Lithium ion power battery echelon utilization auxiliary analysis method and system

Abstract

The invention relates to the technical field of lithium ion power battery recycling, and discloses a lithium ion power battery echelon utilization auxiliary analysis method and system, and the method comprises the steps: obtaining the operation parameter data of a decommissioned battery pack, analyzing the characteristics of a single battery, and confirming an echelon utilization mode (capacity recombination or internal resistance equalization); performing initial capacity and internal resistance grouping based on the mode, and generating adjustment data through capacity attenuation trajectory matching and internal resistance drift trend analysis; performing electrochemical state degradation prediction on the internal resistance adjustment data, and constructing an echelon utilization comprehensive strategy in combination with the adjustment data; and collecting strategy execution feedback data, evaluating a performance matching degree and iteratively optimizing the strategy. The system comprises a mode identification module, a capacity recombination module, an internal resistance balancing module, a strategy optimization module and a feedback execution module. According to the method, through multi-dimensional feature analysis, dynamic grouping, accurate prediction and closed-loop optimization, the efficiency and safety of echelon utilization of the retired batteries are improved, and the method is suitable for a recycling scene of the lithium ion power batteries.

Description

Technical Field

[0001] The present invention relates to the technical field of lithium-ion power battery recycling, and in particular to a lithium-ion power battery recycling auxiliary analysis method and system. Background Art

[0002] With the rapid development of the new energy vehicle industry, the application scale of lithium-ion power batteries is expanding. However, when power batteries reach the retirement standards in electric vehicles, how to efficiently, safely and economically reuse them has become a major issue facing the industry. Currently, the cascade utilization of retired lithium-ion power batteries mainly faces the following key issues:

[0003] In the process of formulating grouping strategies, most of the existing capacity grouping and internal resistance grouping methods are based on static parameter thresholds, which fail to fully consider the dynamic characteristics of battery parameters that change over time and usage conditions. For example, when grouping capacity, simple grouping is performed based only on the current capacity value, while ignoring the differences in the historical capacity decay trajectories of each single battery. Different decay trajectories may mean that the performance decay speed and life cycle of battery cells in the cascade utilization scenario are different. The static grouping method can easily lead to the problem of asynchronous capacity decay of battery cells in the same group during subsequent use, reducing the rationality of grouping and the overall efficiency of the cascade utilization system. In terms of internal resistance grouping, traditional methods usually do not conduct in-depth analysis of the drift trend of internal resistance, and cannot predict in advance the impact of internal resistance changes on battery pack performance. It is difficult to achieve dynamic internal resistance balancing control, resulting in the gradual expansion of internal resistance differences in the long-term use of the battery pack, further exacerbating the performance degradation of the battery pack.

[0004] In terms of strategy optimization and feedback mechanisms, existing technologies lack the ability to monitor and dynamically optimize the execution effect of cascade utilization strategies in real time. In actual application, cascade utilization strategies are affected by various factors such as ambient temperature and load conditions, and their execution effects may deviate from expectations. If feedback data from strategy execution cannot be collected in a timely manner and the strategy is iteratively optimized, the cascade utilization system will not always be able to maintain its optimal working state, affecting its economic benefits and reliability. For example, when a cascade utilization system is applied to an energy storage scenario, if the strategy cannot be optimized and adjusted based on actual charge and discharge data, it may lead to a decrease in the charge and discharge efficiency of the energy storage system, accelerated storage capacity decay, and failure to meet the long-term stability requirements of the battery pack performance in the energy storage scenario. Summary of the Invention

[0005] The object of the present invention is to provide a lithium-ion power battery recycling auxiliary analysis method and system to solve the problems raised in the above background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a lithium-ion power battery cascade utilization auxiliary analysis method, the method comprising:

[0007] Step S1: Obtaining operating parameter data of a retired battery pack; performing battery cell characteristic analysis on the operating parameter data of the retired battery pack to generate battery cell characteristic data; and confirming a cascade utilization mode of the battery pack based on the battery cell characteristic data to obtain cascade utilization mode data, wherein the cascade utilization mode includes a capacity reorganization mode and an internal resistance balancing mode;

[0008] Step S2: performing initial capacity grouping on the battery cell characteristic data based on the capacity reorganization mode to obtain initial capacity grouping data; performing capacity decay trajectory matching on the initial capacity grouping data to generate capacity reorganization adjustment data; performing initial internal resistance grouping on the battery cell characteristic data based on the internal resistance balancing mode to obtain initial internal resistance grouping data; performing internal resistance drift trend analysis on the initial internal resistance grouping data to generate internal resistance balancing adjustment data;

[0009] Step S3: Predicting electrochemical state degradation of the internal resistance balancing adjustment data to generate internal resistance degradation prediction data; performing dynamic internal resistance balancing control adjustment on the initial internal resistance grouping data based on the internal resistance degradation prediction data to generate dynamic internal resistance balancing strategy data; and constructing a strategy fusion for the cascade utilization mode data using the dynamic internal resistance balancing strategy data and the capacity reorganization adjustment data to generate a comprehensive cascade utilization strategy.

[0010] Step S4: collecting strategy execution feedback data for the cascade utilization comprehensive strategy to obtain a strategy execution feedback data set; performing battery pack performance matching evaluation on the strategy execution feedback data set to generate performance matching evaluation data; iteratively optimizing the cascade utilization comprehensive strategy through the performance matching evaluation data to generate a cascade utilization optimization strategy.

[0011] Preferably, step S1 includes the following steps:

[0012] Step S11: Obtaining operating parameter data of retired battery packs;

[0013] Step S12: performing parameter preprocessing on the retired battery pack operating parameter data to generate standard battery pack parameter data, wherein the parameter preprocessing includes voltage calibration, temperature compensation, data normalization, and outlier removal;

[0014] Step S13: extracting single cell characteristics from the standard battery pack parameter data to obtain single cell characteristic data; calculating the battery pack dispersion based on the single cell characteristic data to generate battery pack dispersion distribution data;

[0015] Step S14: confirming the cascade utilization mode of the battery pack according to the battery pack dispersion distribution data to obtain cascade utilization mode data, wherein the cascade utilization mode includes a capacity reorganization mode and an internal resistance balancing mode.

[0016] Preferably, step S14 includes the following steps:

[0017] Step S141: performing capacity discrete threshold analysis on the battery pack discreteness distribution data to generate capacity discreteness evaluation data; performing internal resistance discrete threshold analysis on the battery pack discreteness distribution data to generate internal resistance discreteness evaluation data;

[0018] Step S142: Calculating a mode selection weight based on the capacity dispersion evaluation data and the internal resistance dispersion evaluation data to generate mode selection weight data; wherein the parameters for calculating the mode selection weight include the capacity decay rate, the internal resistance growth slope, and the number of cycles;

[0019] Step S143: comparing the mode selection weight data with a preset mode switching threshold. When the capacity decay rate weight is higher than the preset threshold, the capacity reconfiguration mode of the battery pack is confirmed.

[0020] Step S144: When the internal resistance growth slope weight is higher than the preset threshold, the internal resistance balancing mode of the battery pack is confirmed; the capacity reorganization mode and the internal resistance balancing mode are prioritized to generate cascade utilization mode data.

[0021] Preferably, step S2 includes the following steps:

[0022] Step S21: performing initial capacity grouping on the single cell characteristic data based on the capacity reorganization mode to obtain initial capacity grouping data; performing capacity clustering on the battery pack dispersion distribution data based on the initial capacity grouping data to generate capacity distribution cluster data;

[0023] Step S22: using the capacity distribution cluster data to match the initial capacity grouping data with historical capacity attenuation trajectories to generate capacity reorganization adjustment data;

[0024] Step S23: performing initial internal resistance grouping on the characteristic data of the single cells based on the internal resistance balancing mode to obtain initial internal resistance grouping data; generating an internal resistance distribution heat map for the battery pack dispersion distribution data based on the initial internal resistance grouping data;

[0025] Step S24: performing internal resistance drift trend analysis on the internal resistance distribution heat map to generate internal resistance balance adjustment data.

[0026] Preferably, step S22 includes the following steps:

[0027] Step S221: Using the capacity distribution cluster data to perform capacity decay curve screening on the characteristic data of the single battery, to obtain a set of historical capacity decay curves;

[0028] Step S222: extracting curve morphological features from the historical capacity decay curve set to generate capacity decay morphological feature data; and calculating intra-cluster similarity of the initial capacity grouping data based on the capacity decay morphological feature data;

[0029] Step S223: When the intra-cluster similarity is lower than a preset threshold, the cluster boundary of the initial capacity grouping data is adjusted according to the capacity decay morphological feature data;

[0030] Step S224: When the intra-cluster similarity is higher than a preset threshold, the initial capacity grouping data is optimized for inter-cluster spacing; the cluster boundary adjustment data and the inter-cluster spacing optimization data are integrated to generate capacity reorganization adjustment data.

