Method, device and equipment for predicting remaining service life of battery and readable storage medium

Abstract

The invention discloses a battery remaining service life prediction method, device and equipment and a readable storage medium, and the method comprises the steps: obtaining an impedance data set through the electrochemical impedance spectroscopy test of a battery in multiple aging states, synchronously constructing an equivalent circuit model of the battery, and carrying out the modeling in combination with relaxation time distribution; extracting mechanism features from the electrochemical impedance spectroscopy to form a mechanism feature data set; and after fusing the impedance data set and the mechanism feature data set, screening features strongly related to battery performance through correlation analysis to obtain a target feature data set, and inputting the target feature data set into a preset prediction model to output a prediction result of the remaining service life of the battery. By means of combination of electrochemical impedance spectroscopy testing and relaxation time distribution modeling, conversion from original impedance data to mechanism characteristics is achieved, the characteristics are directly associated with battery aging essence, dependence on pre-calibration is eliminated, and the problem that an empirical model is insufficient in adaptability under complex working conditions is solved; and the model input quality is optimized through feature fusion and correlation screening.

Description

Technical Field

[0001] This application relates to the field of battery remaining life prediction, specifically to a method, apparatus, device, and readable storage medium for predicting battery remaining life. Background Technology

[0002] In fields such as new energy vehicles and energy storage systems, batteries are core energy storage components, and accurate prediction of their remaining useful life (RUL) is a key technical requirement for ensuring safe system operation and optimizing maintenance strategies. Accurate RUL prediction can effectively avoid safety hazards caused by battery overcharging and over-discharging, while improving battery efficiency and reducing replacement costs. Therefore, it has significant application value in the fields of power battery management systems (BMS) and energy storage battery monitoring.

[0003] In related technologies, to predict the remaining lifespan of a battery, an empirical model is usually established by pre-calibrating the battery's capacity decay curve under standard operating conditions and combining parameters such as cycle count and depth of charge / discharge, thereby estimating the remaining lifespan of the battery.

[0004] However, the battery aging process is affected by a variety of complex factors such as charge / discharge rate, ambient temperature, and storage time. Furthermore, different individual batteries have manufacturing differences and aging paths, which leads to low prediction accuracy of pre-calibrated empirical models under actual complex operating conditions. These models are difficult to dynamically adapt to the nonlinear changes in battery aging characteristics, and are prone to large deviations between prediction results and actual lifespan. This makes it impossible to meet the accuracy requirements for predicting remaining lifespan in high-reliability scenarios. Summary of the Invention

[0005] This application provides a method, apparatus, device, and readable storage medium for predicting the remaining lifespan of a battery. It can solve the technical problems in related technologies, such as the low prediction accuracy of traditional pre-calibrated empirical models under actual complex working conditions, the difficulty in dynamically adapting to the nonlinear changes in battery aging characteristics, and the tendency for large deviations between the predicted results and the actual lifespan.

[0006] In a first aspect, embodiments of this application provide a method for predicting the remaining lifespan of a battery, the method comprising: Based on electrochemical impedance spectroscopy testing of batteries under multiple aging states, impedance datasets are obtained, an equivalent circuit model of the battery is constructed, and relaxation time distribution modeling is combined with the modeling. Mechanistic features of the battery are extracted from the electrochemical impedance spectroscopy to form a mechanism feature dataset. By integrating impedance datasets and mechanistic feature datasets, correlation analysis is used to select features that are strongly correlated with battery performance, thus obtaining the target feature dataset. Input the target feature dataset into the preset prediction model, and output the battery remaining life prediction result. In conjunction with the first aspect, in one embodiment, the electrochemical impedance spectroscopy test of the battery under multiple aging states to obtain an impedance data set includes: Based on the battery aging test, complex impedance values ​​of different frequency bands were collected at multiple aging stages; An impedance dataset is constructed by integrating multi-band complex impedance values ​​under multiple aging conditions.

[0007] In conjunction with the first aspect, in one implementation, the construction of an equivalent circuit model of the battery, combined with relaxation time distribution modeling, extracts the mechanistic features of the battery from the electrochemical impedance spectroscopy to form a mechanistic feature dataset, including: Construct a second-order equivalent circuit model to characterize the electrochemical impedance process of a battery; A relationship function between the electrochemical impedance spectrum and the relaxation time distribution is established based on the relaxation time distribution. The fitted impedance data is obtained by optimization through discretization and regularization. The component parameters of the second-order equivalent circuit model are calculated based on the fitted impedance data, and the peak information of the relaxation time distribution curve is combined to form a mechanism feature dataset.

[0008] In conjunction with the first aspect, in one implementation, obtaining the fitted impedance data through discretization and regularization optimization includes: Using the Gaussian function as the kernel function and the radial basis functions as the basis functions, the objective function is constructed based on the discretization formula; After optimizing the polarization ratio and time constant of the discretized formula using an intelligent optimization algorithm, the optimal polarization ratio and time constant are obtained by solving the objective function. By substituting the optimal polarization ratio and time constant into the discretization formula, the real and imaginary parts of the impedance corresponding to the measured frequency band are calculated, thus obtaining the fitted impedance data.

[0009] In conjunction with the first aspect, in one embodiment, the component parameters for calculating the second-order equivalent circuit model based on the fitted impedance data include ohmic resistance, solid electrolyte interface resistance, and charge transport polarization resistance.

[0010] In conjunction with the first aspect, in one implementation, the fused impedance dataset and the mechanistic feature dataset are used to screen out features strongly correlated with battery performance through correlation analysis to obtain a target feature dataset, including: By merging the impedance dataset and the mechanism feature dataset, a complete impedance feature set is obtained. Voltage data of the battery under multiple aging conditions were collected synchronously to construct a voltage dataset. The correlation between each feature in the complete impedance feature set and the voltage data was calculated using the Pearson correlation coefficient formula. By setting a correlation factor, weakly correlated features with an absolute correlation value less than the correlation factor are removed, and the remaining features constitute the target feature dataset.

