Battery health state detection method, device and equipment and storage medium

Abstract

The invention discloses a battery health state detection method, device and equipment and a storage medium, which are applied to the technical field of battery health states, and align a voltage sequence and a current sequence by acquiring the voltage sequence, the current sequence, temperature distribution data and electrochemical impedance spectroscopy parameters of a to-be-detected lithium battery to obtain aligned data; the method comprises the following steps: extracting temperature distribution data to obtain gradient change characteristics, extracting electrochemical impedance spectroscopy parameters to obtain a relaxation time constant, constructing a multi-dimensional degradation index by fusing the gradient change characteristics and the relaxation time constant, and inputting aligned data and the multi-dimensional degradation index into a target integrated learning model for processing. According to the method, the estimation precision of the health state of the battery is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of battery health status detection, and in particular to a battery health status detection method, device, equipment and storage medium. Background Art

[0002] As a complex electrochemical system, the state of health (SOH) of power lithium-ion batteries will gradually degrade with use. It is very important to accurately grasp the SOH of lithium batteries during the use of lithium-ion batteries. There are two main methods to measure the SOH of lithium batteries: one is the ratio of the maximum available capacity of the lithium battery during use to the rated maximum available capacity of the lithium battery; the other is the ratio of the impedance of the lithium battery during use to the impedance of the lithium battery when it is healthy.

[0003] However, in current testing methods, obtaining capacity requires charging the lithium battery to a fully charged state using constant current and constant voltage under relatively stable conditions, then discharging the lithium battery to a cutoff voltage using constant current. After multiple cycles, multiple sets of measurements are obtained, and the average value is taken as the current available capacity of the lithium battery. However, the multi-physics field coupled aging effects caused by the dynamic working conditions during the actual service life of the battery make it impossible for a single parameter to fully represent the health state. Moreover, the generalization ability of the model trained based on a specific battery chemistry system is greatly reduced when migrating across different types, which seriously restricts its engineering applicability. Therefore, it is necessary to break through the traditional single-parameter detection method and construct a detection model that integrates multiple parameters to improve the accuracy of lithium battery SOH detection. Summary of the Invention

[0004] In order to solve the above technical problems, embodiments of the present invention provide a battery health status detection method, apparatus, device and storage medium to solve the technical problem of low accuracy of health status detection of lithium-ion batteries in the prior art.

[0005] A first aspect of an embodiment of the present invention provides a method for detecting a battery health status, the method comprising:

[0006] Obtain voltage sequence, current sequence, temperature distribution data and electrochemical impedance spectroscopy parameters of the lithium battery to be tested;

[0007] Align the voltage series and current series to obtain aligned data, extract the temperature distribution data to obtain gradient change characteristics, and extract the electrochemical impedance spectroscopy parameters to obtain the relaxation time constant;

[0008] A multidimensional degradation index is constructed based on the gradient change characteristics and relaxation time constant. The aligned data and the multidimensional degradation index are input into the target ensemble learning model for processing to obtain the health status estimation value.

[0009] In a possible implementation of the first aspect, obtaining electrochemical impedance spectroscopy parameters of the lithium battery to be tested includes:

[0010] Apply an excitation signal to the lithium battery to be tested;

[0011] The current phase angle of the lithium battery to be tested under the excitation signal is collected, and the characteristic frequencies of the real impedance and imaginary impedance in the current phase angle are extracted to obtain the electrochemical impedance spectroscopy parameters.

[0012] In a possible implementation of the first aspect, aligning the voltage sequence and the current sequence to obtain aligned data includes:

[0013] Using a preset sliding window method to perform local segmentation on the voltage sequence and the current sequence, respectively, to obtain a first segmentation result and a second segmentation result, wherein the first segmentation result and the second segmentation result respectively include a plurality of windows, and each window includes a plurality of voltage data or current data;

[0014] Matching the windows in the first partition result and the second partition result to obtain multiple window pairs, and using the constraint path weight function to obtain the constraint parameters between each window pair;

[0015] According to the constraint parameters, the paths of each window pair are optimized to obtain the aligned data.

