A high-adaptability lithium-ion battery state-of-health estimation method and device

Abstract

This invention proposes a highly adaptive method and apparatus for estimating the health status of lithium-ion batteries. The invention includes: S1 extracting health indicators from voltage, current, temperature, and time-series data of battery charging, and filtering them using a smoothing algorithm; S2 evaluating the filtered health indicators using correlation analysis and selecting the optimal health features; S3 modeling the joint distribution of the optimal health features and battery health status based on a Gaussian mixture model, and training the model parameters using an iterative optimization algorithm; S4 adaptively updating the model parameters based on new observation data from the test battery using Bayesian inference, estimating the health status based on conditional probability distribution, and outputting the uncertainty quantification result. This invention can capture multimodal degradation characteristics of batteries, achieve online adaptive model updates, improve cross-battery generalization ability, quantify uncertainty estimation, and accurately and reliably estimate the battery health status.

Description

Technical Field

[0001] This invention relates to the field of energy storage technology, and in particular to a highly adaptable method and apparatus for estimating the health status of lithium-ion batteries. Background Technology

[0002] Lithium-ion batteries are the core power source for new energy vehicles, energy storage systems, and portable electronic devices. Estimating their state of health (SOH) is crucial for ensuring safe and reliable operation throughout their entire lifecycle. Traditional SOH methods mainly include physical model-based and data-driven methods. Mechanistic models such as electrochemical models have clear physical meaning, while data-driven methods such as machine learning and deep learning have gained widespread attention due to their powerful feature learning and nonlinear mapping capabilities. Specifically, this estimation method encompasses a complete technical system, from extracting health indicators from voltage, current, temperature, and time-series data during the charging phase, to using statistical modeling and Bayesian inference for SOH prediction.

[0003] However, existing technologies have significant limitations. Traditional machine learning and deep learning models are typically based on the assumption that degradation features follow a unimodal distribution. This assumption is insufficient for generalizing complex datasets consisting of multiple types, batches, and operating conditions, easily leading to biased prediction results. Furthermore, existing methods generally lack endogenous uncertainty quantification mechanisms, making it difficult to provide reliable confidence interval estimates, thus limiting their practical application value in battery management systems. Summary of the Invention

[0004] The present invention aims to at least partially solve one of the technical problems in the related art.

[0005] Therefore, the first objective of this invention is to propose a highly adaptable method for estimating the health status of lithium-ion batteries.

[0006] Another object of the present invention is to provide a highly adaptable lithium-ion battery health status estimation device.

[0007] The third objective of this invention is to provide a computer device.

[0008] A fourth objective of this invention is to provide a non-transitory computer-readable storage medium.

[0009] To achieve the above objectives, a first aspect of the present invention proposes a highly adaptable method for estimating the state of health of lithium-ion batteries, comprising:

[0010] S1 extracts health indicators from voltage, current, temperature and time series data during the battery charging process, and uses a smoothing algorithm to filter the raw data; S2. The correlation analysis method is used to evaluate the correlation of the filtered health indicators and select the optimal health features with the highest correlation to the battery health status. S3, The joint distribution of the optimal health feature and the battery health state is modeled based on the Gaussian mixture model, and the model parameters are trained using an iterative optimization algorithm; S4. Based on new observation data of the test battery, the parameters of the trained Gaussian mixture model are adaptively updated using the Bayesian inference method. The battery health status is estimated based on the conditional probability distribution, and the uncertainty quantification result is output.

[0011] In one embodiment of the present invention, the step of extracting health indicators from voltage, current, temperature, and time series data during the battery charging process, and filtering the original data using a smoothing algorithm, includes: Four statistical characteristics were calculated for the extracted voltage, current, temperature, and time data, including mean, variance, standard deviation, and range. The voltage, current, temperature, and time series data are smoothed using a Savitzky-Golay filter to suppress noise interference and improve the quality of health indicators.

[0012] In one embodiment of the present invention, the step of using correlation analysis to evaluate the correlation of the filtered health indicators and selecting the optimal health feature with the highest correlation to the battery health status includes: The correlation score between each health indicator and the battery health status was calculated based on the grey relational analysis method. The correlation score is between 0 and 1, where 0 represents no correlation and 1 represents perfect correlation. Based on the correlation score calculated by grey relational analysis, the health indicator with the highest correlation score is selected as the optimal health feature.