[0011] In conjunction with the first aspect, in one embodiment, the battery remaining lifespan prediction method further includes the following steps: By collecting multi-dimensional data and lifespan labels of sample batteries throughout their entire life cycle to construct a training set, and after preprocessing and feature selection, the initial neural network model is trained with the target feature data as input and the lifespan label as output, and the hyperparameters are optimized to obtain a converged preset prediction model.

[0012] In conjunction with the first aspect, in one implementation, the step of constructing a training set by collecting multi-dimensional data and lifetime labels throughout the entire life cycle of sample batteries includes: The system collects impedance and mechanistic characteristics data of sample batteries under multiple aging states throughout their entire life cycle; it also simultaneously collects the actual remaining service life tags corresponding to each aging state. Impedance data, mechanistic feature data and corresponding actual remaining useful life labels are associated to construct a training sample set including input features and output labels.

[0013] In conjunction with the first aspect, in one implementation, the step of training an initial neural network model and optimizing hyperparameters using target feature data as input and lifetime labels as output after preprocessing and feature selection to obtain a converged preset prediction model includes: The training sample set is preprocessed by data normalization and outlier removal. Then, the correlation between each feature and the battery performance index of the sample is calculated by the Pearson correlation coefficient formula. The correlation factor is set and weakly correlated features with an absolute correlation value less than the correlation factor are removed to obtain the target feature dataset for training. Using the target feature dataset for training as input and the actual remaining lifespan label as output, the initial neural network model is trained. Hyperparameters such as learning rate and number of convolution kernels are optimized through cross-validation to obtain a converged preset prediction model.

[0014] Secondly, embodiments of this application provide a battery remaining life prediction device, the battery remaining life prediction device comprising: The mechanism feature extraction module is used to obtain impedance data sets based on electrochemical impedance spectroscopy testing under multiple aging states of the battery, construct the equivalent circuit model of the battery and combine it with relaxation time distribution modeling, extract the mechanism features of the battery from the electrochemical impedance spectrum, and form a mechanism feature dataset. The feature selection module is used to fuse impedance datasets and mechanism feature datasets, and select features that are strongly correlated with battery performance through correlation analysis to obtain the target feature dataset. The lifespan prediction module is used to input the target feature dataset into a preset prediction model and output the prediction result of the remaining battery lifespan.

[0015] In conjunction with the second aspect, in one embodiment, the mechanism feature extraction module is further used to perform an aging test based on the battery, collect complex impedance values ​​of different frequency bands at multiple aging stages, and integrate the complex impedance values ​​of multiple frequency bands under multiple aging states to construct an impedance dataset.

[0016] In conjunction with the second aspect, in one embodiment, the mechanism feature extraction module is further used to construct a second-order equivalent circuit model characterizing the battery electrochemical impedance process; establish a relationship function between the electrochemical impedance spectrum and the relaxation time distribution based on the relaxation time distribution, and obtain the fitted impedance data through discretization and regularization optimization; calculate the component parameters of the second-order equivalent circuit model based on the fitted impedance data, and integrate them with the peak information of the relaxation time distribution curve to form a mechanism feature dataset.

[0017] In conjunction with the second aspect, in one implementation, the mechanism feature extraction module is further used to construct an objective function based on the discretization formula, using the radial basis function with the Gaussian function as the kernel function as the basis function; after optimizing the polarization ratio and time constant of the discretization formula using an intelligent optimization algorithm, the optimal polarization ratio and time constant are obtained by solving the objective function; the optimal polarization ratio and time constant are substituted into the discretization formula to calculate the real and imaginary parts of the impedance corresponding to the measured frequency band, that is, to obtain the fitted impedance data.

[0018] Thirdly, embodiments of this application provide a battery remaining life prediction device, which includes a processor, a memory, and a battery remaining life prediction program stored in the memory and executable by the processor. When the battery remaining life prediction program is executed by the processor, it implements the steps of the battery remaining life prediction method as described in some of the above embodiments.

[0019] Fourthly, embodiments of this application provide a computer-readable storage medium storing a battery remaining life prediction program, wherein when the battery remaining life prediction program is executed by a processor, it implements the steps of the battery remaining life prediction method as described in some of the above embodiments.

[0020] The beneficial effects of the technical solutions provided in this application include: This method acquires impedance data by performing electrochemical impedance spectroscopy (EIS) tests on batteries under multiple aging conditions. Simultaneously, an equivalent circuit model of the battery is constructed, and relaxation time distribution modeling is incorporated. Mechanistic features are extracted from the EIS to form a mechanistic feature dataset. After fusing the impedance and mechanistic feature datasets, correlation analysis is used to filter features strongly correlated with battery performance, resulting in a target feature dataset. This target feature dataset is then input into a pre-defined prediction model to output the battery's remaining lifespan prediction. This method, by combining EIS testing with relaxation time distribution modeling, achieves the transformation from raw impedance data to mechanistic features, directly linking features to the essence of battery aging. It eliminates reliance on pre-calibration and effectively improves the adaptability of empirical models under complex operating conditions. Furthermore, feature fusion and correlation filtering optimize the model input quality, and combined with the computational processing of the pre-defined prediction model, enhances the accuracy of remaining lifespan prediction. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating an embodiment of the battery remaining life prediction method of this application; Figure 2 This is a flowchart illustrating another embodiment of the battery remaining life prediction method of this application; Figure 3 This is the second-order equivalent circuit diagram of the electrochemical impedance spectroscopy used in the battery remaining life prediction method of this application. Figure 4 This is a schematic diagram of the hardware structure of the battery remaining life prediction device involved in the embodiments of this application. Detailed Implementation

[0022] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0023] It's important to understand that in fields like new energy vehicles and energy storage systems, batteries, as core energy storage components, require accurate prediction of their Remaining Useful Life (RUL) as a key technical requirement for ensuring safe system operation and optimizing maintenance strategies. Accurate RUL prediction can effectively avoid safety hazards caused by battery overcharging and over-discharging, while also improving battery efficiency and reducing replacement costs. Therefore, it has significant application value in the fields of battery management systems (BMS) and energy storage battery monitoring.