[0016] In a possible implementation of the first aspect, constructing a multidimensional degradation index based on gradient change characteristics and relaxation time constants includes:

[0017] According to the relaxation time constant, the diffusion coefficient attenuation factor is derived using the Arrhenius equation and is used as an aging indicator.

[0018] According to the gradient change characteristics and aging indicators, a multidimensional degradation index is constructed.

[0019] In a possible implementation of the first aspect, the target ensemble learning model is obtained by training using battery sample data, including:

[0020] Acquiring battery sample data, wherein the battery sample data includes voltage sequence sample data, current sequence sample data, temperature distribution sample data, and electrochemical impedance spectroscopy parameter sample data;

[0021] The battery sample data is processed using a generative adversarial network to obtain expanded battery sample data, and the expanded battery sample data is extracted to obtain an extraction result;

[0022] The extraction results are screened using the Shapley value algorithm to obtain the screening results, an initial ensemble learning model is constructed, and the screening results are used to train the initial ensemble learning model to obtain the target ensemble learning model.

[0023] In a possible implementation of the first aspect, the extraction results are screened using a Shapley value algorithm to obtain a screening result, including:

[0024] Calculate the contribution of each extraction result;

[0025] The expanded battery sample data having a contribution greater than a preset contribution threshold is screened out to obtain a screening result.

[0026] In a possible implementation of the first aspect, after obtaining the estimated health status value, the method further includes:

[0027] Calculate the confidence level of the health status estimate. If the confidence level is less than a preset threshold, perform a pulse current injection test on the lithium battery to obtain a transient voltage response curve.

[0028] The transient voltage response curve is analyzed to obtain an impedance increment, and the health state estimation value is corrected according to the impedance increment to obtain a corrected health state estimation value.

[0029] In order to solve the same technical problem, a second aspect of an embodiment of the present invention provides a battery health status detection device, including an acquisition module, a processing module and a prediction module, wherein:

[0030] The acquisition module is used to obtain the voltage sequence, current sequence, temperature distribution data and electrochemical impedance spectroscopy parameters of the lithium battery to be tested;

[0031] The processing module is used to align the voltage sequence and the current sequence to obtain the aligned data, extract the temperature distribution data to obtain the gradient change characteristics, and extract the electrochemical impedance spectroscopy parameters to obtain the relaxation time constant;

[0032] The prediction module is used to construct a multidimensional degradation index based on the gradient change characteristics and relaxation time constant. The aligned data and the multidimensional degradation index are input into the target ensemble learning model for processing to obtain the health status estimation value.

[0033] A third aspect of an embodiment of the present invention provides a computer device, including:

[0034] memory for storing computer programs;

[0035] The processor is configured to implement the steps of the battery health status detection method of the first aspect when executing a computer program.

[0036] A fourth aspect of an embodiment of the present invention provides a storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the battery health status detection method of the first aspect are implemented.

[0037] The technical solution of the present invention has the following advantages:

[0038] The battery health status detection method provided by the embodiment of the present invention obtains the voltage sequence, current sequence, temperature distribution data and electrochemical impedance spectrum parameters of the lithium battery to be detected, aligns the voltage sequence and current sequence to obtain aligned data, extracts the temperature distribution data to obtain gradient change characteristics, extracts the electrochemical impedance spectrum parameters to obtain relaxation time constants, and then constructs a multidimensional degradation index by fusing the gradient change characteristics and the relaxation time constant. The aligned data and the multidimensional degradation index are input into the target integrated learning model for processing to obtain a health status estimation value. The above method effectively improves the estimation accuracy of the battery health status. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific implementation methods or the description of the prior art. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0040] Figure 1 This is a flow chart of a battery health status detection method according to an embodiment of the present invention;

[0041] Figure 2 4 is a structural block diagram of a battery leakage identification device in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0043] The battery health status detection method provided by the embodiment of the present invention is as follows: Figure 1 As shown, Figure 1 This is a flow chart of a battery health status detection method, including steps S101 to S103, each of which is as follows:

[0044] S101. Obtain voltage sequence, current sequence, temperature distribution data and electrochemical impedance spectroscopy parameters of the lithium battery to be tested.

