Battery health state estimation method and system

Abstract

The invention discloses a battery health state estimation method and system, and the method comprises the steps: extracting the strong correlation features of a battery health state through fusing a physical model and relaxation time based on electrochemical impedance spectroscopy data, and carrying out the feature screening and reconstruction through a random forest algorithm, and then optimizing network parameters of the multi-scale kernel extreme learning machine by using a sparrow algorithm so as to estimate the state of health of the battery. The estimation method provided by the invention can capture data features from multiple angles, has good generalization ability, and has significant advantages in the aspects of precision and efficiency compared with an existing estimation method.

Description

Technical Field

[0001] The present invention relates to a method and system for estimating the state of health of a battery, belonging to the technical field of battery state of health estimation. Background Art

[0002] The energy crisis and environmental deterioration are important challenges faced by the sustainable development of human society. Worldwide, governments have successively introduced relevant policies to accelerate the replacement of fossil energy and achieve the low-carbon transformation of the energy structure. In the field of transportation, the electrification transformation of vehicles can reduce 11.9% of global greenhouse gas emissions, so electric vehicles have developed rapidly. Lithium-ion batteries (LIBs) are the main power sources of current electric vehicles because of their high energy density, high power density, long cycle life and other advantages. Accurate assessment of the state of health of the battery is a key technology to ensure the efficient and safe operation of the battery and improve the vehicle use experience. However, under complex road conditions and disordered charging and discharging conditions, the battery aging attenuation mechanism is complex and highly non-linear. Therefore, the estimation of the state of health (SOH) of electric vehicle batteries is a current research hotspot.

[0003] Electric vehicles usually use a battery management system (BMS) to monitor the battery voltage and current states in real time, and can evaluate state parameters such as battery SOH, realize battery charge and discharge control, and ensure the reliable operation of the battery. Battery SOH is a key indicator to quantify the degree of battery aging. Accurate SOH estimation will improve the maintenance efficiency of the battery and enhance the reliability of the vehicle.

[0004] Electrochemical impedance spectroscopy (EIS), as an effective non-destructive testing technology, can provide in-depth information on the internal state of the battery. By analyzing the EIS data, the electro-chemical reaction kinetics and aging mechanism of the battery can be mapped, thus providing a reliable characteristic basis for SOH estimation. Extreme learning machine (ELM), as an efficient estimation model, has achieved remarkable results in multi-field research, but its application in the field of battery SOH is relatively less. Existing research usually ignores the automation problem of its hyperparameter tuning and has less research on variant extreme learning machines. Summary of the Invention

[0005] The present invention provides a method and system for estimating the state of health of a battery, aiming to combine electrochemical impedance spectroscopy and extreme learning machine to achieve more accurate and effective estimation of the state of health of the battery.

[0006] To achieve the above object, the present invention is implemented by adopting the following technical solutions:

[0007] In a first aspect, the present invention provides a method for estimating the state of health of a battery, including the following steps:

[0008] Collect the electrochemical impedance spectroscopy data of the battery;

[0009] Perform physical model modeling on the electrochemical impedance spectroscopy data and obtain the relaxation time distribution curve, and extract the component parameter characteristics and curve characteristics;

[0010] Use the random forest algorithm to screen the parameter characteristics and curve characteristics, and obtain the key characteristics strongly related to the battery health state;

[0011] Use the multi-scale kernel extreme learning machine optimized by the sparrow algorithm to construct a battery health state estimation model, take the key characteristics as the input, and estimate the battery health state.

[0012] Furthermore, the physical model modeling of the electrochemical impedance spectroscopy data and the extraction of component parameter characteristics include: passing the electrochemical impedance spectroscopy data through physical model simulation to synthesize a second-order equivalent circuit model to form a physical model feature set;

[0013] The physical model features include: the resistance values of resistors R0, R1, and R2 , , , the admittance coefficients Q1, Q2 and phase characteristic values α1, α2 of constant phase elements CPE1, CPE2, and the resistance value Wo1-R and time constant Wo1-T related to the diffusion process of Warburg resistance W.

