A rapid battery consistency sorting method based on electrochemical impedance spectroscopy

Abstract

The present invention discloses a method for rapid battery consistency sorting based on electrochemical impedance spectroscopy, comprising: S1, obtaining the electrochemical impedance spectra of all batteries to be sorted; S2, inputting the electrochemical impedance spectra of all the batteries to be sorted into a pre-trained principal component analysis conversion matrix, so that the high-dimensional impedance characteristics are reduced to one-dimensional impedance characteristics; S3, inputting the one-dimensional impedance characteristics of all the batteries to be sorted into a greedy sliding window sorting algorithm to obtain the consistency sorting results of the batteries to be sorted. According to the present invention, the sorting feature measurement time is short, the sorting efficiency is high, and the battery modules obtained by sorting have higher capacity utilization efficiency and longer service life, which is of great significance for improving the efficiency and effect of battery consistency sorting.

Description

Technical Field

[0001] The present invention relates to the technical field of battery management, and in particular to a method for rapid battery consistency sorting based on electrochemical impedance spectroscopy. Background Art

[0002] Battery consistency sorting technology is one of the key technologies for the cascade utilization of retired power batteries. It is of great significance for improving the charge and discharge performance, service life, and safety of battery modules. Series-connected battery modules exhibit a capacity bucket effect, where only the lowest-capacity battery in the module can be fully charged or discharged. This means that the effective capacity of the battery module is equal to the capacity of the lowest-capacity battery cell, and the capacity of the remaining batteries cannot be fully utilized, resulting in a decrease in the economic benefits of the battery module. Therefore, improving the consistency of batteries in the module is of great significance for improving the capacity utilization efficiency of the battery cells in the module and extending the service life of the battery module.

[0003] Invention patent CN115248393A discloses a battery consistency sorting method. By performing charge and discharge tests, measuring the battery internal resistance, and measuring the open-circuit voltage on battery cells, multiple sorting variables, voltage curves, and energy curves corresponding to each battery cell are obtained; multi-parameter sorting is performed on the battery cells based on the multiple sorting variables to obtain multi-parameter sorting results; based on the multi-parameter sorting results, the batteries are sorted into first-category batteries and second-category batteries; and the second-category batteries are sorted based on the voltage curve and energy curve to obtain battery sorting results. This method takes into account multiple consistency evaluation dimensions of batteries and can evaluate battery consistency more comprehensively; however, performing multi-parameter measurements on batteries is time-consuming, and it is difficult to compare the sorting effects of different sorting schemes on the same batch of batteries to prove that this method can achieve the best sorting effect.

[0004] Similarly, invention patents CN115889245A, CN112666471A, and CN107597619A each disclose a battery consistency sorting method. These methods use battery voltage, internal resistance, and discharge capacity as sorting parameters to evaluate battery consistency from multiple dimensions, thereby performing consistency sorting. However, these methods also suffer from the problems of lengthy measurement times, complex operations, and difficulty in demonstrating the superiority of sorting results through rigorous comparative experiments. Summary of the Invention

[0005] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a method for rapid battery consistency sorting based on electrochemical impedance spectroscopy, which has a short time consumption for sorting feature measurement, high sorting efficiency, and higher capacity utilization efficiency and longer service life of the sorted battery modules, which is of great significance for improving the efficiency and effect of battery consistency sorting. In order to achieve the above-mentioned purpose and other advantages according to the present invention, a method for rapid battery consistency sorting based on electrochemical impedance spectroscopy is provided, comprising the following steps:

[0006] S1. Obtain the electrochemical impedance spectra of all batteries to be sorted;

[0007] S2. Inputting the electrochemical impedance spectra of all the batteries to be sorted into a pre-trained principal component analysis conversion matrix to reduce the high-dimensional impedance features to one-dimensional impedance features;

[0008] S3. Input the one-dimensional impedance characteristics of all batteries to be sorted into the greedy sliding window sorting algorithm to obtain consistent sorting results for the batteries to be sorted.

[0009] Preferably, in step S1, electrochemical impedance spectra of all batteries to be sorted under a frequency set f are obtained, and the frequency range of the frequency set f is determined by a permutation feature importance method.