[0045] In this embodiment, during the static phase of charge and discharge of a lithium battery, an AC excitation signal in the frequency range of 0.1-10kHz is applied. Simultaneously, a battery management system or other data acquisition device collects voltage, current, and surface temperature distribution data of the lithium battery under test in real time at a set sampling frequency, thereby obtaining voltage and current series and temperature distribution data. The response current phase angle is measured in real time, and the inflection point frequencies of the Nyquist curves of the real impedance Z' and imaginary impedance Z" are extracted as characteristic parameters.

[0046] It should be noted that the voltage sequence refers to a set of continuous voltage data arranged in chronological order, and the current sequence refers to a set of continuous current data arranged in chronological order.

[0047] In one embodiment, obtaining electrochemical impedance spectroscopy parameters of a lithium battery to be tested includes:

[0048] Apply an excitation signal to the lithium battery to be tested;

[0049] The current phase angle of the lithium battery to be tested under the excitation signal is collected, and the characteristic frequencies of the real impedance and imaginary impedance in the current phase angle are extracted to obtain the electrochemical impedance spectroscopy parameters.

[0050] In this embodiment, an AC excitation signal is applied to the detected lithium battery under constant temperature conditions. In actual applications, the excitation signal configuration and acquisition device settings when applying the AC excitation signal can be set according to actual needs, and this embodiment is no longer limited here.

[0051] Then, current phase angle extraction and impedance calculation are performed. Specifically, an orthogonal reference signal with the same frequency as the excitation signal is first constructed. When the excitation signal is applied, the lithium battery to be tested generates a voltage signal and a current signal. The voltage signal is multiplied by the reference signal and then low-pass filtered to obtain the real and imaginary voltage parts. Similarly, the current signal is multiplied by the reference signal and then low-pass filtered to obtain the real and imaginary current parts. The phase angle is calculated based on the real and imaginary voltage parts, the real and imaginary current parts, and the real and imaginary impedance parts are calculated at each frequency point based on the phase angle, the real and imaginary voltage parts, the real and imaginary current parts, and the Nyquist curve inflection point frequency is extracted based on the real and imaginary impedance parts. The Nyquist curve inflection point frequency is used as an electrochemical impedance spectroscopy parameter.

[0052] It should be noted that the calculation process of the phase angle, the real part and the imaginary part of the impedance can be performed according to existing commonly used methods, and the specific calculation method is not limited in this embodiment.

[0053] S102 , aligning the voltage sequence and the current sequence to obtain aligned data, extracting the temperature distribution data to obtain gradient change characteristics, and extracting electrochemical impedance spectroscopy parameters to obtain a relaxation time constant.

[0054] In this embodiment, the voltage sequence and the current sequence are dynamically time-warped and aligned to generate a time-synchronized voltage and current feature matrix. At the same time, the gradient change characteristics of the temperature distribution data and the relaxation time constant of the electrochemical impedance spectroscopy parameters are extracted to construct a multidimensional degradation index. By adopting a dynamic time warping algorithm to eliminate the timing offset caused by the change in charge and discharge rate, the matching accuracy is improved by constraining the path weight function, and the influence of the fluctuation of the charge and discharge working conditions is effectively overcome. In addition, by fusing the temperature gradient distribution and the EIS frequency domain parameters to construct a multidimensional degradation index, the SOH estimation accuracy can be significantly improved.

