Battery life prediction method and device, terminal equipment and storage medium

Abstract

The invention is suitable for the technical field of battery life prediction, and provides a battery life prediction method, terminal equipment and a storage medium, and the method comprises the steps: obtaining first feature data of a to-be-detected battery; the first characteristic data is collected working condition data of the battery in a charging and discharging process; extracting second feature data related to life attenuation of the battery according to the first feature data; wherein the dimension of the second feature data is lower than that of the first feature data; and predicting the service life of the to-be-detected battery according to the second feature data. The method can improve the prediction precision of the battery life.

Description

Technical Field

[0001] The present application belongs to the technical field of battery life prediction, and in particular relates to a battery life prediction method, terminal device and storage medium. Background Art

[0002] With the widespread adoption of lithium-ion batteries in electric vehicles, energy storage systems, and other fields, accurately predicting the remaining battery life is crucial for battery health management, maintenance decisions, and system optimization. The battery degradation process involves dynamic changes in multiple dimensions, including current, voltage, capacity, and internal resistance. As the number of feature dimensions increases, the redundancy between these features significantly impacts model stability.

[0003] At the same time, batteries have multi-physical field coupling effects such as electricity, heat, chemistry, and mechanics. Traditional methods find it difficult to effectively quantify the synergy and conflict between features, resulting in the inability of the prediction model to fully explore the potential correlation information of multi-dimensional features, affecting the prediction accuracy of the remaining battery life. Summary of the Invention

[0004] The embodiments of the present application provide a battery life prediction method, apparatus, terminal device, and storage medium, which can improve the prediction accuracy of battery life.

[0005] In a first aspect, an embodiment of the present application provides a battery life prediction method, comprising:

[0006] Acquire first characteristic data of the battery to be tested; the first characteristic data is collected operating condition data of the battery during a charge and discharge process;

[0007] Extracting second feature data related to battery life degradation based on the first feature data; wherein the dimension of the second feature data is lower than the dimension of the first feature data;

[0008] The life of the battery to be tested is predicted according to the second characteristic data.

[0009] In an embodiment of the present application, as the battery completes a complete charge-discharge cycle (from the start of charging to the end of discharging), operating data such as current, voltage, and charge-discharge capacity are recorded in real time. Because the collected operating data contains high-dimensional redundant data, dimensionality reduction is performed on the collected operating data to remove redundant information, thereby extracting key aging characteristic data corresponding to the battery. Finally, the battery life is predicted based on this key aging characteristic data. In this method, by performing dimensionality reduction on high-dimensional data, multiple significant key aging characteristics can be obtained, which can improve the accuracy of battery life prediction.

[0010] In a possible implementation of the first aspect, obtaining first characteristic data of a battery to be tested includes:

[0011] Collecting first sub-data and second sub-data; wherein the first sub-data is the real-time working condition of the battery during the charging and discharging process, and the second sub-data is the statistical value of the working condition data of the battery during one charging and discharging process;

[0012] The first sub-data and the second sub-data are fused to obtain first feature data.

[0013] In an embodiment of the present application, by integrating the real-time charging and discharging conditions of the battery (first sub-data) and the statistical values ​​of the single charging and discharging conditions (second sub-data), the global features (such as the cycle average capacity decay rate) can be integrated while retaining the instantaneous details of the decay process (such as the voltage mutation point), thereby obtaining the first characteristic data that more comprehensively characterizes the health status of the battery, laying a data foundation for improving the accuracy of life prediction and the ability to analyze the decay mechanism.

[0014] In a possible implementation of the first aspect, extracting second feature data related to battery life degradation based on the first feature data includes:

[0015] Acquire a local change characteristic during the battery life degradation process according to the first characteristic data to obtain third characteristic data;

[0016] Obtaining a time series dependency relationship between various operating condition data during the battery life degradation process according to the third characteristic data to obtain fourth characteristic data;

[0017] Second feature data related to battery life degradation is extracted based on the fourth feature data.

[0018] In an embodiment of the present application, by sequentially extracting local change features from the first feature data to obtain third feature data, obtaining timing dependencies from the third feature data to obtain fourth feature data, and then screening out second feature data related to life decay from the fourth feature data, it is possible to achieve progressive feature extraction from capturing instantaneous anomalies in battery operating condition data to mining full-cycle decay laws, thereby obtaining key features that combine local details with global timing correlations, effectively improving the accuracy of battery life prediction and the depth of decay mechanism analysis.

[0019] In a possible implementation of the first aspect, extracting second feature data related to battery life degradation based on the fourth feature data includes:

[0020] Obtaining a first parameter and a second parameter corresponding to the fourth characteristic data; wherein the first parameter is the linear correlation degree between the fourth characteristic data and the battery life, and the second parameter is the importance degree of the fourth characteristic data in affecting the battery life;

[0021] The second characteristic data is filtered out from the fourth characteristic data according to the first parameter and the second parameter.

[0022] In an embodiment of the present application, by combining the fourth characteristic data with the linear correlation degree of battery life (first parameter) and the importance of influence (second parameter) for double screening, features that are only statistically relevant but have weak actual contributions can be accurately eliminated, and key features with both strong correlation and high contribution can be retained, thereby ensuring that the second characteristic data can reflect both the linear correlation of life attenuation and the actual physical impact, effectively improving the accuracy and interpretability of the battery life prediction model.

[0023] In a possible implementation of the first aspect, filtering out the second feature data from the fourth feature data according to the first parameter and the second parameter includes:

[0024] Acquire fifth feature data that satisfies a first condition from the fourth feature data, wherein the fifth feature data satisfies the first condition if an absolute value of a first parameter corresponding to the fifth feature data is greater than a first preset threshold;

[0025] Second characteristic data satisfying a second condition is obtained from the fifth characteristic data, wherein the second characteristic data satisfies the second condition if a second parameter corresponding to the second characteristic data is greater than a second preset threshold.