[0013] In one embodiment of the present invention, the method of modeling the joint distribution of the optimal health feature and the battery health state based on a Gaussian mixture model and training the model parameters using an iterative optimization algorithm includes: The joint random variable consisting of the optimal health feature matrix and the battery health state vector is modeled as a weighted combination of multiple Gaussian components; The parameters of a Gaussian mixture model are estimated by maximizing the log-likelihood of a given set of parameters using the expectation-maximization algorithm. The mixing coefficients of each Gaussian component, the mean vector of each Gaussian component, and the covariance matrix of each Gaussian component.

[0014] In one embodiment of the present invention, the joint distribution expression of the Gaussian mixture model is:

[0015] in, It is the number of Gaussian components. It is the first Mixing coefficients ( ), The mean is Covariance is Multivariate Gaussian distribution.

[0016] In one embodiment of the present invention, the step of adaptively updating the parameters of the trained Gaussian mixture model using a Bayesian inference method based on new observation data of the test battery, estimating the battery health status based on the conditional probability distribution, and outputting the uncertainty quantification result includes: Based on the newly observed best health indicator vector of the test battery, calculate the posterior probability of each Gaussian component, and select the most likely Gaussian component for health status estimation. The unknown health status is estimated as a normal distribution with the conditional mean and conditional variance of the most likely Gaussian components as parameters, and the health status estimate and its corresponding variance are output as the uncertainty quantification result.

[0017] In one embodiment of the present invention, the optimal health indicator vector newly observed from the test battery includes: The estimated state of health of the battery in the i-th cycle is: ; The variability in health status estimates is: ; Normal distribution Output the conditional probability distribution and uncertainty quantification results of the battery health status, where zi is the health status value of the test battery in the i-th cycle.

[0018] To achieve the above objectives, a second aspect of the present invention provides a highly adaptable lithium-ion battery health status estimation device, comprising: The indicator extraction and data filtering module is used to extract health indicators from voltage, current, temperature and time series data during the battery charging process, and to filter the raw data using a smoothing algorithm. The Health Indicator Correlation Analysis and Optimal Feature Screening Module is used to evaluate the correlation of filtered health indicators using correlation analysis methods, and to screen out the optimal health features with the highest correlation to the battery health status. The model modeling and parameter training module is used to model the joint distribution of the optimal health features and battery health status based on the Gaussian mixture model, and to train the model parameters using an iterative optimization algorithm. The model parameter update and health status estimation and quantification module is used to adaptively update the parameters of the trained Gaussian mixture model based on new observation data of the test battery using the Bayesian inference method, and to estimate the battery health status based on the conditional probability distribution and output the uncertainty quantification results.

[0019] This invention provides a highly adaptive method and apparatus for estimating the health status of lithium-ion batteries. By using a Gaussian mixture model, it effectively captures multimodal degradation features in multi-battery datasets, providing reliable uncertainty quantification while achieving accurate health status estimation. It also features online adaptive updates and cross-battery generalization capabilities.

[0020] To achieve the above objectives, a third aspect of this application provides a computer device, including a processor and a memory; wherein the processor runs a program corresponding to the executable program code by reading executable program code stored in the memory, for implementing the method described in the first aspect embodiment.

[0021] To achieve the above objectives, a fourth aspect of this application provides a non-transitory computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method described in the first aspect.

[0022] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0023] Figure 1 This is a flowchart of a highly adaptable lithium-ion battery health state estimation method according to an embodiment of the present invention; Figure 2 This is a structural diagram of a highly adaptable lithium-ion battery health status estimation device according to an embodiment of the present invention; Figure 3 It is a computer device according to an embodiment of the present invention. Detailed Implementation

[0024] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0025] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0026] The following description, with reference to the accompanying drawings, describes a highly adaptable method and apparatus for estimating the health status of a lithium-ion battery according to an embodiment of the present invention.