[0024] To predict the remaining lifespan of a battery, an empirical model is typically established by pre-calibrating the battery's capacity decay curve under standard operating conditions and combining parameters such as cycle count and depth of charge / discharge, thereby estimating the remaining lifespan of the battery.

[0025] However, the battery aging process is affected by a variety of complex factors such as charge / discharge rate, ambient temperature, and storage time. Furthermore, different individual batteries have manufacturing differences and aging paths, which leads to low prediction accuracy of pre-calibrated empirical models under actual complex operating conditions. These models are difficult to dynamically adapt to the nonlinear changes in battery aging characteristics, and are prone to large deviations between prediction results and actual lifespan. This makes it impossible to meet the accuracy requirements for predicting remaining lifespan in high-reliability scenarios.

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0027] In a first aspect, embodiments of this application provide a method for predicting the remaining lifespan of a battery.

[0028] In one embodiment, reference is made to Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the battery remaining lifespan prediction method of this application. Figure 1 As shown, the method for predicting the remaining battery life includes: S100: Based on the electrochemical impedance spectroscopy test of the battery under multiple aging states, the impedance data set is obtained, the equivalent circuit model of the battery is constructed and combined with the relaxation time distribution modeling, the mechanism characteristics of the battery are extracted from the electrochemical impedance spectrum, and the mechanism characteristic data set is formed. S200: By integrating impedance datasets and mechanistic feature datasets, and using correlation analysis to select features that are strongly correlated with battery performance, a target feature dataset is obtained. S300: Input the target feature dataset into the preset prediction model and output the prediction result of the remaining battery life.

[0029] In this embodiment, an impedance dataset is obtained by performing electrochemical impedance spectroscopy tests on the battery under multiple aging states. Simultaneously, an equivalent circuit model of the battery is constructed and combined with relaxation time distribution modeling. Mechanistic features are extracted from the electrochemical impedance spectrum to form a mechanism feature dataset. After fusing the impedance dataset and the mechanism feature dataset, features strongly correlated with battery performance are selected through correlation analysis to obtain a target feature dataset. The target feature dataset is then input into a preset prediction model to output the battery remaining life prediction result. This method combines electrochemical impedance spectroscopy testing with relaxation time distribution modeling to transform raw impedance data into mechanistic features, directly linking these features to the essence of battery aging. This eliminates reliance on pre-calibration and effectively improves the adaptability of empirical models under complex operating conditions. By optimizing model input quality through feature fusion and correlation screening, and combining this with the computational processing of a pre-defined prediction model, the accuracy of remaining lifespan prediction is enhanced. This reduces the deviation between prediction results and actual lifespan, meeting the needs of high-reliability scenarios. Furthermore, it provides support for the safe operation of batteries in new energy vehicles, energy storage systems, and other fields, helping to optimize operation and maintenance strategies, improve battery utilization efficiency, and reduce replacement costs. It has application value in the fields of power battery management systems and energy storage battery monitoring.

[0030] Furthermore, in one embodiment, S100 includes S101, which includes the following steps: S101-1: Based on battery aging test, collect complex impedance values ​​of different frequency bands at multiple aging stages; S101-2: Integrate multi-band complex impedance values ​​under multiple aging conditions to construct an impedance dataset.

[0031] In this embodiment, in S101-1, an aging test is performed on the solid-state battery. Complex impedance values ​​at different frequency bands are collected at multiple stages of the battery aging process. These complex impedance values ​​include both real and imaginary parts, reflecting the electrochemical response characteristics of the battery under different frequency excitations. In S101-2, the multi-frequency complex impedance values ​​collected under multiple aging states are integrated to construct an impedance dataset. This step, by collecting and integrating full-frequency complex impedance data across multiple aging stages, can comprehensively capture the dynamic changes in the battery's electrochemical characteristics as aging progresses. It avoids the shortcomings of single aging states or limited frequency band data failing to fully reflect aging characteristics. This provides a reliable data foundation for subsequently combining equivalent circuit models and relaxation time distribution modeling to extract mechanistic features from the electrochemical impedance spectrum, thereby supporting the improvement of the accuracy of battery remaining life prediction.

[0032] Furthermore, in one embodiment, S100 includes S102, which includes the following steps: S102-1: Construct a second-order equivalent circuit model to characterize the electrochemical impedance process of a battery; S102-2: Establish the relationship function between electrochemical impedance spectrum and relaxation time distribution based on relaxation time distribution, and obtain the fitted impedance data by optimization through discretization and regularization; S102-3: Calculate the component parameters of the second-order equivalent circuit model based on the fitted impedance data, and integrate them with the peak information of the relaxation time distribution curve to form a mechanism feature dataset.

[0033] In this embodiment, in S102-1, a second-order equivalent circuit model characterizing the electrochemical impedance process of a solid-state battery is constructed. The model includes an ohmic resistor, a solid electrolyte interface resistance, a charge transport polarization resistance, and constant phase angle elements. In S102-2, a relationship function between the electrochemical impedance spectrum and the relaxation time distribution is established based on the relaxation time distribution (DRT). The relationship function is discretized to obtain a discretized formula. Since the measured impedance only contains a portion of the frequencies, it is difficult to obtain an accurate analytical solution. Therefore, a regularization method is used to optimize the solution of the discretized formula to obtain the fitted impedance data. In S102-3, the component parameters of the second-order equivalent circuit model are calculated based on the fitted impedance data. The peak information of the relaxation time distribution curve is extracted simultaneously, and the component parameters and peak information are integrated to form a mechanism feature dataset. This step combines a second-order equivalent circuit model with relaxation time distribution modeling to transform the original impedance data into quantifiable aging mechanism features. This solves the problem that the original impedance data is difficult to directly correlate with the essence of aging. The resulting mechanism feature dataset can accurately reflect the changing patterns of the internal electrochemical processes of the battery, laying a mechanistic foundation for subsequent screening of strongly correlated features through correlation analysis and improving the accuracy of remaining service life prediction.

