New energy battery aging state prediction method, device and storage medium

Abstract

The application provides a new energy battery aging state prediction method, device and storage medium. The method determines a battery aging feature sequence through the charging and discharging data of the target battery, inputs the battery aging feature sequence into a pre-trained aging state prediction model, determines local time sequence features through a convolution module, determines the bidirectional dependency relationship of the local time sequence features in the time dimension through a bidirectional long short-term memory module, obtains time sequence dynamic change features, adjusts the weight of each time step in the time sequence dynamic change features through an attention module, determines the current aging state, realizes the prediction of the battery aging state, can accurately capture the local time sequence features, improves the feature expression ability, accurately captures the bidirectional dependency relationship in the time dimension, enhances the analysis ability of the battery aging trend, can focus on the effective information of the key time points, improves the prediction accuracy, realizes low cost and simple process, and is suitable for various types of batteries and complex working conditions.

Description

Technical Field

[0001] This application relates to the field of new energy battery technology, specifically to a method, device and storage medium for predicting the aging state of new energy batteries. Background Technology

[0002] Currently, calibration methods for calendar aging models of new energy batteries can be mainly categorized into two types: (1) electrochemical models and (2) data-driven models. Among them, electrochemical models use physical mechanism-based models to characterize the aging mechanism inside the battery at the microscopic level. Data-driven models predict the aging state by mining aging patterns in the battery's historical operating data.

[0003] However, electrochemical mechanism models rely on a large number of microscopic parameters that are difficult to measure directly, leading to complex calibration processes, limited versatility, and high computational costs under complex operating conditions, making real-time applications difficult, especially in meeting the needs of embedded systems or battery management systems for rapid response and online monitoring. Data-driven models generally suffer from insufficient generalization ability, making it difficult to adapt to different types of batteries, complex operating conditions, and diverse aging mechanisms, affecting the model's predictive accuracy and stability. Summary of the Invention

[0004] In view of the above-mentioned defects or deficiencies in the prior art, this application aims to provide a method, device and storage medium for predicting the aging state of new energy batteries, so as to solve the problems of high cost, complex calculation, poor adaptability and insufficient accuracy in predicting battery aging state.

[0005] This application provides a method for predicting the aging state of new energy batteries, the method comprising: Acquire charge and discharge data of the target battery, and determine the battery aging characteristic sequence based on the charge and discharge data; The battery aging feature sequence is input into a pre-trained aging state prediction model, wherein the aging state prediction model includes a convolution module, a bidirectional long short-term memory module, and an attention module. The convolution module determines the local temporal features in the battery aging feature sequence, and the bidirectional long short-term memory module determines the bidirectional dependency of the local temporal features in the time dimension to obtain the temporal dynamic change features. The current aging state of the target battery is determined by adjusting the weights of each time step in the time-series dynamic change features using the attention module.

[0006] Optionally, a battery aging characteristic sequence is determined based on the charge / discharge data, including: The battery aging characteristics at each time point are determined based on the charge and discharge data, wherein the battery aging characteristics include constant voltage charging time, constant current charging time, average internal resistance, terminal voltage change rate, discharge cutoff voltage, discharge time, and maximum charging current. A battery aging feature sequence is constructed based on the battery aging characteristics at each time point.

[0007] Optionally, a battery aging feature sequence is constructed based on the battery aging characteristics at each time point, including: Some battery aging features are removed based on a pre-built aging state association relationship, wherein the aging state association relationship is used to describe battery aging features associated with battery aging state. A battery aging feature sequence is constructed based on the remaining battery aging features at each time point.

[0008] Optionally, the construction of the aging state association includes: Acquire sample data, and extract the feature sequence of each sample and the corresponding sample aging state sequence from the sample data; For each of the sample feature sequences, determine the correlation between the sample feature sequence and the sample aging state sequence; Based on the correlation degree corresponding to each of the sample feature sequences, the aging state correlation relationship is constructed.

[0009] Optionally, the correlation includes curve trend similarity, determining the correlation between the sample feature sequence and the sample aging state sequence, including: The sample feature sequence and the sample aging state sequence are normalized respectively, and the absolute difference sequence between the normalized sample feature sequence and the sample aging state sequence is determined. Determine the minimum and maximum differences in the absolute difference sequence, and determine the correlation coefficient of each absolute difference in the absolute difference sequence based on the minimum and maximum differences; Based on the correlation coefficient of each absolute difference in the absolute difference sequence, the curve trend similarity between the sample feature sequence and the sample aging state sequence is determined.

[0010] Optionally, the correlation includes numerical linear similarity and ordinal monotonic similarity. Determining the correlation between the sample feature sequence and the sample aging state sequence includes: Determine the mean of the feature sequence of the sample feature sequence and the mean of the state sequence of the sample aging state sequence, and determine the sum of the products of the deviations from the mean of the feature sequence and the mean of the state sequence, as well as the sum of the squares of the deviations from the mean; The numerical linear similarity between the sample feature sequence and the sample aging state sequence is determined based on the sum of the products of the deviations from the mean and the sum of the squares of the deviations from the mean. The sample feature sequence and the sample aging state sequence are numerically sorted respectively, and the rank difference value of the sample feature sequence and the sample aging state sequence is determined according to the sorting results. The sorting monotonic similarity between the sample feature sequence and the sample aging state sequence is determined based on the rank difference value.