[0027] Figure 1 This is a flowchart of a highly adaptable lithium-ion battery health state estimation method according to an embodiment of the present invention, such as... Figure 1 As shown, it includes: S1 extracts health indicators from voltage, current, temperature and time series data during the battery charging process, and uses a smoothing algorithm to filter the raw data; Furthermore, in battery health state estimation, the first step is to extract characteristic information reflecting the battery's degradation state from the time-series data of the battery charging process. This step extracts health indicators from the voltage, current, temperature, and time-series data during the battery charging phase. This time-series data contains the electrochemical behavior characteristics of the battery during charging and discharging. By extracting meaningful statistical features, a correlation mapping between these features and the battery's health state can be established. To improve the quality of the extracted health indicators and suppress noise interference, a smoothing algorithm is used to filter the original data to eliminate high-frequency noise caused by measurement errors and environmental disturbances, while retaining key characteristic information that characterizes the battery degradation trend. As a specific implementation method, statistical characteristics such as mean, variance, standard deviation, and range can be calculated for voltage, current, temperature, and time series data, and the original data can be smoothed using a Savitzky-Golay filter. This step provides a high-quality data foundation for subsequent health indicator screening and health state estimation, effectively improving the signal-to-noise ratio of the characteristic signals and creating conditions for accurately capturing battery degradation patterns.

[0028] Furthermore, four statistical features were calculated for the extracted voltage (V), current (I), temperature (T), and time (T) data, including mean (A), variance (V), standard deviation (S), and range (D). Statistical Feature Set Defined in equation (1).

[0029] (1) in, , , .

[0030] For a given battery cycle, It is the average value of the applied current sequence extracted during constant voltage charging mode. It is the variance of the temperature sequence extracted during constant current charging.

[0031] Furthermore, to improve the quality of the extracted health indicators and suppress noise interference, a Savitzky-Golay filter was introduced to smooth the original data. Subsequently, grey relational analysis was used to screen the processed health indicators to identify the key indicators most correlated with the battery health status, thereby improving the accuracy of health status estimation. Grey relational analysis measures the geometric similarity between a specific health indicator curve and the battery health status curve, calculating a correlation score between 0 and 1: where 0 represents no correlation and 1 represents perfect correlation.

[0032] S2. The correlation analysis method is used to evaluate the correlation of the filtered health indicators and select the optimal health features with the highest correlation to the battery health status. Furthermore, after smoothing and filtering the raw data, the extracted health indicators need to be relevance evaluated to select the optimal health features with the highest correlation to the battery health status. The core purpose of this step is to identify the key features that best characterize the battery degradation state from numerous candidate health indicators, thereby improving the accuracy and computational efficiency of subsequent health status estimation.

[0033] Furthermore, the correlation analysis method quantitatively assesses the statistical correlation between various health indicators and battery health status, thereby enabling feature selection of health indicators. Specifically, based on the principle of geometric similarity, this method measures the closeness between the health indicator curve and the battery health status reference curve, calculating a correlation score between zero and one; the closer the score is to one, the higher the correlation between the health indicator and the battery health status. By setting a reasonable correlation threshold or employing a ranking selection strategy, the optimal subset of health features with the highest correlation can be selected from the original set of health indicators, serving as input for subsequent Gaussian mixture modeling.

[0034] Furthermore, as a specific implementation method, grey relational analysis can be used to screen the filtered health indicators. This method calculates the geometric similarity between a specific health indicator curve and the battery health state curve, obtaining a correlation score between zero and one, where zero represents no correlation and one represents perfect correlation. This allows the selection of the key indicators most correlated with the battery health state as the optimal health features.

[0035] By using the correlation analysis method described above to select the optimal health features, we can effectively reduce the feature dimensionality, eliminate redundant features that are weakly correlated with the battery degradation state, significantly improve the accuracy and generalization ability of the health state estimation model, and reduce computational complexity, thus laying a data foundation for achieving efficient and accurate battery health state estimation.

