An on-line battery equalization method and system

Abstract

The application discloses a kind of battery online equalization method and system, collect battery pack operating data and strip polarized voltage interference extraction steady-state deviation characteristics, combine capacity attenuation index to construct single state matrix;Then, through cluster analysis, identify high-energy single and low-energy single, according to its position distribution and current distribution situation, analyze energy mismatch degree and demarcate stratified equalization domain;Further, the deterioration evolution law of outlier single is mined to form a consistent deterioration trajectory, combined with the internal resistance change characteristics to generate internal resistance compensation coefficient to determine the equalization critical position, establish the charge equalization correlation table;Subsequently, the low-loss energy transfer path is planned to develop the basic equalization list;Finally, according to the imbalance degree and thermal constraint condition, form the equalization order table, determine the equalization start threshold and output the equalization execution command.The application can accurately distinguish the real state difference and transient fluctuation of single, predict the equalization demand in advance, orderly dispatch equalization operation, improve the equalization efficiency and operation reliability of battery pack.

Description

Technical Field

[0001] This invention relates to the field of energy storage system management technology, and in particular to an online battery balancing method and system. Background Technology

[0002] As a core component of energy storage systems, battery packs experience gradual deviations in the state of charge (SOC) of individual cells during long-term cyclic use due to factors such as differences in manufacturing processes, operating environments, and varying aging rates. If these deviations are not addressed promptly, they can lead to overcharging or over-discharging of some cells, reducing the overall usable capacity of the battery pack, accelerating performance degradation, and even posing safety hazards. Therefore, online balancing technology has become a crucial means to extend the lifespan of battery packs and ensure operational safety.

[0003] Existing battery equalization methods mostly employ fixed threshold triggering or simple voltage difference comparison strategies. In assessing the condition of individual cells, they lack the ability to effectively distinguish between transient and steady-state components in voltage fluctuations, easily misinterpreting transient deviations caused by polarization voltage as genuine state differences. Regarding equalization timing, existing methods typically react passively based on the current state, lacking the ability to predict cell aging trends. Equalization is often initiated only after cells have already severely deteriorated, missing the optimal intervention window. In terms of equalization execution scheduling, existing methods rarely consider the risks of thermal accumulation and mutual interference between equalization operations. Continuously equalizing the same area may lead to excessively high local temperatures, and the lack of systematic prioritization prevents equalization resources from being allocated preferentially to the most urgent cells. Summary of the Invention

[0004] This invention discloses an online battery balancing method and system. By collecting the operating data of the battery pack, a single-cell state matrix is ​​constructed. Based on inconsistent clustering, high-energy cells and low-energy cells are identified and hierarchical balancing domains are defined. Combined with single-cell aging trend analysis, the balancing critical position is determined and a charge balancing association table is established. Energy transfer paths are planned and a basic balancing list is formulated. A balancing sequence table is formed and the balancing start threshold is determined to output the balancing execution command, thereby realizing online dynamic balancing of the battery pack's charge state.

[0005] The first aspect of this invention proposes an online battery balancing method, comprising the following steps:

[0006] Acquire voltage discrete characteristics, capacity deviation characteristics, and outlier cell data of the battery pack under operating conditions, and perform consistency offset verification on the voltage discrete characteristics and the capacity deviation characteristics to generate a cell state matrix;

[0007] Using the monomer state matrix, inconsistency clustering analysis is performed to identify high-energy monomers and low-energy monomers. The energy mismatch degree is analyzed based on the positional distribution of the high-energy monomers and the low-energy monomers. Based on the energy mismatch degree, a hierarchical equilibrium domain is delineated.

[0008] Trend analysis is performed on the outlier data to form a consistent degradation trajectory. Internal resistance deviation features are extracted from the consistent degradation trajectory to generate an internal resistance compensation coefficient. Based on the consistent degradation trajectory and the internal resistance compensation coefficient, degradation rate is evaluated to determine the equilibrium critical position. A charge balance correlation table is established according to the equilibrium critical position and the hierarchical equilibrium domain.

[0009] Based on the energy mismatch, the transfer loss rate is extracted. Based on the transfer loss rate and the internal resistance compensation coefficient, an energy transfer path is generated. Based on the energy transfer path and the charge balance association table, a basic balance list is formulated.

[0010] The basic balance list is used to perform balance priority allocation to form a balance order table. Based on the balance order table and the balance critical position, the balance start threshold is determined. The balance order table is then subjected to step verification according to the balance start threshold, and a balance execution command is output.

[0011] A second aspect of this invention provides an online battery balancing system, comprising:

[0012] The data acquisition module is used to acquire voltage discrete characteristics, capacity deviation characteristics, and outlier cell data of the battery pack under operating conditions, and to perform consistency offset verification on the voltage discrete characteristics and the capacity deviation characteristics to generate a cell state matrix.

[0013] The mismatch analysis module is used to perform inconsistent clustering analysis using the monomer state matrix to identify high-energy monomers and low-energy monomers, analyze the energy mismatch degree based on the positional distribution of the high-energy monomers and low-energy monomers, and delineate a hierarchical equilibrium domain based on the energy mismatch degree.

[0014] The trend analysis module is used to perform trend analysis on the outlier data to form a consistent degradation trajectory, extract internal resistance deviation characteristics from the consistent degradation trajectory to generate an internal resistance compensation coefficient, evaluate the degradation rate based on the consistent degradation trajectory and the internal resistance compensation coefficient to determine the equilibrium critical position, and establish a charge balance correlation table according to the equilibrium critical position and the hierarchical equilibrium domain.

[0015] The energy transfer module is used to extract the transfer loss rate based on the energy mismatch, generate an energy transfer path based on the transfer loss rate and the internal resistance compensation coefficient, and formulate a basic balance list by combining the energy transfer path with the charge balance association table.

[0016] The equalization output module is used to perform equalization priority allocation using the basic equalization list to form an equalization order table, determine the equalization start threshold based on the equalization order table and the equalization critical bit, and perform step verification on the equalization order table according to the equalization start threshold to output the equalization execution command.

