Battery capacity degradation knee point detection method and device, equipment and storage medium

Abstract

The invention provides a battery capacity degradation knee point detection method and device, equipment and a storage medium. The method comprises the steps that key features in current and capacity data of a battery in a constant-voltage charging stage are extracted; wherein the key features comprise a current attenuation coefficient, a current drop average rate, a target time point current ratio, a current change curvature and a charging electric quantity in a constant-voltage charging stage; performing data preprocessing on the key features to determine a plurality of key feature sequences; and inputting the plurality of key feature sequences into a pre-trained degraded knee point prediction model, carrying out self-attention processing and degraded knee point detection processing on the plurality of key feature sequences, and predicting a capacity degraded knee point of the battery. Key feature vectors are mined from current and capacity data in a constant-voltage charging stage, and high-quality data support is provided for model analysis. The knee points can be preliminarily identified by using the anomaly detection capability, and reconstruction errors are processed by using smooth filtering and a dynamic threshold method, so that the knee points can be detected more accurately.

Description

Technical Field

[0001] The present application relates to the technical field of battery capacity detection, and in particular to a method, device, equipment and storage medium for detecting a battery capacity degradation knee point. Background Art

[0002] With age and the number of charge and discharge cycles, batteries age, leading to a decline in capacity and performance. This capacity degradation exhibits a nonlinear trend. This is the result of the combined effects of external cycling stresses (such as charge and discharge rate and temperature) and complex internal degradation mechanisms (such as SEI film growth and rupture, and electrode material phase transitions). The nonlinear degradation trajectory of lithium-ion batteries occurs in two stages: initially, the capacity declines slowly and approximately linearly, and then the degradation rate accelerates after passing the knee point. Accurately detecting the knee point is crucial for battery cell development and battery management systems (BMS), and can be used for remaining service life prediction and fault diagnosis. For example, in electric vehicles, this can optimize battery management strategies and avoid catastrophic failures. Currently, physical and chemical modeling methods are the primary approach to detecting the knee point, but these methods suffer from numerous challenges: first, measuring the complex physical and chemical parameters within the battery is extremely difficult, as these parameters vary with usage and the external environment; second, they require significant computational resources and suffer from poor real-time performance. Therefore, accurately determining the knee point of battery capacity degradation has become a significant technical challenge. Summary of the Invention

[0003] In light of this, the purpose of this application is to provide a method, apparatus, device, and storage medium for detecting the knee point of battery capacity degradation. By mining key feature vectors from current and capacity data during the constant-voltage charging phase, this method provides high-quality data support for model analysis. Anomaly detection capabilities can be used to initially identify the knee point, while smoothing filtering and dynamic thresholding methods are used to address reconstruction errors, enabling more accurate knee point detection.

[0004] The present application provides a method for detecting a battery capacity degradation knee point, the method comprising: Extract key features from the battery's current and capacity data during the constant-voltage charging phase; the key features include current attenuation coefficient, average current drop rate, current ratio at the target time point, current change curvature, and charge capacity during the constant-voltage charging phase; Performing data preprocessing on the key features to determine multiple key feature sequences; The multiple key feature sequences are input into a pre-trained degradation knee point prediction model, and self-attention processing and degradation knee point detection processing are performed on the multiple key feature sequences to predict the capacity degradation knee point of the battery; wherein the degradation knee point prediction model is obtained by iteratively training a deep learning model of the self-attention mechanism.

[0005] In a possible implementation manner, performing data preprocessing on the key features to determine multiple key feature sequences includes: Performing high-frequency noise removal processing on each of the key features based on Gaussian filtering to determine a plurality of key features after denoising; performing normalization processing on the plurality of key features after denoising to determine the plurality of normalized key features; The multiple normalized key features are split into fixed-length sequences to determine multiple key feature sequences.

[0006] In a possible implementation manner, the data preprocessing of the key features to determine multiple key feature sequences includes: Performing high-frequency noise removal processing on each of the key features based on Gaussian filtering to determine a plurality of key features after denoising; performing normalization processing on the plurality of key features after denoising to determine the plurality of normalized key features; The multiple normalized key features are split into fixed-length sequences to determine multiple key feature sequences.