[0026] In an embodiment of the present application, by first screening out the fifth feature data that is strongly linearly correlated with the battery life using the absolute value of the first parameter greater than a preset threshold, and then screening out the second feature data that is important for the lifespan using the second parameter greater than the preset threshold, a progressive feature screening from statistical correlation to actual contribution is achieved, which not only avoids feature redundancy caused by relying solely on linear correlation, but also eliminates pseudo features with high correlation but low impact, ensuring that the final second feature data has both strong correlation and high explanatory power, effectively improving the accuracy and robustness of the battery life prediction model.

[0027] In a possible implementation of the first aspect, the method further includes:

[0028] Obtain a trained preset model; wherein the preset model includes a feature extraction network and a prediction network; the feature extraction network is used to extract second feature data related to the battery life attenuation based on the first feature data; the prediction network is used to predict the life of the battery to be tested based on the second feature data.

[0029] In an embodiment of the present application, through the collaborative work of the feature extraction network and the prediction network in the trained preset model, key features that are both relevant and important can be automatically extracted from the battery operating condition data and the life can be accurately predicted, realizing end-to-end automated analysis from raw data to life prediction, and providing efficient and accurate decision support for battery health management.

[0030] In a possible implementation of the first aspect, the feature extraction network of the preset model includes a convolutional neural network and a long short-term memory network;

[0031] The convolutional neural network is used to obtain local change characteristics during the battery life degradation process according to the first feature data to obtain third feature data;

[0032] The long short-term memory network is used to obtain the temporal dependency between various operating condition data during the battery life degradation process according to the third characteristic data to obtain the fourth characteristic data.

[0033] In an embodiment of the present application, through the synergistic effect of the convolutional neural network and the long short-term memory network in the feature extraction network, the convolutional neural network can be used to capture the local instantaneous change characteristics during the battery life degradation process, and the long short-term memory network can be used to mine the temporal dependency relationship between different working condition data, thereby realizing deep extraction of battery degradation features from local space to global time, and providing multi-dimensional feature support for accurate prediction of battery life.

[0034] In a possible implementation of the first aspect, the step of training the preset model includes:

[0035] Acquire multiple fifth characteristic data of the training battery; the fifth characteristic data is historical operating condition data collected from the training battery during multiple charge and discharge processes;

[0036] Inputting the fifth feature data into a feature extraction network in a preset model to be trained to extract sixth feature data related to the life attenuation of the training battery;

[0037] Inputting the sixth feature data into a prediction network in a preset model to be trained to obtain a first loss;

[0038] When the first loss does not reach the convergence condition, the parameters of the preset model to be trained continue to be iteratively trained according to multiple fifth feature data until the first loss reaches the convergence condition. The training is then terminated, and the preset model to be trained with converged output parameters is the pre-trained preset model.

[0039] In an embodiment of the present application, by inputting the historical operating condition data (fifth characteristic data) of multiple charge and discharge of the training battery into the feature extraction network and prediction network of the preset model, combined with the iterative optimization mechanism of the first loss, end-to-end model parameter fine-tuning from the original operating condition data to life prediction is achieved, so that the preset model can adaptively capture the local characteristics and temporal dependencies of battery degradation, and finally output a pre-trained model with both high prediction accuracy and robustness.

[0040] In a second aspect, an embodiment of the present application provides a terminal device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, a battery life prediction method as described in any one of the first aspects above is implemented.

[0041] In a third aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the battery life prediction method as described in any one of the first aspects above.

[0042] In a fourth aspect, an embodiment of the present application provides a computer program product, which, when executed on a terminal device, enables the terminal device to execute any one of the battery life prediction methods of the first aspect.

[0043] It can be understood that the beneficial effects of the second to fourth aspects mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0045] Figure 1 1 is a flow chart of a method for predicting remaining battery life provided by an embodiment of the present application;

[0046] Figure 2 This is a schematic diagram of a process for obtaining first characteristic data provided in an embodiment of the present application;

[0047] Figure 3 This is a schematic diagram of the process of obtaining the second characteristic data provided in the embodiment of the present application Figure 1 ;

[0048] Figure 4 This is a schematic diagram of the process of obtaining the second characteristic data provided in the embodiment of the present application Figure 2 ;

[0049] Figure 5 This is a flow chart of the training preset model provided in the embodiment of the present application;

[0050] Figure 6 This is a data comparison and analysis chart of the maximum discharge capacity of the battery provided in the embodiment of the present application versus the number of cycles;

[0051] Figure 7 This is a data change analysis diagram of the voltage-capacity difference provided in an embodiment of the present application;

[0052] Figure 8 This is a data analysis chart showing changes in battery internal resistance versus cycle number provided in an embodiment of the present application;

[0053] Figure 9 This is an analysis chart of the importance of the battery's original characteristic data versus the internal resistance change data provided by an embodiment of the present application;

[0054] Figure 10 This is a schematic diagram of a high-dimensional aging feature correlation heat map provided in an embodiment of the present application;

[0055] Figure 11 This is a schematic diagram of a heat map of the correlation between low-dimensional aging characteristics of different batteries provided in the embodiments of the present application;

[0056] Figure 12 Schematic diagram of the contribution of low-dimensional aging characteristics of different batteries provided in the embodiments of the present application;

[0057] Figure 13 This is a schematic diagram of the overall structure of the battery life prediction method provided by the embodiment of the present application;

[0058] Figure 14 This is a schematic diagram comparing the predicted value of the remaining battery life and the actual value provided by the embodiment of the present application;

[0059] Figure 15 This is a structural diagram of the terminal device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0060] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0061] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0062] It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0063] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0064] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0065] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with the embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in yet other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized.

[0066] With the widespread adoption of lithium-ion batteries in electric vehicles, energy storage systems, and other fields, accurately predicting the remaining battery life is crucial for battery health management, maintenance decisions, and system optimization. The battery degradation process involves dynamic changes in multiple dimensions, including current, voltage, capacity, and internal resistance. As the number of feature dimensions increases, the redundancy between these features significantly impacts model stability.