[0011] Optionally, the battery aging characteristics at each time point are determined based on the charge and discharge data, including: Based on the charge and discharge data, determine the measured voltage and current values ​​for each charging cycle, as well as the voltage change rate, internal resistance, and measured current values ​​for each discharging cycle. Based on the measured voltage and current values ​​during each charging cycle, as well as the measured voltage change rate, internal resistance, and current values ​​during each discharging cycle, the battery aging characteristics at each time point are determined.

[0012] Optionally, the bidirectional long short-term memory module is used to determine the bidirectional dependency relationship of the local temporal features in the time dimension to obtain the temporal dynamic change features, including: The forward processing unit in the bidirectional long short-term memory module processes the data from the first time step to the last time step in the local time sequence feature to obtain the preceding time sequence information. The reverse processing unit in the bidirectional long short-term memory module processes the data from the last time step in the local time sequence feature to the data from the first time step, thus obtaining the subsequent time sequence information. The dynamic change characteristics of the time series are determined based on the preceding and following time series information.

[0013] This application embodiment also provides an electronic device, the electronic device comprising: Processor and memory; The processor executes the steps of the new energy battery aging state prediction method provided in any embodiment of this application by calling the program or instructions stored in the memory.

[0014] This application also provides a computer-readable storage medium storing a program or instructions that cause a computer to execute the steps of the new energy battery aging state prediction method provided in any embodiment of this application.

[0015] In summary, this application proposes a method for predicting the aging state of new energy batteries. This method acquires the charge-discharge data of the target battery, determines the battery aging feature sequence based on the data, and inputs this sequence into a pre-trained aging state prediction model. The convolution module in this model determines the local temporal features of the battery aging feature sequence, and the bidirectional long short-term memory module determines the bidirectional dependency of these local temporal features in the time dimension, obtaining the dynamic temporal change features. The attention module in the model adjusts the weights of each time step in the dynamic temporal change features to determine the current aging state of the target battery, thus predicting the battery aging state. This method integrates convolution, bidirectional long short-term memory, and attention modules, accurately capturing the local temporal features of the battery aging feature sequence, improving feature representation capabilities, accurately capturing the bidirectional dependency in the time dimension, enhancing the model's ability to analyze battery aging trends, and enabling the model to focus more on effective information at key time points, improving prediction accuracy. It eliminates the need to build an electrochemical model, resulting in low cost, simple process, applicability to various types of batteries and complex operating conditions, and ensuring the stability of predictions for various complex aging behaviors. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a flowchart of a method for predicting the aging state of a new energy battery provided in an embodiment of this application; Figure 2 This is a schematic diagram of an aging state prediction model provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0018] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0019] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] As mentioned in the background section, this application proposes a method for predicting the aging state of new energy batteries to address the problems in existing technologies and achieve calendar aging assessment of new energy batteries. Specifically, calendar aging assessment of new energy batteries refers to evaluating the performance degradation of new energy batteries (such as lithium-ion batteries) under non-consumable conditions, caused solely by the passage of time and the influence of environmental factors (such as temperature and state of charge).

[0021] Figure 1 This is a flowchart illustrating a method for predicting the aging state of a new energy battery, as provided in an embodiment of this application. See also... Figure 1 The method for predicting the aging status of new energy batteries specifically includes: S110. Obtain the charge and discharge data of the target battery, and determine the battery aging characteristic sequence based on the charge and discharge data.

[0022] The charge / discharge data describes the target battery's operational information during the charge / discharge phase, such as charge / discharge voltage, charge / discharge current, internal resistance, and capacity. In this embodiment, charge / discharge data can be collected from the battery at multiple charge / discharge phases.

[0023] The battery aging characteristic sequence can include battery aging characteristics at different times, and these characteristics can reflect the degradation of battery performance. Specifically, battery aging characteristics include constant current charging time, average internal resistance, terminal voltage change rate, discharge cutoff voltage, discharge time, and maximum charging current.

[0024] In some implementations, determining the battery aging characteristic sequence based on charge and discharge data includes the following steps: Step 11: Determine the battery aging characteristics at each time point based on the charge and discharge data. The battery aging characteristics include constant voltage charging time, constant current charging time, average internal resistance, terminal voltage change rate, discharge cutoff voltage, discharge time, and maximum charging current. Step 12: Construct a battery aging feature sequence based on the battery aging characteristics at each time point.

[0025] Among them, constant voltage charging time refers to the time from the start of constant voltage charging to the satisfaction of charging termination conditions after the target battery enters the constant voltage charging stage. Constant current charging time refers to the time required for the target battery to charge from a preset initial capacity or voltage to a preset target capacity or voltage under a constant charging current. Average internal resistance refers to the average value of the equivalent DC internal resistance of the target battery under preset operating conditions, preset SOC (State of Charge), and preset temperature range. Specifically, it can be the statistical average value of the internal resistance under a complete charge and discharge stage.

[0026] In addition, the rate of change of terminal voltage can refer to the change in terminal voltage of the target battery per unit time during constant current charging and discharging. The discharge cutoff voltage can be the minimum permissible terminal voltage for effective discharge of the target battery. The discharge time can refer to the time consumed by the target battery to continuously discharge from a preset initial state of charge (SOC) to the cutoff voltage under specified discharge current, ambient temperature, and termination conditions. The maximum charging current can be the highest instantaneous or continuous charging current that is allowed to be applied to the target battery.

