Storage battery capacity attenuation trend prediction method

Abstract

The invention discloses a storage battery capacity attenuation trend prediction method, and belongs to the technical field of storage battery prediction. By collecting voltage, current and temperature data of each monomer in real time and combining historical capacity attenuation and internal resistance growth data, the method identifies a voltage and capacity difference value, evaluates cyclic stress non-uniform distribution, and determines a current sharing proportion and a load unbalance degree. Identifying an abnormal mode of new battery overload and aged battery deep discharge, constructing a mixing abnormal working condition identification mode, if the unbalance degree exceeds the standard, adaptively adjusting the charging and discharging time and the current switching frequency, establishing a load balance control framework, predicting the capacity attenuation rate and the residual cycle index of each monomer, and determining the capacity matching degree and the life matching degree; finally, a comprehensive residual life estimation value and a credibility interval are generated through fusion; the performance balance of the mixed battery pack is remarkably improved, the overall service life is prolonged, and the method is suitable for real-time monitoring of a battery management system.

Description

TECHNICAL FIELD

[0001] The application relates to a battery capacity attenuation trend prediction method and belongs to the technical field of battery prediction. BACKGROUND

[0002] As an important energy storage device in the new energy field, a battery pack plays a key role in energy management in scenarios such as electric vehicles and energy storage power stations. Its reliability and service life directly affect the operating efficiency and safety of the system.

[0003] With the rapid growth of global demand for clean energy, the performance optimization of battery packs has become a research hotspot, especially how to deal with performance degradation under complex working conditions in actual operation. The efficient operation of the battery pack not only concerns the stability of the device itself, but also has important significance for reducing operating costs and improving energy utilization.

[0004] However, in actual use, the battery pack often faces the complex situation of mixing new and old batteries, which poses a severe challenge to traditional management methods.

[0005] Currently, the management method for the battery pack mainly relies on parameter matching at the initial grouping, such as capacity, internal resistance and other static indicators.

[0006] The existing scheme often assumes that the performance of each single battery in the battery pack is consistent, ignoring the performance differences caused by different aging degrees during operation. The differences between the single batteries in the battery pack will be continuously amplified due to the charge and discharge cycles in long-term operation. This amplification effect makes it impossible for the traditional method to accurately capture the real state of the battery pack under complex working conditions, thereby affecting the reliability of the life prediction. In the scene of mixing new and old batteries, the aging states of the single batteries in the battery pack are different, resulting in significant differences in voltage response, current distribution and temperature change during charging and discharging.

[0007] These differences will cause uneven distribution of cycle stress in the battery pack, accelerating the capacity attenuation of some single batteries. For example, in a mixed battery pack, new batteries may bear more current load due to their low internal resistance, while aged batteries may enter deep discharge state prematurely due to capacity decline. SUMMARY

[0008] According to the problems described in the background, the application aims to provide a battery capacity attenuation trend prediction method to solve the problems existing in the prior art.

[0009] To achieve the above-mentioned purpose, the application provides the following technical scheme: a battery capacity attenuation trend prediction method, comprising the following steps:

[0010] (1) Obtain the voltage response data and current distribution data of each cell in the new and old mixed battery pack during the charging and discharging process, and determine the voltage difference and capacity difference between each cell.

[0011] (2) Based on the voltage difference and capacity difference, calculate the inconsistency index of the aging degree of the new and old mixed cells, determine the actual current sharing ratio of each cell, and generate the load distribution imbalance.

[0012] (3) Based on the load distribution imbalance, determine the new battery overload mode and the aged battery deep discharge mode to form a mixed-mix abnormal working condition identification mode.

[0013] (4) Based on the mixed-battery abnormal operating condition identification mode, adjust the current distribution ratio of new battery and aged battery, generate charging and discharging time interval and current switching frequency, and construct a load balancing control framework.

[0014] (5) Based on the load balancing control framework, determine the capacity decay path and internal resistance growth path of the new and old mixed battery pack under multiple cycle cycles, and generate capacity matching degree and lifetime matching degree.

[0015] (6) Based on the capacity matching degree and life matching degree, generate the final remaining life estimate and confidence range, and output them to the battery pack management system.

[0016] Preferably, step (1) includes the following steps:

[0017] (1.1) Obtain the voltage change curves of each cell terminal of the mixed old and new battery pack under constant current charging and discharging conditions, record the charging start voltage, charging cut-off voltage, discharging start voltage and discharging cut-off voltage, collect the real-time current value of each cell branch, and calculate the voltage change rate and current distribution ratio.

[0018] (1.2) Extract the capacity retention rate sequence and internal resistance measurement value sequence of each cell in recent charge-discharge cycles from the battery management system, calculate the capacity decay rate and internal resistance growth rate, and generate an aging state feature vector group.

[0019] (1.3) Based on the aging state feature vector group, calculate the Euclidean distance between the feature vectors of adjacent cells, generate the difference metric between cells, and determine the voltage difference distribution map and the capacity difference distribution map.

[0020] Preferably, step (2) includes the following steps:

[0021] (2.1) Calculate the mean and standard deviation of the voltage difference and capacity difference between all units, generate the voltage difference dispersion and capacity difference dispersion, calculate the weighted sum of the voltage difference dispersion and capacity difference dispersion, and generate the aging inconsistency comprehensive index.

[0022] (2.2) Based on the aging inconsistency comprehensive index and the unit location number, generate the cyclic stress distribution sequence, calculate the deviation rate between the measured current value of each unit and the average distribution value of the total current, and determine the actual current sharing ratio.

[0023] (2.3) Based on the sum of squares of the differences between the actual current sharing ratio and the ideal uniform sharing ratio, generate the load distribution variance and calculate the load distribution imbalance.

[0024] Preferably, step (3) includes the following steps:

[0025] (3.1) Determine the temperature sensor acquisition interval based on the load distribution imbalance, obtain the relative temperature rise value between the surface temperature of each unit and the ambient temperature, and determine the unit in thermal anomaly state.

[0026] (3.2) Based on the temperature anomaly index and current sharing ratio of the cells in the thermal anomaly state, determine the overload characteristics of the new battery and the deep discharge characteristics of the aged battery, and generate an overload anomaly mode set and a deep discharge anomaly mode set through a clustering algorithm.

[0027] (3.3) Calculate the corrected temperature distribution sequence and voltage fluctuation amplitude based on the overload anomaly mode set and deep discharge anomaly mode set;

[0028] (3.4) Based on the corrected temperature distribution sequence and voltage fluctuation amplitude, generate temperature distribution non-uniformity and voltage fluctuation frequency, and construct a mixed-mixing abnormal working condition identification mode.

[0029] Preferably, step (4) includes the following steps:

[0030] (4.1) Based on the mixed abnormal operating condition identification mode, extract the overloaded cell and deep discharge cell numbers, calculate the current reduction coefficient of the new battery and the current increase coefficient of the aged battery, and generate the adjusted current distribution ratio sequence.

[0031] (4.2) Based on the adjusted current distribution ratio sequence, generate charging pause intervals and discharging rest intervals, and construct a charging and discharging time control sequence;

[0032] (4.3) Based on the charge and discharge time control sequence, calculate the basic switching frequency and the temperature correction switching frequency, and determine the actual current switching frequency;

[0033] (4.4) Based on the adjusted current distribution ratio sequence, the charge and discharge time control sequence and the actual current switching frequency, a control structure including a current distribution layer, a time control layer and a frequency adjustment layer is constructed as a load balancing control framework.

[0034] Preferably, step (5) includes the following steps:

[0035] (5.1) Based on the load balancing control framework, record the capacity retention rate and internal resistance measurement values ​​of each cell, generate the capacity decay rate sequence and the internal resistance growth rate sequence, and determine the differentiated decay path of new and old batteries.