[0017] The beneficial effects of this invention are reflected in the following points: 1. By performing polarization voltage stripping processing on the battery pack operating data, the voltage fluctuations caused by transient current are separated from the intrinsic state differences of individual cells. Based on steady-state deviation characteristics and capacity decay indicators, a cell state matrix is ​​constructed. Then, high-energy cells and low-energy cells are identified through inconsistent clustering. Based on their topological location and current distribution relationship, hierarchical equilibrium domains are delineated, enabling the equilibrium operation to be performed preferentially in relatively independent regions, reducing mutual interference between regions. 2. By performing trend analysis on outlier cell data to form a consistent degradation trajectory, two different types of state changes, gradual aging and sudden degradation, are distinguished. Combined with the difference in internal resistance growth rate, an internal resistance compensation coefficient is generated. The time when a cell reaches the degradation threshold is predicted and corrected to determine the equilibrium critical position. A charge equilibrium correlation table is established to realize the correlation mapping between equilibrium timing and equilibrium region, enabling the equilibrium system to actively intervene before the cell degradation accelerates, avoiding intervention lag caused by passive response. 3. By analyzing the energy loss and temperature sensitivity characteristics of each transfer channel, the transfer loss rate is extracted, and energy transfer paths with high efficiency and good stability are selected and a basic balance list is formulated. The balance tasks are sorted according to the degree of imbalance, and the balance sequence table is formed by inserting division points according to the heat accumulation risk. Combined with the dynamically adjusted balance start threshold, the trigger conditions are verified step by step to output the balance execution command, so that the balance operation is executed in an orderly manner under the premise of ensuring thermal safety, thereby improving the utilization efficiency of balance resources. Attached Figure Description

[0018] The accompanying drawings illustrate specific examples of the technical solutions described in this invention and, together with the detailed embodiments, form part of the specification, serving to explain the technical solutions, principles, and effects of this invention.

[0019] Figure 1 This is a schematic flowchart of an online battery balancing method according to the present invention.

[0020] Figure 2 This is a structural block diagram of an online battery balancing system according to the present invention. Detailed Implementation

[0021] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0022] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0023] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0024] The technical solutions of the embodiments of this application will be described below.

[0025] like Figure 1 As shown, this embodiment of the invention provides an online battery balancing method, including the following steps S110-S150:

[0026] Step S110: Obtain the voltage discrete characteristics, capacity deviation characteristics, and outlier cell data of the battery pack under operating conditions, and perform consistency offset verification on the voltage discrete characteristics and capacity deviation characteristics to generate a cell state matrix.

[0027] Specifically, the voltage dispersion characteristics, capacity deviation characteristics, and outlier cell data of the battery pack under operating conditions are acquired. During the charge-discharge cycle, the terminal voltage, cumulative discharge amount, and internal resistance of each cell in the battery pack will show differentiated changes. The terminal voltage of each cell is collected and its deviation from the mean of the group is calculated. The deviation is arranged by cell number to form voltage dispersion characteristics. At the end of the discharge, the terminal voltage of cell number 5 in the battery pack is 12.3V, while the mean of the group is 12.6V. The voltage deviation of this cell is -0.3V and is recorded in the voltage dispersion characteristics. The voltage dispersion characteristics can show the voltage differentiation of each cell during the charge-discharge process. The cumulative discharge capacity is obtained by integrating the discharge current of each cell over time. The ratio of the cumulative discharge capacity of each cell to its nominal capacity is used as the capacity decay index. The capacity decay index is arranged by cell number to form the capacity deviation characteristic. After 500 cycles, the cumulative discharge capacity of cell number 12 in the battery pack decreased by 18% compared to its nominal capacity, while the average decrease in the group was only 10%. The capacity decay index of this cell in the capacity deviation characteristic is significantly higher than that of other cells. The capacity deviation characteristic can show the difference in actual usable capacity caused by different degrees of aging of each cell. In the battery pack, some cells have voltage, current, or internal resistance parameters that deviate significantly from the normal distribution range of the group. The multi-dimensional parameter records of these cells are extracted to form outlier cell data. The internal resistance of cell number 23 in the battery pack is 15mΩ, while the average value in the group is only 9mΩ, which is more than twice the normal fluctuation range. The historical records of voltage, current, internal resistance, and temperature of this cell are included in the outlier cell data. The outlier cell data provides raw material for subsequent trend analysis and degradation trajectory analysis.

[0028] In some embodiments, the step of performing consistency offset verification on the voltage discrete features and the capacity deviation features to generate a single-unit state matrix includes: performing polarization voltage stripping on the voltage discrete features to obtain steady-state discrete components; classifying the steady-state discrete components according to the deviation magnitude to form a deviation classification spectrum; performing time-series correlation analysis on the deviation classification spectrum and the capacity deviation features to obtain offset verification factors; and using the offset verification factors to construct a single-unit state matrix.

[0029] The steady-state discrete components are obtained by polarization voltage stripping of the voltage discrete characteristics. The instantaneous voltage deviation in the voltage discrete characteristics consists of two parts: one part is the steady-state deviation caused by the difference in intrinsic state of the individual cells, and the other part is the dynamic polarization voltage drop caused by the current flowing through the internal resistance. The total deviation formed by the superposition of the two parts cannot directly reflect the true degree of degradation of the individual cells. The polarization voltage of each cell is calculated based on the current value and internal resistance value at the corresponding moment of the voltage discrete characteristics: polarization voltage V_p = I × R, where I is the cell current and R is the cell internal resistance. The steady-state voltage deviation of each cell is obtained by subtracting the corresponding polarization voltage from the instantaneous voltage deviation of the voltage discrete characteristics. When the battery pack switches to a high-current discharge state, the voltage discrete characteristics of cell No. 5 show a sharp increase in deviation to 120mV. The polarization voltage at this moment is calculated to be 95mV. After stripping, the steady-state voltage deviation is only 25mV, which is within the normal range, indicating that the voltage fluctuation is caused by the internal resistance voltage drop rather than intrinsic degradation. The steady-state voltage deviation values ​​of each cell are organized into steady-state discrete components according to the cell number. The steady-state discrete components eliminate the dynamic polarization effect caused by current changes and record the intrinsic voltage deviation of each cell in the form of a numerical sequence, providing clean input data for subsequent classification according to the deviation magnitude.