[0014] Furthermore, the acquisition of the electrochemical impedance spectroscopy data of the battery includes: using an electrochemical workstation to perform electrochemical impedance spectroscopy tests on the battery, and collecting electrochemical impedance spectroscopy data according to the frequency range, frequency resolution, voltage range, and current range information.

[0015] Furthermore, the obtaining of the relaxation time distribution curve for the collected electrochemical impedance spectroscopy data and the extraction of curve characteristics include:

[0016] Pass the electrochemical impedance spectroscopy data through the open-source software DRTtools to obtain the relaxation time distribution curve, and extract the peak height, peak position, valley height, valley position, half-peak area, the ratio of each peak height to the sum of peak heights, and the ratio of each valley height to the sum of valley heights from the relaxation time distribution curve as the relaxation time feature set; among them, the relaxation time distribution calculation formula is as follows:

[0017] ;

[0018] In the formula, L, , j, ω, and τ are inductance, ohmic resistance, complex unit, angular frequency, and characteristic time constant respectively, is the time constant distribution of different reaction processes.

[0019] Furthermore, the use of the random forest algorithm to screen the parameter characteristics and curve characteristics includes:

[0020] According to the component parameter characteristics and curve characteristics, the random forest algorithm is used to capture the interaction between features;

[0021] The random forest algorithm automatically considers the interaction relationship between different features through the splitting mechanism of decision trees. By combining multiple trees with different features, complex feature interactions are found and their importance is evaluated;

[0022] Rank the features according to their importance, select the features with higher importance rankings in each battery, and select their intersection as the input of the battery health state estimation model.

[0023] Furthermore, the kernel of the multi-scale kernel extreme learning machine is a composite kernel formed by weighting multiple RBF kernels, and the output of the multi-scale kernel extreme learning machine is expressed as:

[0024] ;

[0025] ;

[0026] ;

[0027] In the formula: is the multi-scale composite kernel matrix of weighted combination, is the output weight vector considering the regularization parameter, is the regularization parameter, I is the identity matrix, T is the target output matrix of the training data, is the kernel matrix, is the weight coefficient, N represents the number of kernels in the multi-scale kernel extreme learning machine, and i is the index of the kernel function, which is used to traverse all kernel functions.

[0028] Furthermore, the output of the RBF kernel extreme learning machine is:

[0029]

[0030] In the formula, λ is the regularization parameter, I is the identity matrix, T is the target output matrix of the training data, and H is the kernel matrix, where H is defined as:

[0031] ;

[0032] In the formula: and are the samples of the input data, and γ is the kernel parameter.

[0033] Furthermore, the multi-scale kernel extreme learning machine uses multiple regularization parameters; among them, the final output result is expressed as:

[0034] ;

[0035] In the formula, is the number of estimations, that is, the number of regularization parameters. After times of estimation, the average value is taken as the final output of the model.

[0036] Furthermore, the hyperparameters of the battery state of health estimation model are optimized using the sparrow algorithm:

[0037] The hyperparameters optimized by the sparrow algorithm are: kernel parameter , , …, ; regularization parameter , , …, ; weight coefficient , , , …, ; where is the number of RBF kernels, is the number of estimations; the sparrow algorithm takes the root mean square error (RMES) as the optimization objective, and obtains the optimal parameters of the model by discovering the position of the discoverer, updating the position of the follower and calculating the fitness in each iteration; in the sparrow algorithm, the position of the discoverer in each iteration follows the formula:

[0038] ;

[0039] In the formula, represents the position of the th sparrow in the th dimension; represents the current iteration number; represents the upper limit of the iteration number; represents a random number; represents the warning value;; ST represents the safety threshold; Q represents a random number that follows a normal distribution with a mean of 0 and a variance of 1; L represents a 1×d-dimensional vector with a value of 1, and d represents the dimension of the space;

[0040] The description formula for updating the follower position is:

[0041] ;

[0042] In the formula, represents the best position at the current moment, represents the position with the worst fitness;; A represents a d×d-dimensional matrix; L represents a 1×d-dimensional vector with a value of 1; i≤n / 2 means that the

[0043] th individual encounters danger and updates its position through anti-predation behavior as described below:

[0044] ;

[0045] In the formula, represents the parameter for adjusting the step size, represents the parameter for controlling the step size, represents the fitness value of the represents the best fitness value in the current sparrow population, represents the worst fitness value in the current sparrow population, and a represents the smallest constant to avoid a zero denominator.

