A sorting method for cascade utilization of retired lithium-ion power batteries

Abstract

The present invention relates to the field of power lithium-ion battery recycling, and specifically to a retired lithium-ion power battery cascade utilization sorting method, comprising the following steps: selecting three measurable battery parameters, namely, the maximum available capacity of the battery, the internal resistance of the battery, and the voltage difference of 3 to 8 minutes in the discharge curve, as sorting parameters, and testing each sorting parameter; determining the optimal k value by the contour coefficient method and the elbow rule, and then clustering and sorting the battery by the k-means++ algorithm according to the tested sorting parameters; combining the CRITIC weighting method with the gray correlation analysis method, matching the batteries with a correlation degree less than a threshold in each group of batteries after cluster sorting, and assembling them into a battery pack. The present invention can improve the sorting accuracy and sorting efficiency, and make the values ​​of the sorting characteristic parameters between the sorted single cells as close as possible, thereby minimizing the impact of the inconsistency between the batteries from all aspects to the greatest extent, and thus increasing the efficiency and life of the battery pack.

Description

Technical Field

[0001] The present invention relates to the field of power lithium-ion battery recycling, and in particular to a method for sorting retired lithium-ion power batteries for cascade utilization. Background Art

[0002] For pure electric vehicles, their power source is a power battery pack, which is usually composed of a large number of single cells connected in series and parallel. There are many types of power batteries, such as lithium-ion batteries, lead-acid batteries, nickel-metal hydride batteries, etc. Compared with the latter, lithium-ion batteries generally have higher voltage, greater energy, and longer cycle life during operation. In addition, due to their low self-discharge rate, when the battery is not used, there is less loss, which reduces a lot of costs. Therefore, lithium-ion batteries have broad application potential.

[0003] Generally speaking, lithium-ion batteries installed on electric vehicles can only meet the daily use requirements of new energy vehicles when their maximum available capacity is above about 80%. When it is lower than this value, the power battery will often face retirement. With the continuous increase in the number of new energy vehicles, the number of retired batteries will also increase. For retired lithium-ion batteries, the treatment methods are mainly divided into re-grouping for cascade utilization and direct decomposition to extract the useful or relatively expensive ones, but after grouping, they can still be used in other occasions with low power requirements. If the decomposition treatment is blindly carried out, it will often cause huge losses. And because the battery contains heavy metals, the recycling process also has certain damage to the environment. Therefore, according to actual needs, recombining retired batteries into battery packs for recycling can save a lot of costs. However, due to differences in manufacturers, specifications, manufacturing processes, use environments, etc., there are more or less inconsistencies between each single cell. If the single cells with large inconsistencies are combined into a battery pack for use, as the use time increases, the inconsistency will be magnified, which will reduce the life of the battery pack, and in serious cases, it will also cause certain safety hazards.

[0004] There have been relevant studies. For example, Chinese patent application CN110135490B is a method for classifying retired lithium batteries based on sample labels. The invention adopts a label-based screening mode, which requires battery recycling stations to set different performance labels according to the different requirements of retired battery performance in various utilization scenarios. Then, the parameter data of each retired battery is matched with various labels. This sorting method is very time-consuming and cumbersome in practice.

[0005] Another example is a method for fast estimation of SOH of power batteries based on Kmeans clustering in Chinese patent application CN112305442B. This invention uses Kmeans clustering to cluster the charge and discharge voltage, current and temperature, and obtains the weights for calculating the health status value by calculating the Euclidean distance of each sample. Compared with the present invention, the K-means clustering based on data mining technology needs to determine the k value in advance and the selection of the initial cluster center is random, which can easily affect the iteration speed of the algorithm. In the actual sorting process of retired batteries, the more iterations there are, the more computational effort the model will have, which will take more time and increase the sorting cost.

[0006] Another example is Chinese patent application CN114833097A, which discloses a sorting method and device for the cascade utilization of retired power batteries. The invention screens and groups the batteries. Each battery cell needs to undergo 9 rounds of charge and discharge tests, with the test time being more than 10 hours. After each round of charge and discharge, screening is required. Combined with voltage screening, the total number of screenings is more than 10 times.

[0007] Another example is the screening and reorganization method for recycling power batteries disclosed in patent CN116651773A. The testing time is more than 10 hours, which is time-consuming, occupies a large amount of testing resources and human resources, and the screening efficiency is not high.

[0008] In summary, the sorting accuracy and sorting speed of the existing power lithium battery recycling sorting methods are difficult to reconcile, and a power lithium battery recycling sorting method that is compatible with high accuracy and high efficiency is urgently needed. Summary of the invention

[0009] The purpose of the present invention is to provide a method for sorting retired lithium-ion power batteries for cascade utilization, so as to solve the problems raised in the background technology.

[0010] In order to achieve the above object, the technical solution of the present invention is as follows: A method for recycling and sorting retired lithium-ion power batteries comprises the following steps:

[0011] Three measurable battery parameters, namely, the maximum available capacity of the battery, the internal resistance of the battery, and the voltage difference between 3 and 8 minutes in the discharge curve, were selected as sorting parameters, and each sorting parameter was tested;

[0012] The optimal k value is determined by the silhouette coefficient method and the elbow rule, and the batteries are clustered and sorted by the k-means++ algorithm according to the tested sorting parameters;

[0013] The CRITIC weighting method is combined with the grey correlation analysis method, and the batteries with correlation degrees less than the threshold in each group of batteries after cluster sorting are matched together and assembled into a battery pack.

[0014] Furthermore, the test method for the maximum available capacity of the battery is: through constant current or constant voltage discharge, the battery's state of charge is reduced from 100% to 0%, and then the maximum available capacity is obtained by the ampere-hour integration method or by extracting the cumulative amount of discharged electricity; the calculation expression of the ampere-hour integration method is as follows:

[0015]

[0016] Where C is the battery capacity, T is the discharge time, I is the discharge current, t∈[0,T], and d represents the differential.

[0017] Further, the test method of the battery internal resistance is: the battery is tested by HPPC pulse discharge. In the HPPC test, the battery is discharged for a period of time and then its output voltage is measured; the internal resistance of the battery is obtained by measuring the voltage response under different discharge currents; the internal resistance is calculated using the following formula:

[0018] R=(V1-V2) / I

[0019] Where R is the internal resistance of the battery, V1 is the voltage of the battery at the beginning of discharge, V2 is the voltage of the battery at the end of discharge, and I is the discharge current.

[0020] Furthermore, the test method for the voltage difference between 3 and 8 minutes in the discharge curve is:

[0021] The battery is discharged from 100% SOC to 0% SOC by constant current or constant voltage discharge, and the time and voltage values ​​are measured to obtain a discharge curve, and the voltage difference of 3 to 8 minutes is obtained according to the discharge curve.

