Lithium ion battery health state estimation method using partial constant current charging data

Abstract

The invention relates to a lithium ion battery health state estimation method using partial constant current charging data. The method comprises the following steps: acquiring constant current charging data of a target lithium ion battery and carrying out feature extraction; the extracted feature data are input into a pre-trained health state estimation model, a health state estimation result of the target lithium ion battery is output, the health state estimation model is obtained based on training of a training set, the training set comprises constant-current charging data of a plurality of lithium ion batteries and extracted corresponding features, and the constant-current charging data of the lithium ion batteries and the extracted corresponding features are obtained. The health state estimation model is obtained by building a mapping relation between the health characteristics and the health state of the battery through a multi-layer perception mechanism. According to the method, high-quality features capable of reflecting the health condition of the lithium battery are extracted from voltage and current signals in part of constant-current charging data, and accurate SOH estimation of the lithium battery is completed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of lithium battery state of health estimation, and particularly relates to a lithium ion battery state of health estimation method using partial constant current charging data. BACKGROUND

[0002] Lithium ion batteries are widely used in electric vehicles, portable electronic devices and large-scale energy storage systems due to their high energy density, long life, low cost and low self-discharge rate. As a complex electrochemical system, various side reactions inevitably occur during repeated charging and discharging cycles, including electrolyte decomposition, solid electrolyte interface film growth and electrode material degradation. These side reactions gradually reduce battery performance, reduce the operating efficiency of related equipment, and in severe cases, can cause thermal runaway. Therefore, developing an efficient and accurate battery SOH estimation method has important research value for ensuring the safety and stability of LIB operation.

[0003] The estimation method of lithium ion battery SOH is mainly divided into model-based method and data-driven method. Among them, the model-based estimation method mainly includes electrochemical model and equivalent circuit model. The electrochemical model reveals the degradation mechanism of the battery by describing the physical and chemical reaction process inside the battery. The equivalent circuit model is to convert the lithium ion battery into a circuit system composed of voltage source, resistance, capacitance and inductance, etc. However, the model-based method usually contains multiple highly coupled partial differential equations, the parameters are complex and difficult to obtain, and the calculation cost is high, which is limited in practical application to a certain extent.

[0004] The data-driven method does not need to model and analyze the internal working mechanism of the battery, but extracts features from a large amount of battery data through machine learning and deep learning algorithms, and constructs the mapping relationship between the features and the battery SOH. Although the data-driven method has made remarkable achievements in the field of battery SOH, it still has some limitations. In actual application scenarios, it is challenging to obtain complete charging and discharging data of the battery to realize feature extraction. In addition, the extraction process of most battery features is highly dependent on expert knowledge and experience, which limits the universality and automation degree of the method. Therefore, the present application proposes a lithium ion battery state of health estimation method using partial constant current charging data. SUMMARY

[0005] The purpose of the present application is to provide a lithium ion battery state of health estimation method using partial constant current charging data, which extracts high-quality features reflecting the health status of lithium batteries from voltage and current signals in partial constant current charging data, and completes accurate SOH estimation of lithium batteries.

[0006] To achieve the above purpose, the present application provides the following scheme:

[0007] A lithium ion battery state of health estimation method using partial constant current charging data, comprising:

[0008] Obtaining constant current charging data of a target lithium ion battery and extracting features;

[0009] Input the extracted feature data into a pre-trained state of health estimation model, and output the state of health estimation result of the target lithium ion battery, wherein the state of health estimation model is obtained based on a training set, the training set includes constant current charging data and extracted corresponding features of several lithium ion batteries, and the state of health estimation model obtains the mapping relationship between health features and battery state of health through a multilayer perception mechanism.

[0010] Optionally, the constant current charging data of several lithium ion batteries in the training set is obtained by performing a cycle aging experiment, extracting voltage, current and capacity data during the battery charging and discharging process, and processing the voltage and current signals in the constant current charging stage.

[0011] Optionally, the feature extraction of the constant current charging data of the target lithium ion battery and the feature extraction of the constant current charging data of several lithium ion batteries in the training set both use the CMFCre algorithm, and the CMFCre algorithm comprises:

[0012] Segmenting the voltage signal and the current signal in the constant current charging stage to obtain local voltage signals and local current signals;

[0013] Using a composite multiscale decomposition method to reconstruct the sequence of the local voltage signals and the local current signals to obtain reconstructed signals;

[0014] Extracting the frequency spectrum information of the reconstructed signals by a Hamming window method and a discrete Fourier transform, and introducing a cumulative residual entropy theory to obtain voltage entropy features and current entropy features.