[0007] In one possible implementation, the inputting of the multiple key feature sequences into a pre-trained degradation knee point prediction model, performing self-attention processing and outlier detection processing on the multiple key feature sequences, and predicting the capacity degradation knee point of the battery includes: Performing self-attention processing on the plurality of key feature sequences based on the attention network layer of the degradation knee prediction model to determine the plurality of key feature sequences after the attention processing; The automatic encoder based on the degradation knee point prediction model performs outlier detection processing, reconstruction error calculation processing and dynamic threshold detection processing on the key feature sequences after multiple attention processing to predict the capacity degradation knee point of the battery.

[0008] In one possible implementation, the autoencoder based on the degradation knee prediction model performs outlier detection processing, reconstruction error calculation processing, and dynamic threshold detection processing on the key feature sequences after multiple attention processing to predict the capacity degradation knee point of the battery, including: Performing feature encoding processing on the key feature sequences after the multiple attention processes, performing feature decoding processing on the key feature sequences after the feature encoding, and determining a reconstructed key feature sequence; Performing reconstruction error calculation processing on each of the key feature sequences after attention processing and the corresponding reconstructed key feature sequence to determine a reconstruction error sequence between each of the key feature sequences after attention processing and the corresponding reconstructed key feature sequence; Determining at least one abnormal key feature sequence from a plurality of the key feature sequences after attention processing based on the plurality of the reconstructed error sequences; Dynamic threshold detection processing is performed on the target reconstruction error sequence of each abnormal key feature sequence to predict the capacity degradation knee point of the battery.

[0009] In one possible implementation, performing dynamic threshold detection processing on the target reconstruction error sequence of each abnormal key feature sequence to predict the capacity degradation knee point of the battery includes: Performing Gaussian filtering on the target reconstruction error sequence of the abnormal key feature sequence to determine the target reconstruction error sequence after Gaussian filtering; Performing smoothing filtering on the target reconstruction error sequence after Gaussian filtering and performing first-order difference calculation on the smoothed error to determine the rate of change of each target reconstruction error sequence at each time step; If the change rate is greater than or equal to a preset change rate threshold, the time step corresponding to the change rate is used as the capacity degradation knee point of the battery.

[0010] In one possible implementation, the degradation knee point prediction model is determined by the following steps: Inputting a plurality of sample key feature sequences into a deep learning model, reconstructing the sample key feature sequences, and determining a reconstructed sample key feature sequence; Determining a loss value between the reconstructed sample key feature sequence and the corresponding sample key feature sequence; If the loss value is greater than a preset threshold, the network parameters of the deep learning model are changed, and the deep learning model with the changed network parameters is continued to be iteratively trained. If the loss value is less than or equal to the preset threshold, the deep learning model is determined as the degraded knee point prediction model.

[0011] The present application also provides a device for detecting a battery capacity degradation knee point, the device comprising: A feature extraction module is used to extract key features from the current and capacity data of the battery during the constant voltage charging phase; wherein the key features include the current attenuation coefficient, the average rate of current decrease, the current ratio at the target time point, the curvature of the current change, and the charging capacity during the constant voltage charging phase; A data processing module, configured to perform data preprocessing on the key features to determine a plurality of key feature sequences; A determination module is used to input multiple key feature sequences into a pre-trained degradation knee point prediction model, perform self-attention processing and degradation knee point detection processing on the multiple key feature sequences, and predict the capacity degradation knee point of the battery; wherein the degradation knee point prediction model is obtained by iteratively training a deep learning model of the self-attention mechanism.

[0012] In a possible implementation manner, when the data processing module is used to perform data preprocessing on the key features to determine multiple key feature sequences, the data processing module is specifically used to: Performing high-frequency noise removal processing on each of the key features based on Gaussian filtering to determine a plurality of key features after denoising; performing normalization processing on the plurality of key features after denoising to determine the plurality of normalized key features; The multiple normalized key features are split into fixed-length sequences to determine multiple key feature sequences.

[0013] An embodiment of the present application also provides an electronic device, including: a processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are executed by the processor, the steps of the battery capacity degradation knee point detection method as described above are performed.

[0014] An embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the above-mentioned method for detecting the knee point of battery capacity degradation are executed.