[0022] Furthermore, the optimal k value is determined by the silhouette coefficient method, including:

[0023] Calculate the silhouette coefficient of each sample and calculate the dissimilarity between the sample and other samples in the same cluster. The dissimilarity is measured by the squared Euclidean distance, which is expressed as follows:

[0024]

[0025] In the formula, a i is the dissimilarity between samples in the same cluster, a i The larger the value is, the more discrete the sample is from other samples in the same cluster. n is the number of samples in the nth cluster, x i and x j are the i-th and j-th samples respectively, S n is the nth cluster, d is the Euclidean distance;

[0026] The minimum value of the distance between the sample and other samples in different clusters is used to characterize the inter-cluster dispersion. The expression is as follows:

[0027]

[0028] Where b i is the inter-cluster dispersion of the i-th sample, S m is the mth cluster, N m is the number of samples in the mth cluster;

[0029] The calculation formula of the silhouette coefficient of a single sample that combines the intra-cluster similarity and inter-cluster dispersion is as follows:

[0030]

[0031] Calculate the mean silhouette coefficient of all sample points after clustering as the total silhouette coefficient under the current k value:

[0032]

[0033] In the formula, SC is the total silhouette coefficient, and N is the total number of cluster sample points;

[0034] By comparing the SC values ​​under different k values, the optimal k value is determined.

[0035] Furthermore, the optimal k value is determined by the elbow rule, including:

[0036] Through the SSE-based elbow rule, its expression is as follows:

[0037]

[0038] In the formula, x i is the i-th sample, S j is the jth cluster, e j is the cluster center of the jth cluster;

[0039] Traverse different k values ​​and calculate the SSE values ​​under different k values. The SSE value decreases as the k value increases. The k value corresponding to the inflection point where the SSE value decreases is the optimal k value.

[0040] Furthermore, before clustering and sorting the batteries using the k-means++ algorithm, the dimensions of the sorting parameters tested are unified, including:

[0041] Through the maximum-minimum dimensionless method, the expression is as follows:

[0042]

[0043] In the formula, x stdis the sorting parameter of the test, max is the maximum value of a certain parameter in each sample, min is the minimum value of the same parameter in each sample, x is the original data of the sorting parameter of the test, and the data after dimensionless processing are all in the range of [0,1].

[0044] Furthermore, the CRITIC weighting method is combined with the grey relational analysis method, including:

[0045] (1) Establish the original data matrix and reference vector:

[0046]

[0047] Where D is an original data matrix with m rows and n columns, indicating that the number of retired lithium-ion batteries in the original data D is m, and there are n selected evaluation indicators; d m is the mth retired battery; d 1n is the original data of the nth evaluation index corresponding to the first battery, d m1 is the original data of the first evaluation index corresponding to the mth battery, d mn is the original data of the corresponding nth evaluation index of the mth battery, and so on;

[0048] d * is the reference vector, which consists of the optimal values ​​of each parameter in m retired batteries; is the optimal value of retired batteries on evaluation parameter 1; is the optimal value of the retired battery on the evaluation parameter n, and so on;

[0049] (2) The values ​​in the original data matrix D are dimensionless. The dimensionless data processing is as follows:

[0050] For positive indicators:

[0051]

[0052] For negative indicators:

[0053]

[0054] Where, d' ij is a positive indicator, which means the jth positive indicator value of the i-th battery after dimensionless processing. The larger the value of the positive indicator, the better it is; d" ij is the reverse index, which represents the jth reverse evaluation index value of the i-th battery after dimensionless processing. The larger the value of the reverse index, the worse it is. For the convenience of description, the dimensionless data is represented by d ij Indicates; d in the formula ij Represents the original data of the jth evaluation index of the ith battery, maxdj is the maximum value in the original data matrix D, mind j is the minimum value in the original data matrix D;

[0055] (3) Obtaining the volatility between data:

[0056] The CRITIC algorithm uses the size of the standard deviation to measure the volatility of the data within each indicator. The subsequent weighting step will increase or decrease the weight of the indicator accordingly based on the volatility difference calculated here; the calculation formula for the standard deviation is as follows:

[0057]

[0058] Among them, M j is the standard deviation of the jth indicator, m is the number of samples, Represents the average value of the jth indicator, and the calculation formula is:

[0059]

[0060] (4) Obtain the Pearson correlation coefficient between indicators:

[0061] The CRITIC algorithm involves the determination of the correlation coefficient between indicators when calculating the conflict. Here, the Pearson correlation coefficient is used for calculation. The Pearson correlation coefficient is a statistic used to determine the degree of correlation between two variables. Its definition is the ratio of the covariance between the two variables to the product of the standard deviation, expressed as r. ij Represents the Pearson coefficient between the i-th and j-th indicators:

[0062]

[0063] In the formula, cov(I, J) is the covariance between the i-th index and the j-th index, I and J are the i-th and j-th indexes respectively, and the symbol σ I and σ J are the standard deviations of the i-th and j-th indicators respectively, E is the mathematical expectation; μ I and μ J are the mathematical expectations of I and J respectively. For the convenience of calculation, assuming that each variable has m samples, the simplified calculation formula of the Pearson correlation coefficient is as follows:

[0064]

[0065] In the formula, r' ij is the Pearson correlation coefficient obtained using the simplified calculation formula, represents the mean of the i-th index and the j-th index; I k , J krepresents the mth sample index value of the Ith and Jth indexes; from the mathematical expression, we can know that r' ij The value interval is [-1, 1], when r' ij When the value is greater than 0, it means that the two variables are positively correlated. ij When the value is less than 0, it means that the two variables are negatively correlated. ij The closer the absolute value of is to 1, the greater the correlation between the two variables; if it is closer to 0, the smaller the correlation between the two variables; when r' ij When the value is equal to ±1, the data points of variable x and variable y on the two-dimensional plane will fall on a straight line, that is, there is a complete linear correlation between the two variables;

[0066] (5) Obtain conflicts between data:

[0067] The definition of indicator conflict in the CRITIC algorithm is:

[0068]

[0069] A j is the index conflict of the jth index, where r' ij is the Pearson correlation coefficient between indicator i and indicator j. When the correlation between the two indicators is strong, r' ij The larger the value of ;

[0070] (6) Obtaining information carrying capacity

[0071] The information load of the defined indicators is as follows:

[0072] B j =M j ·A j

[0073] B j is the information carrying capacity of the jth indicator. The information carrying capacity of the indicator comprehensively considers the internal correlation and difference of the data. The larger its value is, the more important the indicator is in the evaluation process, and the greater the weight it should be given.