[0067] At the same time, batteries have multi-physical field coupling effects such as electricity, heat, chemistry, and mechanics. Traditional methods find it difficult to effectively quantify the synergy and conflict between features, resulting in the inability of the prediction model to fully explore the potential correlation information of multi-dimensional features, affecting the prediction accuracy of the remaining battery life.

[0068] In order to solve the problems in the above-mentioned related technologies, the present application provides a battery life prediction method based on a method of decoupling deep convolutional timing networks from multi-physical field coupling features. By collecting instant and cumulative features such as current and voltage in the battery cycle and fusing them into a comprehensive matrix, after standardization and principal component analysis dimensionality reduction, 177 high-dimensional decay features are extracted using a hybrid model combining convolutional neural networks and long-short-term memory networks. Then, six key features that are strongly correlated with the remaining life are screened out using the Pearson correlation coefficient and Shapley Additive exPlanations (AHAP) value. Finally, a feedforward neural network is used for modeling and prediction. The effectiveness of this scheme has been verified in 16,000 cycle data, which solves the problems of feature redundancy and insufficient quantification of multi-physical field coupling effects in traditional methods, and improves the prediction accuracy of battery life.

[0069] See also Figure 1 , is a flow chart of a method for predicting remaining battery life provided in an embodiment of the present application. As an example and not a limitation, the method may include the following steps:

[0070] S101, obtaining first characteristic data of a battery to be tested; the first characteristic data is collected operating condition data of the battery during a charge and discharge process.

[0071] In the embodiment of the present application, the operating condition data, i.e., the first characteristic data, refers to the electrical parameters generated in real time by the lithium-ion battery during the charging and discharging process, including voltage (V), current (A), internal resistance (mΩ), and charge and discharge capacity (Ah), etc.

[0072] Specifically, voltage, current and other sensors can be used to collect operating condition data of lithium-ion batteries at a certain frequency. During the collection process, it is necessary to focus on monitoring the late stage of constant current charging (voltage close to the cutoff value) and the constant voltage charging stage (current continues to decay), because the risk of thermal runaway is higher in this stage. During the discharge process, it is necessary to pay special attention to large current discharge (such as when electric vehicles accelerate) and deep discharge (voltage close to the cutoff value), which can easily expose internal defects of the battery.

[0073] In one embodiment, see Figure 2 , is a schematic diagram of the process of obtaining the first feature data provided in an embodiment of the present application, such as Figure 2 As shown, step S101 includes:

[0074] S201, collecting first sub-data and second sub-data; wherein the first sub-data is the real-time working condition of the battery during the charging and discharging process, and the second sub-data is the statistical value of the working condition data of the battery during one charging and discharging process.

[0075] In the embodiment of the present application, the first sub-data is the electrical parameters of the battery that change in real time during the charging and discharging process, reflecting the dynamic state within a single cycle, and belongs to the category of "instant features", such as the real-time voltage of the battery during one charge and discharge (such as the voltage fluctuation per second from 3.7V to 4.2V during charging), real-time current (such as the current is 1.5A at the beginning of discharge and drops to 0.8A in the later stage), real-time capacity release / charge amount (such as 500mAh has been released at the 10th minute of discharge) and real-time change of internal resistance (such as the internal resistance rises from 20mΩ to 25mΩ in the later stage of charging).

[0076] The second sub-data is a statistical summary of the operating data during a single charge and discharge process, reflecting the overall characteristics of a single cycle, which belongs to the "cumulative characteristics", such as the total charge / discharge capacity of the battery during a single charge and discharge process (such as the total discharge capacity of this time is 2000mAh), maximum / minimum voltage / current (such as the maximum charging voltage of 4.2V and the minimum discharge voltage of 3.2V), average voltage / current (such as the average discharge current of 1.2A) and voltage-capacity difference (the integral difference between the discharge voltage and capacity of this cycle, reflecting the polarization effect), etc.

[0077] Real-time sampling at a frequency of 10Hz-100Hz can be performed through high-precision sensors (such as Hall current sensors and voltage sampling circuits) and stored as time series data (such as CSV format, where each row corresponds to a timestamp + parameter value), namely the first sub-data. The second sub-data can perform statistical calculations on the first sub-data, for example: traversing all real-time voltage values ​​and taking extreme values ​​to obtain the maximum or minimum voltage value, current value, etc.

[0078] S202: Fusing the first sub-data and the second sub-data to obtain first feature data.

[0079] In the embodiment of the present application, the first sub-data is second-level dynamic time series data (such as real-time fluctuations in voltage and current), reflecting the instantaneous state within a single cycle. The second sub-data is the global statistical characteristics of a single cycle (such as average voltage and total capacity), which characterizes the overall performance of the cycle. Combining "dynamic details" with "global summary" forms a more complete single-cycle feature representation. The technical goal of "comprehensively characterizing the battery degradation trajectory through multi-feature fusion" is achieved to avoid information loss of a single data type.

[0080] Specifically, the first sub-data (i.e., the instant feature) can be formed into a time series matrix (such as 1000 time points × 3 columns: time, current, and voltage), and the second sub-data (i.e., the cumulative feature) can be formed into a one-dimensional vector (such as the cumulative discharge capacity of the number of cycles and the cumulative growth of the internal resistance). Then, the two can be aligned based on the time scale to ensure that the cycle number of the instant feature is consistent with the cycle number of the cumulative feature (such as the instant feature of the nth cycle corresponds to the cumulative feature of the nth cycle). Finally, based on the feature dimension, they are integrated into a fused feature, i.e., the first feature data.