[0031] Preferably, step S224 includes the following steps:

[0032] When the intra-cluster similarity is higher than the preset threshold, the inter-cluster cross-interference analysis is performed on the capacity distribution cluster data to generate the inter-cluster interference coefficient; the inter-cluster distance of the initial capacity grouping data is reset according to the inter-cluster interference coefficient;

[0033] Perform capacity gradient verification on the reset inter-cluster distance to generate capacity gradient verification data; perform dynamic weight correction on the inter-cluster interference coefficient based on the capacity gradient verification data to generate a corrected interference coefficient;

[0034] The inter-cluster spacing is optimized by using the corrected interference coefficient and the capacity gradient verification data to generate optimized capacity grouping data; the optimized capacity grouping data is fused with the cluster boundary adjustment data to generate capacity reorganization adjustment data.

[0035] Preferably, step S24 includes the following steps:

[0036] Step S241: extracting regional extreme resistance values ​​from the internal resistance distribution heat map to generate regional extreme value distribution data;

[0037] Step S242: Identify the internal resistance drift direction based on the regional extreme value distribution data to generate an internal resistance drift vector diagram; perform trend line fitting on the internal resistance drift vector diagram to generate internal resistance drift trend line data;

[0038] Step S243: Calculating the drift compensation amount for the initial internal resistance group data according to the internal resistance drift trend line data to generate a drift compensation parameter set;

[0039] Step S244: dynamically adjusting the group boundaries of the initial internal resistance group data using the drift compensation parameter set to generate internal resistance balance adjustment data.

[0040] Preferably, step S3 includes the following steps:

[0041] Step S31: extracting electrochemical impedance spectrum characteristics from the internal resistance balance adjustment data to generate impedance spectrum characteristic data;

[0042] Step S32: performing electrochemical state degradation prediction based on the impedance spectrum characteristic data to generate internal resistance degradation prediction data;

[0043] Step S33: Calculating dynamic compensation coefficients for the initial internal resistance group data according to the internal resistance degradation prediction data to generate a dynamic compensation coefficient set;

[0044] Step S34: constructing a strategy by integrating the dynamic compensation coefficient set with the internal resistance balance adjustment data to generate a comprehensive strategy for cascade utilization.

[0045] Preferably, step S32 includes the following steps:

[0046] Step S321: performing frequency domain response feature decomposition on the impedance spectrum characteristic data to generate a frequency domain feature vector set;

[0047] Step S322: performing degradation feature recognition on the frequency domain feature vector set through a convolutional neural network to generate degradation feature labeling data;

[0048] Step S323: Predicting an electrochemical state degradation curve based on the degradation feature marker data to generate degradation curve prediction data;

[0049] Step S324: performing matching verification based on the degradation curve prediction data and the historical degradation database to generate internal resistance degradation prediction data.

[0050] Preferably, the present invention further includes a lithium-ion power battery cascade utilization auxiliary analysis system for executing the lithium-ion power battery cascade utilization auxiliary analysis method as described above, the system comprising:

[0051] A pattern recognition module is used to obtain operating parameter data of retired battery packs; perform battery cell feature analysis on the operating parameter data of retired battery packs to generate battery cell feature data; and perform cascade utilization mode confirmation on the battery pack based on the battery cell feature data to obtain cascade utilization mode data;

[0052] A capacity reorganization module is used to perform initial capacity grouping on the battery cell characteristic data based on the capacity reorganization mode to obtain initial capacity grouping data; perform capacity attenuation trajectory matching on the initial capacity grouping data to generate capacity reorganization adjustment data;

[0053] The internal resistance balancing module is used to perform initial internal resistance grouping on the battery cell characteristic data based on the internal resistance balancing mode to obtain initial internal resistance grouping data; perform internal resistance drift trend analysis on the initial internal resistance grouping data to generate internal resistance balancing adjustment data;

[0054] The strategy optimization module is used to predict electrochemical state degradation based on internal resistance balancing adjustment data and generate internal resistance degradation prediction data; dynamically adjust the initial internal resistance grouping data based on the internal resistance degradation prediction data to generate dynamic internal resistance balancing strategy data; and integrate the dynamic internal resistance balancing strategy data with the capacity reorganization adjustment data to construct a strategy for cascade utilization mode data to generate a comprehensive cascade utilization strategy;

[0055] The feedback execution module is used to collect strategy execution feedback data for the cascade utilization comprehensive strategy to obtain a strategy execution feedback data set; perform battery pack performance matching evaluation on the strategy execution feedback data set to generate performance matching evaluation data; and iteratively optimize the cascade utilization comprehensive strategy using the performance matching evaluation data to generate a cascade utilization optimization strategy.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] In terms of battery feature analysis and mode confirmation, the accuracy and reliability of the data are ensured by obtaining the operating parameter data of retired battery packs and performing parameter preprocessing (including voltage calibration, temperature compensation, data normalization, and outlier removal). On this basis, the single-cell feature extraction and battery pack dispersion calculation are performed on the standard battery pack parameter data, which can accurately analyze the degree of dispersion of key parameters such as battery pack capacity and internal resistance. Through capacity discrete threshold analysis and internal resistance discrete threshold analysis, the mode selection weight is calculated by combining parameters such as capacity decay rate, internal resistance growth slope, and number of cycles, and compared with the preset mode switching threshold, a scientific and reasonable confirmation of the battery pack cascade utilization mode (capacity reorganization mode or internal resistance balancing mode) is achieved. This refined feature analysis and mode confirmation method changes the limitations of traditional methods that rely solely on experience or a single parameter for mode judgment. It can select the most suitable cascade utilization mode based on the actual performance status of the battery pack, laying a solid foundation for the subsequent grouping strategy formulation and improving the pertinence and effectiveness of cascade utilization.

[0058] In the process of formulating the capacity reorganization and internal resistance balancing grouping strategy, after the initial capacity grouping is set for the single cell characteristic data based on the capacity reorganization mode, the capacity distribution cluster data is used to match the historical capacity attenuation trajectory of the initial capacity grouping data. Specifically, the capacity distribution cluster data is used to screen the historical capacity attenuation curve set, extract the capacity attenuation morphological characteristic data, and calculate the intra-cluster similarity based on this. According to the similarity results, the cluster boundaries of the initial capacity grouping data are adjusted or the inter-cluster spacing is optimized. This dynamic capacity grouping adjustment method fully considers the differences in the historical capacity attenuation trajectories of battery cells, so that the battery cells in the same group have similar capacity attenuation trends, effectively improves the rationality of capacity grouping, and reduces the problem of battery pack performance degradation caused by asynchronous capacity attenuation. In the internal resistance balancing mode, by generating an internal resistance distribution heat map and performing internal resistance drift trend analysis, the regional internal resistance extreme values ​​are extracted, the internal resistance drift direction is identified and the trend line is fitted, and then the drift compensation amount is calculated and the grouping boundary is dynamically adjusted. This method can track the changing trend of internal resistance in real time, predict the impact of internal resistance differences on battery pack performance in advance, realize dynamic optimization of internal resistance grouping, ensure the internal resistance balance of battery packs during cascade utilization, and reduce safety risks caused by excessive internal resistance differences.

[0059] In terms of electrochemical state degradation prediction and strategy fusion construction, by extracting electrochemical impedance spectroscopy features from internal resistance balancing adjustment data, using convolutional neural networks to identify degradation features of frequency domain feature vector sets, and combining them with historical degradation databases for matching verification, accurate prediction of battery internal resistance degradation is achieved. Based on the internal resistance degradation prediction data, a dynamic compensation coefficient set is calculated, and a strategy fusion is constructed with the internal resistance balancing adjustment data and capacity reorganization adjustment data to generate a comprehensive strategy for cascade utilization. This predictive control method based on advanced algorithms and big data analysis can deeply explore the electrochemical changes in the battery, formulate targeted internal resistance balancing control strategies in advance, effectively slow down the degradation rate of battery performance, and improve the scientific nature and foresight of the cascade utilization strategy. Through strategy fusion construction, the capacity reorganization and internal resistance balancing strategies are organically combined to achieve comprehensive optimization of battery pack performance and further improve the overall efficiency of the cascade utilization system.