[0036] (5.2) Based on the differentiated degradation paths of the new and old batteries, generate capacity degradation curves for the new and aged batteries, determine the capacity convergence cycle number and predict the remaining cycle number;

[0037] (5.3) Calculate the capacity synchronization coefficient and lifetime retention coefficient based on the capacity convergence cycle number and the predicted remaining cycle number, and generate a comprehensive performance index;

[0038] (5.4) Determine the capacity matching degree and lifetime matching degree based on the comprehensive performance index and the predicted remaining number of cycles.

[0039] Preferably, step (5) further includes the following steps:

[0040] (5.5) Based on the load balancing control framework, generate a capacity time series and construct a capacity decay curve;

[0041] (5.6) Based on the capacity decay curve, calculate the number of cycles required for the new battery to decay to the current capacity of the aged battery and the number of cycles required for the aged battery to reach the failure threshold;

[0042] (5.7) Based on the number of cycles, generate the capacity decay rate ratio relationship and determine the capacity convergence time node and failure time node;

[0043] (5.8) Generate the shortest cycle lifetime prediction value based on the capacity convergence time node and the failure time node.

[0044] Preferably, step (6) includes the following steps:

[0045] (6.1) Normalize the capacity decay and internal resistance growth of each cell to generate a normalized capacity decay sequence and a normalized internal resistance growth sequence, and calculate the comprehensive aging index sequence.

[0046] (6.2) Based on the capacity matching degree and lifetime matching degree, allocate capacity decay weight and internal resistance growth weight to generate a comprehensive score for mixing inconsistency;

[0047] (6.3) Based on the comprehensive score of the mixing inconsistency and the predicted shortest cycle lifetime, generate the remaining lifetime probability distribution and extract the final remaining lifetime estimate and confidence interval.

[0048] (6.4) Based on the final remaining life estimate and confidence interval, generate a monitoring data set and output it to the battery pack management system.

[0049] The beneficial effects of this invention are:

[0050] 1. Traditional methods, based on static initial parameter matching, cannot address the dynamic performance differentiation caused by aging differences during operation. This invention, by collecting voltage response and current distribution data in real time during charging and discharging, dynamically calculates the aging inconsistency index and load distribution imbalance, accurately capturing the performance differences between new and old batteries under actual operating conditions. This fundamentally solves the core problems of uneven current distribution and unbalanced cyclic stress distribution caused by mixing batteries.

[0051] 2. Existing technologies can only perform condition monitoring and lack effective control methods. This invention innovatively constructs a mixed-battery abnormal operating condition identification mode and dynamically adjusts the current distribution ratio between new and aged batteries based on this mode to generate an optimized charging and discharging strategy. This proactive load balancing control framework can effectively suppress the overload mode of new batteries and the deep discharge mode of aged batteries, thereby slowing down the overall battery pack degradation rate and extending the service life of the mixed-battery pack.

[0052] 3. Traditional methods suffer from significant prediction bias due to neglecting the amplification effect of inconsistent aging among individual cells. This invention simulates and predicts the differentiated degradation path of the battery pack across multiple cycles through load balancing control, and introduces multi-dimensional indicators such as capacity matching and lifetime matching. By comprehensively evaluating these indicators, the final estimated remaining lifetime and its reliability range more closely match the actual degradation pattern of the hybrid battery pack, significantly improving prediction accuracy and reliability.

[0053] 4. The final output of this invention is a specific remaining lifespan estimate, a confidence range, and key parameters such as the charge / discharge time interval and current switching frequency generated to achieve balanced control. This structured and quantified data can be directly embedded into the control logic of the battery pack management system, enabling it to perform intelligent charge / discharge management, early warning, and maintenance scheduling, greatly improving the system's intelligence level and operational economy. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0055] The embodiments of the present invention will be further described below with reference to the accompanying drawings:

[0056] Example 1

[0057] like Figure 1 As shown, the present invention provides a method for predicting the capacity degradation trend of a battery, comprising the following steps:

[0058] (1) Obtain the voltage response data and current distribution data of each cell in the new and old mixed battery pack during the charging and discharging process, and determine the voltage difference and capacity difference between each cell.

[0059] Step (1) includes the following steps:

[0060] (1.1) Obtain the voltage change curves of each cell terminal of the mixed old and new battery pack under constant current charging and discharging conditions, record the charging start voltage, charging cut-off voltage, discharging start voltage and discharging cut-off voltage, collect the real-time current value of each cell branch, and calculate the voltage change rate and current distribution ratio.

[0061] Specifically, the voltage change curves of each individual cell in the mixed old and new battery pack under constant current charging and discharging conditions are obtained. The voltage at the start of charging, the end of charging, the start of discharging, and the end of discharging are recorded. The real-time current value of each cell branch is collected by a Hall current sensor. The rate of change of voltage of each cell is obtained by dividing the voltage difference between adjacent times by the time interval. The current distribution ratio of each cell is obtained by the ratio of the current of the cell branch to the total current.

[0062] Among them, by installing a high-precision voltage acquisition module at both ends of each individual cell in the mixed old and new battery pack, the terminal voltage data during the charging and discharging process can be obtained in real time.

[0063] When the battery pack enters constant current charging mode, the terminal voltage value of each cell is recorded every 10 seconds, forming a voltage time series. The voltage change rate is obtained by dividing the voltage difference between two adjacent sampling times by the sampling time interval, reflecting the voltage response characteristics of a single battery cell during a specific charge and discharge phase. Hall current sensors are installed in series in the branches of each cell to accurately measure the actual current flowing through each cell. In mixed battery packs, due to the difference in internal resistance between new and old batteries, even in parallel connection, the current borne by each cell is not the same. By calculating the ratio of the current in the branch of a single cell to the total current of the battery pack, the current distribution ratio is obtained. This ratio directly reflects the phenomenon that the newer battery, due to its lower internal resistance, bears a larger current load.

[0064] (1.2) Extract the capacity retention rate sequence and internal resistance measurement value sequence of each cell in recent charge-discharge cycles from the battery management system, calculate the capacity decay rate and internal resistance growth rate, and generate an aging state feature vector group.

[0065] Specifically, the capacity retention rate sequence and internal resistance measurement value sequence of each cell in recent charge-discharge cycles are extracted from the battery management system database. The capacity retention rate sequence is differentially processed to obtain the capacity decay rate, and the internal resistance measurement value sequence is linearly fitted to obtain the internal resistance growth rate. Based on the voltage change rate, current distribution ratio, capacity decay rate and internal resistance growth rate, they are arranged according to the cell number to form an aging state feature vector group.

[0066] The capacity retention rate sequence extracted from the battery management system database contains data points from the most recent 30 charge-discharge cycles.

[0067] The capacity retention rate sequence is subjected to first-order differencing, i.e., the change in capacity retention rate between adjacent cycles is calculated to obtain the capacity decay rate. The internal resistance measurement value sequence is linearly fitted using the least squares method, and the slope of the fitted line is the internal resistance growth rate. The voltage change rate, current distribution ratio, capacity decay rate, and internal resistance growth rate are arranged in order of individual cell number to form a four-dimensional feature vector. The set of feature vectors of all individual cells constitutes the aging state feature vector group.

[0068] (1.3) Based on the aging state feature vector group, calculate the Euclidean distance between the feature vectors of adjacent cells, generate the difference metric between cells, and determine the voltage difference distribution map and the capacity difference distribution map.

[0069] Specifically, the difference between individual cells is calculated by the Euclidean distance between the feature vectors of adjacent cells in the aging state feature vector group. When the difference exceeds the first preset threshold, it is determined that there is a significant aging difference between adjacent cells. All cells are compared pairwise to obtain voltage difference distribution map and capacity difference distribution map, and the maximum voltage difference, the maximum capacity difference and the corresponding cell number are determined.

[0070] Euclidean distance is used to quantify the degree of aging difference between adjacent monomers.