[0027] In the embodiments of this application, it is considered that as the number of charge-discharge cycles increases, the constant current charging time, constant voltage charging time, terminal voltage change rate, discharge time, and maximum charging current will gradually decrease. Among them, as the number of charge-discharge cycles gradually increases, the decreasing trend of the terminal voltage change rate will gradually strengthen. In addition, as the number of charge-discharge cycles increases, the average internal resistance and discharge cutoff voltage will gradually increase.

[0028] Therefore, constant voltage charging time, constant current charging time, average internal resistance, terminal voltage change rate, discharge cutoff voltage, discharge time, and maximum charging current can all indirectly reflect the aging state of the target battery. The above parameters are selected as battery aging characteristics, and the above battery aging characteristics are extracted from the charge and discharge data of the target battery to predict the battery aging state.

[0029] In one example, determining the battery aging characteristics at various times based on charge and discharge data includes the following steps: Step 111: Determine the measured voltage and current values ​​for each charging cycle, as well as the voltage change rate, internal resistance, and measured current values ​​for each discharging cycle based on the charging and discharging data. Step 112: Determine the battery aging characteristics at each time point based on the measured voltage and current values ​​in each charging cycle, as well as the measured voltage change rate, internal resistance, and current values ​​in each discharging cycle.

[0030] The charge / discharge data may include measured voltage and current values ​​for each charging cycle, as well as measured voltage change rate, internal resistance, and current values ​​for each discharging cycle. Additionally, the charge / discharge data may include timestamps corresponding to the measured values.

[0031] Specifically, the start and end times of charging can be determined based on the measured voltage and current values ​​and corresponding timestamps of the charging cycle in the charging and discharging data. This allows for the identification of the charging stage, and further, the constant current and constant voltage stages can be distinguished within the charging stage based on the measured voltage and current values. Moreover, the constant current charging time can be determined based on the timestamps corresponding to the constant current stage, and the constant voltage charging time can be determined based on the timestamps corresponding to the constant voltage stage.

[0032] Furthermore, for each charging cycle, all valid measured current values ​​within that cycle can be selected, and the largest measured current value can be chosen to obtain the maximum charging current. To avoid spike interference, a sliding window filter can also be used to select the largest measured current value.

[0033] In this embodiment, the start and end times of discharge can be determined based on the voltage change rate, measured current value, and corresponding timestamp of the discharge cycle in the charge / discharge data. This allows for the location of the discharge stage, and the discharge time is determined based on the timestamp corresponding to that stage. Furthermore, by iterating through the measured voltage values ​​at the end of each discharge stage, the average of the measured voltage values ​​from multiple consecutive times is selected as the discharge cutoff voltage.

[0034] In addition, a stable discharge range can be selected during the discharge phase. For example, the range from the initial value to the end value of the SOC (e.g., 30%~70%) can be selected as the stable discharge range. Then, the average value of all internal resistances within the stable discharge range can be calculated to obtain the average internal resistance.

[0035] Furthermore, the rate of change of terminal voltage can include both the rate of change of charging terminal voltage and the rate of change of discharging terminal voltage. Specifically, during the discharging phase, a mid-discharge interval can be selected, and the rate of change at each time point within that interval can be calculated, then the average value can be calculated to obtain the rate of change of discharging terminal voltage. Alternatively, the rate of change at each time point within the constant current segment of the charging phase can be calculated, then the average value can be calculated to obtain the rate of change of charging terminal voltage.

[0036] The above examples demonstrate how to extract battery aging characteristics based on measured voltage and current values ​​during charging cycles, as well as measured voltage change rate, internal resistance, and current values ​​during discharging cycles, ensuring the accuracy of these characteristics.

[0037] After extracting the battery aging characteristics at each time point based on the charge and discharge data, further, in step 12, a battery aging characteristic sequence can be constructed based on the battery aging characteristics at each time point. For example, battery aging characteristics at multiple consecutive time points can be selected according to a sliding time window to form a battery aging characteristic sequence.

[0038] The above implementation considers that the average resistance and discharge cutoff voltage are related to the static health characteristics of the target battery, the discharge time and constant current charging time are related to the capacity-related characteristics of the target battery, the terminal voltage change rate and constant voltage charging time are related to the dynamic response characteristics of the target battery, and the maximum charging current is related to the power performance characteristics of the target battery. By extracting constant voltage charging time, constant current charging time, average internal resistance, terminal voltage change rate, discharge cutoff voltage, discharge time, and maximum charging current, a battery aging feature sequence is constructed. This sequence can accurately predict the early aging (such as changes in terminal voltage change rate and average internal resistance), mid-term aging (such as significant decay of constant voltage charging time and constant current charging time), and late aging (such as significant abnormalities in discharge cutoff voltage and maximum charging current) of the target battery. It covers the aging state prediction of the target battery throughout its entire life cycle and can provide stable and reliable feature inputs throughout the entire battery life cycle, ensuring prediction accuracy.

[0039] In the embodiments of this application, considering that there may be redundant features in constant voltage charging time, constant current charging time, average internal resistance, terminal voltage change rate, discharge cutoff voltage, discharge time and maximum charging current, some redundant features can be eliminated according to the correlation between each feature and the battery aging state, so as to improve the prediction efficiency while ensuring the prediction accuracy.