[0034] Furthermore, in one embodiment, S102-2 includes the following steps: S102-2-1: Using the radial basis functions with the Gaussian function as the kernel function as the basis functions, the objective function is constructed based on the discretization formula; S102-2-2: After optimizing the polarization ratio and time constant of the discretized formula using an intelligent optimization algorithm, the optimal polarization ratio and time constant are obtained by solving the objective function. S102-2-3: Substitute the optimal polarization ratio and time constant into the discretization formula to calculate the real and imaginary parts of the impedance corresponding to the measured frequency band, thus obtaining the fitted impedance data.

[0035] In this embodiment, in S102-2-1, a radial basis function with a Gaussian function as the kernel function is selected as the basis function. An objective function is constructed based on the discretization formula of the electrochemical impedance spectrum and relaxation time distribution. In S102-2-2, an intelligent optimization algorithm is used to optimize the polarization ratio and time constant in the discretization formula. Then, the optimal polarization ratio and time constant are obtained by solving the objective function. In S102-2-3, the optimal polarization ratio and time constant are substituted into the discretization formula to calculate the real and imaginary parts of the impedance corresponding to the measured frequency band, i.e., the fitted impedance data. This step constructs an objective function using specific basis functions and combines intelligent optimization algorithms with parameter optimization using discretization formulas. This effectively solves the problem of finding analytical solutions when the measured impedance only contains a portion of the frequencies. At the same time, regularization constraints (implicit in the objective function construction logic) prevent overfitting, ensuring that the fitted impedance data closely approximates the measured results and accurately reflects the distribution of the polarization process inside the battery. This provides a high-precision data foundation for subsequent calculation of second-order equivalent circuit element parameters and extraction of relaxation time distribution curve features, thereby ensuring the reliability of the mechanism feature dataset and supporting the improvement of the accuracy of remaining service life prediction.

[0036] Furthermore, in one embodiment, in S102-3, the component parameters for calculating the second-order equivalent circuit model based on the fitted impedance data include ohmic resistance, solid electrolyte interface resistance, and charge transport polarization resistance.

[0037] In this embodiment, the fitted impedance data (including the real and imaginary parts of the impedance corresponding to the measured frequency band) obtained by discretization formula and intelligent optimization algorithm in S102-2 is used as input. Combined with the topological relationship of the second-order equivalent circuit model that characterizes the electrochemical impedance process of the battery, the core resistance parameters reflecting different electrochemical processes inside the battery are quantitatively solved through parameter identification and fitting calculation. Among them, the ohmic resistance characterizes the basic resistance characteristics of the electronic conductor inside the battery, the solid electrolyte interface resistance reflects the resistance characteristics of the solid electrolyte interface film formed between the electrode and the electrolyte, and the charge transport polarization resistance corresponds to the resistance characteristics of the charge transfer process at the electrode / electrolyte interface. The peak information of the relaxation time distribution curve is extracted simultaneously, and the three types of resistance parameters obtained above are integrated with the peak information to form a mechanism feature dataset. This process achieves quantitative capture of key electrochemical characteristic changes during battery aging by accurately solving the core resistance parameters of the equivalent circuit. Combined with the characteristics of the relaxation time distribution curve, the mechanism feature dataset can more comprehensively reflect the internal aging mechanism of the battery. This solves the problem that the original impedance data is difficult to directly correlate with the essence of aging, and provides a physically meaningful and reliable foundation for subsequent feature selection and remaining service life prediction model input, thereby supporting the improvement of prediction accuracy.

[0038] Furthermore, in one embodiment, step S200 includes the following steps: S200-1: Merge the impedance dataset and the mechanism feature dataset to obtain a complete impedance feature set; S200-2: Synchronously collect voltage data of the battery under multiple aging conditions to construct a voltage dataset, and use the Pearson correlation coefficient formula to calculate the correlation between each feature in the complete impedance feature set and the voltage data; S200-3: Set a correlation factor, remove weakly correlated features whose absolute correlation value is less than the correlation factor, and the remaining features constitute the target feature dataset.

[0039] In this embodiment, in S200-1, the impedance dataset (containing multi-band complex impedance values ​​under multiple aging states) constructed in step S10 is merged with the mechanism feature dataset (containing second-order equivalent circuit element parameters and relaxation time distribution curve peak information) formed in step S102 to obtain a complete impedance feature set covering the original response data and mechanism quantification features; in S200-2, during the synchronous aging stage of collecting impedance data and extracting mechanism features, the battery voltage data is collected synchronously and a voltage dataset is constructed. The Pearson correlation coefficient formula is used to calculate the correlation between each feature in the complete impedance feature set and the voltage data. The voltage data is directly related to the battery performance state, and its correlation can characterize the feature's ability to reflect the battery performance; in S200-3, a correlation factor δ is set, and weakly correlated features with an absolute correlation value < δ are eliminated. The remaining features constitute the target feature dataset. This step compensates for the incompleteness of single-type feature information by integrating the original impedance data with mechanistic features. By using quantitative analysis of Pearson correlation coefficient and removal of weakly correlated features, redundant information is effectively filtered out, allowing the target feature dataset to focus on core features strongly correlated with battery performance. This reduces the computational complexity of subsequent prediction models and improves the effectiveness of model input features, providing reliable support for the preset prediction model to output accurate remaining service life results.