[0030] A deviation grading spectrum is formed based on the deviation magnitude of steady-state discrete components. The standard deviation of all steady-state voltage deviations in the steady-state discrete components is calculated as the grading benchmark. The normalized deviation magnitude K = ΔV_s / σ is obtained by dividing the steady-state voltage deviation of each individual cell in the steady-state discrete components by the grading benchmark, where ΔV_s is the steady-state voltage deviation and σ is the standard deviation. The normalized deviation magnitude eliminates the dimensional influence caused by differences in voltage levels among different battery packs, making the grading standard universal. The deviation levels are divided according to the numerical range of the normalized deviation magnitude: Level 1 (minor deviation) is less than one standard deviation; Level 2 (moderate deviation) is between one and two standard deviations; and Level 3 (severe deviation) is greater than two standard deviations. The standard deviation of the steady-state discrete components of the battery pack is 18mV. Cell #23 has a steady-state voltage deviation of 42mV, with a normalized deviation of 2.33, exceeding twice the standard deviation and thus classified as Level 3. Cell #11 has a steady-state voltage deviation of 15mV, with a normalized deviation of 0.83, less than one standard deviation and thus classified as Level 1. A deviation classification spectrum is formed by recording the cell numbers and their corresponding deviation levels. This spectrum converts continuous voltage deviation values ​​into discrete level labels. Cells with Level 3 deviations will be the first to reach the cutoff voltage during deep discharge of the battery pack, limiting the overall usable capacity of the pack. These cells require close monitoring during equalization operations.

[0031] A time-series correlation analysis was conducted on the deviation grading spectrum and capacity deviation characteristics to obtain the offset verification factor. The deviation grading spectrum records the time-series changes in the deviation level of each cell, while the capacity deviation characteristics record the time-series changes in the capacity degradation index of each cell. Pearson correlation coefficients were calculated for the deviation level sequence and capacity degradation index sequence of the same cell. The Pearson correlation coefficient measures the degree of synchronous change between voltage deviation and capacity degradation, and its value ranges from -1 to +1. Cells in the battery pack with internal structural damage due to plate sulfation or active material shedding exhibit the characteristic of synchronous deterioration of voltage deviation and capacity degradation. These cells typically have high Pearson correlation coefficients and are considered degraded cells requiring priority intervention. Cell number 8 in the battery pack showed a level 2 deviation in the deviation grading spectrum for three consecutive months, but its capacity degradation index remained stable, with a calculated Pearson correlation coefficient of only 0.15. Investigation revealed that the cell's location near the air conditioning vent caused a localized low temperature. The level fluctuations in the deviation grading spectrum coincided with the air conditioning start-stop cycle, indicating a temperature-induced transient deviation rather than true degradation. The Pearson correlation coefficients of each monomer in the deviation grading spectrum and capacity deviation characteristics are normalized to obtain the offset verification factor. The offset verification factor is used to distinguish between the real deteriorated monomers and the transient interference monomers, providing a basis for determining the equilibrium priority in the subsequent process.

[0032] A single-cell state matrix was constructed using offset verification factors. The verification coefficient value for each cell was determined based on the magnitude of the offset verification factor. A higher verification coefficient value indicates a stronger correlation between voltage deviation and capacity degradation, classifying the cell as a true faulty cell with simultaneous voltage and capacity degradation. A lower verification coefficient value indicates that the voltage deviation is affected by external factors, classifying the cell as a transient abnormality that does not require immediate intervention. In the battery pack, cell number 12 is marked as level three deviation in the deviation grading spectrum, but its offset verification factor is only 0.18. Cell number 7 is marked as level two deviation, but its offset verification factor reaches 0.87. When sorted according to the offset verification factor, cell number 7 has a higher priority than cell number 12, indicating that although cell number 7 currently has a lower deviation level, its voltage deviation and capacity degradation show a strong correlation and a continuously worsening trend, requiring priority treatment. The individual unit state matrix uses the unit number as the row index and assembles the deviation level, capacity degradation level, and offset verification factor values ​​as column fields to form a structured two-dimensional table M=[n,L_d,L_c,γ], where n is the unit number, L_d is the deviation level, L_c is the capacity degradation level, and γ is the verification factor value. The individual unit state matrix integrates the scattered deviation level, capacity degradation level, and verification factor values ​​into structured data, supporting sorting and filtering by any field, facilitating the rapid identification of deteriorated units requiring priority equalization.

[0033] Step S120: Using the monomer state matrix, perform inconsistent clustering analysis to identify high-energy monomers and low-energy monomers, analyze the energy mismatch degree based on the positional distribution of high-energy monomers and low-energy monomers, and delineate the hierarchical equilibrium domain based on the energy mismatch degree.

[0034] Specifically, the individual cell state matrix records the deviation level, capacity decay level, and verification coefficient value for each cell. Weighted clustering is performed on these three fields to identify cells with abnormal energy states. The deviation level and capacity decay level are converted into numerical form: Level 1 corresponds to a value of 1, Level 2 to 2, and Level 3 to 3. Higher values ​​indicate more severe deviations or decays. The weights are determined based on the influence of each field on the energy state: deviation level has a weight of 0.3, capacity decay level has a weight of 0.5, and verification coefficient value has a weight of 0.2. The sum of the weighted values ​​of the three fields for each cell is calculated as a comprehensive energy index. A higher comprehensive energy index indicates more severe cell degradation and lower usable energy. Cells with a comprehensive energy index below one standard deviation of the group mean are marked as high-energy cells, and cells with a comprehensive energy index above one standard deviation of the group mean are marked as low-energy cells. In the battery pack, cell number 7 deviates from grade 2, but its verification coefficient value is as high as 0.87. Its overall energy index exceeds the upper threshold, so it is marked as a low-energy cell, indicating that the cell's voltage and capacity deteriorate simultaneously, resulting in lower actual usable energy. Cell number 3 deviates from grade 1 with a verification coefficient value of 0.22. Its overall energy index is below the lower threshold, so it is marked as a high-energy cell, indicating that its voltage deviation is slight, its capacity decay is minimal, and its actual usable energy is sufficient. In the series structure of the battery pack, high-energy cells are prone to overcharging at the end of charging, while low-energy cells are prone to over-discharging at the end of discharging. The presence of these two types of cells limits the overall usable capacity of the pack to the weakest cell.