[0015] The present application provides a method, apparatus, device, and storage medium for detecting the battery capacity degradation knee point. The method comprises: extracting key features from the current and capacity data of the battery during the constant voltage charging phase; wherein the key features include the current attenuation coefficient, the average rate of current drop, the current ratio at the target time point, the curvature of the current change, and the charge capacity during the constant voltage charging phase; performing data preprocessing on the key features to determine multiple key feature sequences; inputting the multiple key feature sequences into a pre-trained degradation knee point prediction model, performing self-attention processing and degradation knee point detection processing on the multiple key feature sequences, and predicting the capacity degradation knee point of the battery; wherein the degradation knee point prediction model is obtained by iteratively training a deep learning model with a self-attention mechanism. By mining key feature vectors from the current and capacity data during the constant voltage charging phase, high-quality data support is provided for model analysis. The knee point can be preliminarily identified by utilizing the anomaly detection capability, and the reconstruction error is processed by using smoothing filtering and dynamic thresholding methods to more accurately detect the knee point.

[0016] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. 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 relevant drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 A flowchart of a method for detecting a battery capacity degradation knee point provided in an embodiment of the present application; Figure 2 This is a schematic diagram of the structure of a device for detecting a knee point of battery capacity degradation provided by an embodiment of the present application; Figure 3 This is a second structural diagram of a device for detecting a battery capacity degradation knee point provided in an embodiment of the present application; Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for which protection is claimed, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, each other embodiment obtained by those skilled in the art without making creative work falls within the scope of protection of the present application.

[0020] First, the application scenarios to which this application is applicable are introduced. This application can be applied in the field of battery capacity detection technology.

[0021] Research has found that batteries age with age and the number of charge and discharge cycles, leading to a decline in capacity and performance. This capacity degradation exhibits a nonlinear trend. This is the result of the combined effects of external cycling stresses (such as charge and discharge rate and temperature) and complex internal degradation mechanisms (such as SEI film growth and rupture, and electrode material phase transitions). The nonlinear degradation trajectory of lithium-ion batteries occurs in two stages: initially, the capacity declines slowly and approximately linearly, and then the degradation rate accelerates after passing the knee point. Accurately detecting the knee point is crucial for battery cell development and battery management systems (BMS), and can be used for remaining service life prediction and fault diagnosis. For example, in electric vehicles, it can optimize battery management strategies and avoid catastrophic failures. Currently, physical and chemical modeling methods are the primary approach to detecting the knee point, but these methods suffer from numerous challenges: first, measuring the complex physical and chemical parameters within the battery is extremely difficult, as these parameters vary with usage and the external environment; second, they require significant computational resources and suffer from poor real-time performance. Therefore, accurately determining the knee point of battery capacity degradation has become a significant technical challenge.

[0022] Based on this, the present invention provides a method for detecting the knee point of battery capacity degradation. By mining key feature vectors from current and capacity data during the constant-voltage charging phase, this method provides high-quality data support for model analysis. This method leverages anomaly detection capabilities to initially identify the knee point, and employs smoothing filtering and dynamic thresholding to address reconstruction errors, enabling more accurate knee point detection.

[0023] See also Figure 1 , Figure 1 This is a flow chart of a method for detecting the knee point of battery capacity degradation provided by an embodiment of the present application. Figure 1 As shown in , the detection method provided in the embodiment of the present application includes: S101: Extract key features from the current and capacity data of the battery during the constant voltage charging stage; wherein the key features include the current attenuation coefficient, the average rate of current decrease, the current ratio at the target time point, the curvature of the current change, and the charging capacity during the constant voltage charging stage.

[0024] In this step, key features are extracted from the current and capacity data of the battery during the constant voltage charging stage.

[0025] Here, the current attenuation coefficient is: In the constant voltage charging stage, according to the Nernst equation, the change of the current of a normal battery with time usually follows the exponential decay law, that is, ,in is the constant voltage charging starting current, K Is the attenuation coefficient. This coefficient is related to factors such as the battery's material, structure, and internal electrochemical reaction. When the battery deteriorates, the actual measured current attenuation coefficient The attenuation coefficient is obtained by fitting the current and time data of the constant voltage charging stage. The nonlinear least squares fitting function is used. To the actual measured current data Minimize the objective function , adjust the parameters through the Gauss-Newton method iterative algorithm and K The value of S To achieve the minimum, we get k The value is the current attenuation coefficient. Among them, the average rate of current decrease: The average rate of current decrease can reflect the overall trend of current change over a period of time from a macro perspective. It can help determine whether the electrochemical reaction rate inside the battery is normal. If the battery has an abnormality, such as a micro short circuit or aging of the electrode material, the average rate of current decrease may change. Select a time interval for the constant voltage charging stage. , calculate the average rate of current drop . 10 seconds after charging starts , and the current drops to 0.2 times the initial current, which is approximately In this interval, record the starting current and end current , calculate the average rate according to the above formula.