[0040] Furthermore, in one embodiment, prior to S100, the following steps are included: S000: By collecting multi-dimensional data and lifespan labels of the entire life cycle of sample batteries, a training set is constructed. After preprocessing and feature selection, the initial neural network model is trained with the target feature data as input and the lifespan label as output, and the hyperparameters are optimized to obtain a converged preset prediction model.

[0041] In this embodiment, multi-dimensional data and corresponding lifetime labels are collected from the solid-state battery samples under various aging states throughout their entire life cycle. The multi-dimensional data includes impedance data and mechanistic feature data, which are correlated with the lifetime labels to construct a training set. The training set is preprocessed and feature-selected. Using the target feature dataset for training as input and the lifetime labels as output, an initial neural network model is trained. Simultaneously, hyperparameters are optimized to achieve model convergence, resulting in a preset prediction model. This step constructs a training set using full-life-cycle sample data, combines preprocessing and feature selection to optimize input quality, and the converged preset prediction model obtained through model training and hyperparameter optimization can fully learn the correlation between features and remaining lifespan, avoiding the problems of insufficient prediction accuracy and weak generalization ability of the initial model. This provides reliable support for subsequently inputting the target feature dataset of actual batteries into the model to output accurate prediction results, ensuring the practicality and accuracy of the entire prediction method.

[0042] Furthermore, in one embodiment, S000 includes the following steps: S000-1: Collect impedance data and mechanism characteristic data of sample batteries under multiple aging states throughout their entire life cycle; simultaneously collect the actual remaining service life tags corresponding to each aging state; S000-2: Associate impedance data, mechanistic feature data, and corresponding actual remaining useful life labels to construct a training sample set that includes input features and output labels.

[0043] In this embodiment, in step S000-1, for the sample solid-state battery, impedance data and mechanistic feature data are collected under multiple aging states throughout its entire life cycle. The impedance data consists of multi-band complex impedance values, and the mechanistic feature data includes second-order equivalent circuit element parameters and relaxation time distribution curve peak information. Simultaneously, the actual remaining service life label corresponding to each aging state is collected. In step S000-2, the collected impedance data, mechanistic feature data, and the actual remaining service life label of the corresponding aging state are correlated and matched to construct a training sample set containing input features (impedance data and mechanistic feature data) and output labels (actual remaining service life). This step, by collecting feature data from multiple aging states throughout the sample battery's entire life cycle and establishing a correlation with the actual service life label, ensures that the training sample set can cover the complete aging process and multi-dimensional features of the battery. This solves the problem that single-state or missing label data cannot support effective model learning, providing a comprehensive and realistic sample foundation for subsequent training of the initial neural network model. This ensures that the model can accurately capture the intrinsic relationship between features and remaining service life, thereby supporting the convergence of the preset prediction model and improving prediction accuracy.

[0044] Furthermore, in one embodiment, S000 includes the following steps: S000-3: Perform data normalization and outlier removal preprocessing on the training sample set, then calculate the correlation between each feature and the battery performance index of the sample using the Pearson correlation coefficient formula, set the correlation factor and remove weakly correlated features whose absolute correlation value is less than the correlation factor, and obtain the target feature dataset for training. S000-4: Using the target feature dataset for training as input and the actual remaining lifespan label as output, train the initial neural network model, optimize hyperparameters such as learning rate and number of convolution kernels through cross-validation, and obtain a converged preset prediction model.

[0045] In this embodiment, in steps S000-3, preprocessing operations such as data normalization and outlier removal are performed on the training sample set constructed in S000-2. Then, the correlation between each preprocessed feature and the performance indicators of the sample battery is calculated using the Pearson correlation coefficient formula. A correlation factor is set, and weakly correlated features with an absolute correlation value less than the correlation factor are removed to obtain the target feature dataset for training. In step S000-4, the initial neural network model is trained using the target feature dataset for training as input and the actual remaining lifespan label as output. During the process, hyperparameters such as the learning rate and the number of convolutional kernels are optimized through cross-validation until the model converges to obtain the preset prediction model. This step removes data noise and abnormal interference through preprocessing, focuses on core effective features through correlation screening, reduces redundant model calculations, and improves the model training effect by combining hyperparameter optimization. This enables the converged preset prediction model to accurately learn the intrinsic relationship between features and remaining lifespan, avoiding prediction bias caused by poor data quality or unreasonable model parameters, and providing a reliable model foundation for subsequent prediction of the remaining lifespan of actual batteries.

[0046] In summary, referring to Figure 2 As shown, this embodiment of the application achieves lifetime prediction through four core stages: "data acquisition → mechanism modeling → feature selection → model prediction". The specific steps are as follows: Step 1: Multi-dimensional data collection The core objective of this stage is to obtain impedance and voltage data during the battery aging process, providing a foundation for subsequent modeling.

[0047] 1. Aging test and electrochemical impedance spectroscopy (EIS) test Accelerated aging tests (such as cycle charge-discharge aging, high-temperature storage aging, etc.) are performed on the target battery, and EIS tests are conducted at multiple aging stages (such as capacity decay to 95%, 90%, 85% of the initial capacity, etc.). The test frequency band covers low frequency (0.01Hz) to high frequency (100kHz), and the complex impedance value at each frequency point is recorded. , including the real part and the imaginary part; Building an impedance dataset ,in 2b represents the number of aging stages, and a represents the number of test frequency points in a single stage.

[0048] 2. Voltage data acquisition Synchronously collect battery stack voltage data at each aging stage to construct a voltage dataset Y={y i},in , where n is the number of voltage sampling points, used for subsequent feature correlation screening.

