cell to be tested during the constant current and constant voltage charging stage, the static stage, and the constant current discharge stage.

[0074] Finally, step S300 is executed, that is, according to each of the feature data, a self-discharge anomaly detection is performed on each of the battery cells to be tested using an anomaly detection model, wherein the anomaly detection model is obtained by training a self-attention enhanced generative adversarial autoencoder model.

[0075] In an exemplary embodiment of the present application, in step S300, performing self-discharge abnormality detection on each of the tested cells using an abnormality detection model according to each of the characteristic data further includes the following steps:

[0076] S301: performing data processing on each of the feature data using the anomaly detection model to obtain reconstructed data corresponding to each of the feature data;

[0077] S302: Calculating a reconstruction difference according to each of the characteristic data and the corresponding reconstruction data, and obtaining an abnormality indicator of each of the lithium battery cells according to the reconstruction difference.

[0078] It should be noted that the output result of the abnormality detection model includes the abnormality index of each battery cell to be tested. In this embodiment, the mean square error (MSE) and the mean absolute error (MAE) are used as the abnormal screening indexes of the battery cells to be tested, and the mean square error and the mean absolute error of each battery cell to be tested are calculated. All the battery cells to be tested are arranged in descending order according to the size of the mean square error value and the mean absolute error value, and the lithium battery cells that are most likely to have self-discharge abnormalities are screened out according to the preset abnormal percentage threshold. In this embodiment, the abnormal percentage threshold is set to 30%, that is, the battery cells to be tested in the top 30% of the mean square error value and the top 30% of the mean absolute error value are screened out as lithium battery cells that may have self-discharge abnormalities. It is worth noting that in other embodiments, the mean square error value and the mean absolute error value of the battery cell to be tested can be normalized and then weighted averaged to obtain the comprehensive abnormality index of the battery cell to be tested, and then the lithium battery cells that may have self-discharge abnormalities can be screened out according to the comprehensive abnormality index of the battery cell to be tested. In addition, if the method of weighted normalization is adopted, the weight ratio of the mean square error and the mean absolute error can be adaptively set according to the actual application requirements.

[0079] See also Figure 2 and Figure 3 As shown, the specific establishment process of the anomaly detection model will be described in detail below:

[0080] The first layer of the encoder encodes the input raw data x for the first time through the activation function to obtain the first layer of encoding with different dimensions. The encoding process of the encoder can be described as:

[0081] h e =σ e (W1x+b1) (1)

[0082] Among them, W1, b1 are encoding weight and bias respectively, σ e is the activation function of the encoder.

[0083] The first layer encoding result h e Input self-attention mechanism for linear transformation and activation, h eThe data shape is usually [batch_size, sequence_length, features], where batch_size represents the traversal batch, sequence_length represents the traversal sequence length, and features represents the number of traversal features. An intermediate representation u is obtained through a linear transformation and tanh activation function:

[0084] u = tanh(h e W SA +b SA )

[0085] Among them, W SA is the self-attention mechanism weight matrix, whose shape is (input_shape[-1],input_shape[-1]), b SA is the bias term, whose shape is (input_shape[-1]), and their shapes are determined by the input h e The last dimension [features] determines

[0086] Then use a weight vector u w Perform dot product operation with u, u w Updated and optimized by back propagation and gradient descent during training. Apply the softmax activation function to get the attention score α, which is a two-dimensional tensor of shape [batch_size, sequence_length]:

[0087] α=softmax(u·u w )

[0088] In particular, the softmax activation function ensures that the sum of the attention scores at all positions is 1 and the score at each position is in the range [0, 1].

[0089] Finally, the attention score α is combined with the original input h e Perform element-by-element multiplication to obtain the attention-weighted feature representation, which is a three-dimensional tensor of shape [batch_size, sequence_length, features]:

[0090] h' e =h ijk α ij

[0091] Among them, i is the traversal batch, j is the traversal sequence length, and k is the number of traversal features.

[0092] Will h' eContinue to input the subsequent layers of the encoder. The encoding process of each layer is shown in formula (1). After the encoding is completed, the output is h′.

[0093] Use a generative adversarial network for adversarial training and define a discriminator to distinguish between real samples and fake samples. The discriminator accepts the potential representation h′ as input and outputs a probability value between 0 and 1, indicating the probability that the input is a real sample.

[0094] For the real sample h real , the probability that the discriminator outputs is close to 1:

[0095] D(h real )≈1

[0096] For the generated fake sample h fake , the probability that the discriminator outputs is close to 0:

[0097] D(h fake )≈0(6)

[0098] Among them, h real The random sample conforms to the Gaussian distribution. The Gaussian distribution is selected as the prior distribution because the charging and discharging characteristic index of the lithium battery cell is close to the Gaussian distribution. fake It is obtained by encoding the training data through the encoder.