[0060] During the strategy execution feedback and iterative optimization phase, the strategy execution feedback dataset is collected and the battery pack performance matching is evaluated. Based on the evaluation results, the comprehensive cascade utilization strategy is iteratively optimized. This closed-loop feedback optimization mechanism monitors the effectiveness of strategy execution in real time, promptly identifies problems in the strategy's actual application, and adjusts and optimizes it based on actual conditions. This allows the cascade utilization strategy to consistently adapt to varying environmental conditions and load conditions, ensuring the stability and reliability of the cascade utilization system. Through continuous iterative optimization, the performance of the cascade utilization system can be continuously improved, extending the cascade utilization cycle of retired batteries and increasing resource utilization and economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 This is a working principle diagram of the auxiliary analysis method for lithium-ion power battery recycling according to the present invention;

[0062] Figure 2 Flowchart for parameter preprocessing and feature extraction of retired battery packs;

[0063] Figure 3 Flowchart for decision making on cascade utilization model selection. DETAILED DESCRIPTION

[0064] 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.

[0065] See also Figure 1-Figure 3 The present invention relates to an auxiliary analysis method for the cascade utilization of lithium-ion power batteries, and the specific implementation steps are as follows:

[0066] Step S1: Obtaining operating parameter data of a retired battery pack, and generating battery cell characteristic data by performing battery cell characteristic analysis on the data; confirming a cascade utilization mode of the battery pack based on the characteristic data, and obtaining cascade utilization mode data including a capacity reorganization mode and an internal resistance balancing mode.

[0067] Step S2: In the capacity reorganization mode, the battery cell characteristic data is initially grouped into capacity groups to obtain initial capacity grouping data, and then the data is matched with the capacity decay trajectory to generate capacity reorganization adjustment data; in the internal resistance balancing mode, the battery cell characteristic data is initially grouped into internal resistance groups to obtain initial internal resistance grouping data, and then the data is analyzed for internal resistance drift trends to generate internal resistance balancing adjustment data.

[0068] Step S3: Predict electrochemical state degradation based on the internal resistance balance adjustment data to generate internal resistance degradation prediction data; perform dynamic internal resistance balance control adjustment on the initial internal resistance grouping data based on the prediction data to generate dynamic internal resistance balance strategy data; and construct a comprehensive strategy for cascade utilization by integrating the dynamic internal resistance balance strategy data with the capacity reorganization adjustment data.

[0069] Step S4: Collect the strategy execution feedback data of the cascade utilization comprehensive strategy to form a strategy execution feedback data set; perform battery pack performance matching evaluation on the data set to generate performance matching evaluation data; iteratively optimize the cascade utilization comprehensive strategy based on the evaluation data to generate a cascade utilization optimization strategy.

[0070] The technical solution of the present invention is further described in detail below with reference to specific embodiments.

[0071] Example 1:

[0072] This embodiment involves a detailed implementation of step S1, specifically including obtaining operational parameter data for retired battery packs, performing parameter preprocessing on the data, extracting cell characteristics and calculating dispersion, and determining a cascade utilization model based on the dispersion data. The following describes each sub-step in detail, including its specific execution, data processing logic, and interactions.

[0073] In step S11, the operating parameter data of the retired battery pack is obtained. The data can be directly read from the battery management system (BMS) through a wired or wireless communication interface, or by importing a historical operating data file from a storage medium. The data content includes but is not limited to the voltage sequence of the single cell, the charge and discharge current curve, the temperature distribution data, the cycle number statistics, the cumulative charge and discharge capacity and other real-time or historical operating parameters. The data acquisition frequency must meet the accuracy requirements of subsequent analysis, for example, collecting voltage data once per second, recording the average temperature value once per minute, etc. The acquired data is stored in a structured format (such as CSV, database table) to facilitate subsequent processing.

[0074] Execute step S12 to perform parameter preprocessing on the raw data to generate standard battery pack parameter data. The preprocessing process includes four core links: voltage calibration, temperature compensation, data normalization and outlier removal. In the voltage calibration link, by comparing the voltage data of all single cells at the same time, the single cells with voltage jumps or continuous deviations from the group mean are identified, and the abnormal voltage data are corrected using polynomial fitting or Kalman filtering algorithm. For example, if the voltage of a single cell suddenly jumps and exceeds the normal fluctuation range of the adjacent cells during the charging process, a reasonable voltage curve is fitted according to the voltage trend of the previous and next moments to replace the abnormal data points. In the temperature compensation link, based on the battery thermodynamic model, a mapping relationship between temperature and parameters such as capacity and internal resistance is established. For the capacity data measured at different temperatures, they are uniformly converted to equivalent capacity values ​​at standard temperature (such as 25°C). The specific conversion formula can use the Arrhenius equation correction coefficient to eliminate the interference of temperature on capacity characterization. During data normalization, the Z-score method is used to convert parameters of different dimensions, such as voltage, current, and internal resistance, into dimensionless values ​​with a mean of 0 and a standard deviation of 1. For example, for voltage parameters, the mean μ and standard deviation σ of the voltage of all individual cells are calculated. The normalized value of a cell voltage value x is (x-μ) / σ, ensuring that different parameters are on the same order of magnitude, facilitating subsequent feature extraction and comparative analysis. During outlier removal, a density-based outlier detection algorithm (such as DBSCAN) is used to independently detect each parameter dimension and identify abnormal points that deviate from the main data distribution. For example, in the capacity data distribution, if the capacity value of a cell falls outside three standard deviations of the global data, it is identified as an outlier and removed. The location and type of the outlier are also recorded for reference in subsequent analysis. After this preprocessing, standard battery pack parameter data with a uniform format and no significant outliers is generated.

[0075] In step S13, single-cell feature extraction and dispersion calculation are performed on the standard battery pack parameter data. In the feature extraction stage, statistical features, trend features, and cycle features are extracted from the time series data for each single cell. Statistical features include voltage mean, variance, maximum, and minimum values; initial value, current value, and attenuation of capacity; initial value, current value, and growth of internal resistance; trend features are calculated through linear regression to calculate voltage attenuation slope, capacity attenuation rate, and internal resistance growth slope; and cycle features include number of cycles, charge and discharge depth distribution, and the proportion of high and low temperature cycles. Through the above feature extraction, single-cell feature data containing dozens of dimensions is generated, with each cell corresponding to a feature record. Fields include cell number, voltage mean, capacity attenuation rate, and internal resistance growth slope. In the dispersion calculation stage, based on the single-cell feature data, the dispersion of the battery pack in key parameters such as capacity, internal resistance, and voltage is calculated. Capacity dispersion is measured by calculating the ratio of the standard deviation of all single-cell capacity values ​​to the mean (coefficient of variation), reflecting the uniformity of the capacity distribution within the battery pack. Internal resistance dispersion uses the same coefficient of variation calculation method to evaluate the consistency of internal resistance parameters. Voltage dispersion is characterized by the difference between the maximum and minimum voltages (voltage range) and the standard deviation. In addition, the distribution entropy value of each parameter can be calculated to measure the degree of disorder in the data distribution. The resulting battery pack dispersion distribution data includes statistical quantities such as the coefficient of variation, range, and entropy value of each parameter, as well as visual descriptions such as histograms or box plots of the parameter distribution, which intuitively reflect the degree of performance differences between single cells within the battery pack.

[0076] Step S14 determines the cascade utilization mode based on the battery pack dispersion distribution data. This includes sub-steps such as threshold analysis, weight calculation, mode selection, and priority sorting. In step S141, capacity dispersion threshold analysis and internal resistance dispersion threshold analysis are first performed. For capacity dispersion, a preset capacity dispersion threshold range (e.g., coefficient of variation 0.1-0.3) is used. Based on the battery pack's usage scenario and cascade utilization objectives (e.g., energy storage systems with high capacity consistency requirements), a determination is made as to whether the current capacity dispersion exceeds a reasonable range. If the coefficient of variation is greater than 0.3, the capacity distribution is too dispersed, and a capacity reorganization mode should be prioritized. If it is less than 0.1, the capacity consistency is good, and other modes may be more suitable. The internal resistance dispersion threshold analysis uses similar logic, with a preset internal resistance coefficient of variation threshold (e.g., 0.2-0.4). Combined with the battery pack's state of health (SOH) assessment results, a determination is made as to whether the internal resistance dispersion exceeds the normal aging range. Through this analysis, capacity dispersion assessment data and internal resistance dispersion assessment data are generated, containing information such as the actual dispersion value, a comparison with the threshold, and the dispersion level (e.g., high, medium, low).

[0077] In step S142, based on the capacity dispersion evaluation data and the internal resistance dispersion evaluation data, the mode selection weight is calculated in combination with parameters such as the capacity decay rate, the internal resistance growth slope, and the number of cycles. The mode selection weight adopts a multi-factor weighted summation method, and the weight coefficient of each factor is pre-set according to the application scenario of the battery cascade utilization. For example, in the energy storage scenario, capacity consistency has a greater impact on system performance, so the weight coefficient of the capacity decay rate is set to 0.5, the weight of the internal resistance growth slope is 0.3, and the weight of the number of cycles is 0.2; while in the low-speed electric vehicle scenario, internal resistance balance is more important for safety, and the weight distribution can be adjusted to 0.3 for capacity decay rate, 0.5 for internal resistance growth slope, and 0.2 for number of cycles. During the specific calculation, each parameter is first normalized, and the capacity decay rate (unit: % / cycle), internal resistance growth slope (unit: mΩ / cycle), number of cycles, etc. are converted into dimensionless values ​​between 0 and 1. Then, the comprehensive weight value is calculated according to the weight coefficient to generate the mode selection weight data, which includes the capacity-related weight, internal resistance-related weight, and comprehensive weight value.