[0015] Optionally, the composite multiscale decomposition method is:

[0016] ;

[0017] Wherein, is the kth subsequence of , represents the tth value in , N is the data length of the current signal or the voltage signal, τ is the scale factor, represents the ith current signal or voltage signal, and T is the data length of the kth subsequence.

[0018] Optionally, the Hamming window method is:

[0019] ​;

[0020] ;

[0021] wherein w t is a weight coefficient of the window function, represents a signal processed by the Hanning window.

[0022] Optionally, the discrete Fourier transform is:

[0023] ;

[0024] wherein represents the frequency spectrum information extracted from the signal , j is an imaginary unit, and h is a data length of the frequency spectrum information.

[0025] Optionally, the cumulative residual entropy theory is used to quantify the frequency spectrum complexity of the signal, and the quantification method is:

[0026] ;

[0027] wherein CDF(.) is a cumulative distribution function, is a voltage signal or a current signal decomposes the frequency spectrum complexity of the signal in the τ scale.

[0028] Optionally, the health state estimation model is obtained by constructing a mapping relationship between the health feature and the battery health state through the multi-layer perception mechanism, and the mapping relationship comprises:

[0029] The Pearson correlation coefficient is used as an index for measuring the correlation between the extracted feature and the battery health state, a mapping relationship is constructed, and the health state estimation model is obtained.

[0030] Optionally, the Pearson correlation coefficient is:

[0031] ;

[0032] wherein r represents a correlation coefficient, En i and SOH i respectively represent an i-th entropy feature and a corresponding battery health state, and represent corresponding average values, and n represents the number of entropy features.

[0033] Optionally, after the health state estimation model is trained, the health state estimation model is tested, and the regression performance on the battery health state is quantitatively evaluated by a determination coefficient during the testing process, and the determination coefficient is:

[0034] ;

[0035] wherein R 2 is a determination coefficient, and y i are estimated and actual values, respectively, is an average value of y i .

[0036] The beneficial effects of the present application are:

[0037] (1) The present application proposes a composite multiscale decomposition method, which can fully extract the potential state of health information in lithium battery data.

[0038] (2) The present application uses the Hanning window method and discrete Fourier transform to extract the frequency spectrum information of the signal from the reconstructed voltage or current signal, and quantifies the frequency spectrum information through the cumulative residual entropy theory to obtain the health characteristics of the battery.

[0039] (3) The present application proposes a CMFCre method, which can directly extract key features for characterizing the SOH of the battery from part of the voltage and current signals in the constant current charging process.

[0040] (4) The present application can avoid the complex manual feature extraction process, and can realize high-quality health feature extraction without relying on professional knowledge, only using part of the constant current charging data, which can accurately characterize the SOH of the battery and has significant practical value. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0042] Figure 1 is a flow chart of a lithium ion battery state of health estimation method using part of the constant current charging data according to an embodiment of the present application;

[0043] Figure 2 is a flow chart of the CMFCre algorithm according to an embodiment of the present application;

[0044] Figure 3 is a schematic diagram of a battery aging test platform according to an embodiment of the present application;

[0045] Figure 4 is a schematic diagram of the current-voltage interval division of the constant current charging data according to an embodiment of the present application;

[0046] Figure 5 is an entropy value thermodynamic diagram of the CMFCre algorithm in each interval current-voltage according to an embodiment of the present application;

[0047] Figure 6 This is a schematic diagram of the correlation coefficients for each lithium battery feature in an embodiment of the present invention;

[0048] Figure 7 This is a schematic diagram of the MLP estimation results for the SOH of a lithium battery according to an embodiment of the present invention. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 are within the scope of protection of the present invention.

[0050] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0051] This embodiment provides a method for estimating the health status of a lithium-ion battery using partial constant current charging data, including:

[0052] Acquire constant current charging data of the target lithium-ion battery and extract its features;

[0053] The extracted feature data is input into a pre-trained health status estimation model, and the health status estimation result of the target lithium-ion battery is output. The health status estimation model is obtained by training based on a training set, which includes constant current charging data of several lithium-ion batteries and the corresponding extracted features. The health status estimation model is obtained by constructing a mapping relationship between health features and battery health status through a multilayer perceptron.

[0054] Specifically, such as Figure 1 As shown, this embodiment includes the following:

[0055] Step 1: Set up an experimental platform and collect aging cycle experimental data of lithium-ion batteries using charging and discharging equipment. Extract data such as voltage, current, and capacity during the battery charging and discharging process, and randomly divide the collected data into training set, validation set, and test set according to a ratio of 2:1:7.