[0040] In one example, a battery aging feature sequence is constructed based on the battery aging characteristics at each time point, including: Based on the pre-built aging state correlation, some battery aging features are removed, and a battery aging feature sequence is constructed based on the remaining battery aging features at each time point.

[0041] The aging state correlation is used to describe the battery aging characteristics associated with the battery aging state. For example, the correlation degree between each battery aging characteristic and the aging state can be determined in advance, and the aging state correlation can be constructed based on the correlation degree.

[0042] Specifically, based on the pre-built aging state correlation, battery aging characteristics associated with the battery aging state can be filtered out, thereby retaining these battery aging characteristics and eliminating other battery aging characteristics.

[0043] After removing some battery aging features, a battery aging feature sequence can be constructed based on the remaining battery aging features at each time point. This example demonstrates how redundant features can be removed, significantly improving prediction efficiency while maintaining accuracy.

[0044] In some implementations, the construction of aging state associations includes the following steps: Step 21: Obtain sample data and extract the feature sequences of each sample and the corresponding sample aging state sequences from the sample data; Step 22: For each sample feature sequence, determine the correlation between the sample feature sequence and the sample aging state sequence; Step 23: Construct the aging state association relationship based on the correlation degree corresponding to the feature sequences of each sample.

[0045] In step 21, sample data can be obtained by collecting battery operation information during the cyclic charge-discharge phase based on experimental data or real-time monitoring data. To ensure the accuracy and reliability of subsequent feature extraction and analysis, the sample data can be preprocessed, including noise reduction, missing value imputation, outlier removal, and data smoothing.

[0046] Furthermore, the values ​​of each battery aging feature at different times and the corresponding battery aging state values ​​at different times can be extracted from the sample data. Each sample feature sequence is constructed based on the values ​​of each battery aging feature at different times, and a sample aging state sequence is constructed based on the values ​​of the battery aging state at different times. The sample feature sequence consists of sample aging features from multiple times, and the sample aging state sequence consists of sample aging states from multiple times.

[0047] After obtaining the feature sequences of each sample and the corresponding sample aging state sequences, in order to improve the accuracy of data analysis, feature dimensionality reduction and centering can be performed on the feature sequences of each sample.

[0048] Furthermore, in step 22, for each sample feature sequence, the correlation between the sample feature sequence and the sample aging state sequence can be calculated. For example, the correlation between the sample feature sequence and the sample aging state sequence can be determined by grey relational analysis, or by Pearson correlation coefficient analysis or Spearman rank correlation coefficient analysis.

[0049] Regarding step 22 above, in one specific implementation, the correlation degree includes curve trend similarity. Determining the correlation degree between the sample feature sequence and the sample aging state sequence includes the following steps: Step 221: Normalize the sample feature sequence and the sample aging state sequence respectively, and determine the absolute difference sequence between the normalized sample feature sequence and the sample aging state sequence. Step 222: Determine the minimum and maximum differences in the absolute difference sequence, and determine the correlation coefficient of each absolute difference in the absolute difference sequence based on the minimum and maximum differences; Step 223: Determine the curve trend similarity between the sample feature sequence and the sample aging state sequence based on the correlation coefficient of each absolute difference in the absolute difference sequence.

[0050] In step 221, the battery aging feature sequence and the battery aging state sequence can be normalized to convert the values ​​of the battery aging features and the battery aging states in the sequence to the range [0,1]. For example, the maximum value in the sequence can be searched, and then the ratio between each value in the sequence and the maximum value can be calculated, with the ratio used as the normalization result.

[0051] Furthermore, after normalization is completed, the absolute difference between the sample feature sequence and the sample aging state sequence at corresponding times can be calculated, thereby obtaining the absolute difference sequence.

[0052] Furthermore, in step 222, the minimum and maximum differences in the absolute difference sequence can be searched, i.e., the global minimum and global maximum values ​​can be searched. Then, the correlation coefficient of each absolute difference in the absolute difference sequence can be calculated using the minimum and maximum differences, as shown in the following formula: ; In the formula, The k-th absolute difference in the absolute difference sequence. To be the minimum difference, For the maximum difference, The correlation coefficient is the correlation coefficient of the k-th absolute difference. The resolution coefficient, It is usually taken as 0.5. The smaller the size, the stronger the resolution.

[0053] After obtaining the correlation coefficient, in step 223, the correlation coefficients of each absolute difference in the absolute difference sequence can be averaged, or a weighted sum of the correlation coefficients of each absolute difference in the absolute difference sequence can be performed to obtain the curve trend similarity. The curve trend similarity reflects the degree of similarity between the curve trends of the sample feature sequence and the sample aging state sequence over time.

[0054] Steps 221-223 above determine the correlation between the sample feature sequence and the sample aging state sequence by analyzing the similarity of the curve trends between each sample feature sequence and the sample aging state sequence. This can measure the correlation between the two parameters from the curve trends and ensure the accuracy of the correlation analysis.