[0099] Real sample loss function definition:

[0100]

[0101] Fake sample loss function definition:

[0102]

[0103] The total loss of the discriminator is the combination of the real and fake encoding losses:

[0104] L D =L real +L fake

[0105] By training the encoder (as a generator) to minimize the loss between the discriminator's output of the fake encoding and the true label:

[0106]

[0107] In the generative adversarial network, the discriminator and encoder are updated alternately, that is, L is minimized. D and L E, so that the encoding layer of the autoencoder and the generative adversarial network are jointly trained, ensuring that while each round of training the autoencoder, the encoder output is continuously adjusted through the feedback of the adversarial training. After the generative adversarial training is completed, the encoder outputs the new code as the decoder input for reconstruction.

[0108] The decoding process can be described as:

[0109] x R =σ d (W2h new +b2)

[0110] Among them, w2, b2 are encoding weights and biases, σ d is the activation function of the decoder.

[0111] After decoding, the decoder obtains reconstructed data with the same shape as the original lithium battery cell characteristic data.

[0112] The reconstruction difference between the reconstructed sample and the original sample is calculated. The larger the reconstruction difference, the more likely it is that there is a self-discharge anomaly. The mean square error and mean absolute error are used as indicators to measure the reconstruction difference.

[0113]

[0114] in, represents the reconstructed value of the jth feature of the i-th lithium battery cell in the test set, x ij It represents the true value of the jth feature of the ith lithium battery cell in the test set. When the reconstructed value is completely consistent with the true value, it is equal to 0, that is, the possibility of self-discharge anomaly in the lithium battery cell is smaller. It is worth noting that the larger the error value, the larger the reconstruction difference.

[0115] In addition, in this embodiment, the self-attention enhanced generative adversarial autoencoder model is optimized for hyperparameters by using a Bayesian optimization algorithm. The Bayesian optimization is a method for searching for global extreme values ​​of a function based on the Bayesian method, which has significant advantages in dealing with black box problems and non-convex problems, and can globally search for the optimal hyperparameter combination of the anomaly detection model without being limited to local optimal solutions. In this embodiment, Bayesian optimization is used to search for the maximum value of the objective function.

[0116] The traditional Bayesian optimization based on Gaussian process regression (GPR) probability model needs to establish a Gaussian process model in the entire parameter space, which faces the problem of high computational complexity when processing large-scale data sets. In this embodiment, Bayesian optimization based on TPE modeling strategy is used, and EI (Expected Improvement) is used as the acquisition function. EI takes into account the current value y that is slightly lower than the optimal value. *and the improvement that might be obtained at some point x.

[0117]

[0118] Where y is the value of the objective function at point x, y * is a threshold (slightly worse than the current best result). When x is given, it is y relative to the threshold y * The average improvement of threshold y * The selection conforms to the following formula:

[0119] p(y>y * )=γ

[0120] Hyperparameter set x = [σ ed ,σ ac1 ,σ ac2 ,σ bs ,σ ep ], the objective function is the AUC (Area Under the ROC Curve) value of the prediction result.

[0121] It should be noted that the AUC value is obtained by calculating the area under the ROC curve. The ROC curve uses the true positive rate (TPR) as the vertical axis and the false positive rate (FPR) as the horizontal axis, showing the performance of the model at different thresholds.

[0122] First, a set of hyperparameter configurations are randomly selected for evaluation in the initialization phase, and the objective function value is calculated. The TPE proxy model is constructed using the evaluated hyperparameter configurations and the corresponding objective function values. Unlike the probability model based on Gaussian process regression, which directly models p(y|x), the TPE strategy models p(x|y) and p(y). TPE uses the following density definition p(x|y):

[0123]

[0124] TPE constructs different distributions for observation points on both sides of the threshold y*, and then we can get:

[0125]

[0126] From this we can deduce:

[0127]

[0128] Thus we get:

[0129]

[0130] The next hyperparameter configuration to be evaluated is selected by maximizing the indicator of the TPE model, the AUC is evaluated using the selected hyperparameter configuration, the results are recorded, the proxy model is updated using the new evaluation results, and the above steps are repeated until the maximum number of evaluations is reached.

[0131] This paper focuses on the electrochemical mechanism analysis of abnormal self-discharge of lithium-ion batteries. By deeply mining the data of lithium-ion capacity distribution stage, it successfully extracts a series of electrochemical features closely related to battery self-discharge characteristics, such as voltage, current, temperature and capacity. Compared with traditional mechanical features such as diaphragm thickness, the extraction process of these electrochemical features is simpler, more direct and effective.