[0046] In a second aspect, the present invention provides a battery health state estimation system for performing any of the above battery health estimation methods, including:

[0047] A battery data acquisition module for acquiring electrochemical impedance spectroscopy data;

[0048] A data processing module for performing physical model modeling on the electrochemical impedance spectroscopy data, extracting component parameter features and obtaining relaxation time distribution curve features, and screening and obtaining key features strongly related to the battery health state by using a random forest algorithm;

[0049] A battery health state estimation model, which is constructed by a multi-scale kernel extreme learning machine optimized by a sparrow algorithm, for estimating the health state of the battery according to the key features.

[0050] Compared with the prior art, the beneficial effects achieved by the present invention:

[0051] The present invention provides a battery health state estimation method. Through a comprehensive feature extraction method that combines a physical model and relaxation time feature analysis, it well compromises the interpretability of the battery aging mechanism and the estimation accuracy of the battery health state.

[0052] In addition, the present invention uses a multi-scale kernel extreme learning machine optimized by a sparrow algorithm as the estimation model, enabling the model to capture complex non-linear relationships in the data from multiple perspectives and having a more stable estimation ability. Brief Description of the Drawings

[0053] Figure 1 is a flowchart of a battery health estimation method provided in Embodiment 1 of the present invention;

[0054] Figure 2 is a schematic diagram of an equivalent circuit (ECM) used for extracting physical model parameter features based on electrochemical impedance spectroscopy (EIS) in Embodiment 1 of the present invention;

[0055] Figure 3It is a schematic diagram of the distribution of relaxation times (DRT) curve used to extract relaxation time characteristics based on electrochemical impedance spectroscopy (EIS) in Embodiment 1 of the present invention;

[0056] Figure 4 It is the structural diagram of the estimation model in Embodiment 1 of the present invention;

[0057] Figure 5 It is a schematic diagram of the estimation results for battery samples 25C01, 25C02, 25C05, and 25C06 in Embodiment 1 of the present invention;

[0058] Figure 6 It is a schematic diagram of the structure of a computer device in Embodiment 4 of the present invention. Detailed implementation manners

[0059] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and cannot be used to limit the protection scope of the present invention.

[0060] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "center", "longitudinal", "lateral", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be understood as a limitation of the present invention. In addition, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined with "first", "second", etc. may explicitly or implicitly include one or more of such features. In the description of the present invention, unless otherwise specified, the meaning of "plurality" is two or more.

[0061] In the description of the present invention, it should be noted that unless otherwise clearly defined and limited, the terms "mounted", "connected", "coupled" should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection, an electrical connection; it may be directly connected, or indirectly connected through an intermediate medium, and it may be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0062] Embodiment 1

[0063] As Figure 1 shown, this embodiment introduces a battery health estimation method, including the following steps:

[0064] S1, collect the electrochemical impedance spectroscopy (EIS) data of the battery.

[0065] Specifically, it includes: using an electrochemical workstation to conduct the electrochemical impedance spectroscopy test of the battery, and collecting data through the frequency range, frequency resolution, voltage range, and current range information.

[0066] S2, perform physical model modeling and relaxation time distribution curve calculation on the collected electrochemical impedance spectroscopy data, and extract the component parameter characteristics and curve characteristics.

[0067] As Figure 2 shown, specifically, it includes: passing the electrochemical impedance spectroscopy data through physical model simulation to synthesize a second-order equivalent circuit model, forming a physical model feature set;

[0068] Through equivalent circuit (ECM) fitting, the physical model characteristics extracted are respectively: the resistance values of resistors R0, R1, and R2 , , , the admittance coefficients Q1, Q2 and phase characteristic values α1, α2 of constant phase elements CPE1, CPE2, and the resistance value Wo1-R and time constant Wo1-T related to the diffusion process of Warburg resistor W.