[0055] Regarding step 22 above, in another specific implementation, the correlation degree includes numerical linear similarity and ordinal monotonic similarity. Determining the correlation degree between the sample feature sequence and the sample aging state sequence includes the following steps: Step 224: Determine the mean of the feature sequence of the sample feature sequence and the mean of the state sequence of the sample aging state sequence, and determine the sum of the products of the deviations of the feature sequence mean and the mean of the state sequence from the mean, as well as the sum of the squares of the deviations from the mean; Step 225: Determine the numerical linear similarity between the sample feature sequence and the sample aging state sequence based on the sum of the products of deviations from the mean and the sum of squares of deviations from the mean. Step 226: Sort the sample feature sequence and the sample aging state sequence numerically, and determine the rank difference between the sample feature sequence and the sample aging state sequence based on the sorting results. Step 227: Determine the sorted monotonic similarity between the sample feature sequence and the sample aging state sequence based on the rank difference value.

[0056] In step 224, the mean of each value in the sample feature sequence can be calculated to obtain the mean of the feature sequence, and the mean of each value in the sample aging state sequence can be calculated to obtain the mean of the state sequence.

[0057] Furthermore, in step 224, the sum of the products of the deviations from the mean of the feature sequence and the mean of the state sequence can be calculated. That is, the difference between each value in the sample feature sequence and the mean of the feature sequence, and the difference between each value in the sample aging state sequence and the mean of the state sequence are calculated. Then, the product of the two differences is calculated and the products obtained at all time points are summed. The specific calculation process can be referred to the following formula: ; In the formula, It is the sum of the products of the differences from the mean. Let be the value at time i in the sample feature sequence. Let be the value at time i in the sample aging state sequence. The mean of the characteristic sequence, The mean of the state sequence. This represents the summation of multiplications over all time points.

[0058] Furthermore, in step 224, the sum of squared deviations from the mean of the feature sequence and the mean of the state sequence can be calculated. This involves summing the squared differences between each value in the sample feature sequence and the mean of the feature sequence, and summing the squared differences between each value in the sample aging state sequence and the mean of the state sequence. The square roots of these two sums are then calculated separately and multiplied. The specific calculation process can be found in the following formula: ; In the formula, The sum of squared deviations from the mean. This represents the summation of the squared differences over all time points.

[0059] Furthermore, in step 225, the product of deviations from the mean can be divided by the sum of squared deviations from the mean to obtain the numerical linear similarity. The numerical linear similarity describes the degree of linear correlation between the sample feature sequence and the sample aging state sequence, and its value range can be [-1, 1], where -1 represents a perfect negative correlation, 1 represents a perfect positive correlation, and 0 represents no correlation.

[0060] Furthermore, in step 226, the sample feature sequence and the sample aging state sequence can be numerically sorted separately, and the rank of each value in the sorting result can be determined. For example, they can be sorted in ascending order and assigned ranks 1, 2, 3, ... according to the sorting result. Then, the difference between the ranks of the values ​​in the sample feature sequence and the sample aging state sequence at the same time can be calculated to obtain the rank difference value.

[0061] Furthermore, in step 227, the sum of squares of all rank differences can be calculated to obtain the sorted monotonic similarity. The sorted monotonic similarity describes the degree of monotonic correlation between the sample feature sequence and the sample aging state sequence, and its value range can be [-1, 1], where -1 indicates a completely negative correlation, 1 indicates a completely positive correlation, and 0 indicates no correlation.

[0062] Steps 224-227 above determine the correlation between the sample feature sequence and the sample aging state sequence by analyzing the numerical linear similarity and sorting monotonic similarity between each sample feature sequence and the sample aging state sequence. This can measure the correlation between the two parameters from the numerical linear change and sorting monotonicity, ensuring the accuracy of the correlation analysis.

[0063] After calculating the correlation between the sample feature sequence and the sample aging state sequence, some battery aging features can be selected from all battery aging features as battery aging features associated with the battery aging state based on each correlation, thus obtaining the aging state correlation relationship.

[0064] For example, battery aging features with curve trend similarity greater than a preset curve similarity threshold can be selected; or, battery aging features with both numerical linear similarity and sorted monotonic similarity higher than the corresponding thresholds can be selected; or, the first p battery aging features can be selected in descending order of curve trend similarity.

[0065] S120. Input the battery aging feature sequence into the pre-trained aging state prediction model, wherein the aging state prediction model includes a convolutional module, a bidirectional long short-term memory module, and an attention module.

[0066] Specifically, after obtaining the battery aging feature sequence, this sequence can be input into a pre-trained aging state prediction model. This aging state prediction model consists of a convolutional module, a bidirectional long short-term memory module, and an attention module. The convolutional module, bidirectional long short-term memory module, and attention module are connected sequentially.