[0071] For adjacent monomers numbered i and i+1, their feature vectors each contain numerical values ​​in four dimensions. The feature vector V of monomer i is... i =(U i ,I i C i ,R i ), the eigenvector V of single entity i+1 i+1 =(U i+1 ,I i+1 C i+1 ,R i+1 This is achieved by calculating the Euclidean distance between two vectors in four-dimensional space. This yields a comprehensive metric that reflects the differences in voltage, current, capacity, and internal resistance. When this metric exceeds a preset threshold, it indicates a significant difference in the aging levels of the two cells, which can lead to stress concentration during the charging and discharging process.

[0072] The voltage difference distribution map is generated by comparing each cell pair. The horizontal axis represents the cell number pair, and the vertical axis represents the corresponding voltage difference. The capacity difference distribution map is constructed in a similar way. These two distribution maps can visually show which cell pairs in the hybrid battery pack have significant performance differences.

[0073] (2) Based on the voltage difference and capacity difference, calculate the inconsistency index of the aging degree of the new and old mixed cells, determine the actual current sharing ratio of each cell, and generate the load distribution imbalance.

[0074] Step (2) includes the following steps:

[0075] (2.1) Calculate the mean and standard deviation of the voltage difference and capacity difference between all units, generate the voltage difference dispersion and capacity difference dispersion, calculate the weighted sum of the voltage difference dispersion and capacity difference dispersion, and generate the aging inconsistency comprehensive index.

[0076] Specifically, based on the identified voltage and capacity differences, the mean and standard deviation of the voltage differences between all individual units are calculated, as are the mean and standard deviation of the capacity differences between all individual units. The voltage difference dispersion is obtained by comparing the standard deviation of the voltage difference with the rated voltage, and the capacity difference dispersion is obtained by comparing the standard deviation of the capacity difference with the rated capacity. The aging inconsistency comprehensive index is obtained by weighted summing the voltage difference dispersion and the capacity difference dispersion.

[0077] The calculation process of the aging inconsistency index fully considers the differences in voltage and capacity.

[0078] When a mixed battery pack contains cells with different aging levels, the voltage differences between these cells exhibit a specific distribution pattern. The dispersion of voltage differences can be quantified by calculating the standard deviation of the voltage differences between all pairs of cells. This voltage difference dispersion is obtained by ratioing the standard deviation to the rated voltage of the battery pack; this ratio eliminates the influence of battery packs with different voltage levels. A similar method is used to calculate the capacity difference dispersion, but considering that capacity degradation has a more direct impact on battery pack performance, the weighting coefficient for capacity difference dispersion is typically set to 0.6, while the weighting coefficient for voltage difference dispersion is 0.4, when calculating the comprehensive aging inconsistency index. This weighting reflects the dominant role of capacity degradation in the battery aging process.

[0079] (2.2) Based on the aging inconsistency comprehensive index and the unit location number, generate the cyclic stress distribution sequence, calculate the deviation rate between the measured current value of each unit and the average distribution value of the total current, and determine the actual current sharing ratio.

[0080] Specifically, the correspondence between the aging inconsistency comprehensive index and the location number of each cell is used to construct a numerical sequence reflecting the distribution of cyclic stress. Each value in the sequence represents the relative stress level borne by the corresponding cell. The deviation rate is calculated by the difference between the measured current value of each cell and the average current distribution value of the battery pack. The actual current sharing ratio of each cell is determined by the product of the numerical sequence and the deviation rate.

[0081] The numerical sequence of cyclic stress distribution is constructed based on the differences in electrochemical stress experienced by each monomer during charge-discharge cycles.

[0082] The higher the aging inconsistency index, the more significant the aging differences within the battery pack, and the greater the cyclic stress experienced by the corresponding cells. In practical applications, a numerical sequence reflecting stress distribution is formed by establishing a mapping relationship between the aging inconsistency index and cell location numbers. Each value in the sequence represents the stress multiple of the corresponding cell relative to the average level. New batteries have higher stress values ​​because they need to handle more current, while aged batteries experience different types of stress due to deep discharge caused by capacity decay. The determination of the current sharing ratio involves comparing the measured current with the theoretical current sharing value. Ideally, parallel-connected battery cells should evenly share the total current, i.e., the theoretical current value of each cell equals the total current divided by the number of cells. However, due to differences in internal resistance, the actual current distribution deviates from the theoretical value. By calculating the deviation rate between the measured current and the theoretical current sharing value and multiplying it by the cyclic stress numerical sequence, a more accurate current sharing ratio is obtained, which comprehensively reflects the impact of aging differences and internal resistance differences on current distribution.

[0083] (2.3) Based on the sum of squares of the differences between the actual current sharing ratio and the ideal uniform sharing ratio, generate the load distribution variance and calculate the load distribution imbalance.

[0084] Specifically, the load distribution variance is calculated by the sum of the squares of the differences between the actual current sharing ratio of each cell and the preset ideal uniform sharing ratio. The load distribution imbalance is determined by the ratio of the square root of the load distribution variance to the rated current. The load distribution imbalance quantifies the degree of unevenness in the current load distribution in the mixed old and new battery packs.

[0085] The load distribution imbalance is determined by calculating the sum of squares of the differences between the actual current sharing ratio of each unit and the ideal uniform sharing ratio.

[0086] The square root of the load distribution variance reflects the standard deviation of the current distribution. The load distribution imbalance obtained by normalizing it with the rated current can intuitively characterize the uniformity of load distribution in the hybrid battery pack.

[0087] (3) Based on the load distribution imbalance, determine the new battery overload mode and the aged battery deep discharge mode to form a mixed-mix abnormal working condition identification mode.

[0088] Step (3) includes the following steps:

[0089] (3.1) Determine the temperature sensor acquisition interval based on the load distribution imbalance, obtain the relative temperature rise value between the surface temperature of each unit and the ambient temperature, and determine the unit in thermal anomaly state.

[0090] Specifically, the temperature sensor acquisition interval is determined by multiplying the load distribution imbalance degree by the preset sampling period. During the charging and discharging process, the surface temperature of each cell and the ambient temperature are acquired according to the acquisition interval. The difference between the temperature of each cell and the ambient temperature is calculated to obtain the relative temperature rise value. The temperature anomaly index is obtained by multiplying the relative temperature rise value by the load distribution imbalance degree. When the temperature anomaly index exceeds the preset threshold, it is determined that the cell has a thermal anomaly state.

[0091] The dynamic adjustment mechanism for temperature acquisition intervals is based on real-time changes in load distribution imbalance.

[0092] When the load distribution imbalance in a mixed battery pack is high, it indicates a significant performance difference between the new and old batteries. At this time, the rate of temperature change will accelerate, so it is necessary to shorten the sampling interval to capture rapid temperature changes.

[0093] The actual temperature acquisition interval is obtained by multiplying the load distribution imbalance by the reference sampling period.

[0094] The baseline sampling period is typically set to 30 seconds. When the load distribution imbalance is 0.5, the actual sampling interval is 15 seconds, thus achieving accurate tracking of temperature changes. The calculation of relative temperature rise eliminates the influence of ambient temperature fluctuations on the evaluation results. During the actual operation of the battery pack, the ambient temperature varies with the season, day and night, and ventilation conditions. Directly using absolute temperature values ​​can lead to evaluation bias. By calculating the difference between the cell surface temperature and the ambient temperature, the relative temperature rise value can accurately reflect the heat generated by the electrochemical reaction inside the battery. The temperature anomaly index is obtained by multiplying the relative temperature rise value by the load distribution imbalance. This index comprehensively considers the absolute degree of temperature rise and the unevenness of load distribution. When the index exceeds a preset threshold, it indicates that the cell is in an abnormal heating state.

[0095] (3.2) Based on the temperature anomaly index and current sharing ratio of the cells in the thermal anomaly state, determine the overload characteristics of the new battery and the deep discharge characteristics of the aged battery, and generate an overload anomaly mode set and a deep discharge anomaly mode set through a clustering algorithm.