[0069] Use the open-source software DRTtools to calculate the relaxation time distribution curve from the electrochemical impedance spectroscopy data, and extract the peak height, peak position, valley height, valley position, half-peak area, the ratio of each peak height to the sum of peak heights, and the ratio of each valley height to the sum of valley heights from the relaxation time distribution (DRT) curve as the relaxation time feature set; among them, the relaxation time distribution calculation formula is as follows:

[0070] ;

[0071] In the formula, L, , j, ω, and τ are inductance, ohmic resistance, complex unit, angular frequency, and characteristic time constant respectively, is the time constant distribution of different reaction processes.

[0072] S3, use the random forest algorithm to screen the parameter characteristics and curve characteristics, and obtain the key characteristics strongly related to the battery health state.

[0073] Specifically, it includes: according to the component parameter characteristics and curve characteristics, use the random forest algorithm to capture the interaction between features.

[0074] The random forest algorithm automatically considers the interaction relationship between different features through the splitting mechanism of decision trees, and finds complex feature interactions and evaluates their importance by combining multiple trees of different features.

[0075] Rank the importance of features, select the features with higher importance rankings in each battery, select their intersection to obtain the key features, and use the key features as the input of the battery health state estimation model.

[0076] It should be added that the above steps for obtaining key features can be used to generate a training dataset. By obtaining existing electrochemical impedance spectroscopy data including known battery health state information, processing the electrochemical impedance spectroscopy data through the above steps, and using the selected key features and the corresponding battery health state information for training the battery health state estimation model.

[0077] S4. Construct a battery health state estimation model using a multi-scale kernel extreme learning machine (SSA-MSK-ELM) optimized by the sparrow algorithm (SAA), use the key features as the input, and estimate the battery health state.

[0078] Among them, in order to capture data features at multiple scales, a kernel extreme learning machine, a multi-scale kernel extreme learning machine is adopted, and its kernel is set to a composite kernel formed by weighting three RBF kernels. The output of the RBF kernel extreme learning machine is expressed as:

[0079] ;

[0080] ;

[0081] ;

[0082] In the formula: is the weighted combined multi-scale composite kernel matrix, is the output weight vector considering the regularization parameter, is the regularization parameter, I is the identity matrix, T is the target output matrix of the training data, is the kernel matrix, is the weight coefficient, N represents the number of kernels in the multi-scale kernel extreme learning machine, i is the index of the kernel function, used to traverse all kernel functions, and its framework is as shown in Appendix Figure 4 shown.

[0083] In addition, the extreme learning machine is a feedforward neural network. Among them, the output of the RBF kernel extreme learning machine is:

[0084] ;

[0085] In the formula, λ is the regularization parameter, I is the identity matrix, T is the target output matrix of the training data, H is the kernel matrix, and H is defined as:

[0086] ;

[0087] In the formula: and are samples of input data, and γ is the kernel parameter.

[0088] Among them, in order to enable the model to have good generalization ability, the multi-scale kernel extreme learning machine used uses multiple regularization parameters. Among them, the final output result is expressed as:

[0089] ;

[0090] In the formula, is the number of estimations, that is, the number of regularization parameters. After times of estimation, its average value is taken as the final output of the model.

[0091] It should be added that the estimation model accuracy depends to a large extent on the setting of model parameters. Therefore, in this embodiment, the hyperparameters of the battery health state estimation model are optimized using the sparrow algorithm;

[0092] The hyperparameters optimized by the sparrow algorithm are: kernel parameter , , …, ; regularization parameter , , …, ; weight coefficient , , , …, . Among them, is the number of RBF kernels, is the number of estimations. The sparrow algorithm takes the root mean square error (RMES) as the optimization goal, and obtains the optimal parameters of the model by discovering the position of the discoverer, updating the position of the follower and calculating the fitness in each iteration. In the sparrow algorithm, the position of the discoverer in each iteration follows the formula:

[0093] ;

[0094] In the formula, represents the position of the th sparrow in the th dimension; represents the number of this iteration; represents the upper limit of the number of iterations; represents a random number; represents the warning value; ST represents the safety threshold; Q represents a random number that follows a normal distribution with a mean of 0 and a variance of 1; L represents a 1×d-dimensional vector with a value of 1, and d represents the dimension of the space.