[0067] In this embodiment of the application, the training process of the aging state prediction model is as follows: 1. Collect battery operating information during the cyclic charge and discharge phase based on experimental data or real-time monitoring data to obtain sample data; and perform preprocessing operations on the collected sample data, such as noise reduction, missing value imputation, outlier removal, and data smoothing, to ensure the accuracy and reliability of subsequent feature extraction and analysis. 2. Feature extraction is performed on the sample data to obtain the feature sequence of each sample and the corresponding sample aging state sequence. Among them, the ratio of the current cycle battery capacity to the initial rated capacity can be defined as the battery aging state, which directly reflects the degree of battery degradation. ; In the formula, This indicates that the battery is in an aging state. This is the initial rated capacity. This represents the battery capacity for the current cycle. 3. Further process the extracted feature data, including feature dimensionality reduction, centering, and normalization, to improve the accuracy and effectiveness of data analysis; 4. Based on the correlation between aging states, retain the feature data associated with battery aging characteristics, and organize the feature data in multiple consecutive cycles into time series samples according to the sliding time window. Use the sample aging state at the corresponding time as the supervised learning label, construct a data pair with one-to-one correspondence between input and output, and obtain the dataset. 5. Divide the constructed dataset into training and test sets according to a certain ratio for model training and performance evaluation to ensure that the model has good fitting and generalization capabilities; or, divide it into training, validation, and test sets according to a certain ratio (e.g., 70%:10%:20%), where the training set is used to train the model, enabling it to learn the mapping relationship between input data and battery aging characteristics; the validation set is used to fine-tune model hyperparameters (e.g., learning rate, number of network layers) to prevent overfitting; and the test set is used to evaluate the model's generalization performance and verify its practical application capabilities. 6. Build an initial model that includes a convolutional module, a bidirectional long short-term memory module, and an attention module. Train the initial model end-to-end using the training set. Calculate the loss function based on the prediction results output by the initial model and the corresponding labels. Adjust the weight parameters in the initial model using the calculated value of the loss function until the convergence condition is met. 7. Save the network structure and weight parameters in the initial model, and use the test set to verify the model performance in order to evaluate the accuracy and robustness of the trained model. If the verification is successful, the aging state prediction model can be obtained.

[0068] In the aforementioned process, there are several ways to obtain representative parameters of battery aging, such as Principal Component Analysis (PCA), clustering algorithms, and wavelet transform. PCA maps high-dimensional data to a low-dimensional space through linear transformation, which can be used to extract the main change patterns in temperature data. By extracting principal components from various charge and discharge data of the battery, time-series features can be obtained, reflecting the heat generation and transfer capabilities of each modeling node under different operating conditions. Clustering algorithms can divide the modeling nodes in the battery pack into several categories, identifying the characteristic changes of different battery packs or nodes during the aging process. Wavelet transform has the advantage of simultaneously capturing time and frequency domain features, making it suitable for analyzing irregular or transient changes in battery aging. Through wavelet transform, the change patterns of battery temperature and voltage features at different scales and over time can be extracted, thereby deeply exploring the subtle changes in the battery aging process.

[0069] S130. The local temporal features in the battery aging feature sequence are determined by the convolution module, and the bidirectional long short-term memory module is used to determine the bidirectional dependency of the local temporal features in the time dimension, so as to obtain the temporal dynamic change features.

[0070] Specifically, after the battery aging feature sequence of the target battery enters the pre-trained aging state prediction model, it is first processed by the convolution module. The convolution module can be composed of multiple convolutional layers to extract local temporal features in the battery aging feature sequence and improve feature representation ability.

[0071] The convolution module can extract local temporal features during battery aging through convolution operations, such as charging and discharging current, voltage fluctuations, and temperature changes. These local temporal features can reflect the performance degradation patterns of different battery cells during battery aging and capture the performance correlations between different cells within the battery pack, such as capacity changes and increased internal resistance of adjacent cells.

[0072] After the convolutional module completes its processing, it can output local temporal features to the bidirectional long short-term memory module. This allows the bidirectional long short-term memory module to capture the bidirectional dependencies of features in the time dimension, thereby enhancing the ability of the aging state prediction model to model aging evolution trends.

[0073] The bidirectional long short-term memory (LSTM) module captures long-term dependencies in the battery aging process through forward and reverse time-series processing. Forward time-series processing handles data from the past to the present time step, while reverse time-series processing handles data from the future to the present time step. The combination of these two approaches provides a more comprehensive reflection of battery aging trends. The bidirectional LSM module effectively captures the dynamic changes in time-series characteristics during battery aging, such as the gradual decrease in battery capacity and the gradual increase in internal resistance, thereby enhancing the predictive ability of the aging state prediction model for the battery aging process.

[0074] In one specific implementation, the bidirectional dependency relationship of local temporal features in the time dimension is determined by a bidirectional long short-term memory module to obtain the dynamic temporal change features, including the following steps: Step 31: Through the forward processing unit in the bidirectional long short-term memory module, the data from the first time step in the local temporal features is processed to the data from the last time step to obtain the preceding temporal information; Step 32: Through the reverse processing unit in the bidirectional long short-term memory module, the data from the last time step in the local temporal features is processed to the data from the first time step to obtain the subsequent temporal information; Step 33: Determine the dynamic change characteristics of the time series based on the preceding and following time series information.

[0075] The bidirectional long short-term memory module consists of a forward processing unit and a reverse processing unit. The forward processing unit captures past information from the first element to the last element of the local temporal feature, while the reverse processing unit captures future information from the last element to the first element of the local temporal feature.

[0076] Specifically, in step 31, the forward processing unit can process the data from the first time step in the local time series feature to the last time step to obtain the preceding time series information. In step 32, the reverse processing unit can process the data from the last time step in the local time series feature to the first time step to obtain the following time series information.

[0077] Furthermore, in step 33, the outputs from the two directions can be concatenated, that is, the preceding time series information and the following time series information can be concatenated to obtain the time series dynamic change features that simultaneously include the preceding and following texts; or, the preceding time series information and the following time series information can be summed to obtain the time series dynamic change features that simultaneously include the preceding and following texts; or, the preceding time series information and the following time series information can be averaged to obtain the time series dynamic change features that simultaneously include the preceding and following texts.