[0081] In the above method, by integrating the real-time charging and discharging conditions of the battery (first sub-data) and the statistical values ​​of the single charging and discharging conditions (second sub-data), the global features (such as the cycle average capacity decay rate) can be integrated while retaining the instantaneous details of the degradation process (such as the voltage mutation point), thereby obtaining the first characteristic data that more comprehensively characterizes the health status of the battery, laying a data foundation for improving the accuracy of life prediction and the ability to analyze the degradation mechanism.

[0082] S102 , extracting second feature data related to battery life degradation based on the first feature data; wherein the dimension of the second feature data is lower than the dimension of the first feature data.

[0083] In the embodiment of the present application, the first feature data is high-dimensional feature data after feature fusion, which contains a large amount of redundant and coupled information. In order to avoid feature interference from a large amount of redundant data, low-dimensional features that are strongly correlated with battery life degradation need to be screened out from the first feature data. By reducing the dimension and screening, redundancy is eliminated and the core degradation driving factors are focused.

[0084] In one embodiment, see Figure 3 , is a schematic diagram of the process of obtaining the second feature data provided in the embodiment of the present application Figure 1 ,like Figure 3 As shown, step S102 includes:

[0085] S301 , obtaining local change characteristics during the battery life degradation process according to the first characteristic data to obtain third characteristic data.

[0086] In an embodiment of the present application, the first feature data is extracted or screened by constructing a feature extraction model, such as using a convolutional neural network-long short-term memory network (CNN-LSTM) as a feature extraction model. The CNN in the feature extraction model slides the convolution kernel to scan the battery feature matrix (i.e., the first feature data such as voltage-time series) to extract the instantaneous anomalies and local patterns (local change characteristics) of the battery life decay process within a single cycle, such as the inflection point of the voltage curve or the peak of the current waveform (third feature data).

[0087] Specifically, the 3×3 convolution kernel in the CNN network extracts local features, such as identifying the mutation pattern of "the voltage drops by 0.2V in the 100th second". The pooling layer in the CNN network performs maximum pooling to compress the data dimension and retain key local features (such as extracting 1 maximum voltage change rate every 100 time points) to extract the third feature data.

[0088] S302 , obtaining a time sequence dependency relationship between various operating condition data during the battery life degradation process according to the third characteristic data to obtain fourth characteristic data.

[0089] In an embodiment of the present application, the third feature data is the local spatial feature of battery degradation extracted by CNN (such as the voltage drop point and current mutation pattern within a single cycle), which reflects the instantaneous anomaly of a single cycle and belongs to the "spatial local feature". The data is transmitted to LSTM to learn the dependencies between long-term data. Through its memory unit, LSTM can effectively capture the state trends and changes of the battery life degradation process, and finally combine with CNN to output 177 (obtained from experimental data) high-dimensional degradation features related to battery degradation (i.e., the fourth feature data).

[0090] S303: Extract second characteristic data related to battery life degradation based on the fourth characteristic data.

[0091] In an embodiment of the present application, since the fourth feature data is high-dimensional feature data, in order to improve the prediction accuracy of battery life, the principal component analysis method can be used to extract the key aging feature data that has a greater impact on battery life, namely the second feature data, from the first feature data.

[0092] In the above method, by extracting local change features from the first feature data in sequence to obtain the third feature data, obtaining the timing dependency from the third feature data to obtain the fourth feature data, and then screening out the second feature data related to life decay from the fourth feature data, it is possible to achieve progressive feature extraction from capturing instantaneous anomalies in battery operating condition data to mining full-cycle decay laws, thereby obtaining key features that combine local details with global timing correlations, effectively improving the accuracy of battery life prediction and the depth of decay mechanism analysis.

[0093] In one embodiment, see Figure 4 , is a schematic diagram of the process of obtaining the second feature data provided in the embodiment of the present application Figure 2 ,like Figure 4 As shown, step S303 includes:

[0094] S401, obtaining a first parameter and a second parameter corresponding to fourth characteristic data; wherein the first parameter is the linear correlation degree between the fourth characteristic data and battery life, and the second parameter is the importance degree of the fourth characteristic data in affecting battery life.

[0095] In an embodiment of the present application, the first parameter may be the Pearson correlation coefficient r, which is used to measure the degree of linear correlation between the fourth characteristic data and the battery life. The value range of r is [-1, 1]. The closer its absolute value is to 1, the stronger the linear relationship. For example, if the "capacity decay rate time series" in the fourth characteristic and the remaining life r = -0.85, it indicates that for every 1% acceleration of the capacity decay, the life is reduced by approximately 85 cycles (linear approximation).

[0096] The second parameter is the SHAP value, which is used to indicate the degree of influence of the fourth feature data on the remaining battery life prediction. For example, if the SHAP value of a feature is 0.2, it indicates that the feature reduces the predicted life by 20 cycles (relative to the baseline value).

[0097] Since the calculation method of the Piel's correlation coefficient (first parameter) and the SHAP value (second parameter) is pre-configured in the CNN-LSTM model, each first parameter and second parameter corresponding to each fourth feature data will be automatically obtained by the CNN-LSTM model.

[0098] S402: Filter out second characteristic data from the fourth characteristic data according to the first parameter and the second parameter.

[0099] In an embodiment of the present application, by calculating the Pearson correlation coefficient (first parameter) and SHAP value (second parameter) of the fourth characteristic data and the battery life, the features that have a strong linear correlation with the battery life attenuation and contribute highly to the prediction results are screened out to obtain the second characteristic data for accurately predicting the battery life.

[0100] In the above method, by combining the fourth characteristic data with the linear correlation degree of battery life (first parameter) and the importance of influence (second parameter) for double screening, the features that are only statistically relevant but have weak actual contributions can be accurately eliminated, and the key features with both strong correlation and high contribution can be retained, thereby ensuring that the second characteristic data can reflect both the linear correlation of life attenuation and the actual physical impact, effectively improving the accuracy and interpretability of the battery life prediction model.