[0095] The description formula for updating the follower position is:

[0096] ;

[0097] Wherein, represents the best position at the current moment, represents the position with the worst fitness, A represents a d×d dimensional matrix; L represents a 1×d dimensional vector with a value of 1; i≤n / 2 means that the i-th follower is near the best position.

[0098] When an individual encounters danger, the anti-predation behavior updates its position as described below:

[0099] ;

[0100] Wherein, represents the parameter for adjusting the step size, represents the parameter for controlling the step size, represents the fitness value of the th sparrow at this time, represents the best fitness value in the current sparrow population,

[0101] The following combines a specific embodiment to illustrate the content involved in the above embodiments.

[0102] In this embodiment, it includes steps of using an open-source dataset to extract and screen features to estimating the state of health (SOH) of the battery using an estimation model (SSA-MSK-ELM). The results are as shown in the appendix Figure 5 as follows.

[0103] The experimental data of this embodiment comes from the Cavendish Laboratory of the University of Cambridge in the UK. This embodiment uses the data with a state of charge (SOC) of 100% at 25°C, and selects the capacity data and impedance data of four batteries, namely 25C01, 25C02, 25C05, and 25C06, in this open-source dataset.

[0104] Through equivalent circuit (ECM) fitting, the extracted physical model features include: the resistance values of resistors R0, R1, and R2 , , , the admittance coefficients Q1, Q2 and phase characteristic values α1, α2 of constant phase elements CPE1, CPE2, and the resistance value Wo1-R and time constant Wo1-T related to the diffusion process of Warburg resistance W.

[0105] To neutralize the physical explanation of the battery aging mechanism by ECM and the requirement for estimation accuracy, through the distribution of relaxation time (DRT) curve, the peak height, peak position, and half-peak integral of the four highest peaks in the curve are extracted, as well as the valley height and valley position of the four valleys close to the four peaks. The ratios of the four peak heights to the sum of the peak heights and the ratios of the four valley heights to the sum of the valley heights are calculated. Their characteristics are respectively Peak Height 1-4, Peak Position 1-4, Half-Peak Area 1-4, Valley Height 1-4, Valley Position 1-4, Peak-to-Total Peak Ratio 1-4, and Valley-to-Total Valley Ratio 1-4.

[0106] The 9-dimensional features of the ECM and the 28-dimensional features of the DRT extracted in this embodiment together extract 37-dimensional features of four battery samples (25C01, 25C02, 25C05, and 25C06). Through the random forest algorithm, the aging health features extracted from each battery are analyzed for correlation with the SOH during its aging process, and the 37-dimensional features are ranked according to importance. Considering the factor of the dimensionality of the model input features, the top 18 groups of features with the highest importance rankings among the four batteries are selected, and their intersection is selected. A total of 9-dimensional feature data is used as the input of the battery health state estimation model. The finally selected 9-dimensional health features are: R0, Wo1-R, Valley-to-Total Valley Ratio 1, Valley Height 2, Half-Peak Area 2, Peak Height 3, Peak Position 3, Valley Position 3, and Half-Peak Area 3.

[0107] The 9-dimensional input features screened by the above random forest algorithm are input into the estimation model (SSA-MSK-ELM) for battery SOH estimation. Using the method of cross-validation, one of the four battery sample data is selected as the test set, and the remaining three are used as the training set. By taking turns using each battery data as the test set and the remaining data as the training set to train the battery health state estimation model, the SOH of the entire battery life cycle can be estimated, which can ensure the performance of the model on different battery data and thus improve the generalization ability of the model. The obtained estimation results are as shown in the appendix Figure 5 shown, where, in the upper left corner Figure 5 (a) is the estimation result when the 25C01 battery sample is used as the test set, in the upper right corner Figure 5 (b) is the estimation result when the 25C02 battery sample is used as the test set, in the lower left corner Figure 5 (c) is the estimation result when the 25C05 battery sample is used as the test set, in the lower right corner Figure 5(d) Estimation results when the 25C06 battery samples are used as the test set. The MAE and RMSE of the SSA-MSK-ELM model for the SOH estimation values of the four battery samples are 0.00942, 0.00901, 0.01449, 0.02175 and 0.01263, 0.01477, 0.01747, 0.02549 respectively.