[0078] For example, Figure 2This is a schematic diagram of an aging state prediction model provided in an embodiment of this application, where Inputs represent the model's inputs, CNNs represent convolutional modules, BiLSTM represents bidirectional long short-term memory modules, Attentionmechanism represents attention modules, and Outputs represent the model's outputs. Figure 2 As shown, the battery aging feature sequence is first input into the model's convolutional module. After processing by the convolutional module, local temporal features are obtained, which include features at multiple time steps (1~t). ~ Furthermore, the local temporal features output by the convolutional module enter the bidirectional long short-term memory (LSTM) module. The forward processing units within this module can process these features in the forward order of time steps, yielding... ~ The reverse processing unit can process data in reverse order of the time steps to obtain... ~ , and thus ~ and ~ Fusion yields temporal dynamic change characteristics. ~ Finally, the temporal dynamic changes output by the bidirectional long short-term memory module enter the attention module, and through the attention module... Adjust the weights of each time step, i.e. ~ The system then integrates the time-series dynamic change features with weights to obtain the output result. .

[0079] Steps 31-33 above, through forward and reverse time series processing, can more comprehensively reflect the trend of battery aging, effectively capture the dynamic changes of time series characteristics during battery aging, and thus improve the prediction accuracy of the model.

[0080] S140. Adjust the weights of each time step in the time-series dynamic change features through the attention module to determine the current aging state of the target battery.

[0081] After obtaining the time-series dynamic change features through the bidirectional long short-term memory module, the weights of each time step in the time-series dynamic change features can be adjusted through the attention module. Then, the features of each time step in the time-series dynamic change features are weighted according to the weights of each time step, and the current aging state of the target battery is output based on the weighting result.

[0082] The attention module assigns different weights to different time steps, enabling the aging state prediction model to focus more on effective information at key time points and improve prediction accuracy. Specifically, by calculating the weights of different time steps, the attention module dynamically adjusts the focus of the aging state prediction model, enhancing its ability to identify key time steps. For example, when the battery shows significant degradation or reaches a predetermined health threshold, the aging state prediction model can more accurately capture and predict the battery's aging process.

[0083] The new energy battery aging state prediction method provided in this application obtains the charge and discharge data of the target battery, determines the battery aging feature sequence based on the charge and discharge data, and inputs the battery aging feature sequence into a pre-trained aging state prediction model. The convolution module in the model determines the local temporal features in the battery aging feature sequence, and the bidirectional long short-term memory module in the model determines the bidirectional dependency of the local temporal features in the time dimension to obtain the temporal dynamic change features. The attention module in the model adjusts the weights of each time step in the temporal dynamic change features to determine the current aging state of the target battery, thereby achieving prediction of the battery aging state. This method integrates the convolution module, the bidirectional long short-term memory module, and the attention module, which can accurately capture the local temporal features of the battery aging feature sequence, improve the feature expression ability, accurately capture the bidirectional dependency in the time dimension, enhance the model's ability to analyze battery aging trends, and enable the model to focus more on the effective information at key time points, thereby improving prediction accuracy. It does not require the construction of an electrochemical model, is low-cost and simple to implement, and is applicable to various types of batteries and complex operating conditions, ensuring the prediction stability for various complex aging behaviors.

[0084] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. For example... Figure 3 As shown, the electronic device 400 includes one or more processors 401 and memory 402.

[0085] The processor 401 may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 400 to perform desired functions.

[0086] The memory 402 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 401 may execute the program instructions to implement the new energy battery aging state prediction method of any embodiment of this application described above, and / or other desired functions. Various contents such as initial external parameters and thresholds may also be stored in the computer-readable storage medium.

[0087] In one example, the electronic device 400 may further include an input device 403 and an output device 404, these components being interconnected via a bus system and / or other forms of connection mechanisms (not shown). The input device 403 may include, for example, a keyboard, a mouse, etc. The output device 404 may output various information to the outside, including warning messages, braking force, etc. The output device 404 may include, for example, a display, a speaker, a printer, and a communication network and its connected remote output devices, etc.

[0088] Of course, for the sake of simplicity, Figure 3 Only some of the components of the electronic device 400 relevant to this application are shown in this illustration; components such as buses, input / output interfaces, etc., are omitted. In addition, the electronic device 400 may include any other suitable components depending on the specific application.

[0089] In addition to the methods and devices described above, embodiments of this application may also be computer program products, which include computer program instructions that, when executed by a processor, cause the processor to perform the steps of the new energy battery aging state prediction method provided in any embodiment of this application.

[0090] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this application. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0091] Furthermore, embodiments of this application may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the steps of the new energy battery aging state prediction method provided in any embodiment of this application.

[0092] The computer-readable storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0093] It should be noted that the terminology used in this application is for the purpose of describing specific embodiments only and is not intended to limit the scope of this application. As shown in the specification and claims of this application, unless the context clearly indicates otherwise, words such as "a," "an," "an," and / or "the" do not specifically refer to the singular and may also include the plural. The terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, or apparatus. Without further limitations, an element defined by the phrase "comprising an..." does not exclude the presence of other identical elements in the process, method, or apparatus that includes said element.