[0132] In view of the complex interactions among various features in the battery capacity classification process and the differences in their effects on self-discharge anomalies, the present invention innovatively introduces a self-attention mechanism in the model encoding layer, which significantly enhances the encoder's ability to handle complex relationships between features, achieves fast and accurate association of features and effective capture of key features, thereby getting rid of excessive reliance on prior knowledge, greatly optimizing the dimensionality reduction effect of data, and significantly improving the detection accuracy of battery self-discharge anomalies.

[0133] At the same time, in order to further improve the generalization ability and robustness of the model, the present invention cleverly integrates the generative adversarial mechanism into the autoencoder framework. This mechanism ensures that the encoder can effectively retain key features and maintain the diversity and authenticity of data during the low-dimensional data generation process, effectively avoiding the key feature overfitting problem that may be caused by the self-attention mechanism, and making the model show higher flexibility and adaptability when processing complex data.

[0134] In response to the non-convex problems faced in the model optimization process, the present invention adopts an advanced Bayesian optimization algorithm to search for the global optimal solution, and combines it with a highly adaptive and flexible modeling strategy based on TPE, effectively avoiding the trap of local optimal solutions and further improving the accuracy and reliability of the prediction method.

[0135] In summary, the present invention provides a method for detecting abnormalities in lithium battery cells, which records the charging and discharging data of each lithium battery cell in the capacity division stage, including battery voltage data, current data, capacity data and temperature data, extracts a variety of dynamic features from the charging and discharging data, and divides the features into a training set, a validation set and a test set. The training set data is used to train a self-attention enhanced generative adversarial autoencoder model, and the reconstructed mean square error and mean absolute error are used as abnormality screening indicators. The hyperparameters in the self-attention enhanced generative adversarial autoencoder model are optimized using a Bayesian optimization algorithm based on a TPE modeling strategy to improve the accuracy of screening lithium battery cells with self-discharge abnormalities. Subsequently, the new feature data only needs to be input into the model to obtain the abnormal indicators of each lithium battery cell. Finally, the lithium battery cells with self-discharge abnormalities are screened out according to the abnormal percentage threshold. The lithium battery cell self-discharge anomaly detection method can predict the self-discharge anomaly of the cell to be tested based on the charge and discharge data of the cell to be tested in the capacity division stage, thereby greatly reducing the time for subsequent static self-discharge rate testing of the lithium battery cell, shortening the production cycle of the lithium battery and avoiding energy waste.

[0136] Based on the same inventive concept, please refer to Figure 4 As shown, another embodiment of the present invention further provides a lithium battery cell self-discharge abnormality detection device 11, the device comprising:

[0137] The data acquisition module 111 is used to acquire the charge and discharge data of each battery cell to be tested in the capacity division stage;

[0138] A feature extraction module 112 is used to extract corresponding dynamic features according to each of the charging and discharging data, so as to generate feature data of each of the cells to be tested in the capacity division stage;

[0139] The anomaly detection module 113 is used to perform self-discharge anomaly detection on each of the battery cells to be tested according to each of the feature data using an anomaly detection model, wherein the anomaly detection model is obtained by training a self-attention enhanced generative adversarial autoencoder model.

[0140] Based on the same inventive concept, please refer to Figure 5 As shown, another embodiment of the present invention further provides an electronic device 1, which may include a memory 12, a processor 13 and a bus, and may also include a computer program stored in the memory 12 and executable on the processor 13, such as a lithium battery cell self-discharge abnormality detection program.

[0141] Among them, the memory 12 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (for example: SD or DX memory, etc.), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 12 can be an internal storage unit of the electronic device 1, such as a mobile hard disk of the electronic device 1. In other embodiments, the memory 12 can also be an external storage device of the electronic device 1, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device 1. Further, the memory 12 can also include both an internal storage unit of the electronic device 1 and an external storage device. The memory 12 can not only be used to store application software and various types of data installed in the electronic device 1, such as the code for detecting abnormal self-discharge of lithium battery cells, etc., but can also be used to temporarily store data that has been output or is to be output.

[0142] In some embodiments, the processor 13 may be composed of an integrated circuit, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and a combination of various control chips. The processor 13 is the control core (Control Unit) of the electronic device 1, and uses various interfaces and lines to connect the various components of the entire electronic device 1, and executes or executes the programs or modules stored in the memory 12 (such as the lithium battery cell self-discharge abnormality detection program, etc.), and calls the data stored in the memory 12 to execute various functions of the electronic device 1 and process data.

[0143] The processor 13 executes the operating system and various installed application programs of the electronic device 1. The processor 13 executes the application programs to implement the steps in the above-mentioned lithium battery cell self-discharge abnormality detection method.

[0144] Exemplarily, the computer program may be divided into one or more modules, which are stored in the memory 12 and executed by the processor 13 to complete the present application. The one or more modules may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program in the electronic device 1. For example, the computer program may be divided into a data acquisition module 111, a feature extraction module 112, and an anomaly detection module 113.