[0055] In one embodiment, aligning the voltage sequence and the current sequence to obtain aligned data includes:

[0056] Using a preset sliding window method to perform local segmentation on the voltage sequence and the current sequence, respectively, to obtain a first segmentation result and a second segmentation result, wherein the first segmentation result and the second segmentation result respectively include a plurality of windows, and each window includes a plurality of voltage data or current data;

[0057] Matching the windows in the first partition result and the second partition result to obtain multiple window pairs, and using the constraint path weight function to obtain the constraint parameters between each window pair;

[0058] According to the constraint parameters, the paths of each window pair are optimized to obtain the aligned data.

[0059] In this embodiment, a sliding window mechanism is used to locally segment the voltage sequence and the current sequence to obtain a first segmentation result and a second segmentation result. The first segmentation result can be understood as a plurality of windows after the voltage sequence is segmented, and the second segmentation result can be understood as a plurality of windows after the current sequence is segmented, each window including a plurality of voltage data or current data.

[0060] The similarity matching of non-equal length sequences is achieved by constraining the path weight function, eliminating the timing offset error caused by the change of charge and discharge rate. Specifically, the sliding window mechanism is used to divide the voltage sequence and current sequence into local segments respectively, and the similarity matching of non-equal length sequences is achieved by constraining the path weight function, eliminating the timing offset error caused by the change of charge and discharge rate. For example, the window division is performed using the offline mode bidirectional window generation of the window division algorithm. Specifically, according to the battery charge and discharge characteristics, N = 50 sampling points are set, the initial step size S = 5, and it is automatically shortened to S = 2 when the voltage inflection point is detected. The default overlap rate = (NS) / N = 90% ensures the continuity of local features. The improved median filtering algorithm is used to process the voltage sequence and current sequence respectively, and then the voltage and current in each window are dynamically scaled. The scaling formula is:

[0061]

[0062] Where V norm is the normalized voltage value, V i is the voltage data in the voltage sequence, I norm is the normalized current value, I rated is the rated current, I t is the normalized current value, and cosθ is the power factor.

[0063] The windows in the first partition result and the second partition result are matched to obtain multiple window pairs. The constraint path weight function is used to set the constraint parameters between each window pair. The constraint path weight function is ω(i, j) = |ij|, which limits the alignment path offset to no more than 20% of the window length. In addition, the charge and discharge stage constraint matrix is ​​introduced, and the cross-stage matching penalty weight is increased by more than 2 times.

[0064] It should be noted that the specific steps for aligning current and voltage series using the improved DTW algorithm include: designing path constraints to limit the path slope to within the ±δ range of the rate ratio to prevent abnormal offsets. Key point anchoring is then performed to identify switching points and enforce alignment between the constant current and constant voltage switching points. The offset is dynamically calculated based on the rate difference. Multi-granularity window partitioning is then performed to separate plateau and abrupt transition regions. This involves applying loose path constraints (e.g., δ = 0.3) in the plateau region, allowing for localized, minute sequence jitter, while applying strict constraints (e.g., δ = 0.1) in the abrupt transition region, along with a path continuity penalty. Because voltage changes more rapidly at high rates, directly calculating the Euclidean distance will amplify errors in abrupt transition regions. Therefore, a rate compensation mechanism is implemented. Specifically, the differential distance is normalized, while the current term retains its original value. Since current is positively correlated with the rate, normalization is not required. For any two window sequences, a distance matrix is ​​calculated. Based on the constraints defined above, the distance matrix of each window is subjected to optimal path search and post-processing to obtain the aligned data.

[0065] In one embodiment, a multi-dimensional degradation index is constructed based on the gradient change characteristics and the relaxation time constant, including:

[0066] According to the relaxation time constant, the diffusion coefficient attenuation factor is derived using the Arrhenius equation and is used as an aging indicator.

[0067] According to the gradient change characteristics and aging indicators, a multidimensional degradation index is constructed.

[0068] In this example, the lithium ion diffusion coefficient attenuation factor is derived from the relaxation time constant and the Arrhenius equation as an aging indicator. The gradient change characteristics and the aging indicator are then integrated to construct a multidimensional degradation index.