[0049] Step 2: Mechanism Feature Extraction This stage transforms the original impedance data into physically meaningful mechanistic features through second-order equivalent circuit modeling and relaxation time distribution (DRT) analysis. The specific steps are as follows: 2.1 Construction of Second-Order Equivalent Circuit Model Model structure: such as Figure 3 As shown, a second-order equivalent circuit characterizing the battery electrochemical process is constructed, including: Ohm resistor : Reflects the resistance of the electronic conductors inside the battery (such as current collectors and electrodes); solid electrolyte interface resistance : Reflects the resistance of the solid electrolyte interface formed between the electrode and the electrolyte; charge transport polarization resistance : Resistance reflecting the charge transfer process at the electrode / electrolyte interface; constant phase angle element , : Corresponding to , The equivalent capacitance characteristic (Q is used to replace the ideal capacitance because the interface is not ideal).

[0050] Impedance expression: Impedance of constant phase angle element: Where Q is the scaling factor. The phase angle index is (0≤φ≤1, φ=1 is an ideal capacitor). Angular frequency; Total battery impedance:

[0051] 2.2 DRT Modeling and Parameter Solving This step uses DRT to decompose EIS data into characteristic parameters of different polarization processes, thus solving the problem of multiple solutions in equivalent circuit parameter fitting.

[0052] 1. Construction of DRT relational functions: Establish the association function between EIS and DRT:

[0053] in, For the ohmic part (corresponding) ), For the polarization part (including) , and , (Comprehensive contribution of components) This represents the proportion of each polarization in the total polarization; It is a time constant; 2. Discretization process: Discretize the continuous DRT relation function into a computable form:

[0054] in, The proportion of the i-th polarization process in the total polarization; : The time constant of the i-th polarization process, and ( For the corresponding polarization resistor, N: Number of discretization points (usually 50-200 to ensure coverage of the main polarization processes).

[0055] 3. Tikhonov Regularization Optimization Solution Since the measured EIS only covers a limited frequency band, directly solving the discretization formula is prone to overfitting and multiple solutions of parameters. Therefore, regularization constraints are introduced. Objective function:

[0056] in, Minimize the error between the calculated impedance (real part + imaginary part) and the measured impedance using the DRT method; Regularization term This is the regularization parameter (value range 0.001-0.1, determined through cross-validation). For parameter smoothing constraint terms; Solution steps: 1. Choosing the Gaussian function as the radial basis function of the kernel function enhances the stability of parameter solving; 2. Optimize initial parameters using intelligent optimization algorithms (such as particle swarm optimization and simulated annealing). , (Initial value), and then solve the objective function using Tikhonov regularization to obtain the optimal value. and ; 3. Substitute the optimal parameters into the discretization formula to calculate the fitted impedance data. , The fitting accuracy was verified by comparing the data with the measured EIS data.

[0057] 2.3 Integration of Mechanism Features Equivalent circuit parameters: Based on the fitted impedance data, calculate the component parameters of the second-order equivalent circuit. , and ; DRT features: Extracting the peak position of the DRT curve (corresponding to...) (Reflecting the time scale of the polarization process), peak intensity (corresponding to) (Reflects the contribution ratio of the polarization process) and half-width at half-maximum (reflects the uniformity of the polarization process distribution). Constructing a mechanistic feature dataset: Integrate the above parameters with DRT features to form a mechanistic feature dataset. S represents the total number of mechanistic features.

[0058] Step 3: Feature Selection and Model Input Construction This stage integrates the original impedance characteristics and mechanistic characteristics, selects features that are strongly correlated with battery performance, reduces model complexity, and improves generalization ability.

[0059] Feature fusion: merging impedance datasets Mechanistic feature dataset To obtain the complete feature set ; Correlation analysis: The correlation between each feature and the voltage dataset Y was calculated using the Pearson correlation coefficient formula. ; Feature screening: Set a correlation factor δ (range 0.6-0.8, adjust according to experimental data) to exclude features. The weakly correlated features with less than δ, and the remaining features constitute the target feature dataset. , as input to the prediction model.

[0060] Step 4: Building and Training the CNN Neural Network Prediction Model This stage uses deep learning technology to build a predictive model to quantitatively output the remaining lifespan.

[0061] 1. Model Structure: Input layer: Receives the target feature dataset The data shape is (length is the length of the feature sequence, and channels are the feature dimensions); Convolutional layer: 3×3 convolutional kernels are used to extract local features and enhance the local correlation of features; Activation layer: via ReLU function Introducing nonlinearity to avoid gradient vanishing; Pooling layer: 2×2 max pooling window is used for dimensionality reduction to retain key features and reduce computational cost; Batch normalization layer: accelerates model training convergence and improves training stability; Dropout layer: The dropout rate is set to 0.3-0.5 to suppress overfitting; Fully connected layer: Flatten the pooling output into a one-dimensional vector, and perform feature fusion and dimensionality reduction through 2-3 fully connected layers; Output layer: Outputs the predicted remaining battery life (usually expressed as the number of cycles required for the capacity to decay to 80% of the initial capacity).

[0062] 2. Model Training and Optimization: Loss function: using mean squared error This measures the deviation between the predicted value and the actual lifespan. Optimizer: The Adam optimizer is used, with the initial learning rate set to 0.001 and adjusted using a decay strategy; Parameter tuning: Optimization using 5-fold cross-validation Hyperparameters such as (32-128), number of convolutional kernels (32-128), and number of network layers; Model validation: Use independent test sets (containing data under different operating conditions) to validate the model's generalization ability and ensure that the prediction error is less than 3%.