[0096] Specifically, based on the temperature anomaly index and current sharing ratio of the individual cells in the thermal anomaly state, the overload bearing characteristics of new batteries with high temperature rise and current sharing ratio exceeding a preset multiple of the average value are identified, and the deep discharge characteristics of aged batteries with voltage drop rate exceeding a preset rate and voltage value close to the discharge cutoff voltage are identified. The overload bearing characteristics and deep discharge characteristics are classified by K-means clustering algorithm to obtain the overload anomaly pattern set of new batteries and the deep discharge anomaly pattern set of aged batteries.

[0097] Among them, the identification of the overload bearing characteristics of the new battery is based on the dual criteria of temperature rise rate and current sharing ratio.

[0098] In hybrid battery packs, newer batteries, due to their lower internal resistance, automatically bear a larger current load in the parallel circuit. This overload phenomenon manifests as a rapid temperature rise and a significantly higher-than-average current sharing ratio. By setting a current sharing ratio exceeding the average by 1.2 times as the overload criterion, combined with a temperature criterion of the temperature rise exceeding the ambient temperature by 10 degrees Celsius, it is possible to accurately identify newer battery cells in an overloaded state. Conversely, the deep discharge characteristics of aged batteries are characterized by a rapid voltage drop approaching the discharge cutoff voltage. When the capacity of an aged battery has severely degraded, it will reach a low voltage state earlier during normal discharge. If discharge continues, it will enter the deep discharge region, accelerating irreversible damage to the battery.

[0099] Among them, the application of the K-means clustering algorithm in anomaly pattern classification makes full use of its ability to cluster multidimensional data.

[0100] The algorithm first takes a vector composed of four features for each individual cell: temperature anomaly index, current sharing ratio, voltage drop rate, and relative capacity. It sets the number of clusters to two, corresponding to overload anomalies and deep discharge anomalies, respectively. By iteratively calculating the Euclidean distance from each data point to the cluster center, the algorithm continuously updates the cluster center positions, ultimately dividing all anomalous cells into two sets. Cells in the overload anomaly set exhibit a combination of high temperature rise and high current, while cells in the deep discharge anomaly set exhibit a combination of low voltage and rapid decay. This clustering-based classification method can adaptively identify different types of anomaly patterns, avoiding the limitations of manually setting fixed thresholds.

[0101] (3.3) Calculate the corrected temperature distribution sequence and voltage fluctuation amplitude based on the overload anomaly mode set and deep discharge anomaly mode set;

[0102] Specifically, the temperature rise slope in the new battery overload anomaly mode set and the voltage drop slope in the aging battery deep discharge anomaly mode set are used to construct a linear correction relationship for temperature data and a piecewise correction relationship for voltage data, respectively. The original temperature data is compensated using the linear correction relationship to obtain the corrected temperature distribution sequence. The difference between the maximum and minimum voltage values ​​during the charging and discharging process is calculated using the piecewise correction relationship to obtain the corrected voltage fluctuation amplitude.

[0103] The linear correction relationship for temperature data is constructed based on the statistical characteristics of the temperature rise slope in the overload anomaly pattern set.

[0104] By linearly fitting the temperature-time curves of all overloaded cells in the dataset, the distribution range of the temperature rise slope is obtained. This slope is used as a correction coefficient to linearly compensate the original temperature data, making the corrected temperature value more accurately reflect the actual thermal state. The piecewise correction relationship for the voltage data considers the nonlinear variation characteristics of voltage in different ranges. In the high-voltage range, the voltage change is relatively gentle, and a smaller correction coefficient is used; in the low-voltage range close to the discharge cutoff voltage, the voltage change is drastic, and a larger correction coefficient is used. This piecewise correction method can more accurately describe the voltage fluctuation characteristics throughout the entire discharge process.

[0105] (3.4) Based on the corrected temperature distribution sequence and voltage fluctuation amplitude, generate temperature distribution non-uniformity and voltage fluctuation frequency, and construct a mixed-mixing abnormal working condition identification mode.

[0106] Specifically, the temperature distribution non-uniformity is obtained by calculating the temperature standard deviation of each individual cell based on the corrected temperature distribution sequence, the voltage fluctuation frequency is obtained based on the frequency of change of the corrected voltage fluctuation amplitude, the number of individual cells in the overload abnormal mode set and the deep amplification abnormal mode set are counted, and a four-dimensional feature vector is constructed by temperature distribution non-uniformity, voltage fluctuation frequency, number of overload individual cells and number of deep amplification individual cells to form a mixed-mixing abnormal operating condition identification mode.

[0107] The temperature distribution non-uniformity is obtained by calculating the standard deviation of the corrected temperature distribution sequence. This index quantifies the degree of temperature difference within the battery pack.

[0108] Voltage fluctuation frequency is obtained by counting the number of times the voltage fluctuation amplitude exceeds the threshold per unit time, reflecting the voltage stability. The number of overloaded cells and the number of deep-discharged cells are obtained directly from two sets of abnormal patterns, representing the severity of different types of abnormalities.

[0109] For example, in a hybrid battery pack containing 20 cells, when the temperature distribution non-uniformity exceeds 5 degrees Celsius, the voltage fluctuation frequency exceeds 3 times per minute, the number of overloaded cells reaches 4, and the number of deeply discharged cells reaches 3, the resulting four-dimensional feature vector is [5, 3, 4, 3]. The hybrid abnormal operating condition identification mode formed by this feature vector can comprehensively characterize the abnormal operating state of the battery pack, not only identifying the existence of abnormalities but also distinguishing the type and degree of abnormalities.

[0110] By continuously monitoring and updating the four-dimensional feature vector, the operating status changes of the hybrid battery pack can be tracked in real time, and potential safety hazards can be detected in a timely manner. This identification mode can also trigger corresponding levels of protection measures according to different degrees of anomalies, ranging from minor load adjustments to emergency charge and discharge interruptions, to achieve graded protection for the hybrid battery pack.

[0111] (4) Based on the mixed-battery abnormal operating condition identification mode, adjust the current distribution ratio of new battery and aged battery, generate charging and discharging time interval and current switching frequency, and construct a load balancing control framework.

[0112] Step (4) includes the following steps:

[0113] (4.1) Based on the mixed abnormal operating condition identification mode, extract the overloaded cell and deep discharge cell numbers, calculate the current reduction coefficient of the new battery and the current increase coefficient of the aged battery, and generate the adjusted current distribution ratio sequence.

[0114] Specifically, if any dimension of the four-dimensional feature vector in the mixed-condition abnormal operation identification mode exceeds a preset threshold, the overload cell number and the deep discharge cell number in the feature vector are extracted. The current reduction coefficient of the new battery is calculated based on the difference between the current sharing ratio of the overload cell and the average value. The current increase coefficient of the aged battery is calculated based on the difference between the voltage value of the deep discharge cell and the discharge cutoff voltage. The adjusted current distribution ratio sequence is obtained through the current reduction coefficient of the new battery and the current increase coefficient of the aged battery.

[0115] Specifically, when the temperature distribution non-uniformity exceeds 0.8, the voltage fluctuation frequency is higher than 15 times per minute, the number of overloaded cells reaches 3, or the number of deep-discharged cells exceeds 2, the load redistribution process is immediately initiated. This threshold setting is based on a large amount of experimental data statistics and can effectively identify early signs of abnormal mixed-use conditions.

[0116] (4.2) Based on the adjusted current distribution ratio sequence, generate charging pause intervals and discharging rest intervals, and construct a charging and discharging time control sequence;

[0117] Specifically, using the adjusted current distribution ratio sequence, the charging pause interval is determined based on the product of the new battery current reduction coefficient and the preset reference time, and the discharge rest interval is determined based on the reciprocal of the aging battery voltage recovery rate. The charging and discharging time control sequence is obtained by alternating the charging pause interval and the discharge rest interval. The control sequence includes the charging duration, pause duration, discharging duration, and rest duration.

[0118] The charging pause interval setting reflects the refined control of thermal management for the new battery.