[0026] Here, the target time point current ratio is used to compare the current ratios at different target time points during constant voltage charging. This provides information about the battery's relative current status at different moments. Different battery states (normal or abnormal) will result in different ratios. Two key time points during constant voltage charging are selected: 10 seconds and 200 seconds after the start of charging. By observing that the current drops to approximately 1 / 10 of the initial current at 200 seconds, the current ratio is calculated to obtain the target time point current ratio.

[0027] Among them, the curvature of current change: The curvature of current change describes the curvature of the current curve over time. The curvature of the current curve of a normal battery is relatively stable, but when the battery fails, the curvature of the current change may change significantly. The central difference method is used to calculate the curvature of current change. For time series , let the time interval be , for the time point , the curvature of the current change The curvature can be calculated at each time point, and then the mean and mean square values ​​of the curvature can be selected as features.

[0028] Among them, the charging capacity in the constant voltage charging stage: In the constant voltage charging stage, the charging capacity growth pattern of a normal battery is relatively stable. When a fault such as a micro short circuit occurs in the battery, the growth pattern of the charging capacity will change due to abnormal internal current distribution, such as a slower growth rate or a reduced final charging capacity. The charging capacity is calculated using the trapezoidal integration method. .

[0029] S102: performing data preprocessing on the key features to determine a plurality of key feature sequences.

[0030] In this step, data preprocessing is performed on the key features to determine multiple key feature sequences.

[0031] In a possible implementation manner, performing data preprocessing on the key features to determine multiple key feature sequences includes: A: Perform high-frequency noise removal processing on each key feature based on Gaussian filtering to determine multiple key features after denoising.

[0032] Here, Gaussian filtering is used to remove high-frequency noise from each key feature, and multiple denoised key features are determined.

[0033] B: performing normalization processing on the plurality of key features after denoising to determine the plurality of normalized key features.

[0034] Here, normalization processing is performed on the multiple denoised key features to determine the multiple normalized key features.

[0035] C: splitting the multiple normalized key features into fixed-length sequences to determine multiple key feature sequences.

[0036] Here, multiple normalized key features are split into fixed-length sequences to determine multiple key feature sequences.

[0037] The continuous time series data is split into sequences of fixed length to be input into the model for training and prediction. Here, the length of each sequence represents several consecutive charging cycles of the battery.

[0038] S103: Input the multiple key feature sequences into a pre-trained degradation knee point prediction model, perform self-attention processing and degradation knee point detection processing on the multiple key feature sequences, and predict the capacity degradation knee point of the battery; wherein the degradation knee point prediction model is obtained by iteratively training a deep learning model of the self-attention mechanism.

[0039] In this step, multiple key feature sequences are input into a pre-trained degradation knee point prediction model, and self-attention processing and degradation knee point detection processing are performed on the multiple key feature sequences to predict the battery capacity degradation knee point.

[0040] In one possible implementation, the inputting of the multiple key feature sequences into a pre-trained degradation knee point prediction model, performing self-attention processing and outlier detection processing on the multiple key feature sequences, and predicting the capacity degradation knee point of the battery includes: a: Based on the attention network layer of the degenerate knee prediction model, self-attention processing is performed on the multiple key feature sequences to determine the key feature sequences after multiple attention processing.

[0041] Here, self-attention processing is performed on multiple key feature sequences according to the attention network layer of the degraded knee prediction model to determine multiple key feature sequences after attention processing.

[0042] b. The autoencoder based on the degradation knee point prediction model performs outlier detection processing, reconstruction error calculation processing, and dynamic threshold detection processing on the key feature sequence after multiple attention processing to predict the capacity degradation knee point of the battery. Here, the autoencoder of the degradation knee point prediction model performs outlier detection processing, reconstruction error calculation processing, and dynamic threshold detection processing on the key feature sequences after multiple attention processing to predict the battery capacity degradation knee point.

[0043] Among them, the battery capacity degradation knee point refers to a key point where the capacity degradation trend of a lithium-ion battery changes significantly during use.