[0069] S103. Construct a multidimensional degradation index based on the gradient change characteristics and the relaxation time constant, input the aligned data and the multidimensional degradation index into the target ensemble learning model for processing, and obtain a health status estimation value.

[0070] In this embodiment, after the temperature distribution data is denoised by Gaussian filtering, the standard deviation of the temperature gradient of each area on the battery surface is calculated. Specifically, the grid area is divided according to the layout of the battery surface temperature sensor, or is divided into a central area, an edge area, and a tab area based on the physical structure, and then the standard deviation of the temperature gradient of each area is calculated. The method for calculating the temperature gradient standard deviation is: first, the temperature distribution data of each area is denoised by Gaussian filtering, and then the data is smoothed.

[0071] In one embodiment, the target ensemble learning model is obtained by training using battery sample data, including:

[0072] Acquiring battery sample data, wherein the battery sample data includes voltage sequence sample data, current sequence sample data, temperature distribution sample data, and electrochemical impedance spectroscopy parameter sample data;

[0073] The battery sample data is processed using a generative adversarial network to obtain expanded battery sample data, and the expanded battery sample data is extracted to obtain an extraction result;

[0074] The extraction results are screened using the Shapley value algorithm to obtain the screening results, an initial ensemble learning model is constructed, and the screening results are used to train the initial ensemble learning model to obtain the target ensemble learning model.

[0075] In this embodiment, an initial ensemble learning model is trained using battery sample data to obtain a target ensemble learning model. During training, the battery sample data is first expanded using a generative adversarial network to obtain expanded battery sample data. This expanded battery sample data includes data on batteries of varying degrees of aging. The battery sample data includes voltage sequence sample data, current sequence sample data, temperature distribution sample data, and electrochemical impedance spectroscopy parameter sample data.

[0076] Data extraction is performed on the expanded battery sample data to obtain extraction results. This can be understood as aligning the voltage sequence sample data and the current sequence sample data to obtain aligned sample data, extracting the temperature distribution sample data to obtain the sample gradient change characteristics, and extracting the electrochemical impedance spectroscopy parameter sample data to obtain the sample relaxation time constant. The extraction results are screened using the Shapley value algorithm to eliminate redundant parameters and optimize the model input dimension, reducing the feature dimension of the input to the initial ensemble learning model to a preset dimension, for example, from 32 dimensions to 18 dimensions. The input features with reduced dimensionality are then input into the initial ensemble learning model for training to obtain the target ensemble learning model.

[0077] In one embodiment, the extraction results are screened using a Shapley value algorithm to obtain screening results, including:

[0078] Calculate the contribution of each extraction result;

[0079] The expanded battery sample data having a contribution greater than a preset contribution threshold is screened out to obtain a screening result.

[0080] In this embodiment, the contribution of each extraction result is calculated, and the expanded battery sample data with a contribution greater than a preset contribution threshold is screened out to obtain a screening result. For example, XGBoost is selected as the base model, and then the SHAP library is used to calculate the marginal contribution of each extraction result, i.e., the Shapley value. The expanded battery sample data with a contribution greater than the preset contribution threshold is screened out to obtain a screening result.

[0081] It should be noted that the preset contribution threshold may be set to 5%.

[0082] In one embodiment, after obtaining the estimated health status value, the method further includes:

[0083] Calculate the confidence level of the health status estimate. If the confidence level is less than a preset threshold, perform a pulse current injection test on the lithium battery to obtain a transient voltage response curve.

[0084] The transient voltage response curve is analyzed to obtain an impedance increment, and the health state estimation value is corrected according to the impedance increment to obtain a corrected health state estimation value.

[0085] In this embodiment, a confidence assessment is performed on the obtained health status estimate. Specifically, an online confidence assessment module is used to determine the confidence level of the health status estimate, which is used to monitor its reliability. When the confidence level falls below a confidence threshold, a pulse current injection test is triggered to obtain a transient voltage response curve. The health status estimate is then dynamically corrected based on the correlation coefficient between the transient voltage response curve and a reference curve.