[0119] In an energy storage system containing a mix of 20 new and old batteries, overload characteristics were detected in new batteries numbered 3, 7, and 11, with current sharing ratios of 1.8, 1.6, and 1.9 times the average, respectively. Meanwhile, aging batteries numbered 15 and 18 showed deep discharge characteristics. Based on the differences, the current reduction coefficient for the new batteries was calculated to be 0.25, and the current increase coefficient for the aging batteries was calculated to be 0.15. This led to the generation of an adjusted current distribution ratio sequence, reducing the current sharing of the overloaded new batteries to 1.2 times the average level and increasing the current sharing of the deeply discharged aging batteries to 1.1 times the average level.

[0120] For example, when the current reduction factor of the new battery is 0.25 and the preset reference time is 120 seconds, the charging pause interval is 30 seconds. This time window allows the overloaded new battery to balance heat dissipation and internal chemical reactions. Correspondingly, the discharge pause interval is calculated based on the voltage recovery characteristics of the aged battery. If the voltage recovery rate of the aged battery is 0.02V / s, the pause interval is 50 seconds to ensure that the aged battery has sufficient time to recover its voltage.

[0121] (4.3) Based on the charge and discharge time control sequence, calculate the basic switching frequency and the temperature correction switching frequency, and determine the actual current switching frequency;

[0122] Specifically, based on the charge and discharge time control sequence, the number of charge and discharge state transitions per unit time is calculated to obtain the basic switching frequency. The temperature correction switching frequency is obtained by multiplying the temperature standard deviation of each cell by the basic switching frequency. If the temperature correction switching frequency exceeds the preset upper limit, the preset upper limit value is used as the actual current switching frequency. The actual current switching frequency controls the load transfer speed between the new and old batteries.

[0123] The design of the charge / discharge time control sequence follows the inherent laws of battery electrochemical characteristics.

[0124] In a typical control cycle, the charging duration is set to 180 seconds, the pause duration to 30 seconds, the discharging duration to 150 seconds, and the rest duration to 50 seconds. This timing arrangement ensures energy transfer efficiency while avoiding excessive stress concentration in individual cells. The current switching frequency control mechanism further optimizes the smoothness of the load transfer process.

[0125] (4.4) Based on the adjusted current distribution ratio sequence, the charge and discharge time control sequence and the actual current switching frequency, a control structure including a current distribution layer, a time control layer and a frequency adjustment layer is constructed as a load balancing control framework.

[0126] Specifically, a control structure including a current distribution layer, a time control layer, and a frequency adjustment layer is established through the adjusted current distribution ratio sequence, the charge and discharge time control sequence, and the actual current switching frequency. Proportional-integral control is used to dynamically update the parameter values ​​of each layer according to the real-time temperature and voltage deviation, forming an adaptive load balancing control framework.

[0127] Specifically, when the base switching frequency is calculated to be 8 times per minute and the standard deviation of the temperature of each unit is 2.3℃, the temperature-corrected switching frequency is 18.4 times per minute. Since this value exceeds the preset upper limit of 15 times per minute, the upper limit is used as the actual switching frequency to prevent instability of the control system and additional energy consumption caused by frequent switching. In the hierarchical design of the control structure, the current distribution layer is responsible for adjusting the current carrying ratio of each unit in real time, the time control layer manages the timing coordination of charging, discharging and rest cycles, and the frequency adjustment layer controls the speed and rhythm of load transfer.

[0128] For example, when an abnormal temperature rise of a new battery is detected, the current distribution layer immediately reduces its share, the time control layer extends its pause time, and the frequency adjustment layer appropriately reduces the switching frequency to slow down the rate of load change.

[0129] The introduction of proportional-integral control algorithm enables dynamic adaptive adjustment of parameters.

[0130] When the real-time temperature deviation reaches 3℃ and the voltage deviation is 0.15V, the proportional circuit responds quickly to the deviation change, and the integral circuit eliminates the steady-state error, ensuring that the parameter values ​​of each layer can be accurately adjusted to follow the changes in system state.

[0131] (5) Based on the load balancing control framework, determine the capacity decay path and internal resistance growth path of the new and old mixed battery pack under multiple cycle cycles, and generate capacity matching degree and lifetime matching degree.

[0132] Step (5) includes the following steps:

[0133] (5.1) Based on the load balancing control framework, record the capacity retention rate and internal resistance measurement values ​​of each cell, generate the capacity decay rate sequence and the internal resistance growth rate sequence, and determine the differentiated decay path of new and old batteries.

[0134] Specifically, based on the current distribution ratio sequence and charge / discharge time control sequence output by the adaptive load balancing control framework, the capacity retention rate and internal resistance measurement value of each cell are recorded in multiple consecutive charge / discharge cycles. The capacity decay rate sequence is obtained by the difference in capacity retention rate between adjacent cycles, and the internal resistance growth rate sequence is obtained by the difference in internal resistance value between adjacent cycles. The differentiated degradation path of new and old batteries is identified based on the changing patterns of the capacity decay rate sequence and the internal resistance growth rate sequence.

[0135] The adaptive load balancing control framework outputs control parameters including a current distribution ratio sequence and a charge / discharge time control sequence, which directly affect the working state of each unit in subsequent cycles.

[0136] By continuously monitoring data from more than 30 charge-discharge cycles, the dynamic response characteristics of new and old batteries under controlled conditions can be captured. Capacity retention is obtained as the ratio of discharged capacity to initial capacity, and internal resistance is measured using AC impedance spectroscopy or DC internal resistance testing. The capacity decay rate sequence reflects the rate of capacity degradation, while the internal resistance growth rate sequence characterizes the deterioration process of the battery's internal electrochemical properties.

[0137] The identification of differentiated degradation paths is based on the essential differences in aging mechanisms between new and old batteries.

[0138] The capacity decay of new batteries mainly stems from the growth of the solid electrolyte interfacial film and the initial loss of active materials. The decay rate gradually slows down with increasing cycle count, exhibiting an exponential decay characteristic. Aging batteries, having already undergone the initial rapid decay phase, primarily experience capacity loss due to the continuous shedding of active materials and corrosion of the current collector, exhibiting a relatively stable linear decay trend. Time-series analysis of the capacity decay rate sequence and the internal resistance growth rate sequence allows for accurate differentiation between these two distinct decay modes.

[0139] (5.2) Based on the differentiated degradation paths of the new and old batteries, generate capacity degradation curves for the new and aged batteries, determine the capacity convergence cycle number and predict the remaining cycle number;

[0140] Specifically, using the differentiated degradation path, the new battery capacity degradation rate sequence is processed by an exponential fitting method to obtain the new battery degradation curve, and the aged battery capacity degradation rate sequence is processed by a linear fitting method to obtain the aged battery degradation curve. The number of capacity convergence cycles is determined based on the intersection of the new battery degradation curve and the aged battery degradation curve. The number of cycles when the two curves reach 80% of the rated capacity is used to obtain the predicted number of remaining cycles.

[0141] Among them, the exponential fitting method uses the least squares method to process the capacity decay rate sequence of the new battery. The fitting function is in the form that the capacity retention rate is equal to the initial value multiplied by the negative decay coefficient with the base of the natural logarithm and the number of cycles raised to the power of the product.

[0142] The exponential function accurately describes the characteristic of a new battery gradually stabilizing after its initial rapid degradation. The linear fitting method, on the other hand, fits the capacity degradation rate sequence of the aged battery using a first-order polynomial; the slope of the resulting straight line represents the stable degradation rate of the aged battery. The intersection of the two fitted curves has significant physical meaning; it indicates the point at which the degradation rate of the new battery decreases to a level comparable to that of the aged battery, at which point the capacity difference between the two types of batteries reaches its minimum, defined as the capacity convergence point.

[0143] The process of determining the number of convergence loops involves numerically solving for the intersection of two fitted curves.