[0036] S3, The joint distribution of the optimal health feature and the battery health state is modeled based on the Gaussian mixture model, and the model parameters are trained using an iterative optimization algorithm; Furthermore, a Gaussian mixture model is used to model the joint distribution of optimal health features and battery health state, and an iterative optimization algorithm is employed to train the model parameters. This step addresses the multimodal and heterogeneous characteristics commonly found in multi-battery training data by using a probabilistic modeling framework to explicitly represent the multimodal characteristics of battery degradation data as a weighted combination of multiple Gaussian components, effectively overcoming the prediction bias caused by the unimodal distribution assumption in traditional estimation methods. During modeling, the optimal health feature matrix and the battery health state vector are treated as joint random variables, and their statistical correlation is captured by modeling this joint distribution. The model parameters are trained using an iterative optimization strategy based on maximum log-likelihood, iteratively updating the mixing coefficients, the mean vector of each Gaussian component, and the covariance matrix until convergence to the optimal parameter set. As a specific implementation, the expectation-maximization algorithm can be used to train the Gaussian mixture model parameters, whose log-likelihood function can be expressed as a weighted combination of Gaussian components. The trained model parameters fully describe the joint probability distribution relationship between the optimal health features and the battery health state.

[0037] Furthermore, Gaussian mixture models, as a probabilistic modeling framework, can explicitly represent the multimodal characteristics of battery degradation data as a weighted combination of multiple Gaussian components, thus effectively overcoming the aforementioned limitations. Gaussian mixture models aim to learn the joint distribution. It is a matrix of selected optimal health indicators based on training data. and battery health state vector A joint random variable.

[0038] The joint distribution is modeled using a Gaussian mixture model as follows: (2) in, It is the number of Gaussian components. It is the first Mixing coefficients ( ), The mean is Covariance is Multivariate Gaussian distribution. Given a set of training data from aged battery cells, use the Expectation-Maximization (EM) algorithm to maximize the given set of unknown parameters. Log-likelihood is used to estimate the parameters of Gaussian mixture models. ,in .

[0039] Furthermore, the trained Gaussian mixture model describes the joint distribution as shown in equation (3). For the k-th element, and They are and The mean vector, and They are and The covariance matrix. yes and The covariance between them.

[0040] (3) ; (4) This invention effectively solves the problem of complex and diverse degradation feature distributions in multi-battery datasets by leveraging the multi-component representation capabilities of Gaussian mixture models, and significantly improves the generalization ability of health status estimation models for different battery types and operating conditions.

[0041] S4. Based on new observation data of the test battery, the parameters of the trained Gaussian mixture model are adaptively updated using the Bayesian inference method. The battery health status is estimated based on the conditional probability distribution, and the uncertainty quantification result is output.

[0042] Furthermore, in the battery health status estimation process, when new observational data of the test battery is obtained, Bayesian inference methods are used to adaptively update the parameters of the trained Gaussian mixture model, thereby achieving health status estimation based on conditional probability distribution and outputting uncertainty quantification results. Specifically, based on the optimal health index vector obtained by the test battery in the current cycle, combined with the parameters of each Gaussian component in the trained Gaussian mixture model, the posterior conditional probability distribution of the health status is calculated using Bayesian statistical methods. This conditional probability distribution is presented in the form of a Gaussian distribution, with its mean reflecting the best estimate of the health status and its variance quantifying the degree of uncertainty of the estimation result. When selecting the Gaussian component for estimation, the maximum log-likelihood criterion is used to determine the best-matching Gaussian component, and then the conditional mean and conditional variance are calculated based on the covariance matrix of this component. Compared with traditional deterministic methods that only output single-point estimates, this method can output complete probability distribution characteristics, characterizing the confidence range of the prediction results in the form of confidence intervals, thus providing a more reliable decision-making basis for battery health management.

[0043] Furthermore, when the test set battery is in the i-th cycle, the optimal health indicator vector is y_i = [y_1, y_2, ..., y_n], where n is the length of the optimal health indicator. The unknown health state value is estimated as a conditional probability f(z_i | y_i) using a Bayesian update. The k-th Gaussian component most likely belonging to the optimal health indicator is used for health state estimation, and the unknown health state is updated to C_{i,k} } = μ_Z|yi, the variability is updated to the variance D_{k} of the most likely Gaussian component. } = Σ_Z,Z|yi.

[0044] Furthermore, for batteries in use during the early stages of their lifespan, the proposed framework can adaptively update the parameters of the Gaussian mixture model trained in each cycle based on newly observed optimal health indicators. When the test set batteries are in the i-th cycle, the optimal health indicator vector is... Where n is the length of the optimal health indicator, the proposed framework aims to estimate health status values ​​( ).