[0108] Example 2

[0109] This example provides a battery health state estimation system, including:

[0110] A battery data acquisition module for acquiring electrochemical impedance spectroscopy data;

[0111] A data processing module for performing physical model modeling on the electrochemical impedance spectroscopy data, extracting component parameter features and obtaining relaxation time distribution curve features, and screening and obtaining key features strongly correlated with the battery health state by using the random forest algorithm;

[0112] A battery health state estimation model, which is constructed by a multi-scale kernel extreme learning machine optimized by the sparrow algorithm. First, the battery health state estimation model is trained using the key features, and the battery health state estimation model can be used to estimate the battery health state after training.

[0113] For the specific function implementation of the above modules, refer to the relevant content in the method of Example 1, which will not be elaborated here.

[0114] Example 3

[0115] This example provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the steps of the method described in any one of Example 1.

[0116] Example 4

[0117] This example provides a computer device, including: a processing unit, a storage system, and a connection bus. The device includes various computer system-readable media, which are available media that can be accessed by the device. The storage system includes computer system-readable media in the form of volatile memory. The processing unit executes various functions to implement the steps of the method described in any one of Example 1 by running the program stored in the storage system.

[0118] Example 5

[0119] This example provides a computer program product, including computer programs / instructions. When the computer programs / instructions are executed by a processor, they implement the steps of the method described in any one of Example 1.

[0120] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art of this technology, without departing from the technical principles of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

[0121] Those skilled in the art should understand that the embodiments of the present disclosure can be provided as methods, systems, or computer program products. Therefore, the present disclosure can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present disclosure can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0122] The present disclosure is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present disclosure. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the specified functions in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0123] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device that implements the specified functions in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0124] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the specified functions in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0125] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present disclosure rather than limit the scope of its protection. Although the present disclosure has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that after reading the present disclosure, various changes, modifications or equivalent substitutions can still be made to the specific implementation manners of the invention, but these changes, modifications or equivalent substitutions are all within the scope of protection of the claims pending for publication.

Claims

1. A method for estimating a battery health state, characterized in that: The following steps are involved: Collect electrochemical impedance spectroscopy data of the battery; Performing physical model building and relaxation time distribution curve calculation on the electrochemical impedance spectroscopy data, and extracting component parameter characteristics and curve characteristics; The parameter characteristics and curve characteristics are screened by using a random forest algorithm to obtain key characteristics that are strongly correlated with the battery health status; A battery health state estimation model is constructed using a multi-scale kernel extreme learning machine optimized by the sparrow algorithm, and the key features are used as input to estimate the battery health state.

2. The battery health status estimation method according to claim 1, characterized in that: The step of performing physical modeling on the electrochemical impedance spectrum data and extracting parameter characteristics of components includes: simulating and synthesizing a second-order equivalent circuit model of the electrochemical impedance spectrum data through a physical model to form a physical model feature set; The physical model features include: the resistance values ​​of resistors R0, R1 and R2 , , , the admittance coefficients Q1, Q2 and phase characteristic values ​​α1, α2 of the constant phase elements CPE1, CPE2, and the resistance value Wo1-R and time constant Wo1-T of the Warburg resistor W related to the diffusion process.

3. The battery health status estimation method according to claim 1, characterized in that: The collecting of the electrochemical impedance spectrum data of the battery comprises: using an electrochemical workstation to perform an electrochemical impedance spectrum test on the battery, and collecting the electrochemical impedance spectrum data according to frequency range, frequency resolution, voltage range and current range information.

4. The battery health status estimation method according to claim 2, characterized in that: The relaxation time distribution curve of the collected electrochemical impedance spectroscopy data is obtained to extract the curve characteristics, including: The relaxation time distribution curve was obtained from the electrochemical impedance spectroscopy data using the open source software DRTtools. The peak height, peak position, valley height, valley position, half-peak area, the ratio of each peak height to the sum of peak heights, and the ratio of each valley height to the sum of valley heights were extracted from the relaxation time distribution curve as the relaxation time feature set; the relaxation time distribution calculation formula is as follows: ； Where L, , j, ω and τ are inductance, ohmic resistance, complex unit, angular frequency and characteristic time constant respectively, is the time constant distribution of different reaction processes.