[0044] In one possible implementation, the autoencoder based on the degradation knee prediction model performs outlier detection processing, reconstruction error calculation processing, and dynamic threshold detection processing on the key feature sequences after multiple attention processing to predict the capacity degradation knee point of the battery, including: (1): Perform feature encoding processing on the key feature sequence after multiple attention processing, perform feature decoding processing on the key feature sequence after feature encoding, and determine the reconstructed key feature sequence.

[0045] Here, feature encoding processing is performed on the key feature sequences after multiple attention processing, and feature decoding processing is performed on the key feature sequences after feature encoding to determine the reconstructed key feature sequence.

[0046] (2): Perform reconstruction error calculation on each key feature sequence after attention processing and the corresponding reconstructed key feature sequence to determine the reconstruction error sequence between each key feature sequence after attention processing and the corresponding reconstructed key feature sequence.

[0047] Here, the reconstruction error sequence is determined by the following formula:

[0048] in, Indicates time The reconstruction error, It's time The actual data, It is to reconstruct the data.

[0049] (3): Determine at least one abnormal key feature sequence from the key feature sequences after multiple attention processing based on the multiple reconstructed error sequences.

[0050] Here, the reconstruction error generally grows steadily over time, but increases sharply at the capacity degradation knee.

[0051] Here, when the error in the reconstruction error sequence increases sharply, it is determined that the key feature sequence corresponding to the reconstruction error sequence is an abnormal key feature sequence.

[0052] (4): Perform dynamic threshold detection processing on the target reconstruction error sequence of each abnormal key feature sequence to predict the capacity degradation knee point of the battery.

[0053] Here, dynamic threshold detection processing is performed on the target reconstruction error sequence of each abnormal key feature sequence to predict the battery capacity degradation knee point.

[0054] In one possible implementation, performing dynamic threshold detection processing on the target reconstruction error sequence of each abnormal key feature sequence to predict the capacity degradation knee point of the battery includes: I: performing Gaussian filtering on the target reconstruction error sequence of the abnormal key feature sequence to determine the target reconstruction error sequence after Gaussian filtering.

[0055] Here, the target reconstruction error sequence of the abnormal key feature sequence is subjected to Gaussian filtering to determine the target reconstruction error sequence after Gaussian filtering.

[0056] II: performing smoothing filtering on the target reconstruction error sequence after Gaussian filtering and performing first-order difference calculation on the smoothed error to determine the rate of change of each target reconstruction error sequence at each time step.

[0057] Here, the first-order difference of the smoothing error is calculated using the following formula:

[0058] in, represents the change of reconstruction error after filtering, is a small value that prevents the denominator from being zero, For time t Smoothing filtering of the reconstruction error under .

[0059] III: If the change rate is greater than or equal to a preset change rate threshold, the time step corresponding to the change rate is used as the capacity degradation knee point of the battery.

[0060] In one possible implementation, the degradation knee point prediction model is determined by the following steps: i: Input multiple sample key feature sequences into a deep learning model, reconstruct the sample key feature sequences, and determine the reconstructed sample key feature sequences.

[0061] ii: Determine the loss value between the reconstructed sample key feature sequence and the corresponding sample key feature sequence.

[0062] Here, the loss value is determined by the following formula:

[0063] Among them, N is the number of sample key feature sequences, It is Sample key feature sequences, It is A reconstructed sample key feature sequence.

[0064] iii: If the loss value is greater than the preset threshold, the network parameters of the deep learning model are changed, and the deep learning model after the network parameters are changed is continued to be iteratively trained. If the loss value is less than or equal to the preset threshold, the deep learning model is determined as the degraded knee point prediction model.

[0065] In this application, Gaussian filtering and normalization are key steps to ensure data stationarity. The temporal transformer, combined with an autoencoder, captures global dependencies through a self-attention mechanism and uses the autoencoder for anomaly detection. The model is trained and parameters are optimized by minimizing the reconstruction error. The knee point is determined by calculating the reconstruction error, with larger reconstruction errors typically corresponding to the knee point. Using smoothing, differencing, rate of change, and dynamic thresholding methods, the knee point of capacity degradation can be detected with high sensitivity.