[0010] Preferably, the permutation feature importance method relies on a multivariate nonlinear regression model such as XGBoost or Gaussian process regression, and the specific steps of the permutation feature importance method are as follows:

[0011] S11. For a data set characterized by the electrochemical impedance spectrum of a battery and labeled by the remaining available discharge capacity of the battery, randomly divide it into a training set and a test set according to a certain ratio;

[0012] S12. Training the selected multivariate nonlinear regression model using the training set to obtain a trained model;

[0013] S13, using the original electrochemical impedance spectroscopy characteristic data in the test set as the model input and the battery remaining available discharge capacity label data in the test set as the true value to calculate the performance index of the model;

[0014] S14. For features in the test set All the corresponding feature data are randomly combined with another feature Perform data exchange, i.e. replace features;

[0015] S15. After completing the replacement feature, the electrochemical impedance spectroscopy feature data in the test set is used as the model input, and the battery remaining available discharge capacity label data in the test set is used as the true value to calculate the performance index of the model;

[0016] S16, repeat steps S14 to S15 several times, so that the feature Exchange data with multiple different features to obtain features The average performance index of the model after multiple permutations of features is calculated, and the difference in the model performance index corresponding to step S13 is calculated, and the difference is used as the feature The permutation feature importance index of ;

[0017] S17. Repeat steps S14 to S16 to calculate the permutation feature importance index of all features, and sort them according to the size of the index to obtain the permutation feature importance ranking results of all features, and determine the features with stronger correlation with the label.

[0018] Preferably, the remaining available discharge capacity of the battery in step S11 is defined as the amount of electricity that can be discharged when the battery is charged and discharged under a specific operating condition starting from the current state until the preset end-of-life state.

[0019] Preferably, the steps of training the principal component analysis conversion matrix in step S2 are as follows:

[0020] S21, obtaining electrochemical impedance spectroscopy data of several batteries of the same model under a frequency set f;

[0021] S22, combining the electrochemical impedance data of all batteries into a two-dimensional matrix and inputting it into a principal component analysis model;

[0022] S23. Calculate and obtain the principal component analysis conversion matrix based on the input data distribution.

[0023] Preferably, the greedy sliding window sorting algorithm in step S3 includes the following steps:

[0024] S31, sorting the one-dimensional impedance characteristics of the batteries to be sorted;

[0025] S32, linearly scanning the one-dimensional impedance characteristic data list with a preset sliding window length, calculating the range of the one-dimensional impedance characteristic in each sliding window, and updating the minimum value of the range and the corresponding sliding window;

[0026] S33. After the linear scan is completed, the battery in the sliding window corresponding to the minimum value of the one-dimensional impedance characteristic range is removed from the battery to be sorted and added to the sorting result list;

[0027] S34. Repeat steps S32 to S33 until the number of batteries to be sorted is less than the sliding window size, or the minimum value of the one-dimensional impedance characteristic range does not meet the optimization constraint; output the sorting result list as the consistency sorting result.

[0028] Compared with the prior art, the present invention has the following advantages: the fast battery consistency sorting method based on electrochemical impedance spectroscopy proposed in the present invention uses the battery broadband impedance as the feature for battery consistency sorting. Due to the different frequency set f scale and electrochemical impedance spectroscopy measurement methods, the characteristic measurement time of a battery can be controlled to less than 1 minute at the shortest; the computational efficiency of the greedy sliding window sorting algorithm is very high, and the sorting of 1,000 battery data can be completed in less than 1 minute.

[0029] The rapid battery consistency sorting method based on electrochemical impedance spectroscopy proposed in the present invention has been scientifically demonstrated to be able to provide a reasonable quantitative definition of battery aging consistency based on the remaining available discharge capacity of the battery. By designing experiments to compare the consistency sorting effects, it is proved that compared with traditional consistency sorting based on capacity calibration values, the battery modules sorted by the present invention have the advantages of higher capacity utilization efficiency and longer service life. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 A flowchart of a method for rapid battery consistency sorting based on electrochemical impedance spectroscopy according to the present invention;

[0031] Figure 2 A diagram showing the importance ranking of substitution features in an embodiment of a method for rapid battery consistency sorting based on electrochemical impedance spectroscopy according to the present invention;

[0032] Figure 3 A schematic diagram of calculating the remaining available discharge capacity of a battery according to the method for rapid battery consistency sorting based on electrochemical impedance spectroscopy of the present invention;

[0033] Figure 4 This is a comparison diagram of electrochemical impedance spectra of the same battery measured under different ambient temperatures and battery terminal voltage conditions according to the battery consistency rapid sorting method based on electrochemical impedance spectroscopy of the present invention;

[0034] Figure 5 This is a graph showing the electrochemical impedance spectroscopy characteristic matching results of a 18650 ternary lithium-ion battery according to the battery consistency rapid sorting method based on electrochemical impedance spectroscopy of the present invention;