[0078] In step S143 and step S144, the mode selection weight data is compared with the preset mode switching threshold for decision. The preset mode switching threshold is pre-set according to the battery type, retirement standard and cascade utilization target. For example, the capacity decay rate weight threshold is set to 0.4, and the internal resistance growth slope weight threshold is set to 0.4. If the calculated capacity decay rate weight is higher than 0.4, it is determined that the main problem of the battery pack is the uneven capacity distribution, and the capacity reorganization mode needs to be confirmed, that is, by regrouping and matching single cells with similar capacity, the capacity consistency of the battery pack is improved; if the internal resistance growth slope weight is higher than 0.4, it is determined that the internal resistance drift is the main problem, and the internal resistance balancing mode needs to be confirmed, that is, by adjusting the grouping strategy to balance the internal resistance difference and reduce the impact of internal resistance on the performance of the battery pack. When both the capacity decay rate weight and the internal resistance growth slope weight do not exceed the threshold, further judgment can be made based on other parameters such as the number of cycles, or a hybrid mode combining capacity reorganization and internal resistance balancing can be adopted by default. Furthermore, the capacity reorganization mode and the internal resistance balancing mode are prioritized. For example, under normal circumstances, the capacity reorganization mode takes precedence over the internal resistance balancing mode. However, if the internal resistance dispersion reaches a critical level (e.g., the coefficient of variation exceeds 0.5), the internal resistance balancing mode is prioritized. The resulting cascade utilization mode data includes information such as the selected mode type, the proportion of mixed modes, and the order of mode priority, providing a clear basis for the formulation of subsequent grouping strategies.

[0079] During the execution of the entire step S1, there is a strict data flow logic between each sub-step: the original data is pre-processed to generate standard data, the standard data is extracted through feature extraction to generate monomer feature data, the monomer feature data is combined with the discreteness calculation to generate discreteness distribution data, the discreteness data is analyzed by threshold value and weight calculation to generate the basis for mode selection, and finally the cascade utilization mode is determined. The output data of each link is used as the input of the next link to form a complete data analysis chain. At the same time, data checkpoints can be set in each link, such as checking the data integrity (missing value ratio) after pre-processing, and verifying the correlation between the feature and the battery status after feature extraction (such as analyzing the correlation between the capacity decay rate and the number of cycles through the Pearson correlation coefficient) to ensure the data quality and the reliability of the analysis results.

[0080] The detailed implementation described above enables intelligent identification of recycling patterns from the raw operating data of retired battery packs, providing a foundation for subsequent operations such as capacity grouping and internal resistance balancing. This process fully considers the standardization of data preprocessing, the comprehensiveness of feature extraction, and the scientific nature of pattern selection, ensuring that the recycling pattern is determined in accordance with the actual performance status of the battery pack, laying the foundation for improving the efficiency and safety of retired battery reuse.

[0081] Example 2:

[0082] This embodiment involves a detailed implementation of the capacity reorganization mode in step S2, specifically including initial capacity grouping, capacity cluster division, historical capacity decay trajectory matching, and group adjustment. The following describes the data processing flow, algorithm logic, and parameter interaction in detail.

[0083] In step S21, the initial capacity grouping is set for the single cell characteristic data based on the capacity reorganization mode. First, the current capacity value of each single cell is extracted from the single cell characteristic data as the basis for grouping. The capacity value can be obtained through standard charge and discharge tests, or calculated based on the difference between the cumulative discharge capacity and the initial capacity in the historical operation data. The initial grouping adopts the equal-interval segmentation method or the adaptive segmentation method: the equal-interval segmentation method divides the fixed interval (such as every 10Ah as a group) according to the range of capacity values ​​(such as 0-100Ah), and classifies the single cells into the corresponding interval according to the capacity value; the adaptive segmentation method reduces the grouping interval in the data-dense area and expands the interval in the sparse area by statistically analyzing the frequency distribution of the capacity data, so that the number of cells in each group is as balanced as possible. After the grouping is completed, the initial capacity grouping data is generated, and each group contains information such as a list of cell numbers, a capacity interval range, and an average capacity value.

[0084] The battery pack dispersion distribution data is divided into capacity clusters based on the initial capacity grouping data. Capacity clustering is achieved using a clustering algorithm, and commonly used algorithms include the K-means algorithm and the hierarchical clustering algorithm. Taking the K-means algorithm as an example, the number of clusters K is first preset (e.g., K = 5, which can be dynamically adjusted based on the capacity dispersion evaluation data), and K initial cluster centers are randomly selected (usually the average capacity value of each initial grouping). The Euclidean distance between each single battery and each cluster center is then calculated, and each battery is assigned to the cluster with the closest distance. The average capacity value of each cluster is then recalculated as the new cluster center, and the iteration is repeated until the cluster center no longer changes significantly or the maximum number of iterations is reached. Through capacity clustering, capacity distribution cluster data is generated. Each cluster contains information such as the cluster number, cluster center capacity value, a list of cells within the cluster, and the capacity distribution standard deviation, which intuitively reflects the distribution of cells at different capacity levels.

[0085] In step S22, the capacity distribution cluster data is used to match the historical capacity decay trajectory of the initial capacity group data. First, step S221 is executed to filter the capacity decay curve that matches the characteristics of each capacity cluster from the historical data. The historical data comes from retired battery packs of the same type or under similar usage conditions. Each curve records the capacity value of a single cell at different cycle times. The screening rules are based on the average capacity value and decay rate range of the current cluster. For example, for a cluster with an average capacity of 80Ah and a decay rate of 0.1% / cycle, the curves with an initial capacity of 75-85Ah and a decay rate between 0.08%-0.12% / cycle in the historical data are filtered to form a set of historical capacity decay curves.

[0086] In step S222, the curve morphology feature extraction is performed on the historical capacity decay curve set. Feature extraction includes time domain features and frequency domain features: the time domain feature extraction curve slope (i.e., decay rate), intercept (initial capacity), inflection point position (such as the number of cycles when the capacity decays to 80% of the initial value), fluctuation amplitude (maximum deviation of the capacity value during the charge and discharge process), etc.; the frequency domain feature converts the curve into frequency components through Fourier transform, extracts the main frequency components and their amplitudes, and reflects the periodic or regular changes in capacity decay. The generated capacity decay morphology feature data is stored in the form of feature vectors, each vector containing the above-mentioned time domain and frequency domain feature parameters.

[0087] Based on the capacity decay morphological feature data, intra-cluster similarity is calculated for the initial capacity grouped data. This similarity is measured using the cosine similarity algorithm or the dynamic time warping (DTW) algorithm. Cosine similarity measures the angle between feature vectors, ranging from -1 to 1, with larger values ​​indicating greater similarity. The DTW algorithm is applicable to time series data, minimizing the morphological differences between two curves by bending the time axis and calculating the cumulative distance as a similarity metric. Using cosine similarity as an example, the similarity between the capacity decay morphological feature vectors of all cells in a cluster and the cluster center feature vector is calculated, and the average value is taken as the intra-cluster similarity value, which reflects the consistency of the decay trends of the cells within the cluster.

[0088] When the intra-cluster similarity falls below a preset threshold (e.g., 0.6), it indicates that the decay morphologies of the cells within the cluster differ significantly, and step S223 needs to be executed to adjust the cluster boundaries. The adjustment method includes: marking cells with similarity below the threshold as "abnormal cells," analyzing the differences between their feature vectors and those of other cells in the cluster (e.g., significantly larger decay slopes), and then reallocating them to the most matching cluster based on the similarity of their feature vectors to the centers of other clusters. Simultaneously, the feature vectors and similarities of each adjusted cluster are recalculated until the similarity of all clusters exceeds the threshold or the maximum number of adjustments is reached.

[0089] When the intra-cluster similarity is higher than the preset threshold, step S224 is executed to optimize the inter-cluster spacing. First, the capacity distribution cluster data is subjected to inter-cluster cross-interference analysis to calculate the inter-cluster interference coefficient. The interference coefficient is measured by the degree of overlap of the capacity distribution of adjacent clusters. For example, for cluster A (capacity range 70-80Ah) and cluster B (80-90Ah), the overlap ratio of the capacity values ​​of the two near 80Ah is calculated. The higher the overlap ratio, the greater the interference coefficient. The inter-cluster distance of the initial capacity grouping data is reset according to the interference coefficient. For adjacent clusters with larger interference coefficients, the capacity interval interval is increased (such as setting the upper limit of cluster A to 78Ah and the lower limit of cluster B to 82Ah) to reduce cross-interference; for clusters with smaller interference coefficients, the interval is maintained or reduced to optimize grouping compactness.