[0144] In practical calculations, by setting the number of cycles as the independent variable, the function values ​​of the exponential curve and the straight line at that point are calculated separately. When the absolute value of the difference between the two function values ​​is less than the preset precision, the corresponding number of cycles is the capacity convergence cycle number. The determination of the predicted remaining cycle number is based on the battery failure criterion: when the capacity retention rate drops to 80% of the rated capacity, the battery is considered to have reached the end of its lifespan. By substituting 80% into the fitted curve equation, the corresponding number of cycles is solved, and the predicted remaining cycle number is obtained by subtracting the number of cycles already completed. This prediction method comprehensively considers the different degradation characteristics of new and old batteries, and can provide a more accurate lifespan estimate.

[0145] (5.3) Calculate the capacity synchronization coefficient and lifetime retention coefficient based on the capacity convergence cycle number and the predicted remaining cycle number, and generate a comprehensive performance index;

[0146] Specifically, the capacity synchronization coefficient is calculated based on the ratio of the capacity convergence cycle number to the current cycle number, and the life retention coefficient is calculated based on the ratio of the predicted remaining cycle number to the rated cycle life. The comprehensive performance index is obtained by multiplying the capacity synchronization coefficient by 0.6 and adding the life retention coefficient by 0.4. The comprehensive performance index quantitatively characterizes the overall matching degree of the hybrid battery pack.

[0147] Among them, the capacity synchronization coefficient reflects the process by which the capacities of new and old batteries tend to be consistent.

[0148] When the capacity synchronization coefficient is close to 1, it indicates that the battery pack is nearing capacity convergence, with minimal capacity differences between individual cells. When the coefficient is much less than 1, it means there is still a long time before capacity convergence, requiring continued adjustment through load balancing control. The lifespan retention coefficient assesses the degree to which the hybrid battery pack retains its design lifespan; a higher coefficient indicates a smaller overall lifespan loss for the battery pack.

[0149] The overall performance index is weighted at a ratio of 0.6 to 0.4, which takes into account the dominant influence of capacity consistency on battery pack performance.

[0150] Capacity differences directly determine uneven current distribution during charging and discharging, which is the main cause of localized overcharging and over-discharging. While lifespan consistency is important, its impact is more evident in long-term use, and therefore it is given relatively low weight. The comprehensive performance index integrates the matching degree of the two dimensions into a single indicator through a weighted summation, facilitating unified evaluation and decision-making.

[0151] (5.4) Determine the capacity matching degree and lifetime matching degree based on the comprehensive performance index and the predicted remaining number of cycles.

[0152] Specifically, the capacity matching degree is determined by comparing the comprehensive performance index with the preset matching threshold. If the comprehensive performance index is higher than the upper threshold, the capacity matching degree is excellent; if it is lower than the lower threshold, the capacity matching degree is poor. The lifetime matching degree is determined by the ratio of the predicted remaining cycle number to the minimum remaining cycle number among all cells. The capacity matching degree and lifetime matching degree together characterize the performance balance state of the hybrid battery pack.

[0153] The capacity matching degree is classified based on the comparison between the comprehensive performance index and the preset threshold.

[0154] When the comprehensive performance index is higher than 0.8, the capacity matching is considered excellent. At this time, the performance difference between the new and old batteries is small, and mixing them will not significantly affect the performance of the battery pack. When the index is between 0.5 and 0.8, the capacity matching is moderate and requires appropriate balancing control. When the index is lower than 0.5, the capacity matching is poor, indicating that the difference between the new and old batteries is too large and they are not suitable for direct mixing.

[0155] The lifetime matching degree is determined by comparing the predicted remaining cycle number with the minimum value among all individual cells.

[0156] This evaluation method follows the "barrel principle," meaning the overall lifespan of a battery pack is limited by the shortest-lived individual cell. A lifespan matching degree close to 1 indicates that the remaining lifespan of each cell is similar, allowing the battery pack to fully utilize the capacity of each cell. A lower lifespan matching degree means that some cells may fail prematurely, affecting the overall lifespan of the pack. Capacity matching degree and lifespan matching degree evaluate the performance balance of a hybrid battery pack from different dimensions, providing a quantitative basis for battery management system decisions.

[0157] Step (5) further includes the following steps:

[0158] (5.5) Based on the load balancing control framework, generate a capacity time series and construct a capacity decay curve;

[0159] Specifically, based on the current distribution data recorded by the adaptive load balancing control framework, the maximum, minimum and average terminal voltage of each cell in each charge and discharge cycle are obtained, a voltage time series is constructed and the voltage decay curve is obtained by fitting, and the actual discharge capacity of each cycle is extracted from the battery management system, a capacity time series is constructed and the capacity decay curve is obtained by fitting.

[0160] In each charge-discharge cycle, the peak, valley, and average terminal voltage data of each cell are collected to form a complete voltage change trajectory.

[0161] For example, in 50 consecutive cycles of monitoring of a hybrid battery pack, the maximum voltage of the new battery cell gradually decreased from 4.15V to 4.08V, the minimum voltage decreased from 3.20V to 3.15V, and the average voltage decreased from 3.68V to 3.62V. These data points constitute a voltage time series reflecting the changes in the internal chemical activity of the battery.

[0162] Among them, the voltage decay curve fitting algorithm can identify the characteristic decay patterns of different types of batteries.

[0163] The voltage decay of new batteries typically exhibits an S-shaped curve, characterized by a slow initial decrease followed by accelerated decay, while aged batteries show a relatively linear decay trend. Using a polynomial fitting method, an accurate voltage decay prediction model can be established. By recording the actual discharge capacity data in each cycle and continuously monitoring, it was found that the capacity of new batteries slowly decayed from 100 Ah to 95 Ah, while the capacity of aged batteries rapidly decreased from 85 Ah to 80 Ah. This differentiated capacity decay pattern reflects the fundamental differences between new and old batteries in terms of chemical activity, internal structural integrity, and ion conductivity.

[0164] (5.6) Based on the capacity decay curve, calculate the number of cycles required for the new battery to decay to the current capacity of the aged battery and the number of cycles required for the aged battery to reach the failure threshold;

[0165] Specifically, based on the capacity decay curve, the current capacity value of the new battery and the current capacity value of the aged battery are extracted. The number of cycles required for the new battery to decay from full capacity to the current capacity level of the aged battery is obtained from the new battery capacity decay curve. The number of cycles required for the aged battery to decay from its current capacity to the failure threshold is estimated from the aged battery capacity decay curve. The failure threshold is set as a preset percentage of the rated capacity.

[0166] Among these, the mathematical modeling of the capacity decay curve lays a scientific foundation for predictive analysis. Through in-depth analysis of historical data, it is possible to accurately calculate the number of cycles required for a new battery to decay from its current capacity level to the same state as an aged battery.

[0167] For example, when the current capacity of a new battery is 95Ah and the capacity of an aged battery is 80Ah, according to the degradation curve equation of the new battery, it is estimated that the new battery needs to undergo approximately 200 additional cycles to reach the current capacity level of the aged battery. A conservative strategy based on a safety margin was adopted in setting the failure threshold.

[0168] The failure threshold is set at 75% of the rated capacity. This standard takes into account the performance requirements in actual applications and reserves sufficient safety buffer space.

[0169] By reverse-engineering the capacity decay curve of the aged battery, the system can accurately predict the critical information that it will take 150 cycles for the battery to decay from its current 80Ah capacity to the 60Ah failure threshold. The establishment of the capacity decay rate ratio provides a quantitative basis for the coordinated control of the blending system.

[0170] (5.7) Based on the number of cycles, generate the capacity decay rate ratio relationship and determine the capacity convergence time node and failure time node;

[0171] Specifically, the ratio of the number of cycles required for the new battery to degrade to the remaining number of cycles for the aged battery is calculated to obtain the capacity decay rate ratio. The capacity convergence time node is determined based on the number of cycles at the intersection of the capacity decay curves, and the failure time node is determined based on the number of cycles for the aged battery to reach the failure threshold. The capacity convergence time node serves as the capacity matching benchmark, and the failure time node serves as the lifetime matching benchmark.