[0094] It should also be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on this application. Unless otherwise expressly specified and limited, the terms "installed," "connected," "linked," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication between two elements. For those skilled in the art, the specific meaning of the above terms in this application can be understood according to the specific circumstances.

[0095] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. The above descriptions are only preferred embodiments of this application. It should be noted that due to the limitations of written expression, while there are objectively infinite specific structures, those skilled in the art can make several improvements, modifications, or changes without departing from the principles of this application, and can also combine the above technical features in an appropriate manner. These improvements, modifications, changes, or combinations, or the direct application of the inventive concept and technical solution to other situations without modification, should all be considered within the scope of protection of this application.

Claims

1. A method for predicting the aging state of new energy batteries, characterized in that, include: Acquire charge and discharge data of the target battery, and determine the battery aging characteristic sequence based on the charge and discharge data; The battery aging feature sequence is input into a pre-trained aging state prediction model, wherein the aging state prediction model includes a convolution module, a bidirectional long short-term memory module, and an attention module. The convolution module determines the local temporal features in the battery aging feature sequence, and the bidirectional long short-term memory module determines the bidirectional dependency of the local temporal features in the time dimension to obtain the temporal dynamic change features. The current aging state of the target battery is determined by adjusting the weights of each time step in the time-series dynamic change features using the attention module.

2. The method according to claim 1, characterized in that, The battery aging characteristic sequence is determined based on the charge and discharge data, including: The battery aging characteristics at each time point are determined based on the charge and discharge data, wherein the battery aging characteristics include constant voltage charging time, constant current charging time, average internal resistance, terminal voltage change rate, discharge cutoff voltage, discharge time, and maximum charging current. A battery aging feature sequence is constructed based on the battery aging characteristics at each time point.

3. The method according to claim 2, characterized in that, A battery aging feature sequence is constructed based on the battery aging characteristics at each time point, including: Some battery aging features are removed based on a pre-built aging state association relationship, wherein the aging state association relationship is used to describe battery aging features associated with battery aging state. A battery aging feature sequence is constructed based on the remaining battery aging features at each time point.

4. The method according to claim 3, characterized in that, The construction of the aging state association includes: Acquire sample data, and extract the feature sequence of each sample and the corresponding sample aging state sequence from the sample data; For each of the sample feature sequences, determine the correlation between the sample feature sequence and the sample aging state sequence; Based on the correlation degree corresponding to each of the sample feature sequences, the aging state correlation relationship is constructed.

5. The method according to claim 4, characterized in that, The correlation includes curve trend similarity, determining the correlation between the sample feature sequence and the sample aging state sequence, including: The sample feature sequence and the sample aging state sequence are normalized respectively, and the absolute difference sequence between the normalized sample feature sequence and the sample aging state sequence is determined. Determine the minimum and maximum differences in the absolute difference sequence, and determine the correlation coefficient of each absolute difference in the absolute difference sequence based on the minimum and maximum differences; Based on the correlation coefficient of each absolute difference in the absolute difference sequence, the curve trend similarity between the sample feature sequence and the sample aging state sequence is determined.

6. The method according to claim 4, characterized in that, The correlation degree includes numerical linear similarity and ordinal monotonic similarity. Determining the correlation degree between the sample feature sequence and the sample aging state sequence includes: Determine the mean of the feature sequence of the sample feature sequence and the mean of the state sequence of the sample aging state sequence, and determine the sum of the products of the deviations from the mean of the feature sequence and the mean of the state sequence, as well as the sum of the squares of the deviations from the mean; The numerical linear similarity between the sample feature sequence and the sample aging state sequence is determined based on the sum of the products of the deviations from the mean and the sum of the squares of the deviations from the mean. The sample feature sequence and the sample aging state sequence are numerically sorted respectively, and the rank difference value of the sample feature sequence and the sample aging state sequence is determined according to the sorting results. The sorting monotonic similarity between the sample feature sequence and the sample aging state sequence is determined based on the rank difference value.

7. The method according to claim 2, characterized in that, Based on the charge and discharge data, the battery aging characteristics at each time point are determined, including: Based on the charge and discharge data, determine the measured voltage and current values ​​for each charging cycle, as well as the voltage change rate, internal resistance, and measured current values ​​for each discharging cycle. Based on the measured voltage and current values ​​during each charging cycle, as well as the measured voltage change rate, internal resistance, and current values ​​during each discharging cycle, the battery aging characteristics at each time point are determined.

8. The method according to claim 1, characterized in that, The bidirectional long short-term memory module determines the bidirectional dependency relationship of the local temporal features in the time dimension, thereby obtaining the dynamic temporal change features, including: The forward processing unit in the bidirectional long short-term memory module processes the data from the first time step to the last time step in the local time sequence feature to obtain the preceding time sequence information. The reverse processing unit in the bidirectional long short-term memory module processes the data from the last time step in the local time sequence feature to the data from the first time step, thus obtaining the subsequent time sequence information. The dynamic change characteristics of the time series are determined based on the preceding and following time series information.

9. An electronic device, characterized in that, The electronic device includes: Processor and memory; The processor executes the steps of the new energy battery aging state prediction method as described in any one of claims 1 to 8 by calling the program or instructions stored in the memory.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program or instructions that cause a computer to perform the steps of the new energy battery aging state prediction method as described in any one of claims 1 to 8.

Images