[0145] The above-mentioned integrated unit implemented in the form of a software function module can be stored in a computer-readable storage medium, and the computer-readable storage medium can be non-volatile or volatile. The above-mentioned software function module is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a computer device, or a network device, etc.) or a processor to perform part of the functions of the lithium battery cell self-discharge abnormality detection method described in each embodiment of the present application.

[0146] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A method for detecting abnormal self-discharge of a lithium battery cell, characterized in that: include: Obtain the charge and discharge data of each battery cell to be tested during the capacity division stage; Extracting corresponding dynamic features according to each of the charge and discharge data to generate feature data of each of the cells to be tested in the capacity division stage; According to each of the feature data, an anomaly detection model is used to perform self-discharge anomaly detection on each of the battery cells to be tested, wherein the anomaly detection model is obtained by training a self-attention enhanced generative adversarial autoencoder model.

2. The method for detecting abnormal self-discharge of a lithium battery cell according to claim 1, characterized in that: The anomaly detection model is obtained in the following way: Establishing the self-attention enhanced generative adversarial autoencoder model; Acquire a standard data set of charging and discharging of a lithium battery in a capacity division stage, wherein the standard data set includes lithium battery cell data with normal self-discharge and lithium battery cell data with abnormal self-discharge; Dividing the standard data set to generate a training set, a validation set, and a test set respectively; According to the training set, the validation set and the test set, the self-attention enhanced generative adversarial autoencoder model is trained, validated and tested to obtain the anomaly detection model.

3. The method for detecting abnormal self-discharge of a lithium battery cell according to claim 2, characterized in that: The dividing the standard data set to generate a training set, a validation set and a test set respectively includes: Randomly divide the lithium battery cell data with normal self-discharge in proportion to generate the training set, positive validation set and positive test set; The lithium battery cell data with abnormal self-discharge is randomly divided in proportion to generate an abnormal verification set and an abnormal test set; Combining the positive validation set and the abnormal validation set to generate the validation set; The positive test set and the abnormal test set are combined to generate the test set.

4. The method for detecting abnormal self-discharge of a lithium battery cell according to claim 1, characterized in that: The extracting corresponding dynamic features according to each of the charge and discharge data to generate feature data of each of the cells to be tested in the capacity division stage includes: According to the process of the lithium battery cell in the capacity division stage, each of the charging and discharging data is divided into constant current and constant voltage charging stage data, static stage data and constant current discharging stage data; The dynamic characteristics of the constant-current and constant-voltage charging stage data, the static stage data, and the constant-current discharging stage data are extracted respectively to generate characteristic data of each of the cells to be tested in the capacity division stage.

5. The method for detecting abnormal self-discharge of a lithium battery cell according to claim 1, characterized in that: The charge and discharge data includes voltage data, current data, capacity data and temperature data.

6. The method for detecting abnormal self-discharge of a lithium battery cell according to claim 1, characterized in that: The performing self-discharge anomaly detection on each of the battery cells to be tested by using an anomaly detection model according to each of the characteristic data comprises: Inputting each of the characteristic data into the abnormality detection model to obtain the abnormality index of each of the battery cells to be tested; According to the abnormal index and the preset threshold, the battery cell to be tested with abnormal self-discharge is detected.

7. The method for detecting abnormal self-discharge of a lithium battery cell according to claim 6, characterized in that: The step of inputting each of the characteristic data into the abnormality detection model to obtain the abnormality index of each of the battery cells to be tested includes: Using the anomaly detection model to process each of the feature data to obtain reconstructed data corresponding to each of the feature data; A reconstruction difference is calculated according to each of the characteristic data and the corresponding reconstruction data, and an abnormality indicator of each of the lithium battery cells is obtained according to the reconstruction difference.

8. A lithium battery cell self-discharge abnormality detection device, characterized in that: The device comprises: A data acquisition module is used to obtain the charge and discharge data of each battery cell to be tested in the capacity division stage; A feature extraction module, used to extract corresponding dynamic features according to each of the charge and discharge data, so as to generate feature data of each of the cells to be tested in the capacity division stage; An anomaly detection module is used to perform self-discharge anomaly detection on each of the battery cells to be tested according to each of the feature data using an anomaly detection model, wherein the anomaly detection model is obtained by training a self-attention enhanced generative adversarial autoencoder model.

9. An electronic device, characterized in that: The electronic device comprises: one or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, enables the electronic device to implement the lithium battery cell self-discharge abnormality detection method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor of a computer, the computer is caused to execute the method for detecting abnormal self-discharge of a lithium battery cell as claimed in any one of claims 1 to 7.

Images