[0035] Figure 6 A diagram showing battery capacity estimation results using feature-matched electrochemical impedance spectroscopy data as input for a method for rapid battery consistency sorting based on electrochemical impedance spectroscopy according to the present invention;

[0036] Figure 7 Graph showing the relationship between the one-dimensional impedance characteristics of the battery consistency rapid sorting method based on electrochemical impedance spectroscopy according to the present invention and the remaining available discharge capacity of the battery. DETAILED DESCRIPTION

[0037] 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. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention. Example 1

[0038] Reference Figure 1 , a battery consistency rapid sorting method based on electrochemical impedance spectroscopy, comprising the following steps:

[0039] 1) Obtain the electrochemical impedance spectra of all batteries to be sorted under the frequency set f;

[0040] 2) The electrochemical impedance spectra of all batteries to be sorted are input into the pre-trained principal component analysis conversion matrix to reduce the high-dimensional impedance features to one-dimensional impedance features;

[0041] 3) Using the one-dimensional impedance characteristic as the sorting parameter, the number of cells in the battery module as the length of the sliding window, and minimizing the range of the one-dimensional impedance characteristic as the optimization goal, the one-dimensional impedance characteristics of all batteries to be sorted are input into the greedy sliding window sorting algorithm to obtain consistent sorting results for the batteries to be sorted.

[0042] In step 1), the frequency range of the frequency set f is determined by the permutation feature importance method. The permutation feature importance method relies on the XGBoost model and includes the following steps:

[0043] For a data set characterized by the electrochemical impedance spectroscopy of a battery and labeled by the remaining available discharge capacity of the battery, it is randomly divided into a training set and a test set according to a certain ratio;

[0044] Use the training set to train the XGBoost model to obtain a trained model;

[0045] The original electrochemical impedance spectroscopy characteristic data in the test set is used as the model input, and the battery remaining available discharge capacity label data in the test set is used as the true value to calculate the model's goodness of fit;

[0046] For the features in the test set All the corresponding feature data are randomly combined with another feature Perform data exchange, i.e. replace features;

[0047] The electrochemical impedance spectroscopy feature data in the test set after feature replacement is used as the model input, and the battery remaining available discharge capacity label data in the test set is used as the true value to calculate the model's goodness of fit;

[0048] Order Features Exchange data with multiple different features to obtain features The average goodness of fit of the model after multiple permutations of features is calculated, and the difference between the goodness of fit of the model before permutation of features is calculated, and the difference is used as the feature The permutation feature importance index of ;

[0049] Calculate the permutation feature importance index of all features and sort them according to the size of the index to obtain the permutation feature importance ranking results of all features and determine the features with stronger correlation with the label. Figure 2 As shown, the top 5 features ranked by permutation feature importance are distributed in the frequency range of 10 Hz to 500 Hz, so the frequency range of the frequency set f is determined to be 10 Hz to 500 Hz.

[0050] In the above steps, the remaining available discharge capacity of the battery is defined as the amount of electricity that can be discharged from the battery starting from the current state, performing charge and discharge cycles under a specific operating condition, until the capacity calibration value decays to 75% of the rated capacity. Figure 3 As shown, the battery Starting from the state, charge and discharge cycles are performed under typical working conditions until the capacity calibration value decays to 75% of the rated capacity, that is, the capacity reaches state, then under typical working conditions, the remaining available discharge capacity of this battery is .

[0051] In step 2), the principal component analysis conversion matrix training method includes the following steps:

[0052] Obtain electrochemical impedance spectroscopy data for several batteries of the same model at a frequency set f;

[0053] The electrochemical impedance data of all batteries were combined into a two-dimensional matrix and input into the principal component analysis model;

[0054] According to the input data distribution, the principal component analysis transformation matrix is ​​calculated.

[0055] In step 3), the greedy sliding window sorting algorithm includes the following steps:

[0056] Sort the one-dimensional impedance characteristics of the batteries to be sorted;

[0057] Perform a linear scan on the one-dimensional impedance characteristic data list with a preset sliding window length, calculate the range of the one-dimensional impedance characteristic in each sliding window, and update the minimum value of the range and the corresponding sliding window;

[0058] After the linear scan is completed, the battery in the sliding window corresponding to the minimum value of the one-dimensional impedance characteristic range is taken out from the battery to be sorted and added to the sorting result list;

[0059] Repeat the above sorting process until the number of batteries to be sorted is less than the sliding window size, or the minimum value of the one-dimensional impedance characteristic range does not meet the optimization constraint; output the sorting result list as the consistency sorting result.

[0060] The process of the greedy sliding window sorting algorithm is as follows Figure 4 shown.