[0090] After resetting the inter-cluster distance, perform capacity gradient verification. Capacity gradient verification evaluates the rationality of inter-cluster capacity differences by calculating the ratio of the average capacity difference of adjacent clusters to the standard deviation within the cluster. For example, if the average capacity of cluster A is 75Ah, the average capacity of cluster B is 85Ah, and the standard deviation within the cluster is 2Ah, then the capacity gradient is 10Ah, which is significantly greater than the standard deviation, indicating that the inter-cluster differences are reasonable; if the average difference is 3Ah, which is close to the standard deviation, there may be a problem of blurred inter-cluster boundaries, and the distance reset strategy needs to be adjusted. The generated capacity gradient verification data contains information such as the average difference, standard deviation, and gradient ratio of each adjacent cluster.

[0091] Based on the capacity gradient verification data, the inter-cluster interference coefficient is dynamically weighted. The correction rule is: if the capacity gradient verification results show that the inter-cluster differences are reasonable, the interference coefficient remains unchanged; if the differences are insufficient, the weight of the interference coefficient is increased to further expand the inter-cluster distance; if the differences are too large, the weight is reduced to allow for an appropriate reduction in distance. The corrected interference coefficient is used in subsequent inter-cluster spacing optimization calculations. By iteratively adjusting the inter-cluster distance, the capacity distribution cluster data maintains reasonable gradient differences while minimizing cross-interference. Finally, the optimized inter-cluster spacing data is fused with the cluster boundary adjustment data to generate capacity reorganization adjustment data, which contains information such as the adjusted capacity intervals of each cluster, the monomer allocation list, and the similarity index.

[0092] Throughout the capacity reorganization model, initial capacity grouping provides a foundational framework for subsequent analysis. Capacity clustering, through clustering algorithms, achieves preliminary data structuring. Matching historical capacity decay trajectories introduces dynamic analysis over time, enabling the grouping strategy to consider not only the current capacity status but also future decay trends. Intra-cluster similarity calculation and inter-cluster spacing optimization form a "cohesive within, divergent outside" grouping optimization mechanism: intra-cluster similarity ensures consistency within a group, while inter-cluster spacing optimization ensures significant differences between groups. Together, these two enhance the refinement of capacity reorganization.

[0093] Furthermore, data interaction between each link is tightly integrated: initial capacity grouping data provides input for capacity clustering, clustering results drive historical curve screening, curve feature extraction supports similarity calculation, and similarity results determine the direction of boundary adjustment or spacing optimization. The resulting adjustment data is then fed back into the grouping strategy. The entire process can be automated through programming, such as using Python's Scikit-learn library to implement K-means clustering and cosine similarity calculation, and using the Pandas library for data cleaning and feature engineering, ensuring the efficiency and repeatability of the analysis process.

[0094] Example 3:

[0095] This embodiment involves a detailed implementation of the internal resistance balancing mode in step S2, specifically including initial internal resistance grouping, internal resistance distribution heat map generation, internal resistance drift trend analysis, and dynamic adjustment of group boundaries. The following describes the data processing logic, visualization analysis method, and dynamic adjustment strategy in detail.

[0096] In step S23, the initial internal resistance grouping is performed on the single cell characteristic data based on the internal resistance balancing mode. First, the current internal resistance value of each cell is extracted from the single cell characteristic data as the basis for grouping. The internal resistance value is obtained through AC impedance measurement or DC internal resistance testing, and includes internal resistance data in the charging state and internal resistance data in the discharging state. The initial grouping adopts a density-based grouping method. For example, based on the kernel density estimation results of the internal resistance data, the grouping boundary is set at the peak of the data distribution, so that the internal resistance data in each group forms a significant density peak. After the grouping is completed, the initial internal resistance group data is generated. Each group contains information such as a list of cell numbers, internal resistance range, average internal resistance value, and internal resistance distribution density.

[0097] The internal resistance distribution heat map of the battery pack discrete distribution data is generated based on the initial internal resistance grouping data. The heat map is presented in the form of a two-dimensional matrix, with the horizontal axis representing the physical position of the single cell (such as the arrangement number in the battery pack), and the vertical axis representing the internal resistance value range. The color depth of each cell corresponds to the internal resistance value of the single cell at that position (the darker the color, the greater the internal resistance). When generating the heat map, the physical structure of the battery pack is first mapped to a coordinate grid, with each single cell corresponding to a node in the grid; then the internal resistance value is normalized to a color value range of 0-255, and a gradient color scheme (such as from blue to red) is used to represent the change in internal resistance from small to large; finally, an interpolation algorithm (such as bilinear interpolation) is used to fill the color transition between adjacent nodes, so that the heat map smoothly displays the spatial continuity of the internal resistance distribution. The internal resistance distribution heat map can intuitively reveal abnormal internal resistance areas within the battery pack. For example, a significant increase in resistance in a local area may indicate problems such as battery aging or poor contact.

[0098] In step S24, the internal resistance drift trend analysis is performed on the internal resistance distribution heat map. First, step S241 is executed to extract the regional internal resistance extreme values ​​in the heat map and generate regional extreme value distribution data. The extreme value extraction adopts the sliding window method to divide the heat map into multiple sub-regions (such as a 3×3 grid window), calculate the maximum internal resistance and the minimum internal resistance in each window, and mark them as regional extreme value points. For each extreme value point, its coordinate position, internal resistance value, extreme value type (maximum or minimum value) and sub-region number are recorded to form regional extreme value distribution data, which can locate the specific location and degree of internal resistance anomaly.

[0099] In step S242, the internal resistance drift direction is identified based on the regional extreme value distribution data, and an internal resistance drift vector diagram is generated. The method for constructing the vector diagram is as follows: for each extreme value point, its current internal resistance value is compared with the internal resistance value of the monomer at the same position in the historical data (the historical data can be taken from the heat map of the previous cycle), and the internal resistance change ΔR is calculated; if ΔR>0, it indicates that the internal resistance has increased and the drift direction is "positive"; if ΔR<0, it indicates that the internal resistance has decreased and the drift direction is "negative". With the extreme value point as the starting point of the vector, a vector with direction (arrow) and amplitude (line segment length) is drawn. The vector length is proportional to |ΔR| to form an internal resistance drift vector diagram. Subsequently, a trend line is fitted to the vector diagram, and the least squares method is used to perform linear regression on the vector data in the same area to fit the internal resistance drift trend line. The slope of the trend line represents the average rate of change of the resistance in the region (unit: mΩ / cycle), and the intercept represents the initial internal resistance level. The generated internal resistance drift trend line data includes the slope, intercept and goodness of fit (R 2 value) and other parameters.

[0100] In step S243, the drift compensation amount of the initial internal resistance group data is calculated based on the internal resistance drift trend line data. The compensation amount calculation is based on the following logic: for each internal resistance group, the trend line slopes of the areas where all monomers in the group are located are counted, and the average slope is calculated as the drift rate prediction value of the group; if the average slope is positive, it indicates that the internal resistance of the group is on an upward trend as a whole, and a positive compensation amount needs to be reserved to adapt to future internal resistance increases; if it is negative, it indicates a downward trend, and a negative compensation amount needs to be reserved. The drift compensation parameter set includes information such as the group number, the current average internal resistance, the predicted drift rate, the compensation coefficient (such as the ratio of the compensation amount to the current internal resistance), and the adjustment period (such as adjustment every 50 cycles). For example, if the current average internal resistance of a group is 50mΩ, the predicted drift rate is 0.1mΩ / cycle, and the compensation coefficient is set to 5%, then the internal resistance upper limit of the group can be adjusted to 50×(1+5%)+0.1×adjustment period during the next adjustment.

[0101] In step S244, the drift compensation parameter set is used to dynamically adjust the group boundaries of the initial internal resistance group data. The adjustment strategy is divided into two cases: for groups with a predicted positive drift rate, the upper limit of its internal resistance interval is expanded, and the calculation formula is: new upper limit = original upper limit × (1 + compensation coefficient) + drift rate × adjustment period; for groups with a predicted negative drift rate, the lower limit of its internal resistance interval is lowered, and the calculation formula is: new lower limit = original lower limit × (1-compensation coefficient) + drift rate × adjustment period. After the adjustment, recheck whether the group boundaries overlap or blank intervals appear. If so, coordinate adjustments are made to the adjacent group boundaries (such as if the upper limit of a group is higher than the lower limit of the next group, the middle value of the two is taken as the new boundary). After the dynamic adjustment is completed, the internal resistance balance adjustment data is generated, which includes information such as the adjusted internal resistance intervals of each group, the monomer reallocation list, and the drift compensation record.