[0056] like Figure 3As shown, an experimental platform was built, and 25 commercial lithium-ion batteries were subjected to cycle aging tests using a CT-4008Tn-5V12A-S1 device. Current, voltage, and capacity data were collected. The lithium battery chemical composition was Li(NiCoMn)O2, with a charge / discharge rate of 1C-1C, charge / discharge cutoff voltages of 4.2V and 2.75V respectively, and a nominal capacity of 1.2 Ah. The collected current and voltage signals were processed, and the current and voltage signals during the constant current charging stage were obtained. To verify the feature extraction capability of the CMFCre algorithm in some constant current charging stages, the current and voltage signals during the constant current charging stage were evenly divided into 10 intervals, such as... Figure 4 As shown.

[0057] Step 2: Extract entropy domain features of current and voltage from constant current charging data using the CMFCre algorithm. First, a composite multi-scale decomposition method is used to reconstruct the original signal sequence to fully extract potential feature information from the current and voltage data. Next, the spectral information of the reconstructed signal is extracted using a Hamming window and discrete Fourier transform. Finally, the cumulative residual entropy theory is introduced to quantify the complexity of the spectral data, thereby constructing health features that can be used to characterize the battery's health status.

[0058] like Figure 2 As shown, the Composite Multi-Scale Frequency Cumulative Residual Entropy Algorithm (CMFCre algorithm) is as follows:

[0059] (1) The composite multi-scale decomposition method decomposes voltage or current signals τ is decomposed into different sequences at multiple scales. This method aims to mine potential feature information in battery signals to achieve more comprehensive feature extraction.

[0060] (1);

[0061] in, yes The k-th subsequence, represent The t-th value in the equation, where N is the data length of the current or voltage signal, and τ is the scaling factor.

[0062] (2) Improve all subsequences by using the Hanning window method The Hanning window improves the accuracy of spectrum analysis and reduces spectral leakage during Fourier transform. By smoothing signal boundaries, it reduces spectral leakage caused by signal discontinuities, thereby enhancing the accuracy of spectrum analysis.

[0063] (2);

[0064] (3);

[0065] where w t is the weight coefficient of the window function, represents the signal after the Hanning window processing.

[0066] (3) Discrete Fourier Transform (DFT) is applied to the signal for frequency spectrum analysis, and its frequency spectrum information is extracted .

[0067] (4);

[0068] (4) Cumulative residual entropy is used to quantify the frequency spectrum complexity of the signal.

[0069] (5);

[0070] where CDF(.) is the cumulative distribution function.

[0071] The CMFCre algorithm is used to extract features from the voltage and current signals of 10 intervals, and the Pearson correlation coefficient is used as an indicator to measure the correlation between the extracted features and the battery SOH. Figure 5 The results of feature extraction in each interval are shown, Figure 6 and the corresponding correlation coefficients are given. In Figure 5 , the vertical coordinate represents the entropy feature data of six scales in different voltage and current regions, and the horizontal coordinate represents the number of batteries in each batch. As shown in Figure 5 , with the increase of charge and discharge cycle number, the color represented by the entropy feature of each region becomes darker. This trend is consistent with the change trend of battery SOH, indicating that the health status of the battery gradually decreases with the increase of charge and discharge cycle. As shown in Figure 6 , the current entropy feature and the voltage entropy feature of each region both show significant positive correlation with the battery SOH, and their correlation coefficients are all over 0.9.

[0072] (6);

[0073] where r represents the correlation coefficient, ranging from 1 to -1. En and SOH represent the entropy feature and the battery SOH, respectively, and and represent their respective average values.

[0074] Step three: input the health features extracted by the CMFCre algorithm into the MLP model and perform 200 iterations of training to achieve high-precision estimation of the lithium-ion battery SOH.

[0075] The SOH estimation results of the extracted CMFCre features using the MLP method are shown in Figure 7 . InFigure 7 In this study, current entropy features and voltage entropy features are used together as inputs to a deep learning model, and the coefficient of determination R is used as the input. 2 To quantitatively assess its impact on the battery's SOH regression performance. For example... Figure 7 As shown, the coefficient of determination R between the predicted results and the battery SOH is [value missing] in 10 different constant current charging intervals. 2 All values ​​exceeded 0.97, indicating that the method has high estimation accuracy. Particularly noteworthy is that, in intervals 1 to 8, the CMFCre+MLP method, using only partial constant current charging data, achieved a high coefficient of determination R between the predicted results and the state of charge (SOH). 2 All values ​​exceeded 0.99, further validating the effectiveness and robustness of the proposed method. These results demonstrate that even using incomplete charging information, high-precision estimation of battery state of health (SOH) can be achieved, providing a feasible solution for rapid battery health status diagnosis in practical applications.