[0045] Furthermore, compared to traditional deterministic methods that only provide single-point estimates, the conditional Gaussian mixture model can output a complete probability estimate, quantifying the uncertainty of the prediction in the form of confidence intervals. This model directly characterizes the possible distribution range of the health state through a probability density function, thus providing a reliability measure for the prediction results. In battery health monitoring and predictive maintenance, mastering the reliability of the estimated value is as important as mastering the actual estimated value; this characteristic gives the conditional Gaussian mixture model a significant application advantage. As shown in equation (5), the Gaussian mixture model algorithm updates the unknown health state value (…) using a Bayesian update method. The estimate is the conditional probability. Use the health indicators most likely to be optimal ( The health status is estimated using the k-th Gaussian component of ). Therefore, the unknown health status ( Updated to The variability is updated to the most probable Gaussian component. variance As shown in equations (6)-(8).

[0046] (5) (6) (7) (8) This step achieves adaptive updating of model parameters through Bayesian inference, enabling the estimation framework to dynamically adjust the prediction strategy based on new observation data, demonstrating a high degree of online adaptability; at the same time, the uncertainty quantification function provides a reliability measure for the estimation results, enhancing the cross-cell generalization ability.

[0047] This invention proposes a conditional Gaussian mixture method for battery health state estimation. This method first extracts key health indicators from voltage, current, temperature, and time-series data during the charging phase. Then, it uses grey relational analysis to assess the correlation between the health indicators and the battery degradation state, selecting the most representative indicators as optimal health features. Based on this, a Gaussian mixture model is used to model the joint distribution of the optimal health features and the battery health state, thereby constructing an effective prediction model. Furthermore, based on new observational data from the test battery, this method can achieve real-time adaptive updates of the parameters of the trained Gaussian mixture model through Bayesian statistics.

[0048] This invention integrates the advantages of statistical hybrid modeling and Bayesian inference, enabling efficient battery state prediction with uncertainty quantification even under limited data conditions. It provides accurate battery health state estimates and simultaneously delivers uncertainty quantification results. Furthermore, it dynamically adapts to changes in real-time operating conditions when new observational data becomes available for the test set batteries, demonstrating high online adaptive capability and cross-battery generalization ability.

[0049] To achieve the above embodiments, such as Figure 2 As shown, this embodiment also provides a highly adaptable lithium-ion battery health status estimation device 10, including: The indicator extraction and data filtering module 100 is used to extract health indicators from voltage, current, temperature and time series data during the battery charging process, and to filter the raw data using a smoothing algorithm. The Health Indicator Correlation Analysis and Optimal Feature Screening Module 200 is used to evaluate the correlation of filtered health indicators using correlation analysis methods, and to screen out the optimal health features with the highest correlation to the battery health status. The model modeling and parameter training module 300 is used to model the joint distribution of the optimal health features and battery health status based on the Gaussian mixture model, and to train the model parameters using an iterative optimization algorithm. The model parameter update and health status estimation and quantification module 400 is used to adaptively update the parameters of the trained Gaussian mixture model based on new observation data of the test battery using the Bayesian inference method, and to estimate the battery health status based on the conditional probability distribution and output the uncertainty quantification result.

[0050] This invention provides a highly adaptive lithium-ion battery health state estimation device that effectively captures multimodal degradation features in multi-battery datasets using a Gaussian mixture model. It achieves accurate health state estimation while providing reliable uncertainty quantification, and possesses online adaptive updating and cross-battery generalization capabilities.

[0051] To implement the methods of the above embodiments, the present invention also provides a computer device, such as... Figure 3 As shown, the computer device 600 includes a memory 601 and a processor 602; wherein, the processor 602 reads the executable program code stored in the memory 601 to run a program corresponding to the executable program code, so as to implement the various steps of the highly adaptable lithium-ion battery health state estimation method described above.

[0052] To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a highly adaptable lithium-ion battery health state estimation method as described in the foregoing embodiments.