[0035] In some embodiments, the step of analyzing the energy mismatch based on the positional distribution of the high-energy cells and the low-energy cells includes: performing topological proximity analysis on the high-energy cells and the low-energy cells to locate neighboring positions and establish a neighboring position table; obtaining the charge / discharge current allocation ratio of the high-energy cells and the low-energy cells in the neighboring position table to generate a current offset; filtering overlapping position pairs of influence domains based on the current offset and the neighboring position table to form an overlap identifier; and determining the energy mismatch based on the overlap degree of the overlap identifier and the current offset.

[0036] Topological proximity analysis is performed on high-energy and low-energy cells to locate their nearest neighbors and establish a proximity table. In a battery bank, cells are connected in series or series-parallel configurations. The relative positions of high-energy and low-energy cells in the electrical topology determine the length of the energy transfer path. Cell pairs with closer topological distances can complete energy transfer through shorter paths, resulting in higher balancing efficiency and lower transmission losses. Cell pairs with greater topological distances require longer paths for energy transfer, leading to increased energy losses during balancing. The topological distance between each pair of high-energy and low-energy cells is calculated. Topological distance is defined as the number of connection nodes traversed between the two cells. When the battery bank uses a four-parallel, eight-series structure, the topological distance between high-energy and low-energy cells located in the same parallel branch is zero, allowing energy to be transferred directly within the branch without passing through external nodes. However, the topological distance between cell pairs distributed at different series levels may reach three or more nodes; the longer the path, the lower the balancing efficiency. High-energy and low-energy cell pairs with a topological distance less than a set threshold are identified as neighboring locations. The high-energy cell number, low-energy cell number, and topological distance of all neighboring locations are recorded to form a neighboring location table. The neighboring location table is arranged in ascending order of topological distance, with cell pairs that are closer in distance appearing first. This facilitates the equilibration strategy to prioritize cell pairs with the shortest transfer paths to improve overall equilibration efficiency.

[0037] The current offset is generated by obtaining the charging and discharging current distribution ratio of high-energy and low-energy cells in the neighboring cell table. The current borne by each pair of high-energy and low-energy cells in the neighboring cell table differs during charging and discharging; cells with lower internal resistance bear a larger current, while cells with higher internal resistance bear a smaller current. This uneven current distribution exacerbates the energy difference between cells, further widening the state of charge difference between high-energy and low-energy cells. For each pair of cells in the neighboring cell table, its charging and discharging current value is read, and the ratio of the high-energy cell current to the low-energy cell current is calculated as the current distribution ratio. The degree to which the current distribution ratio deviates from the ideal value is defined as the current offset, D_I = (I_h - I_l) / (I_h + I_l), where I_h is the high-energy cell current and I_l is the low-energy cell current. Under high-current discharge conditions, the high-energy cell (number 3) in the adjacent cell table has an internal resistance of 8 mΩ and carries a current of 52 A, while its paired low-energy cell (number 7) has an internal resistance of 12 mΩ and carries a current of 48 A. The calculated current offset is 0.04, indicating a slight imbalance in the current distribution of this cell pair. The high-energy cell carries a relatively larger discharge current, causing its state of charge to decrease faster than that of the low-energy cell. The larger the absolute value of the current offset, the more unbalanced the current distribution of the cell pair; the smaller the absolute value, the closer the current distribution is to the ideal state. Subsequent balancing strategies should focus on cell pairs with larger absolute values ​​of current offset.

[0038] The overlapping regions of each cell pair in the adjacent location table are identified by using current offset and proximity location data. Each cell exerts thermal and electromagnetic influences on surrounding cells during charging and discharging, and this influence range constitutes its influence domain. When the influence domains of two adjacent locations overlap, their equalization operations will interfere with each other, requiring coordinated execution rather than independent operation. Otherwise, it may lead to localized heat concentration or superimposed electromagnetic interference. The radius of the influence domain for each cell pair in the adjacent location table is determined based on the magnitude of the current offset. A larger absolute value of the current offset results in a larger influence domain radius, indicating a higher energy transfer intensity during equalization operation and a correspondingly larger range of thermal and electromagnetic influences on surrounding cells. When the battery pack is compactly arranged, the heat dissipation channels between adjacent modules are limited. The first pair of cells in the adjacent location table has a larger absolute value of current offset, corresponding to an influence domain radius of 15mm, while the second pair has a smaller absolute value, corresponding to an influence domain radius of 12mm. The physical distance between the two pairs of cells is only 20mm, resulting in a 7mm overlap area. If both pairs of cells perform equalization operations simultaneously, the superimposed heat may cause the temperature in the overlap area to exceed the safety threshold. Determine whether the influence domains of any two pairs of adjacent positions have spatial overlap. If there is overlap, generate an overlap identifier for the two pairs of positions. The overlap identifier records the numbers of the two pairs of adjacent positions and the ratio of their overlap area. The larger the ratio of the overlap area of ​​the overlap identifier, the stronger the mutual influence between the two pairs of positions. When performing the balancing operation, staggered scheduling is required to avoid interference superposition affecting the balancing security.