[0066] An embodiment of the present application provides a method for detecting the knee point of battery capacity degradation, the method comprising: extracting key features from the current and capacity data of the battery during the constant voltage charging phase; wherein the key features include the current attenuation coefficient, the average rate of current drop, the current ratio at the target time point, the curvature of the current change, and the charge amount during the constant voltage charging phase; performing data preprocessing on the key features to determine multiple key feature sequences; inputting the multiple key feature sequences into a pre-trained degradation knee point prediction model, performing self-attention processing and degradation knee point detection processing on the multiple key feature sequences, and predicting the capacity degradation knee point of the battery; wherein the degradation knee point prediction model is obtained by iteratively training a deep learning model with a self-attention mechanism. By mining key feature vectors from the current and capacity data during the constant voltage charging phase, high-quality data support is provided for model analysis. The knee point can be preliminarily identified by utilizing the anomaly detection capability, and the reconstruction error is processed by using smoothing filtering and dynamic thresholding methods to more accurately detect the knee point.

[0067] See also Figure 2 、 Figure 3 , Figure 2 This is a schematic diagram of the structure of a device for detecting a knee point of battery capacity degradation provided by an embodiment of the present application; Figure 3 This is a second structural diagram of a battery capacity degradation knee point detection device provided in an embodiment of the present application. Figure 2 As shown in , the detection device 200 includes: A feature extraction module 210 is configured to extract key features from the current and capacity data of the battery during the constant voltage charging phase; wherein the key features include the current attenuation coefficient, the average rate of current decrease, the current ratio at the target time point, the curvature of the current change, and the charge capacity during the constant voltage charging phase; The data processing module 220 is used to perform data preprocessing on the key features to determine multiple key feature sequences; The determination module 230 is used to input the multiple key feature sequences into a pre-trained degradation knee point prediction model, perform self-attention processing and degradation knee point detection processing on the multiple key feature sequences, and predict the capacity degradation knee point of the battery; wherein the degradation knee point prediction model is obtained by iteratively training a deep learning model of the self-attention mechanism.

[0068] Furthermore, when the data processing module 220 is used to perform data preprocessing on the key features and determine multiple key feature sequences, the data processing module 220 is specifically used to: Performing high-frequency noise removal processing on each of the key features based on Gaussian filtering to determine a plurality of key features after denoising; performing normalization processing on the plurality of key features after denoising to determine the plurality of normalized key features; The multiple normalized key features are split into fixed-length sequences to determine multiple key feature sequences.

[0069] Furthermore, when the determination module 230 is used to input the multiple key feature sequences into a pre-trained degradation knee point prediction model, perform self-attention processing and outlier detection processing on the multiple key feature sequences, and predict the capacity degradation knee point of the battery, the determination module 230 is specifically used to: Performing self-attention processing on the plurality of key feature sequences based on the attention network layer of the degradation knee prediction model to determine the plurality of key feature sequences after the attention processing; The automatic encoder based on the degradation knee point prediction model performs outlier detection processing, reconstruction error calculation processing and dynamic threshold detection processing on the key feature sequences after multiple attention processing to predict the capacity degradation knee point of the battery.

[0070] Furthermore, when the determination module 230 performs outlier detection processing, reconstruction error calculation processing, and dynamic threshold detection processing on the key feature sequences after multiple attention processing using the autoencoder based on the degradation knee point prediction model to predict the capacity degradation knee point of the battery, the determination module 230 is specifically configured to: Performing feature encoding processing on the key feature sequences after the multiple attention processes, performing feature decoding processing on the key feature sequences after the feature encoding, and determining a reconstructed key feature sequence; Performing reconstruction error calculation processing on each of the key feature sequences after attention processing and the corresponding reconstructed key feature sequence to determine a reconstruction error sequence between each of the key feature sequences after attention processing and the corresponding reconstructed key feature sequence; Determining at least one abnormal key feature sequence from a plurality of the key feature sequences after attention processing based on the plurality of the reconstructed error sequences; Dynamic threshold detection processing is performed on the target reconstruction error sequence of each abnormal key feature sequence to predict the capacity degradation knee point of the battery.

[0071] Furthermore, when the determination module 230 is used to perform dynamic threshold detection processing on the target reconstruction error sequence of each abnormal key feature sequence to predict the capacity degradation knee point of the battery, the determination module 230 is specifically used to: Performing Gaussian filtering on the target reconstruction error sequence of the abnormal key feature sequence to determine the target reconstruction error sequence after Gaussian filtering; Performing smoothing filtering on the target reconstruction error sequence after Gaussian filtering and performing first-order difference calculation on the smoothed error to determine the rate of change of each target reconstruction error sequence at each time step; If the change rate is greater than or equal to a preset change rate threshold, the time step corresponding to the change rate is used as the capacity degradation knee point of the battery.