[0102] During the implementation of the entire internal resistance balancing mode, the initial internal resistance grouping setting combines the data distribution characteristics to avoid the unreasonable division that may be caused by equally spaced grouping; the internal resistance distribution heat map converts the abstract internal resistance data into a spatial distribution form through visualization, facilitating the rapid location of abnormal areas; the internal resistance drift vector diagram and trend line fitting analyze the evolution law of internal resistance from the perspective of dynamic changes, making the grouping strategy forward-looking.

[0103] The data processing of each link has a clear time sequence: the initial grouping data is used as the basis for the generation of the heat map, the heat map guides the extreme value extraction and trend analysis, the trend analysis results drive the calculation of the compensation amount, and the compensation amount ultimately acts on the group boundary adjustment. Among them, the reliability of the trend line fitting is tested by R 2 Value verification, when R 2 When <0.7, the fitting result is considered unreliable and it is necessary to reselect the fitting model (such as using polynomial fitting) or add more historical data.

[0104] Furthermore, the dynamic adjustment strategy incorporates a time-based prediction mechanism. By setting an adjustment period and a compensation coefficient, it balances the stability and adaptability of the grouping strategy. For example, for groups with stable drift rates, the adjustment period can be extended to reduce computational complexity; for groups with large drift rate fluctuations, the period can be shortened to respond to changes in real time. This mechanism avoids group oscillation caused by frequent adjustments while ensuring that group boundaries always adapt to the actual drift trend of the internal resistance.

[0105] Example 4:

[0106] This embodiment involves a detailed implementation of step S3, specifically including electrochemical impedance spectroscopy feature extraction, electrochemical state degradation prediction, dynamic compensation coefficient calculation, and strategy fusion construction. The following details the data acquisition method, model building logic, and strategy generation process.

[0107] In step S31, electrochemical impedance spectroscopy feature extraction is performed on the internal resistance balance adjustment data. The electrochemical impedance spectroscopy (EIS) data is collected by an electrochemical workstation. The specific process is: a small-amplitude sinusoidal AC signal with a frequency range of 10mHz to 10kHz (amplitude is usually 5-10mV) is applied to the battery cell, and the potential response and current response at different frequencies are measured to calculate the impedance modulus |Z| and phase angle During acquisition, the battery must be in a stable state (e.g., fully charged or at a specific state of charge (SOC)) to prevent dynamic changes during the charge and discharge process from interfering with the impedance measurement. After acquisition, the raw impedance data is preprocessed, including removing noise interference (using Fourier transform to filter out high-frequency noise) and filling in missing frequency points (using interpolation) to generate standardized impedance spectrum data.

[0108] The feature extraction step extracts key features reflecting the internal state of the battery from the impedance spectrum data. First, the impedance spectrum is converted into a Nyquist plot (real part Z' is the horizontal axis, imaginary part -Z" is the vertical axis) and a Bode plot (|Z| vs. Curve of change with frequency f). In the Nyquist plot, features such as solution resistance (RΩ, the intersection of the high-frequency region and the real axis), charge transfer resistance (Rct, semicircle diameter), and Warburg impedance (Zw, the slope of the oblique line in the low-frequency region) are extracted; in the Bode plot, parameters such as characteristic frequency (such as the frequency fmax corresponding to the phase angle peak) and low-frequency impedance slope (reflecting the lithium ion diffusion rate) are extracted. In addition, the parameter values ​​of each component are obtained by equivalent circuit fitting (such as Randles model and DFT model) as part of the impedance spectrum characteristic data. The impedance spectrum characteristic data finally generated contains dozens of dimensions, such as RΩ, Rct, Zw, fmax, low-frequency slope, etc. Each feature corresponds to a different physical and chemical process inside the battery (such as Rct reflects the electrochemical reaction kinetics, and Zw reflects the ion diffusion resistance).

[0109] In step S32, electrochemical state degradation prediction is performed based on the impedance spectrum characteristic data. First, step S321 is executed to perform frequency domain response feature decomposition on the impedance spectrum characteristic data. Principal component analysis (PCA) or independent component analysis (ICA) is used to reduce the dimensionality of the high-dimensional feature vector to extract the principal component that best characterizes the degradation trend. For example, the original feature dimension is reduced from 20 dimensions to 3 dimensions through PCA, and the principal components with a cumulative variance contribution rate of more than 90% are retained to generate a frequency domain feature vector set. The vector set contains the score value of each principal component, reflecting the distribution of the original feature in the low-dimensional space.

[0110] In step S322, degradation features are identified for the frequency domain feature vector set using a convolutional neural network (CNN). The CNN model architecture includes an input layer, a convolution layer, a pooling layer, and a fully connected layer: the input layer receives frequency domain feature vectors (such as 3D principal component vectors); the convolution layer uses a 1D convolution kernel to extract local correlation patterns between features, such as capturing feature combinations of different scales through multiple convolution kernels; the pooling layer uses maximum pooling or average pooling to reduce data dimensions and enhance model robustness; the fully connected layer outputs degradation feature labeling data through an activation function (such as ReLU, Sigmoid), and the labeling data is a continuous value between 0 and 1. The larger the value, the higher the degree of degradation. During the model training process, supervised learning is performed using historical impedance spectrum data and corresponding degradation degree labels (such as SOH values ​​based on the number of cycles) to optimize network parameters to minimize prediction errors.

[0111] In step S323, the electrochemical state degradation curve is predicted based on the degradation feature marker data. A long short-term memory (LSTM) network is used to construct a time series prediction model. The input is a sequence of degradation feature markers for multiple consecutive cycles, and the output is the degradation trend prediction value for multiple future cycles. The LSTM model captures long-term dependencies through a gating mechanism. For example, the forget gate determines the proportion of historical information discarded, the input gate controls the influx of new information, and the output gate generates the prediction result. During model training, the historical data is divided into a training set and a validation set. The network weights are optimized using the mean square error (MSE) loss function to generate degradation curve prediction data. This data contains the degradation feature prediction value and corresponding confidence interval for each future cycle.

[0112] In step S324, the predicted degradation curve data is verified for matching against a historical degradation database. The database stores battery impedance spectrum characteristics and degradation curve templates at different aging stages. The matching is calculated using a dynamic time warping (DTW) algorithm, which calculates the cumulative distance between the predicted curve and each template curve. The template with the smallest distance is selected as the matching result to generate internal resistance degradation prediction data. This data includes information such as the matching template's aging stage description (e.g., early degradation, mid-term degradation, and late degradation), the predicted internal resistance growth rate, and the estimated remaining cycle life.

[0113] In step S33, a dynamic compensation coefficient is calculated for the initial internal resistance group data based on the internal resistance degradation prediction data. The compensation coefficient calculation is based on the following rules: for groups predicted to be in the early degradation stage, due to the slow growth of internal resistance, the dynamic compensation coefficient is set to 0.01-0.03 (i.e., the compensation amount is 1%-3% of the current internal resistance); in the mid-stage degradation stage, the compensation coefficient is increased to 0.03-0.05; and in the late-stage degradation stage, the compensation coefficient is set to 0.05-0.10. Simultaneously, the coefficient is adjusted based on the differences in the impedance spectrum characteristics of the cells within the group (e.g., the standard deviation of Rct). The greater the difference, the larger the coefficient, reserving more compensation space. The dynamic compensation coefficient set includes information such as the group number, current average internal resistance, predicted degradation stage, compensation coefficient, and compensation cycle. For example, if a group has an average internal resistance of 60mΩ and is predicted to be in the mid-stage degradation stage, the compensation coefficient is 0.04, and the compensation cycle is 100 cycles, then the compensation amount per cycle is 60×0.04÷100=0.024mΩ.

[0114] In step S34, a strategy fusion is constructed by using a dynamic compensation coefficient set and internal resistance balance adjustment data. The fusion strategy adopts a weighted superposition method to combine the future internal resistance growth prediction value corresponding to the dynamic compensation coefficient with the current internal resistance balance adjustment data (such as group boundaries, monomer allocation). The specific method is: for each internal resistance group, according to the compensation cycle and the predicted rate, the total internal resistance growth in the future compensation cycle is calculated (such as 0.024mΩ / cycle × 100 cycles = 2.4mΩ), and the growth amount is proportionally allocated to the group boundary adjustment (such as the upper limit value increases by 2.4mΩ×0.8, and the lower limit value increases by 2.4mΩ×0.2), and the dynamic weight of the monomers in the group is adjusted at the same time (such as monomers with faster internal resistance growth are given a higher monitoring priority). By iteratively optimizing the fusion parameters (such as the allocation ratio and weight coefficient), the tiered utilization comprehensive strategy includes both the static adjustment of the current internal resistance balance and the dynamic compensation of the future degradation trend, forming a multi-dimensional optimization solution.