[0076] (7);

[0077] in, and y i These are the estimated value and the actual value, respectively. For y i The average value.

[0078] This embodiment proposes a composite multi-scale decomposition method that can fully extract potential health status information from lithium battery data. Utilizing the Hanning window method and discrete Fourier transform, it extracts the spectral information of the reconstructed voltage or current signals and quantifies this spectral information using cumulative residual entropy theory to obtain the battery's health characteristics. Furthermore, it proposes a CMFCre method that can directly extract key features for characterizing the battery's state of health (SOH) from a portion of the voltage and current signals during constant current charging. This embodiment avoids complex manual feature extraction processes and achieves high-quality health feature extraction using only a portion of the constant current charging data without relying on specialized knowledge. It can accurately characterize the battery's SOH and has significant practical value.

[0079] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A lithium-ion battery state-of-health estimation method using partial constant current charging data, characterized by, The method comprises: obtaining and extracting features of constant current charging data of a target lithium ion battery; inputting the extracted feature data into a pre-trained state of health estimation model to output an estimated result of the state of health of the target lithium ion battery, wherein the state of health estimation model is obtained based on a training set, the training set comprises constant current charging data and extracted features of a plurality of lithium ion batteries, and the state of health estimation model is obtained by building a mapping relationship between health features and the state of health of the battery through a multilayer perception mechanism.

2. The lithium-ion battery state-of-health estimation method using partial constant current charging data according to claim 1, characterized by, The constant current charging data of the plurality of lithium ion batteries in the training set is obtained by performing a cycle aging experiment, extracting and processing voltage, current and capacity data in the battery charging and discharging process to obtain voltage signals and current signals in the constant current charging stage.

3. The lithium-ion battery state-of-health estimation method using partial constant current charging data according to claim 1 or 2, characterized by, The feature extraction of the constant current charging data of the target lithium ion battery and the feature extraction of the constant current charging data of the plurality of lithium ion batteries in the training set both adopt a CMFCre algorithm, the CMFCre algorithm comprises: segmenting the voltage signals and current signals in the constant current charging stage to obtain local voltage signals and local current signals; reconstructing sequences of the local voltage signals and local current signals by using a composite multiscale decomposition method to obtain reconstructed signals; extracting frequency spectrum information of the reconstructed signals by using a Hamming window method and a discrete Fourier transform, and introducing a cumulative residual entropy theory to obtain voltage entropy features and current entropy features.

4. The lithium-ion battery state-of-health estimation method using partial constant current charging data according to claim 3, characterized by, The composite multiscale decomposition method is: ； wherein is the kth subsequence of represents the tth value in the sequence, N is the data length of the current signal or voltage signal, and τ is a scale factor, denotes the ith current signal or voltage signal, and T is the data length of the kth subsequence.

5. The lithium-ion battery state-of-health estimation method using partial constant current charging data according to claim 4, characterized in that, The Hamming window method is: ； ； where w t is a weight coefficient of the window function, represents a signal after the Hanning window processing.

6. The lithium-ion battery state-of-health estimation method using partial constant current charging data according to claim 5, wherein, The discrete Fourier transform is: ； wherein represents the spectral information extracted from the signal j is the imaginary unit and h is the data length of the spectral information.

7. The lithium-ion battery state-of-health estimation method using partial constant current charging data according to claim 6, characterized in that, The cumulative residual entropy theory is used to quantify the frequency spectrum complexity of a signal, and the quantification method is: ； where CDF(.) is the cumulative distribution function, is a voltage signal or a current signal The spectral complexity of the signal is decomposed on the scale of τ.

8. The lithium-ion battery state-of-health estimation method using partial constant current charging data according to claim 1, wherein, The state of health estimation model is obtained by building a mapping relationship between health features and the state of health of the battery through a multilayer perception mechanism, which comprises: taking a Pearson correlation coefficient as an index for measuring the correlation between the extracted features and the state of health of the battery to build a mapping relationship and obtain the state of health estimation model.

9. The lithium-ion battery state-of-health estimation method using partial constant current charging data according to claim 8, wherein, The Pearson correlation coefficient is: ； where r represents a correlation coefficient, En i and SOH i respectively represent the i-th entropy feature and the corresponding battery state of health, and represent the corresponding average value, and n represents the number of entropy features.

10. The lithium-ion battery state-of-health estimation method using partial constant current charging data according to claim 9, wherein, The state of health estimation model is tested after training, and the regression performance of the battery state of health is quantitatively evaluated by a determination coefficient during the testing process, and the determination coefficient is: ； where R 2 is the coefficient of determination, and y i are the estimated and actual values, respectively, is the mean of y i .

Images