[0086] The specific steps for dynamically correcting the health status estimate based on the correlation coefficient between the transient voltage response curve and the baseline curve are as follows: extract the time parameter of the second-order derivative extreme point in the voltage recovery phase, and calculate the solid-liquid interface film impedance increment using the equivalent circuit model parameter identification algorithm. For example, when the confidence level is less than 90%, a pulse current injection test is triggered with an amplitude set to 5C and a duration of 2s. The SOH value is corrected using the time difference of the extreme point of the second-order derivative of the transient voltage. Characteristic weight coefficients are set for lithium-ion batteries, lithium iron phosphate batteries, and solid-state batteries respectively. For lithium-ion batteries, the voltage characteristic weight is set to 0.6, and the temperature weight is set to 0.3. For solid-state batteries, the electrochemical impedance spectroscopy parameter characteristic weight is set to 0.7, and the temperature weight is set to 0.2.

[0087] Furthermore, the proposed battery health status detection method is integrated into a battery management system chip. A hardware accelerator is used to parallelly calculate the voltage and current characteristic matrix and the frequency domain characteristics of the electrochemical impedance spectroscopy parameters. The estimated health value is calculated within a single charge and discharge cycle with a response time of less than 50ms. The SOH data frame is then output via the CAN bus with a timestamp.

[0088] This method optimizes the sliding window step size for dynamic time warping and lightweights the ensemble learning model, making it suitable for embedded systems in vehicle-mounted BMSs. The power consumption for a single SOH calculation is less than 10mW. Furthermore, the lightweight model deployed in the embedded BMS is less than 1MB in size and has a response time of ≤200ms.

[0089] The battery health status detection device provided by the embodiment of the present invention is as follows: Figure 2 As shown, Figure 2 The device block diagram of the battery health status detection device 200 includes an acquisition module 201, a processing module 202 and a prediction module 203, wherein:

[0090] The acquisition module 201 is used to obtain the voltage sequence, current sequence, temperature distribution data and electrochemical impedance spectroscopy parameters of the lithium battery to be tested;

[0091] The processing module 202 is used to align the voltage sequence and the current sequence to obtain aligned data, extract the temperature distribution data to obtain gradient change characteristics, and extract the electrochemical impedance spectroscopy parameters to obtain the relaxation time constant;

[0092] The prediction module 203 is used to construct a multidimensional degradation index based on the gradient change characteristics and the relaxation time constant, input the aligned data and the multidimensional degradation index into the target ensemble learning model for processing, and obtain the health status estimation value.

[0093] The specific implementation of the battery health status detection device is basically the same as the specific embodiment of the above-mentioned battery health status detection method, and will not be repeated here.

[0094] In one embodiment of the present application, a computer device is provided, which includes a memory and a processor, wherein a computer program is stored in the memory, and the above steps are implemented when the processor executes the computer program; the computer device provided in this embodiment has an implementation principle and technical effects similar to those of the above method embodiment, and will not be repeated here.

[0095] In one embodiment of the present application, a computer-readable storage medium is provided, on which a computer program is stored, and the above steps are implemented when the computer program is executed by a processor; the computer-readable storage medium provided in this embodiment has an implementation principle and technical effects similar to those of the above method embodiment, and will not be repeated here.

[0096] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0097] The specific embodiments described above further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A battery health status detection method, characterized in that: include: Obtain voltage sequence, current sequence, temperature distribution data and electrochemical impedance spectroscopy parameters of the lithium battery to be tested; Aligning the voltage sequence and the current sequence to obtain aligned data, extracting the temperature distribution data to obtain gradient change characteristics, and extracting the electrochemical impedance spectroscopy parameters to obtain a relaxation time constant; A multidimensional degradation index is constructed according to the gradient change characteristics and the relaxation time constant, and the aligned data and the multidimensional degradation index are input into a target ensemble learning model for processing to obtain a health status estimation value.