[0074] (7) Determination of weights

[0075] The final weight is calculated as follows:

[0076]

[0077] W j The final weight of the objective weighting of the jth evaluation index in the battery sorting process by the CRITIC method;

[0078] (8) Obtain the correlation coefficient between each group of battery parameter data and the corresponding parameters of the reference vector:

[0079]

[0080] In the formula, m is the number of lithium-ion batteries, n is the number of selected evaluation indicators, φ ij is the correlation coefficient between each group of battery parameter data and the corresponding parameters of the reference vector, is the optimal value of the retired battery on the evaluation parameter j, and so on; μ is the resolution coefficient, which is 0.5;

[0081] (9) Obtaining the degree of association:

[0082] The correlation coefficient is weighted by using the weight obtained by the weighting method above to obtain the correlation degree of the battery. The expression is as follows:

[0083]

[0084] In the formula, ψ i The correlation of the battery.

[0085] Furthermore, the reference vector of the grey relational analysis method is determined by the optimal value of each evaluation index, including:

[0086] Determine the optimal values ​​of static parameters: determine the maximum value of the maximum available capacity of the battery as the optimal value, and the minimum value of the battery internal resistance as the optimal value;

[0087] Determine the optimal value of the dynamic parameter: the minimum value of the corresponding evaluation index in each battery data to be evaluated is the optimal value.

[0088] The above scheme has the following beneficial effects:

[0089] 1. The present invention selects sorting characteristic parameters that can characterize the battery state, specifically including the maximum available capacity of the battery, the internal resistance of the battery and the voltage difference of 3 to 8 minutes in the discharge curve. By testing the above battery parameters and obtaining corresponding data, the k-means++ clustering algorithm is used for sorting, thereby improving the sorting accuracy and efficiency, making the values ​​of the sorting characteristic parameters between the sorted single cells as close as possible, thereby minimizing the impact of the inconsistency between batteries from all aspects to the greatest extent, thereby increasing the use efficiency and service life of the battery pack.

[0090] 2. The present invention selects three parameters that have a great influence on battery consistency as sorting parameters, including two static parameters of the battery's maximum available capacity and the battery's internal resistance, and one dynamic parameter of the voltage difference between 3 and 8 minutes in the discharge curve; the static and dynamic parameters of the battery are combined for sorting, which takes into account both the static consistency of the battery and the dynamic consistency of the battery during actual operation.

[0091] 3. The present invention combines the CRITIC weighting method with the grey correlation analysis method to conduct comprehensive performance evaluation and sorting on each group of batteries after preliminary sorting, thereby further reducing the inconsistency between batteries.

[0092] In addition, the grey correlation evaluation method is simple, has few calculation processes, and does not require high original data. Later, an objective weighting step was added to the traditional grey correlation analysis model. The CRITIC method is a method of objective weighting based on the conflict and volatility between the indicator data. The CRITIC objective weighting method is used to objectively weight according to the internal relationship between the data, reducing the impact of subjective weighting.

[0093] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] Figure 1 is a flow chart of an embodiment of the present invention;

[0095] Figure 2 A battery discharge curve diagram of an embodiment of the present invention;

[0096] Figure 3 This is a k-means++ clustering flow chart of an embodiment of the present invention;

[0097] Figure 4 This is a schematic diagram of clustering results when k=2 according to an embodiment of the present invention;

[0098] Figure 5 This is a schematic diagram of clustering results when k=3 according to an embodiment of the present invention;

[0099] Figure 6 This is a schematic diagram of clustering results when k=4 in an embodiment of the present invention;

[0100] Figure 7 This is a schematic diagram of clustering results when k=5 in an embodiment of the present invention;

[0101] Figure 8 This is a schematic diagram of clustering results when k=6 in an embodiment of the present invention;

[0102] Fig. 9 Schematic diagram of total silhouette coefficient corresponding to different k values ​​according to an embodiment of the present invention;

[0103] Fig.10 Schematic diagram of SSE attenuation with different k values ​​according to an embodiment of the present invention;

[0104] Fig.11 It is a flow chart of a grey relational evaluation algorithm based on CRITIC weighting according to an embodiment of the present invention;

[0105] Fig.12 This is an internal resistance life cycle diagram of an embodiment of the present invention;

[0106] Fig.13 It is a capacity life cycle diagram of an embodiment of the present invention;

[0107] Fig.14 A dynamic parameter life cycle diagram of an embodiment of the present invention;

[0108] Fig.15 The overall flow chart of battery sorting according to an embodiment of the present invention;

[0109] Fig.16 This is a standard deviation comparison chart of the first group in an embodiment of the present invention;

[0110] Fig.17 This is a standard deviation comparison diagram of the second group of embodiments of the present invention;

[0111] Fig.18 This is a standard deviation comparison diagram of the third group of embodiments of the present invention;

[0112] Fig.19 A first set of discharge curve comparison diagrams according to an embodiment of the present invention;

[0113] Fig. 20 A second set of discharge curve comparison diagrams according to an embodiment of the present invention;

[0114] Fig.21 This is a comparison diagram of the third set of discharge curves according to an embodiment of the present invention. DETAILED DESCRIPTION

[0115] The technical solution of the present invention will be described clearly and completely below 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 creative work are within the scope of protection of the present invention.

[0116] The following is further described in detail through specific implementation methods:

[0117] Implementation example Figure 1 As shown: A method for sorting retired lithium-ion power batteries for recycling, comprising the following steps:

[0118] S1. Selection of battery sorting parameters: Based on the actual situation of lithium-ion battery sorting, battery parameters that can be directly measured and have a greater impact on the battery pack after grouping are selected as sorting characteristic parameters, namely the maximum available capacity of the battery (C), the internal resistance of the battery (R), and the voltage difference △V (3min~8min) in the discharge curve from 3 to 8 minutes, which are used as two static sorting parameters and one dynamic sorting parameter of the battery.

[0119] The above three measurable battery parameters are used as sorting parameters, and each sorting parameter is tested; the specific test method is as follows:

[0120] (1) The test method for the maximum available capacity (C) of a battery is to reduce the battery's state of charge from 100% to 0% by constant current or constant voltage discharge, and then obtain the maximum available capacity by the ampere-hour integration method or by extracting the cumulative amount of discharged electricity; the calculation expression of the ampere-hour integration method is as follows:

[0121]

[0122] Where T is the discharge time, I is the discharge current, C is the battery capacity, t∈[0, T], and d represents the differential.