5. The battery health status estimation method according to claim 1, characterized in that: The method of screening the parameter characteristics and curve characteristics by using the random forest algorithm includes: According to the parameter characteristics of components and curve characteristics, the random forest algorithm is used to capture the interaction between the characteristics; The random forest algorithm automatically considers the interaction between different features through the splitting mechanism of the decision tree. By combining multiple trees with different features, it finds complex feature interactions and evaluates their importance. The features are ranked by importance, and the features with the highest importance in each battery are taken out, and their intersection is selected as the input of the battery health status estimation model.

6. The battery health status estimation method according to claim 1, characterized in that: The kernel of the multi-scale kernel extreme learning machine is a composite kernel formed by weighting multiple RBF kernels. The output of the multi-scale kernel extreme learning machine is It is expressed as: ； ； ； Where: is the weighted combination of multi-scale composite kernel matrices, is the output weight vector taking into account the regularization parameter, is the regularization parameter, I is the identity matrix, T is the target output matrix of the training data, is the kernel matrix, is the weight coefficient, N represents the number of kernels in the multi-scale kernel extreme learning machine, and i is the index of the kernel function, which is used to traverse all kernel functions.

7. The battery health status estimation method according to claim 6, characterized in that: Output of RBF kernel extreme learning machine for: ； Where λ is the regularization parameter, I is the identity matrix, T is the target output matrix of the training data, and H is the kernel matrix, which is defined as: ； Where: and is a sample of input data, and γ is the kernel parameter.

8. The battery health status estimation method according to claim 6, characterized in that: The multi-scale kernel extreme learning machine uses multiple regularization parameters; wherein the final output result is expressed as: ； In the formula, is the number of estimates, that is, the number of regularization parameters, After the estimation, the average value is taken as the final output of the model.

9. The battery health status estimation method according to claim 6, characterized in that: The hyperparameters of the battery health state estimation model are optimized using the Sparrow algorithm: The hyperparameters optimized by the sparrow algorithm are: kernel parameters , , …, ; Regularization parameter , , …, ; Weight coefficient , , , …, ;in, is the number of RBF kernels, is the number of estimates; the sparrow algorithm takes the root mean square error (RMES) as the optimization target, and obtains the optimal parameters of the model by finding the position of the finder, updating the position of the follower and calculating the fitness in each iteration; in the sparrow algorithm, the position of the finder in each iteration obeys the formula: ； In the formula, Indicates A sparrow in the Location in the dimension; Indicates the number of iterations; Indicates the upper limit of the number of iterations; Represents a random number; represents the warning value; ST represents the safety threshold; Q represents a random number with a mean of 0 and a variance of 1 that follows a normal distribution; L represents a 1×d-dimensional vector with a value of 1, where d represents the dimension of the space; The follower position update description is: ； In the formula, Indicates the best position at the current moment. represents the position with the worst fitness; A represents a d×d dimensional matrix; L represents a 1×d dimensional vector with a value of 1; i≤n / 2 means that the i-th follower is near the optimal position; When an individual encounters danger, it performs anti-predator behavior and updates its position description as follows: ； In the formula, represents the parameter used to adjust the step size, represents the parameter controlling the step size, Indicates The fitness value of the sparrow at this time is represents the best fitness value in the current sparrow population, represents the worst fitness value in the current sparrow population, and a represents the minimum constant to avoid the denominator being 0.

10. A battery health status estimation system, characterized in that: The method for estimating battery health according to any one of claims 1 to 9 comprises: Battery data acquisition module, used to collect electrochemical impedance spectroscopy data; The data processing module is used to build a physical model for electrochemical impedance spectroscopy data, extract component parameter characteristics and obtain relaxation time distribution curve characteristics, and use the random forest algorithm to screen and obtain key features that are strongly related to the battery health status; A battery health state estimation model is constructed by a multi-scale kernel extreme learning machine optimized by a sparrow algorithm, and is used to estimate the health state of the battery according to the key features.

Images