2. The battery health status detection method according to claim 1, wherein: The step of obtaining electrochemical impedance spectroscopy parameters of the lithium battery to be tested includes: Applying an excitation signal to the lithium battery to be tested; The current phase angle of the lithium battery to be tested under the excitation signal is collected, and the characteristic frequencies of the real impedance and the imaginary impedance in the current phase angle are extracted to obtain electrochemical impedance spectroscopy parameters.

3. The battery health status detection method according to claim 1, wherein: The aligning the voltage sequence and the current sequence to obtain aligned data includes: Using a preset sliding window method to perform local segmentation on the voltage sequence and the current sequence, respectively, to obtain a first segmentation result and a second segmentation result, wherein the first segmentation result and the second segmentation result respectively include a plurality of windows, and each window includes a plurality of voltage data or current data; Matching windows in the first partition result and the second partition result to obtain a plurality of window pairs, and obtaining constraint parameters between each of the window pairs using a constraint path weight function; Path optimization is performed on each of the window pairs according to the constraint parameters to obtain aligned data.

4. The battery health status detection method according to claim 1, wherein: The constructing of a multidimensional degradation index according to the gradient change characteristics and the relaxation time constant includes: Derivation of a diffusion coefficient attenuation factor using the Arrhenius equation according to the relaxation time constant, and use of the diffusion coefficient attenuation factor as an aging indicator; A multi-dimensional degradation index is constructed based on the gradient change characteristics and the aging index.

5. The battery health status detection method according to claim 1, wherein: The target ensemble learning model is obtained by training using battery sample data, including: Acquiring battery sample data, wherein the battery sample data includes voltage sequence sample data, current sequence sample data, temperature distribution sample data, and electrochemical impedance spectroscopy parameter sample data; Processing the battery sample data using a generative adversarial network to obtain expanded battery sample data, and extracting the expanded battery sample data to obtain an extraction result; The extraction results are screened using a Shapley value algorithm to obtain screening results, an initial ensemble learning model is constructed, and the initial ensemble learning model is trained using the screening results to obtain a target ensemble learning model.

6. The battery health status detection method according to claim 4, wherein: The extraction results are screened using the Shapley value algorithm to obtain screening results, including: Calculating the contribution of each of the extraction results; The expanded battery sample data whose contribution is greater than a preset contribution threshold is screened out to obtain a screening result.

7. The battery health status detection method according to claim 1, wherein: After obtaining the estimated health status value, the method further includes: Calculating the confidence level of the health status estimation value, and if the confidence level is less than a preset threshold, performing a pulse current injection test on the lithium battery to be calculated to obtain a transient voltage response curve; The transient voltage response curve is analyzed to obtain an impedance increment, and the health state estimation value is corrected according to the impedance increment to obtain a corrected health state estimation value.

8. A battery health status detection device, characterized in that: It includes acquisition module, processing module and prediction module, among which, The acquisition module is used to obtain the voltage sequence, current sequence, temperature distribution data and electrochemical impedance spectroscopy parameters of the lithium battery to be tested; The processing module is used to align the voltage sequence and the current sequence to obtain aligned data, extract the temperature distribution data to obtain gradient change characteristics, and extract the electrochemical impedance spectroscopy parameters to obtain a relaxation time constant; The prediction module is used to construct a multidimensional degradation index based on the gradient change characteristics and the relaxation time constant, input the aligned data and the multidimensional degradation index into a target ensemble learning model for processing, and obtain a health status estimation value.

9. A computer device, characterized in that: include: Memory for storing computer programs; A processor, configured to implement the battery health status detection method according to any one of claims 1 to 7 when executing the computer program.

10. A storage medium, characterized in that: The storage medium stores a computer program, which, when executed by a processor, implements the steps of the battery health status detection method according to any one of claims 1 to 7.

Images