[0115] During the implementation of step S3, electrochemical impedance spectroscopy feature extraction provides a rich physical and chemical basis for degradation prediction. The combination of CNN and LSTM models enables end-to-end modeling from feature recognition to trend prediction. The setting of the dynamic compensation coefficient makes the strategy forward-looking, and the strategy fusion process ensures the organic unity of the current state and future trends.

[0116] Data processing at each stage adheres to strict timing and logical relationships: After feature extraction, impedance spectrum data is fed into a CNN model for degradation feature identification. The identification results are then used by an LSTM model to generate a prediction curve. This prediction curve is then matched with a historical database to output degradation prediction data. This data drives the calculation of dynamic compensation coefficients, which are ultimately fused with internal resistance balancing adjustment data to generate a comprehensive strategy. Model training and validation require regular updates of historical data to accommodate the differences in aging characteristics between battery batches.

[0117] Furthermore, safety thresholds can be set during the strategy fusion process, for example, limiting the adjustment of group boundaries to no more than 15% of the current internal resistance to prevent overcompensation that could lead to failure of the grouping strategy. Furthermore, sensitivity analysis assesses the impact of various feature parameters (such as Rct and fmax) on degradation prediction results, prioritizing significantly impactful features for modeling, thereby improving model efficiency and reliability.

[0118] Through the above implementation method, step S3 realizes the intelligent prediction and dynamic strategy generation of the internal resistance degradation of retired battery packs, combines the electrochemical mechanism analysis with the data-driven model, and provides a scientific time dimension optimization solution for the internal resistance balance control in cascade utilization. It helps to deal with the performance degradation caused by battery aging in advance and improve the long-term stability and safety of the reorganized battery pack.

[0119] Example 5:

[0120] This embodiment relates to the structure and functional implementation of a lithium-ion power battery recycling auxiliary analysis system. This system, used to execute the aforementioned method flow, comprises a pattern recognition module, a capacity reconfiguration module, an internal resistance balancing module, a strategy optimization module, and a feedback execution module. These modules collaborate through data interfaces. The following describes the module composition, data interaction, and specific functional implementation in detail.

[0121] The pattern recognition module is responsible for acquiring the operating parameter data of retired battery packs, analyzing the characteristics of individual cells, and confirming the cascade utilization mode. The data acquisition function is implemented through the integrated CAN bus communication protocol or Ethernet interface, and can read the real-time data such as voltage, current, temperature, etc. in the battery management system (BMS) in real time, or read historical data files (such as CSV, Excel format) through the file import interface. The data preprocessing link has built-in voltage calibration algorithm, temperature compensation model and data normalization tool. The temperature compensation model is based on the thermodynamic equation T norm =T measured -k·Δt, where T norm is the normalized temperature value, T measured is the measured temperature value, k is the temperature compensation coefficient (unit: °C / cycle), and Δt is the temperature change during the charge and discharge process. The preprocessed data is then used to generate single-cell characteristic data using a feature extraction algorithm. This feature extraction includes functional modules such as statistical calculation (such as mean and variance) and trend analysis (such as linear regression slope). The cascade utilization mode confirmation phase is based on dispersion analysis and weight calculation logic. Through a built-in threshold comparator and weight calculator, it outputs the mode selection result (capacity reconfiguration mode, internal resistance balancing mode, or mixed mode) and priority ranking.

[0122] The capacity reorganization module implements grouping settings and adjustments under the capacity reorganization mode. The initial capacity grouping function supports multiple algorithms such as equal-interval segmentation and adaptive segmentation. Users can set the grouping strategy through the parameter configuration interface. The capacity cluster division adopts the K-means clustering algorithm. The algorithm parameters (such as the number of clusters K) can be automatically adjusted or manually set according to the discreteness evaluation results. The historical capacity attenuation trajectory matching function is realized by retrieving the historical database. The database stores the capacity attenuation curves of different aging stages. The curve matching algorithm is based on the dynamic time warping (DTW) distance metric. The intra-cluster similarity calculation adopts the cosine similarity algorithm to output the intra-cluster consistency index; the cluster boundary adjustment and inter-cluster spacing optimization functions are automatically triggered according to the similarity results, and the grouping optimization is achieved by reallocating monomers or adjusting the grouping interval, and finally generating the capacity reorganization adjustment data.

[0123] The internal resistance balancing module handles grouping and drift analysis in the internal resistance balancing mode. Initial internal resistance grouping is based on a density peak clustering algorithm (such as DBSCAN), with adaptive grouping achieved by setting a density threshold and a minimum number of samples. The internal resistance distribution heat map generation function utilizes a data visualization engine to map the internal resistance data into a two-dimensional thermal image, supporting zooming, panning, and extreme value annotation. Internal resistance drift trend analysis includes sub-functions such as extreme value extraction, vector diagram generation, and trendline fitting. Trendline fitting utilizes the least squares method, and the fitting result is expressed as the linear equation R(t) = R0 + α·t, where R(t) is the predicted internal resistance value at time t (in mΩ), R0 is the initial internal resistance value (in mΩ), α is the internal resistance growth slope (in mΩ / cycle), and t is the number of cycles. Drift compensation parameter calculation is based on the trendline slope and grouping characteristics, generating a parameter set containing compensation coefficients and adjustment periods to drive dynamic adjustment of group boundaries.

[0124] The strategy optimization module completes the electrochemical state degradation prediction and strategy fusion. The electrochemical impedance spectrum feature extraction realizes automatic data collection through the integrated electrochemical workstation control interface. The feature extraction algorithm includes Nyquist plot analysis, equivalent circuit fitting and other functions, and outputs characteristic parameters such as RΩ, Rct, and Zw. The degradation prediction link adopts a convolutional neural network (CNN) and a long short-term memory network (LSTM) cascade model. CNN is used for feature dimensionality reduction and degradation feature identification, and LSTM is used for time series prediction. The model training data comes from historical impedance spectra and corresponding SOH labels. The dynamic compensation coefficient calculation is based on the predicted degradation stage and grouping characteristics, and the compensation coefficient is generated using piecewise function rules. The strategy fusion function integrates the capacity reorganization adjustment data and the internal resistance balancing strategy data through a weighted sum algorithm to generate a comprehensive strategy for cascade utilization that includes information such as grouping configuration and compensation parameters.

[0125] The feedback execution module implements strategy execution feedback and iterative optimization. The data acquisition function obtains the operating data of the battery pack during the cascade utilization process (such as the voltage and internal resistance fluctuation after reorganization) in real time through the hardware interface to generate a strategy execution feedback data set. The performance matching evaluation is based on a multi-index comprehensive evaluation system. The evaluation indicators include capacity consistency coefficient, internal resistance dispersion, charge and discharge efficiency, etc. The evaluation results are expressed as a matching score of 0-100. The iterative optimization function triggers the strategy adjustment process by comparing the evaluation results with the preset threshold. The adjustment methods include fine-tuning the capacity grouping boundaries, correcting the internal resistance compensation coefficient, etc., and finally generating a cascade utilization optimization strategy.

[0126] Each module interacts with each other via standardized data interfaces: the cascade utilization pattern data output by the pattern recognition module serves as input to the capacity reorganization module and the internal resistance balancing module; the adjustment data generated by the capacity reorganization module and the internal resistance balancing module is transmitted to the strategy optimization module for integration; the comprehensive strategy output by the strategy optimization module is sent to the feedback execution module for execution; and the feedback data collected by the feedback execution module is fed back to all modules, forming a closed-loop optimization chain. The system hardware architecture utilizes a distributed computing platform that supports multi-threaded data processing and parallel model training. The software layer is designed based on a microservices architecture, which is highly scalable and stable.

Claims

1. A lithium-ion power battery cascade utilization auxiliary analysis method, characterized in that: The following steps are involved: Step S1: Obtaining operating parameter data of retired battery packs; Perform battery cell feature analysis on retired battery pack operating parameter data to generate battery cell feature data; Confirming the battery pack's cascade utilization mode based on the battery cell characteristic data to obtain cascade utilization mode data, where the cascade utilization mode includes a capacity reorganization mode and an internal resistance balancing mode; Step S2: performing initial capacity grouping on the battery cell characteristic data based on the capacity reorganization mode to obtain initial capacity grouping data; performing capacity decay trajectory matching on the initial capacity grouping data to generate capacity reorganization adjustment data; performing initial internal resistance grouping on the battery cell characteristic data based on the internal resistance balancing mode to obtain initial internal resistance grouping data; Perform internal resistance drift trend analysis on the initial internal resistance group data to generate internal resistance balance adjustment data; Step S3: Predicting electrochemical state degradation of the internal resistance balancing adjustment data to generate internal resistance degradation prediction data; performing dynamic internal resistance balancing control adjustment on the initial internal resistance grouping data based on the internal resistance degradation prediction data to generate dynamic internal resistance balancing strategy data; and constructing a strategy fusion for the cascade utilization mode data using the dynamic internal resistance balancing strategy data and the capacity reorganization adjustment data to generate a comprehensive cascade utilization strategy. Step S4: collecting strategy execution feedback data for the cascade utilization comprehensive strategy to obtain a strategy execution feedback data set; Perform battery pack performance matching evaluation on the strategy execution feedback data set to generate performance matching evaluation data; The comprehensive strategy of cascade utilization is iteratively optimized through performance matching evaluation data to generate a cascade utilization optimization strategy.