[0063] Finally, in this embodiment, the original impedance data is converted using DRT analysis. , , Physical parameters such as DRT peak value are used. Among them, the DRT peak value (relaxation time distribution peak value) refers to the iconic data point reflecting the intensity and characteristic time scale of a specific electrochemical polarization process inside the battery, obtained by modeling and calculating the relaxation time distribution (DRT). This solves the "black box" problem of traditional data-driven methods. The features are directly related to the aging mechanism, enabling the model to maintain stable input quality under different operating conditions. To address the multi-solution problem of EIS discretization, Tikhonov regularization and intelligent optimization algorithms are introduced to reduce the fitting error of equivalent circuit parameters and provide a reliable foundation for mechanism feature extraction. The local correlation of features is captured by 3×3 convolution kernels. Combined with batch normalization and Dropout technology, the nonlinear fitting ability of the model to the complex aging process of the battery is improved, and the prediction accuracy is improved compared with traditional machine learning methods. Only conventional EIS testing equipment is required to collect basic data. After the model is trained, it can be quickly deployed to the battery management system (BMS) to provide accurate basis for battery fault early warning and maintenance, and reduce operation and maintenance costs.

[0064] Secondly, embodiments of this application also provide a battery remaining life prediction device, the battery remaining life prediction device comprising: The mechanism feature extraction module is used to obtain impedance data sets based on electrochemical impedance spectroscopy testing under multiple aging states of the battery, construct the equivalent circuit model of the battery and combine it with relaxation time distribution modeling, extract the mechanism features of the battery from the electrochemical impedance spectrum, and form a mechanism feature dataset. The feature selection module is used to fuse impedance datasets and mechanism feature datasets, and select features that are strongly correlated with battery performance through correlation analysis to obtain the target feature dataset. The lifespan prediction module is used to input the target feature dataset into a preset prediction model and output the prediction result of the remaining battery lifespan.

[0065] The functions of each module in the battery remaining life prediction device correspond to the steps in the battery remaining life prediction method embodiment, and their functions and implementation processes will not be described in detail here.

[0066] Thirdly, embodiments of this application provide a battery remaining life prediction device, which can be a personal computer (PC), laptop computer, server, or other device with data processing capabilities.

[0067] Reference Figure 4 , Figure 4 This is a schematic diagram of the hardware structure of the battery remaining life prediction device involved in the embodiments of this application. In the embodiments of this application, the battery remaining life prediction device may include a processor, a memory, a communication interface, and a communication bus.

[0068] The communication bus can be of any type and is used to interconnect the processor, memory, and communication interface.

[0069] The communication interface includes input / output (I / O) interfaces, physical interfaces, and logical interfaces used for interconnecting internal components of the battery remaining life prediction device, as well as interfaces used for interconnecting the battery remaining life prediction device with other devices (such as other computing devices or user equipment). Physical interfaces can be Ethernet interfaces, fiber optic interfaces, ATM interfaces, etc.; user equipment can be displays, keyboards, etc.

[0070] Memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical storage, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.

[0071] The processor can be a general-purpose processor, which can call the battery remaining life prediction program stored in the memory and execute the battery remaining life prediction method provided in the embodiments of this application. For example, the general-purpose processor can be a central processing unit (CPU). The method executed when the battery remaining life prediction program is called can refer to the various embodiments of the battery remaining life prediction method of this application, and will not be repeated here.

[0072] Those skilled in the art will understand that Figure 4 The hardware structure shown does not constitute a limitation of this application and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0073] Fourthly, embodiments of this application also provide a readable storage medium.

[0074] The present application has a readable storage medium storing a battery remaining life prediction program, wherein when the battery remaining life prediction program is executed by a processor, it implements the steps of the battery remaining life prediction method as described above.

[0075] The method implemented when the battery remaining life prediction program is executed can be referred to in various embodiments of the battery remaining life prediction method of this application, and will not be repeated here.

[0076] It should be noted that the sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0077] The terms "comprising" and "having," and any variations thereof, in the specification, claims, and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus. The terms "first," "second," and "third," etc., are used to distinguish different objects, etc., and do not indicate a sequence, nor do they limit "first," "second," and "third" to different types.

[0078] In the description of the embodiments of this application, terms such as "exemplary," "for example," or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplary," "for example," or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary," "for example," or "for instance" is intended to present the relevant concepts in a concrete manner.

[0079] In the description of the embodiments of this application, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The "and / or" in the text is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of this application, "multiple" means two or more.

[0080] In some processes described in the embodiments of this application, multiple operations or steps are included in a specific order. However, it should be understood that these operations or steps may not be executed in the order they appear in the embodiments of this application, or they may be executed in parallel. The sequence number of the operation is only used to distinguish different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed sequentially or in parallel, and these operations or steps may be combined.

[0081] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device to execute the methods described in the various embodiments of this application.

[0082] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A method for predicting the remaining lifespan of a battery, characterized in that, The method for predicting the remaining battery life includes: Based on electrochemical impedance spectroscopy testing of batteries under multiple aging states, impedance datasets are obtained, an equivalent circuit model of the battery is constructed, and relaxation time distribution modeling is combined with the modeling. Mechanistic features of the battery are extracted from the electrochemical impedance spectroscopy to form a mechanism feature dataset. By integrating impedance datasets and mechanistic feature datasets, correlation analysis is used to select features that are strongly correlated with battery performance, thus obtaining the target feature dataset. Input the target feature dataset into the preset prediction model and output the prediction result of the remaining battery life.

2. The battery remaining life prediction method as described in claim 1, characterized in that, The electrochemical impedance spectroscopy test of the battery under multiple aging conditions to obtain impedance data sets includes: Based on the battery aging test, complex impedance values ​​of different frequency bands were collected at multiple aging stages; An impedance dataset is constructed by integrating multi-band complex impedance values ​​under multiple aging conditions.

3. The battery remaining life prediction method as described in claim 1, characterized in that, The equivalent circuit model of the battery is constructed and combined with relaxation time distribution modeling to extract the mechanistic features of the battery from the electrochemical impedance spectroscopy, forming a mechanistic feature dataset, including: Construct a second-order equivalent circuit model to characterize the electrochemical impedance process of a battery; A relationship function between the electrochemical impedance spectrum and the relaxation time distribution is established based on the relaxation time distribution. The fitted impedance data is obtained by optimization through discretization and regularization. The component parameters of the second-order equivalent circuit model are calculated based on the fitted impedance data, and the peak information of the relaxation time distribution curve is combined to form a mechanism feature dataset.