[0123] (2) The test method for the battery internal resistance (R) is as follows: the battery is tested by HPPC pulse discharge. In the HPPC test, the battery is discharged for a period of time and then its output voltage is measured. The internal resistance of the battery is obtained by measuring the voltage response under different discharge currents. The internal resistance is calculated using the following formula:

[0124] R=(V1-V2) / I

[0125] Where R is the internal resistance of the battery, V1 is the voltage of the battery at the beginning of discharge, V2 is the voltage of the battery at the end of discharge, and I is the discharge current.

[0126] (3) The test method for the voltage difference (ΔV(3min~8min)) in the discharge curve for 3 to 8 minutes is as follows: In this embodiment, the open source data set of lithium-ion batteries disclosed by the Toyota Research Institute in cooperation with MIT and Stanford University is used as an example. The first data set is selected. The lithium-ion batteries in the data set are all of the same model and specification, and their basic information is shown in Table 1.

[0127] Table 1 Lithium battery information

[0128]

[0129] In this embodiment, single cells corresponding to test channels CH27, CH28, and CH29 in the experimental batch on June 30, 2017 are selected and recorded as battery 27, battery 28, and battery 29. This data set performs a cyclic discharge aging experiment on each single cell until its capacity decays to the experimental set value. In the experiment, the discharge process of each single cell adopts the same discharge rate (4C) to start constant current discharge from the upper cut-off voltage of 4.2V until the battery voltage drops to the lower cut-off voltage of 2V. The measurement of the capacity and internal resistance of each battery is carried out under the same conditions. The data measured when the three batteries are in different cycle periods are selected for analysis. At this time, the discharge curve diagram (voltage-time) of each battery after testing is as shown in the attached figure. Figure 2 shown.

[0130] Different cycle periods correspond to different aging degrees of the battery. By observing the discharge curves of the three batteries at different aging degrees, we can see that the discharge curves of batteries at different aging degrees have roughly the same trend, that is, at the beginning of discharge, the voltage drops faster, and in the middle period, the drop rate becomes slower, and at the end of discharge, the drop rate begins to accelerate again. Although the trend is the same, it can be clearly seen that the changing trends of each curve are different, and the time when the voltage of batteries with different aging degrees begins to drop rapidly is different.

[0131] Attached Figure 2 The discharge curve trends of battery 27 and battery 29 are quite different, and their voltage difference within 3 to 8 minutes is also quite different. The curve difference between batteries 28 and 29 is relatively small, and their voltage difference within 3 to 8 minutes is also relatively small. It can be seen that the voltage difference during this period of time can be used to approximate the dynamic consistency of the discharge curves between the batteries.

[0132] In addition, the discharge curve can indirectly reflect the change of parameters such as internal resistance over time during the operation of the battery, and can centrally express battery parameters such as battery capacity, internal resistance, and voltage platform. Therefore, ensuring the consistency of the discharge dynamic curve can also comprehensively improve the consistency of other parameters. In summary, the present invention selects △V (3min~8min) as the dynamic sorting parameter of the battery.

[0133] S2. Rapid battery sorting: The optimal k value is determined by the silhouette coefficient method and the elbow rule, and then the batteries are clustered and sorted by the k-means++ algorithm according to the tested sorting parameters. The clustering process of the k-means++ algorithm is shown in the attached figure. Figure 3 As shown in the figure, the optimal K value is determined by introducing the silhouette coefficient (SC) and the elbow method. The specific method is as follows:

[0134] (1) Determine the optimal k value by the silhouette coefficient method, including:

[0135] Calculate the silhouette coefficient of each sample and calculate the dissimilarity between the sample and other samples in the same cluster. The dissimilarity is measured by the squared Euclidean distance, which is expressed as follows:

[0136]

[0137] In the formula, a i is the dissimilarity between samples in the same cluster, a i The larger the value is, the more discrete the sample is from other samples in the same cluster. n is the number of samples in the nth cluster, x i and x j are the i-th and j-th samples respectively, S n is the nth cluster, and d is the Euclidean distance.

[0138] The minimum value of the distance between the sample and other samples in different clusters is used to characterize the inter-cluster dispersion. The expression is as follows:

[0139]

[0140] Where b i is the inter-cluster dispersion of the i-th sample, S m is the mth cluster, N m is the number of samples in the mth cluster.

[0141] The calculation formula of the silhouette coefficient of a single sample that combines the intra-cluster similarity and inter-cluster dispersion is as follows:

[0142]

[0143] Calculate the mean silhouette coefficient of all sample points after clustering as the total silhouette coefficient under the current k value:

[0144]

[0145] Where SC is the total silhouette coefficient and N is the total number of clustering sample points.

[0146] From the above expressions, we can see that the closer a sample point is to other samples in the same cluster, the closer a i The smaller the value of b is, the greater the difference from other cluster samples is. i The larger the k value is, the more similar the sample point is to other sample points in the assigned cluster, and the greater the difference is from other clusters, so the classification effect is good. And because the range of SC value is [-1,1], the closer the SC value is to 1, the better the clustering result is. By comparing the SC values ​​under different k values, the optimal k value can be determined.

[0147] (2) Determine the optimal k value by the elbow rule, including:

[0148] Through the SSE-based elbow rule, its expression is as follows:

[0149]

[0150] In the formula, x i is the i-th sample, S j is the jth cluster, e j is the cluster center of the jth cluster. As can be seen from the expression, when SSE is small, it means that the sample points in each cluster are closer to the cluster center, that is, the clustering effect is good. However, SSE is not the smaller the better. Assume an extreme case: when the k value is set to the number of sample points, each sample point is divided into a cluster. At this time, the SSE value is 0, and the purpose of clustering is not achieved.

[0151] The elbow rule based on SSE takes the SSE value as an indicator, traverses different k values, and calculates the SSE values ​​under different k values. As the k value increases, the more categories are obtained, the less data in each category, and the closer the distance. Therefore, the SSE value will decrease with the increase of the k value. However, when the SSE decreases very slowly, it can be considered that further increasing the number of clusters will not enhance the classification effect. At this time, the k value corresponding to this "inflection point" is the optimal k value.

[0152] In this embodiment, a public lithium-ion battery data set is selected for verification. The selected group is the first group, and the selected batch is the batch on June 30, 2017. In each batch, each test channel corresponds to a lithium-ion battery. In practice, the process of a battery from a brand new state to a retired state is essentially that the battery is continuously charged and discharged, and the capacity decreases little by little until it reaches the retirement standard. Therefore, each battery selects one or two cycles of experimental data for analysis to simulate the sorting of retired batteries. The ratio of the current maximum available capacity to the rated capacity of each battery cycle selected in this paper is about 75% to 83%. The current maximum discharge capacity of the battery is to discharge at a constant current rate of 4C from the fully charged state of the battery until the cut-off voltage is reached, and then keep the voltage unchanged for constant voltage discharge until the current drops to the cut-off current of C / 50. The internal resistance of each battery is measured under the same conditions by applying an average of 10 ±3.6C pulses with a width of 30ms to the battery.