[0039] The energy mismatch is determined based on the degree of overlap of overlapping markers and the current offset. For each pair of adjacent positions in the adjacent position table, the sum of the number of overlapping markers and the percentage of overlapping area is used as the overlap degree index. In the battery pack, the 5th pair of adjacent positions involves 3 overlapping markers with a sum of overlapping area percentages of 0.45, while the 8th pair of adjacent positions involves only 1 overlapping marker with a sum of overlapping area percentages of 0.12. This indicates that the balancing operation of the 5th pair of adjacent positions is more severely affected by surrounding interference and requires coordination and scheduling with multiple pairs of adjacent positions. The overlap degree index is multiplied by the absolute value of the current offset of the position pair to obtain the mismatch index of that position pair. The larger the absolute value of the current offset and the more overlapping markers involved, the higher the mismatch index for that position pair. The overall energy mismatch is obtained by weighted summation of the mismatch indices of all position pairs in the adjacent position table. The energy mismatch E_m = Σ(w_i × |D_I_i| × C_i), where w_i is the weight of the i-th position pair, D_I_i is the current offset of the i-th position pair, and C_i is the overlap index of the i-th position pair. A higher energy mismatch value indicates a more uneven energy distribution in the battery pack and more severe mutual interference during equalization operations. A layered equalization strategy is needed to isolate and process the interfering areas separately. A lower energy mismatch value indicates that the equalization operation of the battery pack can be performed relatively independently without complex coordination and scheduling.

[0040] A tiered equilibrium domain was defined based on energy mismatch. The energy mismatch was divided into tiers according to numerical ranges, with two-tier thresholds (0.5 times and 1.5 times the average energy mismatch) forming a three-tiered structure. The average energy mismatch of the battery pack was 0.28. Based on this, the first tier threshold was set at 0.14, and the second tier threshold at 0.42. Cells with an energy mismatch below 0.14 were assigned to the first tier equilibrium domain (indicating slight mismatch), those between 0.14 and 0.42 were assigned to the second tier equilibrium domain (indicating moderate mismatch), and those above 0.42 were assigned to the third tier equilibrium domain (indicating severe mismatch). Simultaneously, the physical proximity of cells was considered; adjacent cells meeting the same threshold range and with a topological distance of less than three nodes were grouped into the same domain, ensuring shorter energy transfer paths and lower losses between cells within the same domain. Eight cells in the battery pack—cells 5, 8, 12, 15, 18, 19, 21, and 24—have energy mismatch contribution values ​​below 0.14 and are topologically adjacent in the central region of the battery pack. They are assigned to the first hierarchical balancing domain and assigned domain number D_id=1. The remaining cells are assigned to the second and third hierarchical balancing domains according to the same rules and assigned corresponding domain numbers. Assigning a unique domain number D_id to each hierarchical balancing domain facilitates recording the domain affiliation information of each cell in the charge balancing association table. Cells within the same hierarchical balancing domain, due to their similar mismatch levels and proximity, can collaboratively perform balancing operations without mutual interference. However, cells in different domains require coordinated scheduling to avoid overlapping interference that could affect balancing safety.

[0041] Step S130: Perform trend analysis on outlier data to form a consistent degradation trajectory, extract internal resistance deviation characteristics from the consistent degradation trajectory to generate an internal resistance compensation coefficient, evaluate the degradation rate based on the consistent degradation trajectory and the internal resistance compensation coefficient to determine the equilibrium critical position, and establish a charge balance correlation table according to the equilibrium critical position and the hierarchical equilibrium domain.

[0042] In some embodiments, the step of performing trend analysis on the outlier data to form a consistent degradation trajectory includes: performing distributed clustering mining on the outlier data to establish a time-series clustering band; identifying abrupt change anomalies and normal segments based on the time-series clustering band to generate abrupt change marker set and normal degradation segments respectively; performing gradient risk screening on the abrupt change marker set and the normal degradation segments to generate a segmented degradation trajectory; and using the segmented degradation trajectory for continuous extrapolation to form a consistent degradation trajectory.

[0043] Distributed clustering mining is used to establish time-series clustering bands for outlier data. Outlier data contains feature records of multiple outliers at different times. Records with similar features within the same time window often reflect similar degradation mechanisms. Clustering methods are needed to group these records by time period to identify common degradation patterns. Outlier data is divided into consecutive time windows in chronological order. The length of each time window is set according to the charge / discharge cycle of the battery pack, typically using one week as the basic window unit. Cluster analysis was performed on the feature vectors of outlier cells within each time window. The feature vectors included dimensions such as average internal resistance, rate of change of internal resistance, average terminal voltage, and self-discharge rate within that window. Outlier cell data for the battery pack showed that the cluster distance between the internal resistance feature vectors of cells 5, 12, and 18 was less than 0.15 from week 1 to week 3, indicating highly similar degradation characteristics during this period. However, the cluster distance increased to 0.32 from week 4 to week 6, indicating that degradation characteristics began to differentiate. Based on this, the first three weeks were divided into one time-series cluster, and the last three weeks into another, each corresponding to different stages of degradation characteristics. Time windows with similar clustering results were merged to form time-series clusters. These clusters organized discrete time-point data into continuous time-segment features, facilitating the identification of phased changes in the degradation process.

[0044] Based on time-series clustering, jump anomalies and normal segments are identified, generating jump marker sets and normal degradation segments respectively. Two types of segments exist within the time-series clustering: normal degradation segments with gradually changing eigenvalues ​​reflect the natural aging process of the individual cell; and jump anomaly segments with sudden spikes or drops in eigenvalues ​​reflect sudden failures occurring within the individual cell. The degradation mechanisms and treatment strategies for these two types of segments are completely different and require separate labeling. The characteristic value differences between adjacent time periods in the scanning time series cluster were analyzed. A jump detection threshold was set at five times the normal daily fluctuation range. When the characteristic value difference between adjacent time periods exceeded this threshold, it was considered a jump anomaly. In the battery pack's time series cluster, the internal resistance of cell number 12 was 10mΩ on day 15, but suddenly dropped to 6mΩ on day 16, a daily variation of 40%, far exceeding the normal fluctuation range of 5%. A sudden drop in internal resistance usually corresponds to a micro-short circuit within the cell causing current bypass. This time point was marked as a jump anomaly and recorded in the jump marker set. This record in the jump marker set includes the jump time as day 16, internal resistance before the jump as 10mΩ, internal resistance after the jump as 6mΩ, jump direction as decreasing, and suspected fault type as micro-short circuit. Continuous segments in the time series cluster excluding jump anomalies were marked as normal degradation segments. Normal degradation segments reflect the natural aging process of cells under conditions without sudden events, and their degradation rate is relatively stable and predictable.