[0172] Among them, by calculating the ratio of the 200 cycles required for the degradation of a new battery to the remaining 150 cycles for an aged battery, a rate relationship of 1.33 was obtained. This value intuitively reflects the time difference in the synchronization of the performance of the two types of batteries.

[0173] Based on this ratio, the load allocation strategy can be dynamically adjusted to achieve gradual performance matching between new and old batteries. During the determination of time points, the capacity convergence time point and the failure time point constitute the key control points for the lifecycle management of the hybrid system.

[0174] (5.8) Generate the shortest cycle lifetime prediction value based on the capacity convergence time node and the failure time node.

[0175] Specifically, by comparing the capacity convergence time node and the failure time node, the smaller value is selected, and the minimum value is selected as the shortest cycle life prediction value available for the hybrid battery pack based on the number of cycles required for all individual cell capacity decay curves to reach the failure threshold.

[0176] The intersection of the capacity decay curves occurs at the 180th cycle, marking the critical moment when the capacities of the new and old batteries tend to be consistent.

[0177] The determination of the failure time point is based on the predicted time when the aged battery reaches the 75% capacity threshold, which provides a clear time limit for the retirement planning of the entire hybrid system.

[0178] The selection of the shortest cycle life prediction value reflects the specific application of the "weakest link" effect in battery systems.

[0179] By comparing 180 cycles at the capacity convergence point and 150 cycles at the failure point, the system selects the smaller value of 150 cycles as the key control parameter. This conservative prediction strategy ensures that the mixed battery pack avoids systemic risks caused by excessive degradation of individual cells while maintaining overall performance coordination.

[0180] (6) Based on the capacity matching degree and life matching degree, generate the final remaining life estimate and confidence range, and output them to the battery pack management system.

[0181] Step (6) includes the following steps:

[0182] (6.1) Normalize the capacity decay and internal resistance growth of each cell to generate a normalized capacity decay sequence and a normalized internal resistance growth sequence, and calculate the comprehensive aging index sequence.

[0183] Specifically, the capacity degradation of all cells in the mixed old and new battery packs is sorted from largest to smallest, and the internal resistance growth is sorted from highest to lowest. The normalized values ​​are obtained by dividing each value by the maximum value to obtain the normalized capacity degradation sequence and the normalized internal resistance growth sequence. The two normalized sequences are then fused using a weighted average method to obtain the comprehensive aging index sequence of each cell.

[0184] The fusion of capacity decay and internal resistance growth is achieved using a multi-dimensional data standardization method.

[0185] For a mixed battery pack containing 20 cells, the capacity degradation of each cell is obtained by dividing the difference between the current capacity and the initial capacity by the initial capacity, resulting in a percentage value reflecting the degree of capacity loss. The increase in internal resistance is calculated by dividing the difference between the current internal resistance and the initial internal resistance by the initial internal resistance. The sorting process is carried out in descending order of values, so that the cells with the most severe aging are placed at the beginning of the sequence.

[0186] Furthermore, normalization is achieved by dividing by the maximum value, a method that can eliminate the influence of different physical dimensions.

[0187] The maximum capacity decay typically ranges from 20% to 30%, while the maximum internal resistance increase can reach 50% to 100%. Through normalization, the values ​​of both sequences are mapped to a range of 0 to 1, where 1 represents the most severely aged cell in that dimension, and 0 represents the least aged cell. During weighted averaging, the comprehensive aging index for each cell equals the normalized capacity decay value multiplied by the capacity weight, plus the normalized internal resistance increase value multiplied by the internal resistance weight. The resulting comprehensive aging index sequence fully reflects the aging state of each cell. The dynamic weight allocation strategy is based on the relative magnitude of capacity matching and lifetime matching. When the capacity matching value is higher than the lifetime matching value, it indicates that the main problem of the battery pack is capacity inconsistency. In this case, capacity decay should dominate the comprehensive evaluation and is therefore assigned a weight of 0.6, while internal resistance increase is assigned a weight of 0.4. Conversely, when the lifetime matching value is higher than the capacity matching value, it indicates that the difference in remaining lifetime among the cells is the main issue, and internal resistance, as an important indicator of lifetime degradation, should receive a higher weight. This dynamic adjustment mechanism allows the comprehensive score to adaptively reflect the actual state characteristics of the battery pack. The overall score for blending inconsistency is obtained by summing the weighted aging indices of all monomers. The higher the score, the more severe the performance inconsistency caused by blending.

[0188] (6.2) Based on the capacity matching degree and lifetime matching degree, allocate capacity decay weight and internal resistance growth weight to generate a comprehensive score for mixing inconsistency;

[0189] Specifically, based on the numerical values ​​of capacity matching degree and lifetime matching degree, if the capacity matching degree is higher than the lifetime matching degree, the capacity decay weight is set to 0.6 and the internal resistance growth weight is set to 0.4. If the lifetime matching degree is higher than the capacity matching degree, the capacity decay weight is set to 0.4 and the internal resistance growth weight is set to 0.6. The overall score of mismatch inconsistency is obtained by accumulating the products of the capacity decay weight and the internal resistance growth weight with the comprehensive aging index sequence.

[0190] (6.3) Based on the comprehensive score of the mixing inconsistency and the predicted shortest cycle lifetime, generate the remaining lifetime probability distribution and extract the final remaining lifetime estimate and confidence interval.

[0191] Specifically, using the aforementioned mixed-mixing inconsistency comprehensive score and the shortest cycle lifetime prediction value obtained above, the probability distribution of the remaining lifetime is calculated using the Bayesian inference method. The mean is extracted from the probability distribution as the final remaining lifetime estimate, and the range of the mean plus or minus twice the standard deviation is extracted as the confidence interval.

[0192] Among them, the application of Bayesian inference methods in remaining lifetime prediction makes full use of the combination of prior information and observational data.

[0193] The method first establishes a prior distribution of remaining battery life, typically using an empirical distribution based on historical data or a theoretical distribution based on battery aging mechanisms. The parameters of the prior distribution are determined based on factors such as battery type, usage conditions, and manufacturing batch. Observational data includes a comprehensive score for blending inconsistencies and a predicted shortest cycle life. These data are combined with the prior distribution using a likelihood function. The likelihood function describes the probability of observing the current comprehensive score and predicted value given the actual remaining battery life. Using Bayes' theorem, the product of the prior distribution and the likelihood function, after normalization, yields the posterior distribution.

[0194]

[0195] P(R|S,T) represents the posterior distribution of remaining life after observing the composite score S and the predicted value T, P(R) represents the prior distribution of remaining life, L(S,T|R) represents the likelihood function, and the integral term in the denominator represents the normalization constant to ensure that the sum of the posterior distribution probabilities is 1, i.e., the remaining life probability distribution considering all available information. The mean is extracted from the posterior distribution as a point estimate. This mean integrates prior knowledge and measured data, and is more reliable than predictions relying solely on any one information source.

[0196] The determination of the confidence interval is based on the statistical properties of the posterior distribution.

[0197] Standard deviation reflects the degree of uncertainty in a prediction; a larger standard deviation indicates lower prediction reliability. The interval formed by adding or subtracting twice the standard deviation from the mean contains approximately 95% probability quality, meaning there is a 95% probability that the actual remaining lifetime falls within this interval. The width of the reliability interval is influenced by several factors, including the quality of the observed data, the accuracy of prior information, and the degree of consistency within the battery pack.

[0198] (6.4) Based on the final remaining life estimate and confidence interval, generate a monitoring data set and output it to the battery pack management system.

[0199] Specifically, the final remaining life estimate, the upper limit of the confidence interval, and the lower limit of the confidence interval are combined into a monitoring data set, which is then output to the battery pack management system in the format of timestamp and data identifier. The management system performs real-time monitoring and maintenance decisions based on the received monitoring data set.

[0200] The monitoring data groups are organized according to a structured data format.