[0101] In one embodiment, step S402 includes:

[0102] Obtain fifth characteristic data that meets the first condition from the fourth characteristic data, wherein the fifth characteristic data satisfies the first condition if the absolute value of the first parameter corresponding to the fifth characteristic data is greater than a first preset threshold; obtain second characteristic data that meets the second condition from the fifth characteristic data, wherein the second characteristic data satisfies the second condition if the second parameter corresponding to the second characteristic data is greater than a second preset threshold.

[0103] In an embodiment of the present application, the screening of key aging features (second data features) from the fourth feature data can be divided into a two-layer screening process: first, based on the fourth feature data, according to whether the absolute value of the first parameter (Pearson correlation coefficient) is greater than the first preset threshold (such as 0.7), the features with strong correlation are screened out to form the fifth feature data; then from the fifth feature data, according to whether the second parameter (SHAP value) is greater than the second preset threshold (such as 0.05), the features with high contribution to battery life prediction are further screened out, and finally the second feature data closely related to battery life attenuation is obtained, providing accurate and effective feature basis for subsequent battery life prediction and analysis.

[0104] In the above method, the fifth feature data with a strong linear correlation with battery life is firstly screened out by using the absolute value of the first parameter greater than the preset threshold, and then the second feature data with an important impact on life is screened out by using the second parameter greater than the preset threshold. This realizes a progressive feature screening from statistical correlation to actual contribution. It avoids the feature redundancy caused by relying solely on linear correlation, and eliminates pseudo features with high correlation but low impact, ensuring that the final second feature data has both strong correlation and high explanatory power, effectively improving the accuracy and robustness of the battery life prediction model.

[0105] S103: Predict the life of the battery to be tested based on the second characteristic data.

[0106] In an embodiment of the present application, the second characteristic data obtained through screening (these data are screened from the fourth characteristic data through two layers of conditions, satisfying both that the absolute value of the first parameter is greater than the corresponding preset threshold and that the second parameter is greater than another preset threshold, and is closely related to the battery life attenuation) is input into a pre-trained prediction model to estimate and predict life-related indicators such as the time the battery can maintain normal operation in the future and the period during which the capacity decays to a certain degree, thereby providing data support and decision-making basis for battery use, maintenance, and related product design.

[0107] Specifically, the trained feedforward neural network can be used as a prediction model, and a nonlinear activation function (such as ReLU) can be used to adapt to the complex nonlinear mapping relationship between input features and remaining life during battery degradation. The key features after principal component dimensionality reduction (i.e., the second feature data) are input into the feedforward neural network to predict the remaining life of the battery.

[0108] In one embodiment, the method further comprises:

[0109] Obtain a trained preset model; wherein the preset model includes a feature extraction network and a prediction network; the feature extraction network is used to extract second feature data related to the battery life attenuation based on the first feature data; the prediction network is used to predict the life of the battery to be tested based on the second feature data.

[0110] In this embodiment of the present application, a specific model that has been trained is obtained. The model consists of two parts: a feature extraction network and a prediction network. The feature extraction network is a trained feature extraction model. This model acts as a "data filter." It takes the first feature data containing the battery's immediate and cumulative features as input and uses a series of dimensionality reduction and screening algorithms, such as principal component analysis, Pearson correlation coefficient, and SHAP value calculation, to extract the second feature data that is closely related to battery life degradation from the high-dimensional and complex data.

[0111] The prediction network is a trained prediction model, which acts like a "prediction engine". It receives the second feature data as input, uses a feedforward neural network and a nonlinear activation function to establish a complex mapping relationship between the feature and the battery life, and finally outputs an accurate prediction result of the remaining battery life, thereby providing a reliable basis for battery health management and maintenance decisions.

[0112] In the above method, through the collaborative work of the feature extraction network and the prediction network in the trained preset model, key features that are both relevant and important can be automatically extracted from the battery operating condition data and the life can be accurately predicted, realizing end-to-end automated analysis from raw data to life prediction, and providing efficient and accurate decision support for battery health management.

[0113] In one embodiment, the feature extraction network of the preset model includes a convolutional neural network and a long short-term memory network; wherein the convolutional neural network is used to obtain local change characteristics during the battery life decline process based on the first feature data to obtain third feature data; the long short-term memory network is used to obtain the temporal dependency between various operating condition data during the battery life decline process based on the third feature data to obtain fourth feature data.

[0114] In the embodiment of the present application, the feature extraction network in the preset model adopts an architecture that combines a convolutional neural network (CNN) with a long short-term memory network (LSTM). The CNN acts as a "microscopic detector," sliding a convolution kernel across the first feature data (a high-dimensional matrix containing the battery's immediate and cumulative features) to capture local transient changes within a single cycle during the battery's life degradation process, such as inflection points in the voltage curve and sudden changes in the current waveform. These local anomalies and patterns are then refined into third feature data, completing the spatial localization of the degradation characteristics.

[0115] LSTM is like a "time analyst". It takes the time series formed by the third feature data arranged in a cyclic order as input, and uses its unique memory unit and gating mechanism to learn and analyze how various operating data (such as voltage, current, temperature, etc.) between different cycles influence and evolve with time, thereby extracting the timing dependencies across cycles, such as the accelerated trend of capacity decay and the lag effect of internal resistance growth. Finally, it outputs the fourth feature data that reflects the decay timing law. The two work closely together to achieve in-depth extraction of battery decay features from local space to global time.

[0116] In the above method, through the synergistic effect of convolutional neural networks and long short-term memory networks in the feature extraction network, the convolutional neural network can be used to capture the local instantaneous change characteristics during the battery life degradation process, and the long short-term memory network can be used to mine the temporal dependency relationship between different working condition data, thereby realizing deep extraction of battery degradation features from local space to global time, providing multi-dimensional feature support for accurate prediction of battery life.