2. The lithium-ion power battery recycling auxiliary analysis method according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Obtaining operating parameter data of retired battery packs; Step S12: performing parameter preprocessing on the retired battery pack operating parameter data to generate standard battery pack parameter data, wherein the parameter preprocessing includes voltage calibration, temperature compensation, data normalization, and outlier removal; Step S13: extracting single cell characteristics from the standard battery pack parameter data to obtain single cell characteristic data; calculating the battery pack dispersion based on the single cell characteristic data to generate battery pack dispersion distribution data; Step S14: confirming the cascade utilization mode of the battery pack according to the battery pack dispersion distribution data to obtain cascade utilization mode data, wherein the cascade utilization mode includes a capacity reorganization mode and an internal resistance balancing mode.

3. The auxiliary analysis method for lithium-ion power battery recycling according to claim 2, characterized in that: Step S14 includes the following steps: Step S141: performing capacity discrete threshold analysis on the battery pack discreteness distribution data to generate capacity discreteness evaluation data; performing internal resistance discrete threshold analysis on the battery pack discreteness distribution data to generate internal resistance discreteness evaluation data; Step S142: Calculating a mode selection weight based on the capacity dispersion evaluation data and the internal resistance dispersion evaluation data to generate mode selection weight data; wherein the parameters for calculating the mode selection weight include the capacity decay rate, the internal resistance growth slope, and the number of cycles; Step S143: comparing the mode selection weight data with a preset mode switching threshold. When the capacity decay rate weight is higher than the preset threshold, the capacity reconfiguration mode of the battery pack is confirmed. Step S144: When the internal resistance growth slope weight is higher than the preset threshold, the internal resistance balancing mode of the battery pack is confirmed; the capacity reorganization mode and the internal resistance balancing mode are prioritized to generate cascade utilization mode data.

4. The lithium-ion power battery recycling auxiliary analysis method according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: performing initial capacity grouping on the single cell characteristic data based on the capacity reorganization mode to obtain initial capacity grouping data; performing capacity clustering on the battery pack dispersion distribution data based on the initial capacity grouping data to generate capacity distribution cluster data; Step S22: using the capacity distribution cluster data to match the initial capacity grouping data with historical capacity attenuation trajectories to generate capacity reorganization adjustment data; Step S23: performing initial internal resistance grouping on the characteristic data of the single cells based on the internal resistance balancing mode to obtain initial internal resistance grouping data; generating an internal resistance distribution heat map for the battery pack dispersion distribution data based on the initial internal resistance grouping data; Step S24: performing internal resistance drift trend analysis on the internal resistance distribution heat map to generate internal resistance balance adjustment data.

5. The lithium-ion power battery recycling auxiliary analysis method according to claim 4, characterized in that: Step S22 includes the following steps: Step S221: Using the capacity distribution cluster data to perform capacity decay curve screening on the characteristic data of the single battery, to obtain a set of historical capacity decay curves; Step S222: extracting curve morphological features from the historical capacity decay curve set to generate capacity decay morphological feature data; and calculating intra-cluster similarity of the initial capacity grouping data based on the capacity decay morphological feature data; Step S223: When the intra-cluster similarity is lower than a preset threshold, the cluster boundary of the initial capacity grouping data is adjusted according to the capacity decay morphological feature data; Step S224: When the intra-cluster similarity is higher than a preset threshold, the initial capacity grouping data is optimized for inter-cluster spacing; the cluster boundary adjustment data and the inter-cluster spacing optimization data are integrated to generate capacity reorganization adjustment data.

6. The lithium-ion power battery cascade utilization auxiliary analysis method according to claim 5, characterized in that: Step S224 includes the following steps: When the intra-cluster similarity is higher than the preset threshold, the inter-cluster cross-interference analysis is performed on the capacity distribution cluster data to generate the inter-cluster interference coefficient; the inter-cluster distance of the initial capacity grouping data is reset according to the inter-cluster interference coefficient; Perform capacity gradient verification on the reset inter-cluster distance to generate capacity gradient verification data; perform dynamic weight correction on the inter-cluster interference coefficient based on the capacity gradient verification data to generate a corrected interference coefficient; The inter-cluster spacing is optimized by using the corrected interference coefficient and the capacity gradient verification data to generate optimized capacity grouping data; the optimized capacity grouping data is fused with the cluster boundary adjustment data to generate capacity reorganization adjustment data.

7. The lithium-ion power battery recycling auxiliary analysis method according to claim 4, characterized in that: Step S24 includes the following steps: Step S241: extracting regional extreme resistance values ​​from the internal resistance distribution heat map to generate regional extreme value distribution data; Step S242: Identify the internal resistance drift direction based on the regional extreme value distribution data to generate an internal resistance drift vector diagram; perform trend line fitting on the internal resistance drift vector diagram to generate internal resistance drift trend line data; Step S243: Calculating the drift compensation amount for the initial internal resistance group data according to the internal resistance drift trend line data to generate a drift compensation parameter set; Step S244: dynamically adjusting the group boundaries of the initial internal resistance group data using the drift compensation parameter set to generate internal resistance balance adjustment data.

8. The lithium-ion power battery recycling auxiliary analysis method according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: extracting electrochemical impedance spectrum characteristics from the internal resistance balance adjustment data to generate impedance spectrum characteristic data; Step S32: performing electrochemical state degradation prediction based on the impedance spectrum characteristic data to generate internal resistance degradation prediction data; Step S33: Calculating dynamic compensation coefficients for the initial internal resistance group data according to the internal resistance degradation prediction data to generate a dynamic compensation coefficient set; Step S34: constructing a strategy by integrating the dynamic compensation coefficient set with the internal resistance balance adjustment data to generate a comprehensive strategy for cascade utilization.

9. The lithium-ion power battery recycling auxiliary analysis method according to claim 8, characterized in that: Step S32 includes the following steps: Step S321: performing frequency domain response feature decomposition on the impedance spectrum characteristic data to generate a frequency domain feature vector set; Step S322: performing degradation feature recognition on the frequency domain feature vector set through a convolutional neural network to generate degradation feature labeling data; Step S323: Predicting an electrochemical state degradation curve based on the degradation feature marker data to generate degradation curve prediction data; Step S324: performing matching verification based on the degradation curve prediction data and the historical degradation database to generate internal resistance degradation prediction data.

10. A lithium-ion power battery cascade utilization auxiliary analysis system, characterized in that: For executing the auxiliary analysis method for lithium-ion power battery recycling as claimed in claim 1, the system comprises: A pattern recognition module is used to obtain operating parameter data of retired battery packs; perform battery cell feature analysis on the operating parameter data of retired battery packs to generate battery cell feature data; and perform cascade utilization mode confirmation on the battery pack based on the battery cell feature data to obtain cascade utilization mode data; A capacity reorganization module is used to perform initial capacity grouping on the battery cell characteristic data based on the capacity reorganization mode to obtain initial capacity grouping data; perform capacity attenuation trajectory matching on the initial capacity grouping data to generate capacity reorganization adjustment data; The internal resistance balancing module is used to perform initial internal resistance grouping on the battery cell characteristic data based on the internal resistance balancing mode to obtain initial internal resistance grouping data; perform internal resistance drift trend analysis on the initial internal resistance grouping data to generate internal resistance balancing adjustment data; The strategy optimization module is used to predict electrochemical state degradation based on internal resistance balancing adjustment data and generate internal resistance degradation prediction data; dynamically adjust the initial internal resistance grouping data based on the internal resistance degradation prediction data to generate dynamic internal resistance balancing strategy data; and integrate the dynamic internal resistance balancing strategy data with the capacity reorganization adjustment data to construct a strategy for cascade utilization mode data to generate a comprehensive cascade utilization strategy; The feedback execution module is used to collect strategy execution feedback data for the cascade utilization comprehensive strategy to obtain a strategy execution feedback data set; perform battery pack performance matching evaluation on the strategy execution feedback data set to generate performance matching evaluation data; and iteratively optimize the cascade utilization comprehensive strategy using the performance matching evaluation data to generate a cascade utilization optimization strategy.

Images