[0053] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0054] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A highly adaptive lithium-ion battery state-of-health estimation method, characterized in that, include: S1 extracts health indicators from voltage, current, temperature and time series data during the battery charging process, and uses a smoothing algorithm to filter the raw data; S2. The correlation analysis method is used to evaluate the correlation of the filtered health indicators and select the optimal health features with the highest correlation to the battery health status. S3, The joint distribution of the optimal health feature and the battery health state is modeled based on the Gaussian mixture model, and the model parameters are trained using an iterative optimization algorithm; S4. Based on new observation data of the test battery, the parameters of the trained Gaussian mixture model are adaptively updated using the Bayesian inference method. The battery health status is estimated based on the conditional probability distribution, and the uncertainty quantification result is output.

2. The method of claim 1, wherein, The process of extracting health indicators from voltage, current, temperature, and time-series data during battery charging, and then filtering the raw data using a smoothing algorithm, includes: Four statistical characteristics were calculated for the extracted voltage, current, temperature, and time data, including mean, variance, standard deviation, and range. The voltage, current, temperature, and time series data are smoothed using a Savitzky-Golay filter to suppress noise interference and improve the quality of health indicators.

3. The method as described in claim 1, characterized in that, The correlation analysis method is used to evaluate the correlation of the filtered health indicators and select the optimal health features with the highest correlation to the battery health status, including: The correlation score between each health indicator and the battery health status was calculated based on the grey relational analysis method. The correlation score is between 0 and 1, where 0 represents no correlation and 1 represents perfect correlation. Based on the correlation score calculated by grey relational analysis, the health indicator with the highest correlation score is selected as the optimal health feature.

4. The method as described in claim 1, characterized in that, The modeling of the joint distribution of the optimal health features and battery health state based on a Gaussian mixture model, and the training of the model parameters using an iterative optimization algorithm, includes: The joint random variable consisting of the optimal health feature matrix and the battery health state vector is modeled as a weighted combination of multiple Gaussian components; The parameters of a Gaussian mixture model are estimated by maximizing the log-likelihood of a given set of parameters using the expectation-maximization algorithm. The mixing coefficients of each Gaussian component, the mean vector of each Gaussian component, and the covariance matrix of each Gaussian component.

5. The method as described in claim 4, characterized in that, The joint distribution expression of the Gaussian mixture model is: in, It is the number of Gaussian components. It is the first Mixing coefficients ( ), The mean is Covariance is Multivariate Gaussian distribution.

6. The method as described in claim 1, characterized in that, The process involves adaptively updating the parameters of a trained Gaussian mixture model based on new observational data from the test battery using a Bayesian inference method, estimating the battery health status based on a conditional probability distribution, and outputting uncertainty quantification results. This includes: Based on the newly observed best health indicator vector of the test battery, calculate the posterior probability of each Gaussian component, and select the most likely Gaussian component for health status estimation. The unknown health status is estimated as a normal distribution with the conditional mean and conditional variance of the most likely Gaussian components as parameters, and the health status estimate and its corresponding variance are output as the uncertainty quantification result.

7. The method as described in claim 6, characterized in that, The optimal health indicator vector newly observed from the test battery includes: The estimated state of health of the battery in the i-th cycle is: ； The variability in health status estimates is: ； Normal distribution Output the conditional probability distribution and uncertainty quantification results of the battery health status, where zi is the health status value of the test battery in the i-th cycle.

8. A highly adaptable lithium-ion battery health status estimation device, characterized in that, include: The indicator extraction and data filtering module is used to extract health indicators from voltage, current, temperature and time series data during the battery charging process, and to filter the raw data using a smoothing algorithm. The Health Indicator Correlation Analysis and Optimal Feature Screening Module is used to evaluate the correlation of filtered health indicators using correlation analysis methods, and to screen out the optimal health features with the highest correlation to the battery health status. The model modeling and parameter training module is used to model the joint distribution of the optimal health features and battery health status based on the Gaussian mixture model, and to train the model parameters using an iterative optimization algorithm. The model parameter update and health status estimation and quantification module is used to adaptively update the parameters of the trained Gaussian mixture model based on new observation data of the test battery using the Bayesian inference method, and to estimate the battery health status based on the conditional probability distribution and output the uncertainty quantification results.

9. A computer device, characterized in that, Including processor and memory; The processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement a highly adaptable lithium-ion battery health state estimation method as described in any one of claims 1-7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements a highly adaptable lithium-ion battery health state estimation method as described in any one of claims 1-7.

Images