[0117] In one embodiment, see Figure 5 , is a flow chart of the training preset model provided in the embodiment of the present application, such as Figure 5 As shown, including:

[0118] S501, obtaining a plurality of fifth characteristic data of a training battery; the fifth characteristic data is historical operating condition data collected from the training battery during multiple charge and discharge processes.

[0119] In an embodiment of the present application, historical operating condition data such as voltage, current, temperature, and number of cycles are collected from multiple charge and discharge processes of the training battery. The historical operating condition data includes the collected instant feature historical data and cumulative feature historical data. The instant feature data and the cumulative feature historical data are feature fused to obtain a fifth feature number. The fifth feature data is used as a training set to train the preset model to be trained.

[0120] S502: Input the fifth feature data into a feature extraction network in a preset model to be trained to extract sixth feature data related to the life attenuation of the training battery.

[0121] In an embodiment of the present application, through algorithm processing such as convolutional neural network (CNN) and long short-term memory network (LSTM) in the feature extraction network, the temporal dependency between various historical operating condition data during the battery life decline process is excavated, thereby obtaining multiple key historical aging feature data (sixth feature data) that can reflect the changes and trends in the characteristics of the battery in different charging and discharging stages. These data can be used for subsequent analysis of the battery's health status and prediction of the battery life.

[0122] The extraction process of the sixth feature data is the same as that of the second feature data, and reference may be made to steps S301 - S303 and S401 - S402 , which will not be repeated here.

[0123] S503: Input the sixth feature data into a prediction network in a preset model to be trained to obtain a first loss.

[0124] In an embodiment of the present application, in the training process of the battery life prediction model, the first loss is obtained by inputting the sixth feature data (the extracted physical interpretability feature) into the prediction network of the preset model and calculating the error between the predicted value and the actual life value.

[0125] For example, the first loss can be calculated by calculating the mean square error (MSE) between the predicted value and the true lifespan:

[0126]

[0127] The first loss between the preset value of the sixth characteristic data input prediction network and the actual life value is calculated according to the above formula.

[0128] S504: When the first loss does not reach the convergence condition, continue to iteratively train the parameters of the preset model to be trained based on multiple fifth feature data until the first loss reaches the convergence condition. The training is then terminated, and the feature extraction network and prediction network with converged output parameters are the trained preset model.

[0129] In an embodiment of the present application, during the training of the battery life prediction model, when the first loss (i.e., the error between the battery life predicted by the model and the actual life) obtained by inputting the sixth characteristic data into the prediction network has not yet met the pre-set convergence conditions (such as the loss value no longer decreases significantly, the decrease is less than a specific threshold, or the maximum training rounds are reached), multiple fifth characteristic data (collected historical operating condition data) will be used to repeatedly adjust and train the parameters of the preset model (including the feature extraction network and the prediction network) through the back propagation algorithm. This iterative training continuously optimizes the model parameters to make the model prediction more in line with the actual situation and reduce the loss value. The training process will not stop until the first loss meets the convergence conditions, and the feature extraction network and prediction network with stable and optimized parameters will be output as the preset model that has completed training and can be put into use.

[0130] In the above method, by inputting the historical operating condition data (fifth characteristic data) of multiple charge and discharge of the training battery into the feature extraction network and prediction network of the preset model, combined with the iterative optimization mechanism of the first loss, end-to-end model parameter fine-tuning from the original operating condition data to life prediction is achieved, so that the preset model can adaptively capture the local characteristics and temporal dependencies of battery degradation, and finally output a pre-trained model with both high prediction accuracy and robustness.

[0131] It should be noted that the historical operating data collected in this application (real-time feature data and cumulative features are very fragmented) is a series of experiments conducted on a real lithium battery charging data set. Figure 6 , is a data comparison and analysis diagram of the maximum discharge capacity of the battery provided in the embodiment of the present application versus the number of cycles, such as Figure 6 As shown, there are two reference lines. The orange line (EOL: 80% of C_Ni) represents the battery health threshold and is a reference indicator for initial retirement. The red line (EOL: 65% of C_Ni) represents the threshold for complete retirement, representing a state where the battery's energy storage capacity has significantly decreased and its efficiency has been low. The figure shows that the battery capacity decay has obvious nonlinear characteristics. The decay is slow in the early stage and then accelerates, which may be related to material aging and increased internal resistance. The decay curves of different batteries are similar in the early stage, but then diverge, which puts forward requirements for the robustness of the model. For most batteries, the number of cycles is concentrated when the capacity drops to 80%, while the difference increases at 65%, indicating that the randomness of the late decay is enhanced, and higher requirements are placed on the stability of the model.

[0132] See also Figure 7 , is a data change analysis diagram of the voltage-capacity difference provided in the embodiment of the present application, such as Figure 7 As shown, the change in voltage-capacity difference can be divided into three stages: the initial period (0-200 cycles) is characterized by large fluctuations, related to battery activation and interfacial impedance formation; the mid-term (200-800 cycles) is characterized by reduced fluctuations, indicating a relatively stable electrochemical process; and the late period (after 800 cycles) is characterized by increased fluctuations, reflecting increased internal resistance and material degradation. The change in voltage-capacity difference reveals the staged characteristics of battery degradation and is a key indicator for monitoring battery health.

[0133] See also Figure 8 , is a data analysis diagram of battery internal resistance changing with cycle number provided in the embodiment of the present application, such as Figure 8As shown, the battery's internal resistance increases with the number of cycles, which can be divided into three stages: the initial period (0-200 cycles) shows rapid growth, related to SEI film formation; the middle period (200-800 cycles) shows steady growth, reflecting stable operation; and the late period (after 800 cycles) shows accelerated growth with large fluctuations, indicating material degradation and enhanced polarization effects. Changes in internal resistance are highly correlated with capacity decay, especially in the late period, where a rapid increase in internal resistance is often accompanied by a sharp drop in capacity. This is a key indicator of battery degradation and is crucial for predicting battery life.