[0201] The timestamp records the precise moment the data was generated, using a year-month-day-hour-minute-second format with second-level precision. Data identifiers include the battery pack number, data type identifier, and version number, used to accurately identify and process received data in the management system. The final remaining lifespan estimate is in units of cycles, with the upper and lower limits of the confidence interval representing optimistic and conservative estimates, respectively.

[0202] For example, when the estimated final remaining life of a hybrid battery pack is 500 cycles, with a confidence range of 400 to 600 cycles, the management system can develop a maintenance plan accordingly. If 300 cycles have already been used, the system will issue an alert when 350 cycles are reached, reminding maintenance personnel to prepare for battery replacement.

[0203] The battery pack management system receives the monitoring data set and performs real-time monitoring through its built-in decision-making logic.

[0204] The received data is compared with a preset threshold, and an alarm mechanism is triggered when the remaining lifespan falls below the safety threshold. Simultaneously, historical data is stored in a database for trend analysis and continuous optimization of predictive models.

Claims

1. A method of predicting a trend of capacity degradation of a battery, characterized by, The method comprises the following steps: (1) obtaining the voltage response data and current distribution data of each single battery in the new and old mixed battery pack during charging and discharging, determining the voltage difference and capacity difference between each single battery; (2) according to the voltage difference and capacity difference, calculating the inconsistency index of the aging degree of the new and old mixed single battery, determining the actual current sharing ratio of each single battery, and generating the load distribution imbalance degree; (3) according to the load distribution imbalance degree, determining the new battery overload mode and the aging battery deep discharge mode, and forming the mixed abnormal working condition recognition mode; (4) according to the mixed abnormal working condition recognition mode, adjusting the current distribution ratio of the new battery and the aging battery, generating the charging and discharging time interval and the current switching frequency, and constructing the load balancing control framework; (5) according to the load balancing control framework, determining the capacity attenuation path and internal resistance growth path of the new and old mixed battery pack under multiple cycle periods, and generating the capacity matching degree and the life matching degree; (6) according to the capacity matching degree and the life matching degree, generating the final remaining life estimation value and the confidence interval, and outputting to the battery pack management system.

2. The method of claim 1, wherein, The step (1) comprises the following steps: (1.1) obtaining the voltage change curve of each single battery in the new and old mixed battery pack under constant current charging and discharging condition, recording the charging starting voltage, charging cutoff voltage, discharging starting voltage and discharging cutoff voltage, collecting the real-time current value of each single battery branch, calculating the voltage change rate and current distribution ratio; (1.2) extracting the capacity retention rate sequence and internal resistance measurement value sequence of each single battery in recent charging and discharging cycle from the battery management system, calculating the capacity attenuation rate and internal resistance growth rate, and generating the aging state feature vector group; (1.3) according to the aging state feature vector group, calculating the Euclidean distance of adjacent single battery feature vectors, generating the difference measurement value between single batteries, and determining the voltage difference distribution graph and the capacity difference distribution graph.

3. The method of claim 1, wherein the step of predicting the trend of the capacity degradation of the battery is performed by using a regression analysis method. The step (2) comprises the following steps: (2.1) calculating the mean and standard deviation of the voltage difference and capacity difference between all single batteries, generating the voltage difference dispersion and the capacity difference dispersion, calculating the weighted sum of the voltage difference dispersion and the capacity difference dispersion, and generating the aging inconsistency comprehensive index; (2.2) according to the aging inconsistency comprehensive index and the single battery position number, generating the cycle stress distribution sequence, calculating the deviation rate of the measured current value of each single battery and the average distribution value of the total current, and determining the actual current sharing ratio; (2.3) according to the difference square sum of the actual current sharing ratio and the ideal uniform sharing ratio, generating the load distribution variance, and calculating the load distribution imbalance degree.

4. The method of claim 1, wherein the step of predicting the trend of the capacity degradation of the battery is performed by using a function of the form: ###00003### where C is the capacity of the battery, t is the time, and a, b, and c are constants. The step (3) comprises the following steps: (3.1) determining the temperature sensor collection interval according to the load distribution imbalance degree, obtaining the relative temperature rise value of the surface temperature and the environment temperature of each single battery, and determining the thermal abnormal state single battery; (3.2) according to the temperature abnormal index and current sharing ratio of the thermal abnormal state single battery, determining the new battery overload feature and the aging battery deep discharge feature, and generating the overload abnormal mode set and the deep discharge abnormal mode set through clustering algorithm; (3.3) According to the overload abnormal mode set and the deep discharge abnormal mode set, a corrected temperature distribution sequence and a voltage fluctuation amplitude are calculated; (3.4) According to the corrected temperature distribution sequence and the voltage fluctuation amplitude, a temperature distribution unevenness and a voltage fluctuation frequency are generated, and a mixed abnormal working condition recognition mode is constructed.

5. The method of claim 1, wherein, The step (4) comprises the following steps: (4.1) According to the mixed abnormal working condition recognition mode, an overload single cell and a deep discharge single cell number are extracted, a new battery current reduction coefficient and an aged battery current increase coefficient are calculated, and an adjusted current distribution ratio sequence is generated; (4.2) According to the adjusted current distribution ratio sequence, a charging pause interval and a discharging rest interval are generated, and a charging and discharging time control sequence is constructed; (4.3) According to the charging and discharging time control sequence, a basic switching frequency and a temperature corrected switching frequency are calculated, and an actual current switching frequency is determined; (4.4) According to the adjusted current distribution ratio sequence, the charging and discharging time control sequence and the actual current switching frequency, a control structure containing a current distribution layer, a time control layer and a frequency adjustment layer is constructed as a load balancing control framework.

6. The method of claim 1, wherein, The step (5) comprises the following steps: (5.1) According to the load balancing control framework, each single cell capacity retention rate and internal resistance measurement value are recorded, a capacity attenuation rate sequence and an internal resistance growth rate sequence are generated, and a new and old battery differentiated attenuation path is determined; (5.2) According to the new and old battery differentiated attenuation path, a capacity attenuation curve of the new battery and the aged battery is generated, and a capacity convergence cycle number and a predicted remaining cycle number are determined; (5.3) According to the capacity convergence cycle number and the predicted remaining cycle number, a capacity synchronization coefficient and a life retention coefficient are calculated, and a comprehensive performance index is generated; (5.4) According to the comprehensive performance index and the predicted remaining cycle number, a capacity matching degree and a life matching degree are determined.

7. The method of claim 1, wherein the step of predicting the trend of the capacity degradation of the battery is performed by using a regression analysis method. The step (5) further comprises the following steps: (5.5) According to the load balancing control framework, a capacity time sequence is generated, and a capacity attenuation curve is constructed; (5.6) According to the capacity attenuation curve, a cycle number required for the new battery to attenuate to the current capacity of the aged battery and a cycle number of the aged battery to the failure threshold are calculated; (5.7) According to the cycle number, a capacity attenuation rate ratio relationship is generated, and a capacity convergence time node and a failure time node are determined; (5.8) According to the capacity convergence time node and the failure time node, a shortest cycle life prediction value is generated.

8. The method of claim 1, wherein the step of predicting the trend of the capacity degradation of the battery is performed by using a neural network. The step (6) comprises the following steps: (6.1) The capacity attenuation degree and the internal resistance growth amplitude of each single cell are normalized, a capacity attenuation normalized sequence and an internal resistance growth normalized sequence are generated, and a comprehensive aging index sequence is calculated; (6.2) According to the capacity matching degree and the life matching degree, a capacity attenuation weight and an internal resistance growth weight are allocated, and a mixed inconsistency comprehensive score is generated; (6.3) According to the mixed inconsistency comprehensive score and the shortest cycle life prediction value, a residual life probability distribution is generated, and a final residual life estimate value and a credibility interval are extracted; (6.4) Based on the final remaining life estimate and the confidence interval, a monitoring data set is generated and output to the battery pack management system.

Images