[0061] The implementation principle of the present invention is as follows:

[0062] This invention is mainly aimed at the cascade utilization of retired batteries. Due to the relatively poor consistency of retired batteries, parallel connection of batteries poses a safety hazard. Therefore, the reconfiguration modules in the cascade utilization scenario are mainly connected in series.

[0063] Considering the behavior of battery cells in a series module, it is found that due to the series relationship, the current passing through all battery cells in the series module is equal at all times. This means:

[0064] 1) The working conditions experienced by each battery cell are exactly the same;

[0065] 2) From the moment the battery enters the series module, the cumulative discharge capacity of each battery cell is equal at all times.

[0066] Based on the above conclusions, without reorganizing the battery cells to be sorted, each battery cell can be subjected to a single-cell cycle aging test under a typical operating condition until the battery cell capacity decays to the end of its life. Since the available capacity of the battery module is always equal to the capacity of the lowest battery cell, by analyzing the cumulative discharge capacity-capacity curve of each battery cell, the cumulative discharge capacity-capacity curve of the battery module can be calculated when these cells are grouped in different combinations. Figure 5 As shown in the figure, C1, C2 and C3 represent the cumulative discharge capacity-capacity curves of three different battery cells respectively; if these three battery cells are used to form a battery module when the cumulative discharge capacity is 0Ah, then the cumulative discharge capacity-capacity curve of the module can be represented by M; when the cumulative discharge capacity is 0Ah, the remaining available discharge capacity of the battery module is marked in the figure.

[0067] Note that the remaining available discharge capacity of the module is the minimum value of the remaining available discharge capacity of all battery cells. This means that when the battery module reaches the end of its life, the remaining available discharge capacity of the battery cells in the module, except for the battery cell with the lowest capacity, is more or less wasted. The remaining available discharge capacity of a battery cell is , the remaining available discharge capacity of the battery module is , then for the For each battery cell, the remaining available discharge capacity utilization rate Right now

[0068] ;

[0069] The capacity utilization rate of the battery module can be obtained by taking the average of the remaining available discharge capacity utilization rate of all battery cells. This indicator is used as a quantitative indicator to evaluate the aging consistency of battery cells in the module. Obviously, the closer the remaining available discharge capacity of the battery cells is, the higher the capacity utilization of the battery module.

[0070] Since the remaining available discharge capacity of the battery to be sorted under a certain working condition is unknown, it is hoped that a battery state characterization quantity related to the remaining available discharge capacity of the battery can be found, and the remaining available discharge capacity under the working condition can be estimated by the state characterization quantity, or the consistency of the state characterization quantity can be used to approximate the consistency of the remaining available discharge capacity. After trying, it was found that the relationship between the electrochemical impedance spectrum and the remaining available discharge capacity of the battery can be established through the multivariate nonlinear regression model of XGBoost. Figure 6 As shown in the figure, the remaining available discharge capacity of the battery is estimated by electrochemical impedance spectroscopy (EIS) characteristics, and the goodness of fit of the model can reach 0.919, showing a strong correlation; however, the average absolute percentage error of the model reaches 17.5%, indicating that the description of the remaining available discharge capacity of the battery using EIS characteristics is not accurate enough; the larger the actual remaining available discharge capacity, the lower the estimation accuracy.

[0071] It is reasonable for the above two data to show such a relationship, because the greater the remaining available discharge capacity, the farther the distance between the battery and the end of its life is, and the greater the uncertainty. In order to improve the estimation accuracy, the permutation feature importance method is considered to find the frequency range with the strongest correlation with the remaining available discharge capacity in the electrochemical impedance spectrum, and then reduce the feature dimension through the principal component analysis method to better analyze the relationship between the battery impedance and the remaining available discharge capacity. After the above steps, it is found that the one-dimensional impedance feature after dimensionality reduction is exponentially correlated with the remaining available discharge capacity of the battery. Figure 7 As shown in Figure 2, the smaller the one-dimensional impedance feature is, the greater the absolute value of the slope of the remaining available discharge capacity of the battery to the one-dimensional impedance feature is. This results in that when the actual value of the remaining available discharge capacity of the battery is large, the accuracy of the electrochemical impedance feature in describing the remaining available discharge capacity of the battery will inevitably decrease. This finding is consistent with Figure 6 The phenomena shown are exactly the same.

[0072] Despite this, it's still possible to approximate the consistency of a battery's remaining available discharge capacity using the consistency of its one-dimensional impedance characteristics. In an experimental dataset, using the battery's one-dimensional impedance characteristics as a sorting parameter, the average battery module capacity utilization in the sorting results for 4 and 7 battery modules, respectively, was 87.50% and 79.08%.