[0072] Further, such as Figure 3 As shown, the detection device 200 further includes a model training module 240, which is configured to determine the degradation knee point prediction model through the following steps: Inputting a plurality of sample key feature sequences into a deep learning model, reconstructing the sample key feature sequences, and determining a reconstructed sample key feature sequence; Determining a loss value between the reconstructed sample key feature sequence and the corresponding sample key feature sequence; If the loss value is greater than the preset threshold, the network parameters of the deep learning model are changed, and the deep learning model with the changed network parameters is continued to be iteratively trained. If the loss value is less than or equal to the preset threshold, the deep learning model is determined as the degraded knee point prediction model. An embodiment of the present application provides a device for detecting the capacity degradation knee point of a battery. The device comprises: a feature extraction module for extracting key features from the current and capacity data of the battery during the constant-voltage charging phase; wherein the key features include the current attenuation coefficient, the average rate of current drop, the current ratio at the target time point, the curvature of the current change, and the charge capacity during the constant-voltage charging phase; a data processing module for preprocessing the key features to determine multiple key feature sequences; and a determination module for inputting the multiple key feature sequences into a pre-trained degradation knee point prediction model, performing self-attention processing and degradation knee point detection on the multiple key feature sequences, and predicting the capacity degradation knee point of the battery. The degradation knee point prediction model is obtained by iteratively training a deep learning model with a self-attention mechanism. By mining key feature vectors from the current and capacity data during the constant-voltage charging phase, high-quality data support is provided for model analysis. The knee point can be initially identified using anomaly detection capabilities, and reconstruction errors are processed using smoothing filtering and dynamic thresholding methods to more accurately detect the knee point.

[0073] See also Figure 4 , Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 4 As shown in FIG, the electronic device 400 includes a processor 410 , a memory 420 and a bus 430 .

[0074] The memory 420 stores machine-readable instructions executable by the processor 410. When the electronic device 400 is running, the processor 410 communicates with the memory 420 via the bus 430. When the machine-readable instructions are executed by the processor 410, the above-mentioned Figure 1 The steps of the method for detecting the knee point of battery capacity degradation in the illustrated method embodiment and the specific implementation thereof can be found in the method embodiment, which will not be described in detail here.

[0075] The embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the computer program can execute the above-mentioned Figure 1 The steps of the method for detecting the knee point of battery capacity degradation in the illustrated method embodiment and the specific implementation thereof can be found in the method embodiment, which will not be described in detail here.

[0076] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0077] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. There may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed may be through some communication interface, indirect coupling or communication connection of devices or units, which may be electrical, mechanical or other forms.

[0078] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0079] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0080] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0081] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The scope of protection of the present application is not limited thereto. Although the present application has been described in detail with reference to the above-mentioned embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily conceive of changes to the technical solutions described in the above-mentioned embodiments within the technical scope disclosed in the present application, or perform equivalent replacements for some of the technical features thereof. These modifications, changes, or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for detecting a battery capacity degradation knee point, characterized in that: The detection method comprises: Extract key features from the battery's current and capacity data during the constant-voltage charging phase; the key features include current attenuation coefficient, average current drop rate, current ratio at the target time point, current change curvature, and charge capacity during the constant-voltage charging phase; Performing data preprocessing on the key features to determine multiple key feature sequences; The multiple key feature sequences are input into a pre-trained degradation knee point prediction model, and self-attention processing and degradation knee point detection processing are performed on the multiple key feature sequences to predict the capacity degradation knee point of the battery; wherein the degradation knee point prediction model is obtained by iteratively training a deep learning model of the self-attention mechanism.

2. The detection method according to claim 1, wherein The data preprocessing of the key features to determine multiple key feature sequences includes: Performing high-frequency noise removal processing on each of the key features based on Gaussian filtering to determine a plurality of key features after denoising; performing normalization processing on the plurality of key features after denoising to determine the plurality of normalized key features; The multiple normalized key features are split into fixed-length sequences to determine multiple key feature sequences.