[0134] See also Figure 9 , is an analysis diagram of the importance of the original characteristic data of the battery to the internal resistance change data provided by the embodiment of the present application, such as Figure 9 As shown, the lifecycle of battery A is divided into five stages based on internal resistance. The importance of the voltage-capacity difference increases with increasing internal resistance, especially in the later stages, while the importance of internal resistance itself decreases. The importance of features such as the time the battery maintains 4.15V and the charging time is relatively stable. However, the importance of features such as the time it takes for the voltage to drop from 3.6V to 3.4V and the discharge time fluctuates significantly, reflecting their strong predictive power in specific stages.

[0135] See also Figure 10 , is a schematic diagram of a high-dimensional aging feature correlation heat map provided in an embodiment of the present application, such as Figure 10 As shown, it reflects the correlation between the extracted high-dimensional aging features and the predicted remaining battery life, and the color changes reflect the linear and nonlinear relationships between the features.

[0136] In the battery degradation prediction process, principal component analysis is used to extract the most representative and independent features (key aging features) from the numerous aging features. Through principal component analysis, six principal components (PC1 to PC6) were extracted, which can reflect the key changing trends in the battery degradation process.

[0137] See also Figure 11 , is a schematic diagram of the low-dimensional aging feature correlation heat map of different batteries provided in the embodiment of the present application, such as Figure 11 As shown in the figure, through the correlation analysis between the PCA components (PC1 to PC6) of different batteries and the remaining service life, we found that the degradation characteristics of different batteries have different influence patterns. For example, PC1 of battery A and battery F both have a significant negative correlation with RUL, while PC1 of battery C shows a strong positive correlation with RUL. In addition, some principal components of some batteries (such as PC2 and PC5) contribute more to the prediction of RUL, while other principal components have less impact on RUL. Through these analyses, we can deeply understand the inherent mechanism of battery degradation. When training the model, different weights are assigned according to the correlation of different features, providing more accurate feature selection and model optimization basis for subsequent RUL prediction.

[0138] See also Figure 12 , is a schematic diagram of the contribution of low-dimensional aging characteristics of different batteries provided in the embodiments of the present application, such as Figure 12 As shown, the contribution, namely the SAAP value, quantitatively reveals the contribution of the feature to the model prediction and enhances the interpretability of the model. The analysis shows that PC1 and PC2 in battery A have a significant impact on the prediction results, and show a nonlinear change with the increase of internal resistance. In battery C, PC1, PC5 and PC2 have a prominent influence, especially PC1. In battery D, PC1 and PC2 are key features, and PC2 has a greater impact at high internal resistance. In battery E, PC1, PC4 and PC2 play an important role, and PC4 has a significant impact at high internal resistance. In battery F, the influence of PC1 and PC4 is particularly prominent, and PC1 plays a decisive role at high internal resistance. In battery G, PC1, PC2 and PC6 have the greatest impact, and the changes in PC1 and PC2 are strongly related to the model output.

[0139] In summary, SHAP value analysis shows that specific features (such as PC1, PC2, and PC4) have a significant impact on the nonlinear changes in battery degradation, while other features contribute less. These findings provide a reference for assigning model feature weights and help better understand battery degradation characteristics.

[0140] See also Figure 13 , is a schematic diagram of the overall structure of the battery life prediction provided by the embodiment of the present application, such as Figure 13 As shown, the prediction steps include:

[0141] 1) Input the collected battery operating data into the preset model

[0142] 2) Feature extraction network in the preset model extracts aging features

[0143] The feature extraction network is a hybrid CNN+LSTM model framework, with the CNN portion used to extract localized changes during battery degradation. Through a one-dimensional convolutional layer (Conv1D), this layer learns local patterns in the input data. The convolutional layer uses multiple filters to scan input features to extract potential degradation patterns. The pooling layer (MaxPooling1D) reduces the dimensionality of the convolutional output, retaining the most representative local information while reducing computational complexity. The CNN can extract multiple (e.g., 177) high-dimensional aging features.

[0144] The LSTM module is used to capture temporal dependencies in battery degradation data. The local features extracted by the CNN (177 high-dimensional features) are fed into the LSTM layer to learn dependencies between long time steps. Through its memory cells, the LSTM effectively captures long-term trends and changes in the battery's health status. This module typically includes multiple layers of LSTM units to further enhance the model's performance. Principal component analysis is then used to identify key aging features (e.g., six).

[0145] 3) Process of extracting key aging characteristics: Screening can be performed based on the independent aging component feature contribution map (SHAP value can be obtained) and the independent aging feature component correlation heat map (Pearson correlation coefficient can be obtained). Features strongly correlated with remaining life expectancy (RUL) are retained based on the Pearson correlation coefficient (|r| ≥ 0.7). SHAP value is introduced to quantify feature contribution, and minor factors (contribution < 5%) are eliminated, ultimately locking in six key aging characteristics.

[0146] 4) Inputting the key aging characteristics (independent aging characteristic analysis) into the remaining life prediction network model (preset model) to predict the remaining life of the battery and obtain the prediction results.

[0147] 5) Evaluate the prediction performance of the preset model based on the prediction results, and use the predicted value and the actual battery life value to determine the prediction accuracy of the model.

[0148] See also Figure 14 , is a schematic diagram comparing the predicted value of the remaining battery life and the actual value provided in the embodiment of the present application, such as Figure 14 As shown, the preset model used in this application is trained using 6000 battery cycle data from a related research institute, and the results are as follows Figure 14 , it can be seen that the predicted value and the actual value are highly consistent. Although there are slight deviations in some parts with large fluctuations, the overall trend is captured more accurately. Accuracy R 2 The value is 0.9632, which means that the model can explain about 96.32% of the variability between the predicted results and the actual results, further demonstrating the effectiveness of combining deep learning with principal component analysis.