[0153] After removing the abnormal battery experimental data, 37 groups of data were extracted or calculated from the original data, marked as D1, D2, ..., D37. Since the dimensions of the three parameters of capacity C, resistance R, and △V (3-8min) are all different, when performing k-means++ clustering, the calculation part of the Euclidean distance will have problems caused by different dimensions, resulting in large deviations in the clustering results. Therefore, the dimensions of the parameters should be unified before clustering. This embodiment selects the maximum-minimum dimensionless processing method, and its expression is as follows:

[0154]

[0155] In the formula, x std is the sorting parameter of the test, max is the maximum value of a certain parameter in each sample, min is the minimum value of the same parameter in each sample, x is the original data of the sorting parameter of the test, and the data after dimensionless processing are in the range of [0,1], and the dimension is unified; some data before and after dimensionless processing are shown in Tables 2 and 3.

[0156] Table 2 Battery data before dimensionless processing

[0157]

[0158] Table 3 Battery data after dimensionless processing

[0159]

[0160]

[0161] The k-means++ algorithm optimizes the selection of the initial cluster center, which can improve the number of iterations of the algorithm and make it converge faster. This embodiment first uses the above data set to verify the advantages of the k-means++ algorithm.

[0162] (1) In order to avoid contingency, the dimensionless data was clustered 500 times using k-means++ and traditional k-means at different k values, and the number of iterations when each clustering reached the optimal solution was recorded. Finally, the average number of iterations of the two methods after 500 clusterings was calculated for comparison. The results are shown in Table 4.

[0163] Table 4 Comparison of average iteration times of two algorithms

[0164]

[0165] From the results in the table, we can see that when the two algorithms were used to perform 500 repeated clusterings, the average number of iterations of the k-means++ algorithm under the three k values ​​was lower than that of the k-means algorithm. Therefore, the k-means++ algorithm can improve the convergence speed of the k-means clustering algorithm and save time costs in actual battery sorting. The standardized data was imported into the k-means++ algorithm model, and the results obtained for different k values ​​were shown in the attached figure. Figure 4-8 shown.

[0166] (2) Determine the optimal K value. To determine the optimal solution for the k value, calculate the total silhouette coefficient and the sum of squared errors SSE for each k value. The results are shown in the attached figure. Fig. 9 and attached Fig.10 As shown. Fig. 9 and 10 It can be seen that the total silhouette coefficient value is the largest when k = 3, and the corresponding point when k = 3 is the inflection point of the SSE attenuation graph, indicating that the clustering effect is optimal at this time, so k = 3 is selected as the number of clustering categories. The batteries are divided into 3 groups in total, and the results are shown in Table 5.

[0167] Table 5 Classification results

[0168]

[0169] In order to verify the consistency of the batteries in each group after sorting, the performance of each group and the feasibility of the sorting method, this paper chooses the sample standard deviation to measure the discrete degree of each parameter of the single battery in the group after grouping. Its expression is as follows:

[0170]

[0171] Among them, x i is the value of a parameter of the ith battery, and n is the number of batteries in the group. The smaller the standard deviation of the parameters of the batteries in a group, the closer the values ​​of the parameters of the group of batteries are, that is, the smaller the dispersion between the data, the better the consistency after being combined into a battery group, and the higher the efficiency in the subsequent cascade utilization process.

[0172] In order to compare and verify the effectiveness of the method proposed in this paper, the original battery data is randomly divided into three groups according to the size of the number as shown in Table 6, and the standard deviation of the three parameters of each group is calculated and compared with the standard deviation of each group after clustering, as shown in Tables 7 and 8.

[0173] Table 6 Random grouping table

[0174]

[0175] Table 7 Standard deviation after random grouping

[0176]

[0177] Table 8 Standard deviation after clustering grouping

[0178]

[0179] In order to facilitate comparative analysis and make the data more intuitive, the value of S(R) is multiplied by 100 when drawing, and the data in the above table are plotted in a line graph. Based on the line graph, it can be seen that the consistency between the parameters of the three groups after clustering using k-means++ is improved compared with the three groups before clustering. Next, the performance of each group of batteries sorted by the clustering algorithm is analyzed. This study uses the absolute value of the difference between the optimal value of each group of battery parameters minus the average value of each parameter to analyze the battery performance after grouping. Based on the experimental data selected in this article, the optimal values ​​of each parameter are set as follows: capacity 0.9116Ah, internal resistance 0.195Ω, △V 0.1444559V. The calculation results are shown in Table 9.

[0180] Table 9 Absolute value table of parameter mean difference

[0181]

[0182] From the table above, we can see that the average difference of each parameter of the third group of batteries is smaller than that of the other two groups, indicating that its performance is the best. The average difference of the three parameters of the batteries in the second group is larger than that of the other two groups. The performance of each group of batteries is ranked as the third group> the first group> the second group. Therefore, this clustering method can effectively screen out batteries with different performance ranges. Among them, the performance of the second group of batteries is the worst. In practice, they can be eliminated or used as backup power according to the needs of cascade utilization.

[0183] S3. Sorting of retired batteries based on evaluation models: The consistency of each group of batteries after sorting based on the k-means++ algorithm has been improved, but there are still certain differences between some parameters of the batteries in each group, and the required consistency requirements cannot be met in some cascade utilization scenarios, and the weights of various indicators are different in different cascade utilization scenarios.

[0184] Therefore, we choose to use a comprehensive evaluation model to conduct comprehensive performance evaluation and sorting on each group of batteries after preliminary sorting, so as to further reduce the inconsistency between batteries. Grey correlation analysis method is selected as the battery sorting method. The grey correlation evaluation method is simple, has few calculation processes, and does not require high original data. Then, an objective weighting step is added to the traditional grey correlation analysis model. The CRITIC method is a method of objective weighting based on the conflict and volatility between the data of each indicator. The CRITIC objective weighting method is used to objectively weight according to the intrinsic relationship between the data, reducing the impact of subjective weighting.

[0185] Combining the CRITIC weighting method with the grey relational analysis method includes:

[0186] (1) Establish the original data matrix and reference vector:

[0187]

[0188] Where D is an original data matrix with m rows and n columns, indicating that the number of retired lithium-ion batteries in the original data D is m, and there are n selected evaluation indicators; d m is the mth retired battery; d 1n is the original data of the nth evaluation index corresponding to the first battery, d m1 is the original data of the first evaluation index corresponding to the mth battery, d mn is the original data of the corresponding nth evaluation index of the mth battery, and so on;

[0189] d * is the reference vector, which consists of the optimal values ​​of each parameter in m retired batteries; is the optimal value of retired batteries on evaluation parameter 1; is the optimal value of the retired battery in the evaluation parameter n, and so on.