[0045] For example, the step of performing gradient risk screening on the jump marker set and the normal degradation segment to generate segmented degradation trajectories includes: performing distributed gradient detection on the jump marker set and the normal degradation segment to establish a differential coupling band; identifying slope abrupt change locations within the differential coupling band to generate abrupt change warning points; performing accelerated degradation risk assessment on the abrupt change warning points to generate a risk slope ranking table; and selecting high-risk slopes based on the risk slope ranking table to form segmented degradation trajectories.

[0046] Differential coupling bands are established by performing distribution gradient detection on the jump marker set and the normal degradation section. Jump events in the jump marker set and gradual degradation in the normal degradation section alternate on the time axis. The transition region between the two often contains early signals of accelerated degradation. Although the jump event is short-lived, its damage to the internal structure of the monomer will continue to affect the degradation rate for a period of time afterward. This affected transition region needs to be monitored closely. Each time window before and after a jump event in the jump event marker set is taken as the detection range. Within this window, the degradation gradient is compared with that of the normal degradation section to calculate the change in gradient before and after the jump. In the jump event marker set, cell number 12 of the battery pack experienced a sudden drop in internal resistance on day 16. Two weeks before and after this jump event are taken as the detection window. In the normal degradation section, the internal resistance growth gradient in the two weeks before the jump event is 0.05 mΩ / day, indicating that the cell is in a slow aging state. However, the internal resistance growth gradient in the two weeks after the jump event rises to 0.12 mΩ / day, indicating that the micro-short circuit event accelerated the subsequent degradation process. The gradient change amplitude reaches 140%, exceeding the set threshold of 100%. This four-week time window is marked as the differential coupling zone. The degradation gradient in the differential coupling zone exhibits significant nonlinear characteristics, which is different from the linear gradual change pattern of the normal degradation section. The differential coupling zone characterizes the impact range of the jump event on the surrounding normal degradation process. When formulating a balancing strategy, the differential coupling zone should be a key focus area for priority intervention measures.

[0047] Within the differential coupling zone, locations of abrupt slope changes are identified to generate early warning points for these changes. The degradation gradient within the differential coupling zone is not uniformly distributed but exhibits local fluctuations. Secondary abrupt changes in the gradient at some locations form inflection points of accelerated degradation. These locations often indicate that the monomer is about to transition from controllable degradation to uncontrolled degradation, requiring early warning for timely intervention. The second derivative of the degradation gradient sequence within the differential coupling zone is calculated. The second derivative reflects the change in degradation acceleration and can capture inflection points in gradient changes. A change from positive to negative indicates a shift from accelerating to decelerating degradation, suggesting that degradation is stabilizing; a change from negative to positive indicates a shift from decelerating to accelerating degradation, suggesting that degradation is about to worsen. In the differential coupling zone of the battery pack, the second derivative of the degradation gradient of cell number 12 abruptly changed from -0.002 to +0.008 on day 20. The absolute change was 0.01, exceeding the threshold of 0.005, indicating that the degradation rate of this cell began to accelerate on day 20. Day 20 was marked as the abrupt change warning point. The record for this warning point is as follows: abrupt change time of day 20, gradient before the abrupt change of 0.08 mΩ / day, gradient after the abrupt change of 0.15 mΩ / day, and gradient change direction of acceleration. The abrupt change warning point indicates the moment when the degradation rate is about to change significantly. The dense appearance of multiple abrupt change warning points usually indicates that the cell is undergoing a continuous accelerated degradation process.

[0048] An accelerated degradation risk assessment is conducted for mutation warning points, generating a risk slope ranking table. For each mutation warning point, the increase in the degradation gradient after the mutation relative to the pre-mutation level is calculated as an acceleration factor. The acceleration factor equals the post-mutation gradient divided by the pre-mutation gradient. A larger acceleration factor indicates a more severe acceleration in degradation caused by the mutation warning point, a faster rate of monomer performance decline, and the need for more urgent intervention measures. The time urgency is calculated based on the time position of the mutation warning point. The time urgency is calculated using an exponential decay function, giving higher urgency weights to recently occurring mutation warning points. The closer the mutation warning point is to the current time, the higher its time urgency, indicating a more pressing risk. Cell number 12 in the battery pack has three abrupt change warning points. The first occurred 180 days ago with an acceleration factor of 1.2, resulting in a time urgency of 0.3. The second occurred 60 days ago with an acceleration factor of 1.8, resulting in a time urgency of 0.6. The third occurred 7 days ago with an acceleration factor of 2.5, resulting in a time urgency of 0.95. Multiplying the acceleration factor by the time urgency yields comprehensive risk scores of 0.36, 1.08, and 2.38, respectively. The third abrupt change warning point has the highest risk score, indicating that this cell is currently experiencing severe accelerated degradation. A risk slope ranking table is created by sorting the warning points in descending order of risk score. Warning points ranked higher in the risk slope ranking table correspond to the highest risk of accelerated degradation events.

[0049] High-risk slopes are selected based on a risk slope ranking table to form segmented degradation trajectories. A risk score threshold is set as the screening criterion for high-risk warning points. Warning points with risk scores exceeding the threshold are selected from the risk slope ranking table. The time segments of these warning points are extracted and included in the segmented degradation trajectory for focused tracking. Warning points with risk scores below the threshold correspond to relatively mild degradation acceleration and can be temporarily deferred. Each high-risk segment is extended forward and backward for a certain period of time, centered on the corresponding abrupt change warning point. The extension period is determined based on the persistence of the degradation gradient, typically ranging from two weeks to one month, to ensure complete coverage of the initiation and persistence stages of degradation acceleration without missing key degradation evolution information. The battery pack has a risk score threshold of 1.0. In the risk slope ranking table, the second and third abrupt change warning points for cell number 12 have risk scores of 1.08 and 2.38 respectively, both exceeding the threshold and thus classified as high-risk. A first high-risk segment is formed by extending two weeks before and after the second warning point, covering days 46 to 74. A second high-risk segment is formed by extending two weeks before and after the third warning point, covering days 166 to the present. The two segments are merged to form the segmented degradation trajectory of this cell. The segmented degradation trajectory highlights the highest-risk stages in the degradation process of outlier cells. Each segment is labeled with its time range, average degradation gradient, and peak degradation gradient, facilitating subsequent continuous extrapolation to predict the future degradation trend of the cell.