[0149] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0150] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0151] Figure 15 This is a schematic diagram of the structure of the terminal device provided in the embodiment of the present application. Figure 15 As shown, the terminal device 15 of this embodiment includes: at least one processor 150 ( Figure 15 Only one is shown in the figure) a processor, a memory 151, and a computer program 152 stored in the memory 151 and executable on at least one processor 150. When the processor 150 executes the computer program 152, the steps of any of the above-mentioned battery remaining life method embodiments are implemented.

[0152] The terminal device can be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The terminal device may include, but is not limited to, a processor and a memory. Those skilled in the art will understand that Figure 15 It is merely an example of the terminal device 15 and does not constitute a limitation on the terminal device 15. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, it may also include input and output devices, network access devices, etc.

[0153] The processor 150 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor may be a microprocessor or any conventional processor.

[0154] In some embodiments, the memory 151 may be an internal storage unit of the terminal device 15, such as a hard disk or memory of the terminal device 15. In other embodiments, the memory 151 may also be an external storage device of the terminal device 15, such as a plug-in hard disk equipped on the terminal device 15, a smart media card (SMC), a secure digital (SD) card, a flash memory card, etc. Furthermore, the memory 151 may include both an internal storage unit of the terminal device 15 and an external storage device. The memory 151 is used to store an operating system, application programs, a boot loader, data, and other programs, such as the program code of a computer program. The memory 151 may also be used to temporarily store data that has been output or is about to be output.

[0155] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments can be implemented.

[0156] An embodiment of the present application provides a computer program product. When the computer program product is run on a terminal device, the terminal device can implement the steps in the above-mentioned method embodiments when executing the computer program product.

[0157] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor, it can implement the steps of the above-mentioned various method embodiments. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may at least include: any entity or device that can carry the computer program code to the device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electric carrier signal, a telecommunication signal and a software distribution medium. For example, a USB flash drive, a mobile hard disk, a magnetic disk or an optical disk. In some jurisdictions, according to legislation and patent practice, a computer-readable medium cannot be an electric carrier signal or a telecommunication signal.

[0158] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0159] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0160] In the embodiments provided in this application, it should be understood that the disclosed devices / terminal devices and methods can be implemented in other ways. For example, the device / terminal device embodiments described above are merely illustrative. For example, the division of modules or units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0161] Units described as separate components may or may not be physically separate, and 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.

[0162] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A battery life prediction method, characterized in that: The method comprises: Acquire first characteristic data of the battery to be tested; the first characteristic data is collected operating condition data of the battery during a charge and discharge process; Extracting second feature data related to the lifespan degradation of the battery based on the first feature data; wherein the dimension of the second feature data is lower than the dimension of the first feature data; The life of the battery to be detected is predicted according to the second characteristic data.

2. The battery life prediction method according to claim 1, wherein: The obtaining of first characteristic data of the battery to be tested includes: Collecting first sub-data and second sub-data; wherein the first sub-data is the real-time operating condition of the battery during the charging and discharging process, and the second sub-data is the statistical value of the operating condition data of the battery during one charging and discharging process; The first sub-data and the second sub-data are fused to obtain the first feature data.

3. The battery life prediction method according to claim 1, wherein: The extracting, based on the first characteristic data, second characteristic data related to the life attenuation of the battery includes: Acquire local change characteristics during the battery life degradation process according to the first characteristic data to obtain third characteristic data; Acquire a time sequence dependency relationship between various operating condition data during the battery life degradation process according to the third characteristic data to obtain fourth characteristic data; Second feature data related to the life degradation of the battery is extracted based on the fourth feature data.

4. The battery life prediction method according to claim 3, wherein: The extracting second characteristic data related to the life attenuation of the battery according to the fourth characteristic data includes: Obtaining a first parameter and a second parameter corresponding to the fourth characteristic data; wherein the first parameter is the linear correlation between the fourth characteristic data and the battery life, and the second parameter is the importance of the fourth characteristic data in affecting the battery life; The second characteristic data is filtered out from the fourth characteristic data according to the first parameter and the second parameter.

5. The battery life prediction method according to claim 4, wherein: The filtering out the second characteristic data from the fourth characteristic data according to the first parameter and the second parameter includes: Acquire fifth feature data that meets a first condition from the fourth feature data, wherein the fifth feature data meets the first condition if an absolute value of a first parameter corresponding to the fifth feature data is greater than a first preset threshold; Second characteristic data satisfying a second condition is obtained from the fifth characteristic data, wherein the second characteristic data satisfying the second condition is that a second parameter corresponding to the second characteristic data is greater than a second preset threshold.

6. The battery life prediction method according to claim 3, wherein: The method further comprises: Obtain a trained preset model; wherein the preset model includes a feature extraction network and a prediction network; the feature extraction network is used to extract second feature data related to the life attenuation of the battery based on the first feature data; the prediction network is used to predict the life of the battery to be tested based on the second feature data.

7. The battery life prediction method according to claim 6, wherein: The feature extraction network of the preset model includes a convolutional neural network and a long short-term memory network; The convolutional neural network is used to obtain local change characteristics during the battery life degradation process according to the first feature data to obtain third feature data; The long short-term memory network is used to obtain the temporal dependency between various operating condition data during the battery life decay process according to the third characteristic data to obtain fourth characteristic data.

8. The battery life prediction method according to any one of claims 6 or 7, characterized in that: The steps to train a preset model include: Acquire multiple fifth characteristic data of the training battery; the fifth characteristic data is historical operating condition data collected from the training battery during multiple charge and discharge processes; Inputting the fifth feature data into a feature extraction network in a preset model to be trained to extract sixth feature data related to the life attenuation of the training battery; Inputting the sixth feature data into a prediction network in a preset model to be trained to obtain a first loss; When the first loss does not reach the convergence condition, the parameters of the preset model to be trained continue to be iteratively trained according to the multiple fifth feature data until the first loss reaches the convergence condition, and the training ends. The preset model to be trained with converged output parameters is the preset model after pre-training.

9. A terminal device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images