4. The battery remaining life prediction method as described in claim 3, characterized in that, The impedance data obtained through discretization and regularization optimization includes: Using the Gaussian function as the kernel function and the radial basis functions as the basis functions, the objective function is constructed based on the discretization formula; After optimizing the polarization ratio and time constant of the discretized formula using an intelligent optimization algorithm, the optimal polarization ratio and time constant are obtained by solving the objective function. By substituting the optimal polarization ratio and time constant into the discretization formula, the real and imaginary parts of the impedance corresponding to the measured frequency band are calculated, thus obtaining the fitted impedance data.

5. The battery remaining life prediction method as described in claim 3, characterized in that, The component parameters for calculating the second-order equivalent circuit model based on the fitted impedance data include ohmic resistance, solid electrolyte interface resistance, and charge transport polarization resistance.

6. The battery remaining life prediction method as described in claim 1, characterized in that, The fused impedance dataset and mechanistic feature dataset are used to screen out features strongly correlated with battery performance through correlation analysis, resulting in a target feature dataset, including: By merging the impedance dataset and the mechanism feature dataset, a complete impedance feature set is obtained. Voltage data of the battery under multiple aging conditions were collected synchronously to construct a voltage dataset. The correlation between each feature in the complete impedance feature set and the voltage data was calculated using the Pearson correlation coefficient formula. By setting a correlation factor, weakly correlated features with an absolute correlation value less than the correlation factor are removed, and the remaining features constitute the target feature dataset.

7. The battery remaining life prediction method as described in claim 1, characterized in that, The method for predicting the remaining battery lifespan also includes the following steps: By collecting multi-dimensional data and lifespan labels of sample batteries throughout their entire life cycle to construct a training set, and after preprocessing and feature selection, the initial neural network model is trained with the target feature data as input and the lifespan label as output, and the hyperparameters are optimized to obtain a converged preset prediction model.

8. The battery remaining life prediction method as described in claim 7, characterized in that, The process of constructing a training set by collecting multi-dimensional data and lifespan labels throughout the entire lifecycle of sample batteries includes: The system collects impedance and mechanistic characteristics data of sample batteries under multiple aging states throughout their entire life cycle; it also simultaneously collects the actual remaining service life tags corresponding to each aging state. Impedance data, mechanistic feature data and corresponding actual remaining useful life labels are associated to construct a training sample set including input features and output labels.

9. The battery remaining life prediction method as described in claim 7, characterized in that, After preprocessing and feature selection, an initial neural network model is trained and hyperparameters are optimized using target feature data as input and lifetime labels as output to obtain a converged preset prediction model, including: The training sample set is preprocessed by data normalization and outlier removal. Then, the correlation between each feature and the battery performance index of the sample is calculated by the Pearson correlation coefficient formula. The correlation factor is set and weakly correlated features with an absolute correlation value less than the correlation factor are removed to obtain the target feature dataset for training. Using the target feature dataset for training as input and the actual remaining lifespan label as output, the initial neural network model is trained. Hyperparameters such as learning rate and number of convolution kernels are optimized through cross-validation to obtain a converged preset prediction model.

10. A battery remaining life prediction device, characterized in that, The battery remaining lifespan prediction device includes: The mechanism feature extraction module is used to obtain impedance data sets based on electrochemical impedance spectroscopy testing under multiple aging states of the battery, construct the equivalent circuit model of the battery and combine it with relaxation time distribution modeling, extract the mechanism features of the battery from the electrochemical impedance spectrum, and form a mechanism feature dataset. The feature selection module is used to fuse impedance datasets and mechanism feature datasets, and select features that are strongly correlated with battery performance through correlation analysis to obtain the target feature dataset. The lifespan prediction module is used to input the target feature dataset into a preset prediction model and output the prediction result of the remaining battery lifespan.

11. The battery remaining life prediction device as described in claim 10, characterized in that, The mechanism feature extraction module is also used to perform aging tests based on the battery, collect complex impedance values ​​of different frequency bands at multiple aging stages, and integrate the complex impedance values ​​of multiple frequency bands under multiple aging states to construct an impedance dataset.

12. The battery remaining life prediction device as described in claim 10, characterized in that, The mechanism feature extraction module is also used to construct a second-order equivalent circuit model characterizing the electrochemical impedance process of the battery; establish a relationship function between the electrochemical impedance spectrum and the relaxation time distribution based on the relaxation time distribution, and obtain the fitted impedance data through discretization and regularization optimization; calculate the component parameters of the second-order equivalent circuit model based on the fitted impedance data, and integrate them with the peak information of the relaxation time distribution curve to form a mechanism feature dataset.

13. The battery remaining life prediction device as described in claim 12, characterized in that, The mechanism feature extraction module is also used to construct an objective function based on the discretization formula, using the radial basis function with the Gaussian function as the kernel function as the basis function; after optimizing the polarization ratio and time constant of the discretization formula using an intelligent optimization algorithm, the optimal polarization ratio and time constant are obtained by solving the objective function; the optimal polarization ratio and time constant are substituted into the discretization formula to calculate the real part and imaginary part of the impedance corresponding to the measured frequency band, that is, to obtain the fitted impedance data.

14. A battery remaining life prediction device, characterized in that, The battery remaining life prediction device includes a processor, a memory, and a battery remaining life prediction program stored in the memory and executable by the processor, wherein when the battery remaining life prediction program is executed by the processor, it implements the steps of the battery remaining life prediction method as described in any one of claims 1 to 9.

15. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a battery remaining life prediction program, wherein when the battery remaining life prediction program is executed by a processor, it implements the steps of the battery remaining life prediction method as described in any one of claims 1 to 9.

Images