[0190] (2) The values ​​in the original data matrix D are dimensionless. The dimensionless data processing is as follows:

[0191] For positive indicators:

[0192]

[0193] For negative indicators:

[0194]

[0195] Where, d' ij is a positive indicator, which means the jth positive indicator value of the i-th battery after dimensionless processing. The larger the value of the positive indicator, the better it is; d" ij is the reverse index, which represents the jth reverse evaluation index value of the i-th battery after dimensionless processing. The larger the value of the reverse index, the worse it is. For the convenience of description, the dimensionless data is represented by d ij Indicates; d in the formula ij Represents the original data of the jth evaluation index of the ith battery, maxd j is the maximum value in the original data matrix D, mind j is the minimum value in the original data matrix D.

[0196] (3) Obtaining the volatility between data:

[0197] The CRITIC algorithm uses the size of the standard deviation to measure the volatility of the data within each indicator. The subsequent weighting step will increase or decrease the weight of the indicator accordingly based on the volatility difference calculated here; the calculation formula for the standard deviation is as follows:

[0198]

[0199] Among them, M j is the standard deviation of the jth indicator, m is the number of samples, Represents the average value of the jth indicator, and the calculation formula is:

[0200]

[0201] (4) Obtain the Pearson correlation coefficient between indicators:

[0202] The CRITIC algorithm involves the determination of the correlation coefficient between indicators when calculating the conflict. Here, the Pearson correlation coefficient is used for calculation. The Pearson correlation coefficient is a statistic used to determine the degree of correlation between two variables. Its definition is the ratio of the covariance between the two variables to the product of the standard deviation, expressed as r. ij Represents the Pearson coefficient between the i-th and j-th indicators:

[0203]

[0204] In the formula, cov(I, J) is the covariance between the i-th index and the j-th index, I and J are the i-th and j-th indexes respectively, and the symbol σ I and σ J are the standard deviations of the i-th and j-th indicators respectively, E is the mathematical expectation; μ I and μ J are the mathematical expectations of I and J respectively. For the convenience of calculation, assuming that each variable has m samples, the simplified calculation formula of the Pearson correlation coefficient is as follows:

[0205]

[0206] In the formula, r' ij is the Pearson correlation coefficient obtained using the simplified calculation formula, represents the mean of the i-th index and the j-th index; I k , J k represents the mth sample index value of the Ith and Jth indexes; from the mathematical expression, we can know that r' ij The value interval is [-1, 1], when r' ij When the value is greater than 0, it means that the two variables are positively correlated. ijWhen the value is less than 0, it means that the two variables are negatively correlated. ij The closer the absolute value of is to 1, the greater the correlation between the two variables; if it is closer to 0, the smaller the correlation between the two variables; when r' ij When the value of is equal to ±1, the data points of variables x and y on the two-dimensional plane will fall on a straight line, that is, there is a complete linear correlation between the two variables.

[0207] (5) Obtain conflicts between data:

[0208] The definition of indicator conflict in the CRITIC algorithm is:

[0209]

[0210] A j is the index conflict of the jth index, where r' ij is the Pearson correlation coefficient between indicator i and indicator j. When the correlation between the two indicators is strong, r' ij The larger the value of .

[0211] (6) Obtaining information carrying capacity

[0212] The information load of the defined indicators is as follows:

[0213] B j =M j ·A j

[0214] B j is the information carrying capacity of the jth indicator. The information carrying capacity of the indicator comprehensively considers the internal correlation and difference of the data. The larger the value is, the more important the indicator is in the evaluation process and the greater the weight it should be assigned.

[0215] (7) Determination of weights

[0216] The final weight is calculated as follows:

[0217]

[0218] W j It is the final weight of the objective weighting of the j-th evaluation index in the battery sorting process by the CRITIC method.

[0219] (8) Obtain the correlation coefficient between each group of battery parameter data and the corresponding parameters of the reference vector:

[0220]

[0221] In the formula, m is the number of lithium-ion batteries, n is the number of selected evaluation indicators, φ ijis the correlation coefficient between each group of battery parameter data and the corresponding parameters of the reference vector, is the optimal value of the retired battery on the evaluation parameter j, and so on; μ is the resolution coefficient, which is 0.5.

[0222] (9) Obtaining the degree of association:

[0223] The correlation coefficient is weighted by using the weight obtained by the weighting method above to obtain the correlation degree of the battery. The expression is as follows:

[0224]

[0225] In the formula, ψ i The correlation of the battery.

[0226] The model flow chart combining CRITIC weighting method with grey relational analysis method is as follows: Fig.11 shown.

[0227] Determination of reference vector of grey relational analysis method. When using grey relational evaluation algorithm, an important step is to determine the reference vector. This paper selects the optimal value of each indicator to form the reference vector, and then the optimal value of each indicator will be determined.

[0228] (1) Determine the optimal values ​​of static parameters. Capacity and internal resistance are static parameters. According to the internal aging principle of lithium-ion batteries, when lithium-ion batteries are cyclically operated, lithium ions are repeatedly embedded and de-embedded. With continuous use, the battery continues to age, and some crystal cells of its electrode materials will be damaged. This damage is irreversible, and lithium ions cannot be embedded in places where the crystal cells are damaged. This phenomenon will manifest itself as a decrease in capacity on a macro scale. Secondly, when the lithium battery is working, the electrolyte will decompose and form an SEI film at the negative electrode. When the battery is working, some electrons will pass through the SEI film to undergo side reactions and generate lithium salts. After long-term cycles, they accumulate at the SEI film, causing it to thicken, which is manifested as an increase in internal resistance on a macro scale. Select the batteries in the test channels CH27, 28, and 29 in the public data set and record them as battery 1, battery 2, and battery 3. Draw their capacity and internal resistance life cycle diagrams, as shown in the attached figure. Fig.12 and attached Fig.13 shown.

[0229] By the attached Fig.12 and attached Fig.13 It can be seen that as the battery continues to cycle and age, the capacity generally shows a downward trend, while the internal resistance generally shows an upward trend, and as the battery ages more and more seriously, the rate of change increases. Therefore, the optimal value of capacity is determined to be the maximum value, and the optimal value of internal resistance is the minimum value.

[0230] (2) Determine the optimal value of the dynamic parameter. The △V (3-8min) selected by the present invention will be different under different battery cycles. In order to determine whether this parameter is a positive indicator or a negative indicator, the above three batteries are also selected, and the dynamic parameters of each battery are extracted every 100 cycles. The specific results are as follows: Fig.14 shown.