[0073] For comparison purposes, assuming the exact same grouping ratio, using the battery capacity calibration value as the sorting parameter, the average battery module capacity utilization rates for 4 and 7 battery modules, respectively, were 83.07% and 80.25%. Based on these reference indicators, the consistency sorting method using one-dimensional impedance characteristics as the sorting parameter demonstrates significant advantages in both sorting performance and efficiency compared to the traditional consistency sorting method using capacity calibration as the sorting parameter.

[0074] The number of devices and processing scales described herein are intended to simplify the description of the present invention, and applications, modifications, and variations of the present invention will be apparent to those skilled in the art.

[0075] Although the embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the description and implementation methods. They can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A method for rapid battery consistency sorting based on electrochemical impedance spectroscopy, characterized in that: The following steps are involved: S1. Obtain the electrochemical impedance spectra of all batteries to be sorted; S2. Inputting the electrochemical impedance spectra of all the batteries to be sorted into a pre-trained principal component analysis conversion matrix to reduce the high-dimensional impedance features to one-dimensional impedance features; S3. Input the one-dimensional impedance characteristics of all batteries to be sorted into the greedy sliding window sorting algorithm to obtain the consistent sorting results of the batteries to be sorted; In step S1, the electrochemical impedance spectra of all batteries to be sorted are obtained under a frequency set f, and the frequency range of the frequency set f is determined by a permutation feature importance method; The permutation feature importance method relies on the multivariate nonlinear regression model. The specific steps of the permutation feature importance method are as follows: S11. For a data set characterized by the electrochemical impedance spectrum of a battery and labeled by the remaining available discharge capacity of the battery, randomly divide it into a training set and a test set according to a certain ratio; S12. Training the selected multivariate nonlinear regression model using the training set to obtain a trained model; S13, using the original electrochemical impedance spectroscopy characteristic data in the test set as the model input and the battery remaining available discharge capacity label data in the test set as the true value to calculate the performance index of the model; S14. For features in the test set All the corresponding feature data are randomly combined with another feature Perform data exchange, i.e. replace features; S15. After completing the replacement feature, the electrochemical impedance spectroscopy feature data in the test set is used as the model input, and the battery remaining available discharge capacity label data in the test set is used as the true value to calculate the performance index of the model; S16, repeat steps S14 to S15 several times, so that the feature Exchange data with multiple different features to obtain features The average performance index of the model after multiple permutations of features is calculated, and the difference in the model performance index corresponding to step S13 is calculated, and the difference is used as the feature The permutation feature importance index of ; S17. Repeat steps S14 to S16 to calculate the permutation feature importance index of all features, and sort them according to the size of the index to obtain the permutation feature importance ranking results of all features, and determine the features with stronger correlation with the label; The remaining available discharge capacity of the battery in step S11 is defined as the amount of electricity that can be discharged from the current state of the battery through charge and discharge cycles under a specific operating condition until the battery reaches a preset end-of-life state.

2. A method for rapid battery consistency sorting based on electrochemical impedance spectroscopy according to claim 1, characterized in that: The multivariate nonlinear regression model is XGBoost or Gaussian process regression.

3. The method for rapid battery consistency sorting based on electrochemical impedance spectroscopy according to claim 1, characterized in that: The steps of training the principal component analysis conversion matrix in step S2 are as follows: S21, obtain the frequency set of several batteries of the same model f Electrochemical impedance spectroscopy data under; S22, combining the electrochemical impedance data of all batteries into a two-dimensional matrix and inputting it into a principal component analysis model; S23. Calculate and obtain the principal component analysis conversion matrix based on the input data distribution.

4. A method for rapid battery consistency sorting based on electrochemical impedance spectroscopy according to claim 1, characterized in that: The greedy sliding window sorting algorithm in step S3 includes the following steps: S31, sorting the one-dimensional impedance characteristics of the batteries to be sorted; S32, linearly scanning the one-dimensional impedance characteristic data list with a preset sliding window length, calculating the range of the one-dimensional impedance characteristic in each sliding window, and updating the minimum value of the range and the corresponding sliding window; S33. After the linear scan is completed, the battery in the sliding window corresponding to the minimum value of the one-dimensional impedance characteristic range is removed from the battery to be sorted and added to the sorting result list; S34. Repeat steps S32 to S33 until the number of batteries to be sorted is less than the sliding window size, or the minimum value of the one-dimensional impedance characteristic range does not meet the optimization constraint; output the sorting result list as the consistency sorting result.

Images