3. The detection method according to claim 1, wherein Inputting the plurality of key feature sequences into a pre-trained degradation knee point prediction model, performing self-attention processing and outlier detection processing on the plurality of key feature sequences, and predicting the capacity degradation knee point of the battery, includes: Performing self-attention processing on the plurality of key feature sequences based on the attention network layer of the degradation knee prediction model to determine the plurality of key feature sequences after the attention processing; The automatic encoder based on the degradation knee point prediction model performs outlier detection processing, reconstruction error calculation processing and dynamic threshold detection processing on the key feature sequences after multiple attention processing to predict the capacity degradation knee point of the battery.

4. The detection method according to claim 3, characterized in that The autoencoder based on the degradation knee prediction model performs outlier detection processing, reconstruction error calculation processing, and dynamic threshold detection processing on the key feature sequences after multiple attention processing to predict the capacity degradation knee point of the battery, including: Performing feature encoding processing on the key feature sequences after the multiple attention processes, performing feature decoding processing on the key feature sequences after the feature encoding, and determining a reconstructed key feature sequence; Performing reconstruction error calculation processing on each of the key feature sequences after attention processing and the corresponding reconstructed key feature sequence to determine a reconstruction error sequence between each of the key feature sequences after attention processing and the corresponding reconstructed key feature sequence; Determining at least one abnormal key feature sequence from a plurality of the key feature sequences after attention processing based on the plurality of the reconstructed error sequences; Dynamic threshold detection processing is performed on the target reconstruction error sequence of each abnormal key feature sequence to predict the capacity degradation knee point of the battery.

5. The detection method according to claim 4, characterized in that The performing dynamic threshold detection processing on the target reconstruction error sequence of each abnormal key feature sequence to predict the capacity degradation knee point of the battery includes: Performing Gaussian filtering on the target reconstruction error sequence of the abnormal key feature sequence to determine the target reconstruction error sequence after Gaussian filtering; Performing smoothing filtering on the target reconstruction error sequence after Gaussian filtering and performing first-order difference calculation on the smoothed error to determine the rate of change of each target reconstruction error sequence at each time step; If the change rate is greater than or equal to a preset change rate threshold, the time step corresponding to the change rate is used as the capacity degradation knee point of the battery.

6. The detection method according to claim 1, characterized in that The degradation knee point prediction model is determined by the following steps: Inputting a plurality of sample key feature sequences into a deep learning model, reconstructing the sample key feature sequences, and determining a reconstructed sample key feature sequence; Determining a loss value between the reconstructed sample key feature sequence and the corresponding sample key feature sequence; If the loss value is greater than a preset threshold, the network parameters of the deep learning model are changed, and the deep learning model with the changed network parameters is continued to be iteratively trained. If the loss value is less than or equal to the preset threshold, the deep learning model is determined as the degraded knee point prediction model.

7. A device for detecting a battery capacity degradation knee point, characterized in that: The detection device comprises: A feature extraction module is used to extract key features from the current and capacity data of the battery during the constant voltage charging phase; wherein the key features include the current attenuation coefficient, the average rate of current decrease, the current ratio at the target time point, the curvature of the current change, and the charging capacity during the constant voltage charging phase; A data processing module, configured to perform data preprocessing on the key features to determine a plurality of key feature sequences; A determination module is used to input multiple key feature sequences into a pre-trained degradation knee point prediction model, perform self-attention processing and degradation knee point detection processing on the multiple key feature sequences, and predict the capacity degradation knee point of the battery; wherein the degradation knee point prediction model is obtained by iteratively training a deep learning model of the self-attention mechanism.

8. The detection device according to claim 7, characterized in that When the data processing module is used to perform data preprocessing on the key features and determine multiple key feature sequences, the data processing module is specifically used to: Performing high-frequency noise removal processing on each of the key features based on Gaussian filtering to determine a plurality of key features after denoising; performing normalization processing on the plurality of key features after denoising to determine the plurality of normalized key features; The multiple normalized key features are split into fixed-length sequences to determine multiple key feature sequences.

9. An electronic device, characterized in that: include: A processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate via the bus. When the processor runs the machine-readable instructions, the processor executes the steps of the method for detecting the knee point of battery capacity degradation as described in any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for detecting the knee point of battery capacity degradation according to any one of claims 1 to 6 are executed.

Images