[0231] By the attached Fig.14 It can be seen that the overall trend of this dynamic parameter of the three batteries is also rising as the cycle period continues to increase. Therefore, this dynamic parameter, like the internal resistance, is an inverse indicator, that is, its optimal value is the minimum value of this indicator in the data of each battery to be evaluated.

[0232] When determining the sorting process, the initial state of charge (SOC) of each retired battery varies. In the subsequent group utilization process, the inconsistency of the initial SOC will also lead to overcharging and over-discharging. Therefore, when performing a discharge test on retired batteries to extract dynamic parameters, the process no longer discharges the battery for 8 minutes and stops discharging after extracting the dynamic parameters. Instead, the discharge continues after 8 minutes until the battery is fully discharged. At this time, the SOC of each battery reaches about 0%, overcoming the impact caused by the inconsistency of the initial SOC. The overall flow chart is as follows: Fig.15 shown.

[0233] The three groups of batteries were further screened and verified with data. First, the data matrix of each group was dimensionless, and the battery capacity C parameter (positive indicator) was normalized, and the internal resistance R and △V (3-8min) (reverse indicator) were normalized. Then the volatility of the three parameters of capacity C, internal resistance R, and △V (3-8min) was calculated, and the Pearson correlation coefficient between the three parameters was calculated, and then the conflict was obtained. The two were combined to obtain the information carrying capacity as shown in Table 10.

[0234] Table 10 Information carrying scale of each group data

[0235]

[0236] The final CRITIC method weights are shown in Table 11.

[0237] Table 11 CRITIC weights of each group of indicators

[0238]

[0239] The correlation coefficient between the comparison vector and the optimal value vector corresponding to each battery group is calculated, and finally the weighted correlation degree of each battery is obtained. In order to facilitate analysis, the original data of each selected battery parameter and the final correlation degree are placed in a table and sorted based on capacity. The grey analysis correlation degree of each group is sorted.

[0240] Table 13 Grey analysis correlation table of the first group

[0241]

[0242] Table 14 Grey analysis correlation table of the second group

[0243]

[0244] Table 15 Grey analysis correlation table of the third group

[0245]

[0246]

[0247] Analyzing the above table, taking the first group as an example, the battery data numbered D19 ranks second in capacity in the group, but its internal resistance is second to last, which leads to its final comprehensive evaluation correlation ranking sixth. The second and third groups also have this situation. Therefore, if evaluation is based on a single parameter, although the amount of data required to be measured is reduced, it is often difficult to ensure the consistency of the battery pack after evaluation and sorting.

[0248] In order to verify the effect of post-evaluation grouping, several batteries with similar correlations in each group are selected and matched together, further reducing battery inconsistency based on the previous step.

[0249] Table 16 Sorting results after comprehensive evaluation

[0250]

[0251] Finally, the sorting results are analyzed. This embodiment uses a consistency evaluation method, and on this basis, the discharge curve of the battery is drawn for comparison. The standard deviation of the above groups of newly matched batteries is calculated, and the results are shown in Table 17.

[0252] Table 17 Standard deviation of each group after comprehensive evaluation

[0253]

[0254] In order to make a direct comparison with the results of the grouping before and after clustering in the previous step, a three-dimensional radar chart of the data at each stage is drawn. The closer the data is to the center point in the three-dimensional radar chart, the smaller the value of its standard deviation, that is, the smaller the data dispersion, the better the consistency, and the smaller the size of the triangle formed by its connecting lines. In addition, the discharge curves of four, three, and four batteries randomly selected in the preliminary grouping are compared with the discharge curves of the finally sorted batteries. The results are shown in the attached figure. Figure 16-21 shown.

[0255] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the invention.

Claims

1. A method for sorting retired lithium-ion power batteries for cascade utilization. It is characterized in that The steps include: Three measurable battery parameters, namely, the maximum available capacity of the battery, the internal resistance of the battery, and the voltage difference between 3 and 8 minutes in the discharge curve, were selected as sorting parameters, and each sorting parameter was tested; The optimal k value is determined by the silhouette coefficient method and the elbow rule, and the batteries are clustered and sorted by the k-means++ algorithm according to the tested sorting parameters; The CRITIC weighting method is combined with the grey correlation analysis method to match the batteries with a correlation degree less than a threshold in each group of batteries after cluster sorting and assemble them into a battery pack; Among them, the CRITIC weighting method is combined with the grey correlation analysis method, including: (1) Establish the original data matrix and reference vector: Where D is an original data matrix with m rows and n columns, indicating that the number of retired lithium-ion batteries in the original data D is m, and there are n selected evaluation indicators; d m is the mth retired battery; d 1n is the original data of the nth evaluation index corresponding to the first battery, d m1 is the original data of the first evaluation index corresponding to the mth battery, d mn is the original data of the corresponding nth evaluation index of the mth battery, and so on; d * is the reference vector, which consists of the optimal values ​​of each parameter in m retired batteries; is the optimal value of retired batteries on evaluation parameter 1; is the optimal value of the retired battery on the evaluation parameter n, and so on; (2) The values ​​in the original data matrix D are dimensionless. The dimensionless data processing is as follows: For positive indicators: For negative indicators: Where, d' ij is a positive indicator, which means the jth positive indicator value of the i-th battery after dimensionless processing. The larger the value of the positive indicator, the better it is; d" ij is the reverse index, which represents the jth reverse evaluation index value of the i-th battery after dimensionless processing. The larger the value of the reverse index, the worse it is. For the convenience of description, the dimensionless data is represented by d ij Indicates; d in the formula ij Represents the original data of the jth evaluation index of the ith battery, maxd j is the maximum value in the original data matrix D, mind j is the minimum value in the original data matrix D; (3) Obtaining the volatility between data: The CRITIC algorithm uses the size of the standard deviation to measure the volatility of the data within each indicator. The subsequent weighting step will increase or decrease the weight of the indicator accordingly based on the volatility difference calculated here; the calculation formula for the standard deviation is as follows: Among them, M j is the standard deviation of the jth indicator, m is the number of samples, Represents the average value of the jth indicator, and the calculation formula is: (4) Obtain the Pearson correlation coefficient between indicators: The CRITIC algorithm involves the determination of the correlation coefficient between indicators when calculating the conflict. Here, the Pearson correlation coefficient is used for calculation. The Pearson correlation coefficient is a statistic used to determine the degree of correlation between two variables. Its definition is the ratio of the covariance between the two variables to the product of the standard deviation, expressed as r. ij Represents the Pearson coefficient between the i-th and j-th indicators: In the formula, cov(I, J) is the covariance between the i-th index and the j-th index, I and J are the i-th and j-th indexes respectively, and the symbol σ I and σ J are the standard deviations of the i-th and j-th indicators respectively, E is the mathematical expectation; μ I and μ J are the mathematical expectations of I and J respectively. For the convenience of calculation, assuming that each variable has m samples, the simplified calculation formula of the Pearson correlation coefficient is as follows: In the formula, r' ij is the Pearson correlation coefficient obtained using the simplified calculation formula, represents the mean of the i-th index and the j-th index; I k , J k represents the mth sample index value of the Ith and Jth indexes; from the mathematical expression, we can know that r' ij The value interval is [-1, 1], when r' ij When the value is greater than 0, it means that the two variables are positively correlated. ij When the value is less than 0, it means that the two variables are negatively correlated. ij The closer the absolute value of is to 1, the greater the correlation between the two variables; if it is closer to 0, the smaller the correlation between the two variables; when r' ij When the value is equal to ±1, the data points of variable x and variable y on the two-dimensional plane will fall on a straight line, that is, there is a complete linear correlation between the two variables; (5) Obtain conflicts between data: The definition of indicator conflict in the CRITIC algorithm is: A j is the index conflict of the jth index, where r' ij is the Pearson correlation coefficient between indicator i and indicator j. When the correlation between the two indicators is strong, r' ij The larger the value of ; (6) Obtaining information carrying capacity The information load of the defined indicators is as follows: B j =M j ·A j B j is the information carrying capacity of the jth indicator. The information carrying capacity of the indicator comprehensively considers the internal correlation and difference of the data. The larger its value is, the more important the indicator is in the evaluation process, and the greater the weight it should be given. (7) Determination of weights The final weight is calculated as follows: W j The final weight of the objective weighting of the jth evaluation index in the battery sorting process by the CRITIC method; (8) Obtain the correlation coefficient between each group of battery parameter data and the corresponding parameters of the reference vector: In the formula, m is the number of lithium-ion batteries, n is the number of selected evaluation indicators, φ ij is the correlation coefficient between each group of battery parameter data and the corresponding parameters of the reference vector, is the optimal value of the retired battery on the evaluation parameter j, and so on; μ is the resolution coefficient, which is 0.5; (9) Obtaining the degree of association: The correlation coefficient is weighted by using the weight obtained by the weighting method above to obtain the correlation degree of the battery. The expression is as follows: In the formula, ψ i The correlation of the battery.