[0050] A consistent degradation trajectory is formed by continuous extrapolation using segmented degradation trajectories. These segmented trajectories consist of discrete high-risk segments, with time intervals between each segment corresponding to periods of stable degradation with lower risk. Extrapolation methods are used to connect these discrete segments into a continuous trajectory to fully describe the degradation evolution of the individual unit from the present to the future. The degradation gradient trends of each segment in the segmented degradation trajectory are analyzed to identify whether the gradient is increasing or stabilizing. Piecewise linear or polynomial fitting methods are used to interpolate and fill the gaps between adjacent segments. During interpolation, the gradients of the segments at both ends of the interval are referenced for smooth transitions to avoid unreasonable jumps. The segmented degradation trajectory of cell number 12 in the battery pack contains two high-risk segments. The first segment has an average degradation gradient of 0.10 mΩ / day, corresponding to moderate accelerated degradation, while the second segment has an average degradation gradient of 0.18 mΩ / day, corresponding to severe accelerated degradation, showing a clear increasing trend. Linear interpolation is used to fill the gradient between the two segments, gradually changing it from 0.10 mΩ / day to 0.18 mΩ / day. For the open interval after the second segment, the degradation trend of this cell is predicted six months in advance using a gradient of 0.18 mΩ / day. The interpolation and extrapolation results are then merged with the original segments to form a consistent degradation trajectory. This consistent degradation trajectory covers the complete degradation path from the current moment to the predicted time domain endpoint. Each time point on the trajectory corresponds to a specific estimated degradation level, including the estimated internal resistance and the estimated capacity decay rate.

[0051] Internal resistance deviation characteristics are extracted from the consistent degradation trajectory to generate an internal resistance compensation coefficient. Internal resistance values ​​for each stage are extracted along the time axis of the consistent degradation trajectory. The trajectory is divided into multiple time periods by month. The internal resistance increment between adjacent time periods is calculated and divided by the time interval to obtain the internal resistance growth rate for each stage. The internal resistance growth rate reflects the rate of individual cell degradation. The internal resistance growth rate varies among different cells in the battery pack. Some cells experience a rapid increase in internal resistance due to the shedding of positive electrode active material, with a growth rate potentially more than twice that of normal cells. Other cells experience a slow increase in internal resistance due to slight separator damage, with a growth rate only slightly higher than normal. This difference in rate needs to be compensated for in the equalization current calculation. The internal resistance compensation coefficient is obtained by dividing the internal resistance growth rate of each cell by the average value within the group. The internal resistance compensation coefficient is K_r = ΔR_i / ΔR_avg, where ΔR_i is the internal resistance growth rate of the i-th cell, and ΔR_avg is the average internal resistance growth rate within the group. When the battery pack operates for a long time in a high-temperature computer room environment, some individual cells experience accelerated shedding of positive electrode active material due to poor heat dissipation. The consistent degradation trajectory shows that the internal resistance growth rate of these cells is significantly higher than the average level within the group. The calculated internal resistance compensation coefficient is approximately 1.8 to 2.0, indicating that the degradation rate of these cells is close to twice the average level, requiring increased compensation in the balancing current calculation. On the other hand, some cells have good heat dissipation due to their installation location near the ventilation opening. The consistent degradation trajectory shows that their internal resistance growth rate is close to the average level within the group, and the corresponding internal resistance compensation coefficient is close to 1, indicating that no additional compensation is required.

[0052] The degradation rate is assessed and the equilibrium critical point is determined based on the consistent degradation trajectory and internal resistance compensation coefficient. A single-cell degradation threshold is set as the critical standard for determining whether a single cell needs to be taken out of service. When the internal resistance increases to twice the initial value or the capacity decays to 80% of the nominal value, the degradation threshold is considered reached. Cells exceeding the degradation threshold will severely affect the overall performance and safety of the battery pack. The time point when the degradation degree of each cell first reaches the threshold is searched along the consistent degradation trajectory; this time point is the predicted time when the cell reaches its degradation limit. Under continuous high-current discharge conditions, some cells experience a rapid increase in internal resistance due to simultaneous voltage and capacity degradation. Based on the extrapolation results of the consistent degradation trajectory, it is predicted that this cell will reach the degradation threshold in a relatively short period. After correction with the internal resistance compensation coefficient, the equilibrium critical point is determined to be within two weeks, belonging to the emergency level requiring priority intervention. Some cells, due to their installation location near air conditioning vents and long-term exposure to low temperatures, experience a decrease in electrochemical activity. The consistent degradation trajectory shows that their internal resistance growth gradient is at a moderate level. After correction with the internal resistance compensation coefficient, the equilibrium critical point is determined to be one to two months later, belonging to the attention level requiring regular monitoring. Some monomers experienced continued degradation due to minor diaphragm damage. Their degradation trajectory showed a slightly higher rate than the group average. After adjustment based on the internal resistance compensation coefficient, the equilibrium critical point was determined to be two to three months later. These were considered general and could be included in the routine equilibrium program. The equilibrium critical point indicates the latest time for equilibrium intervention for each monomer. Monomers with larger internal resistance compensation coefficients experienced faster-than-expected degradation, thus requiring an earlier equilibrium critical point. If equilibrium was not implemented after the critical point, the monomers might be too deeply degraded to be restored to a usable state through equilibrium measures.