2. The method for recycling and sorting retired lithium-ion power batteries according to claim 1, It is characterized in that The test method for the maximum available capacity of the battery is: through constant current or constant voltage discharge, the battery's state of charge is reduced from 100% to 0%, and then the maximum available capacity is obtained by the ampere-hour integration method or by extracting the cumulative amount of discharged electricity; The calculation expression of the ampere-hour integration method is as follows: Where C is the battery capacity, T is the discharge time, I is the discharge current, t∈[0,T], and d represents the differential.

3. The method for recycling and sorting retired lithium-ion power batteries according to claim 1, It is characterized in that The test method of battery internal resistance is: test the battery through HPPC pulse discharge. In the HPPC test, the battery is discharged for a period of time and then its output voltage is measured; the internal resistance of the battery is obtained by measuring the voltage response under different discharge currents; the internal resistance is calculated using the following formula: R=(V 1 -V 2 ) / I Where R is the internal resistance of the battery, V 1 is the voltage of the battery at the beginning of discharge, V 2 is the voltage of the battery at the end of discharge, and I is the discharge current.

4. The method for recycling and sorting retired lithium-ion power batteries according to claim 1, It is characterized in that The test method for the voltage difference between 3 and 8 minutes in the discharge curve is: The battery is discharged from 100% SOC to 0% SOC by constant current or constant voltage discharge, and the time and voltage values ​​are measured to obtain a discharge curve, and the voltage difference of 3 to 8 minutes is obtained according to the discharge curve.

5. The method for recycling and sorting retired lithium-ion power batteries according to claim 1, It is characterized in that The optimal k value is determined by the silhouette coefficient method, including: Calculate the silhouette coefficient of each sample and calculate the dissimilarity between the sample and other samples in the same cluster. The dissimilarity is measured by the squared Euclidean distance, which is expressed as follows: In the formula, a i is the dissimilarity between samples in the same cluster, a i The larger the value is, the more discrete the sample is from other samples in the same cluster. n is the number of samples in the nth cluster, x i and x j are the i-th and j-th samples respectively, S n is the nth cluster, d is the Euclidean distance; The minimum value of the distance between the sample and other samples in different clusters is used to characterize the inter-cluster dispersion. The expression is as follows: Where b i is the inter-cluster dispersion of the i-th sample, S m is the mth cluster, N m is the number of samples in the mth cluster; The calculation formula of the silhouette coefficient of a single sample that combines the intra-cluster similarity and inter-cluster dispersion is as follows: Calculate the mean silhouette coefficient of all sample points after clustering as the total silhouette coefficient under the current k value: In the formula, SC is the total silhouette coefficient, n is the total number of cluster sample points; By comparing the SC values ​​under different k values, the optimal k value is determined.

6. The method for recycling and sorting retired lithium-ion power batteries according to claim 1, It is characterized in that The optimal k value is determined by the elbow rule, including: Through the SSE-based elbow rule, its expression is as follows: In the formula, x i is the i-th sample, S j is the jth cluster, e j is the cluster center of the jth cluster; Traverse different k values ​​and calculate the SSE values ​​under different k values. The SSE value decreases as the k value increases. The k value corresponding to the inflection point where the SSE value decreases is the optimal k value.

7. The method for recycling and sorting retired lithium-ion power batteries according to claim 1, It is characterized in that Before clustering and sorting the batteries using the k-means++ algorithm, the dimensions of the sorting parameters tested are unified, including: Through the maximum-minimum dimensionless method, the expression is as follows: In the formula, x std is the sorting parameter of the test, max is the maximum value of a certain parameter in each sample, min is the minimum value of the same parameter in each sample, x is the original data of the sorting parameter of the test, and the data after dimensionless processing are all in the range of [0,1].

8. The method for recycling and sorting retired lithium-ion power batteries according to claim 1, It is characterized in that The reference vector of the grey relational analysis method is determined by the optimal value of each evaluation index, including: Determine the optimal values ​​of static parameters: determine the maximum value of the maximum available capacity of the battery as the optimal value, and the minimum value of the battery internal resistance as the optimal value; Determine the optimal value of the dynamic parameter: the minimum value of the corresponding evaluation index in each battery data to be evaluated is the optimal value.

Images