[0053] A charge balancing association table is established based on the equilibrium threshold and hierarchical equilibrium domains. The equilibrium thresholds of each individual cell are sorted chronologically, with cells having higher equilibrium thresholds having higher equilibrium priority. The hierarchical equilibrium domain number of each cell is queried. Cells within the same hierarchical equilibrium domain can perform balancing operations collaboratively without mutual interference. The battery pack is divided into three hierarchical equilibrium domains. The first domain contains eight cells, including cells numbered 5, 8, 12, etc. Cell number 12 has an equilibrium threshold of 2.7 months, and cell number 8 has an equilibrium threshold of 1.5 months. After sorting by equilibrium threshold, cell number 8 has an equilibrium priority of 1, and cell number 12 has an equilibrium priority of 2. This information is written into the corresponding records in the charge balancing association table. Using the individual unit ID as the primary key, and incorporating the equilibrium critical time, the corresponding hierarchical equilibrium domain ID, the equilibrium priority, and the recommended equilibrium method as fields, a charge balance association table is constructed. The charge balance association table is defined as T=[n,t_c,D_id,P,M], where n is the individual unit ID, t_c is the equilibrium critical time, D_id is the hierarchical equilibrium domain ID, P is the equilibrium priority, and M is the recommended equilibrium method. This charge balance association table cross-links the equilibrium timing in the time dimension with the equilibrium region in the spatial dimension, supporting two equilibrium execution modes: batch scheduling by domain or global scheduling based on time urgency.

[0054] Step S140: Extract the transfer loss rate based on the energy mismatch, generate the energy transfer path based on the transfer loss rate and the internal resistance compensation coefficient, and formulate the basic balance list by combining the energy transfer path and the charge balance correlation table.

[0055] Specifically, the transfer loss rate is extracted based on the energy mismatch. The energy mismatch quantifies the complexity of energy transfer between high- and low-energy cells in a battery pack. The calculation of the energy mismatch involves the current offset and overlap of each transfer channel. From these components, the energy loss characteristics of each channel can be extracted. Analyzing the current offset component of each adjacent pair in the energy mismatch, a larger current offset indicates a more unbalanced current distribution in that channel. The heat loss caused by this unbalanced current during energy transfer in that channel is higher. A current offset of +20% in a channel of the battery pack means that the high-energy cell carries 20% more current than the low-energy cell. When transferring energy from the high-energy cell to the low-energy cell, the additional resistance caused by this current distribution difference must be overcome, resulting in some energy being converted into heat and dissipated. Combining the overlap component of each channel in the energy mismatch, a higher overlap indicates stronger mutual interference between that channel and adjacent channels. More coordination margin needs to be reserved during energy transfer to avoid heat superposition when multiple channels are working simultaneously. The transfer loss rate of each channel is obtained by adding the heat loss caused by current offset to the coordination loss caused by overlapping interference. The transfer loss rate η_loss = α × |D_I| + β × C_overlap, where α is the heat loss coefficient (valued at 0.6), |D_I| is the absolute value of the current offset, β is the coordination loss coefficient (valued at 0.4), and C_overlap is the degree of overlap. The transfer loss rate characterizes the efficiency of different transfer channels in energy transport; channels with lower transfer loss rates have higher energy transfer efficiency and should be preferred.

[0056] In some embodiments, generating an energy transfer path based on the transfer loss rate and the internal resistance compensation coefficient includes: performing spatial gradient analysis based on the transfer loss rate to track transferable channels and generate a candidate transfer group; extracting temperature-sensitive features from the internal resistance compensation coefficient to generate a temperature-sensitive correction factor; sorting the candidate transfer groups by efficiency according to the transfer loss rate and the temperature-sensitive correction factor to obtain a preferred transfer combination; and establishing an energy transfer path based on the preferred transfer combination.

[0057] Spatial gradient analysis based on transfer loss rate is used to track transferable channels and generate candidate transfer groups. The transfer loss rate records the proportion of energy transfer loss between adjacent cell pairs in the battery pack. Due to differences in topological distance and connection method, the transfer loss rate of cell pairs at different locations exhibits a spatially uneven distribution. The spatial gradient of the transfer loss rate is calculated along the physical layout direction of the battery pack. Regions with smaller gradient values ​​indicate that the loss difference between adjacent channels is not significant, and energy can flow smoothly along these regions. When the battery pack adopts a modular design, the cells within the same module are connected by low-resistance copper busbars. The spatial gradient of the transfer loss rate between cells 1 to 4 is only 0.2% / cell, indicating that the transfer efficiency within the module is uniform. However, the transfer loss rate gradient between cells 4 and 5, which crosses the module connector, increases sharply to 3.5% / cell, forming a loss jump zone. Continuous channels with a transfer loss rate spatial gradient below the threshold of 1% / cell are marked as transferable channels. High-energy cells and low-energy cells sharing transferable channels are paired to form candidate transfer groups. Each record in the candidate transfer group contains the source cell number, the target cell number, and the transfer loss rate value of the channel. The candidate transfer group provides a set of candidate channels for subsequent efficiency ranking.

[0058] Temperature-sensitive characteristics are extracted from the internal resistance compensation coefficient to generate a temperature-sensitive correction factor. The value of the internal resistance compensation coefficient fluctuates with the operating temperature of the battery pack. At high temperatures, the electrolyte viscosity decreases, reducing ion migration resistance and causing the internal resistance compensation coefficient to decrease. At low temperatures, the electrochemical reactivity decreases, causing the internal resistance compensation coefficient to increase. This temperature response characteristic needs to be considered during path selection. Analyzing the variation of the internal resistance compensation coefficient in different temperature ranges, the internal resistance compensation coefficient of cell number 8 in the battery pack is 2.0 at 25℃ but drops to 1.7 at 40℃, decreasing by 0.02 for every 1℃ increase in temperature. In contrast, the internal resistance compensation coefficient of cell number 12 only decreases from 1.86 to 1.81 under the same temperature change, decreasing by only 0.003 for every 1℃ increase in temperature. The two cells show a significant difference in their temperature sensitivity. After normalizing the response amplitude of the internal resistance compensation coefficient to temperature changes, a temperature-sensitive correction factor is formed for each individual cell. The temperature-sensitive correction factor F_t = ΔK_r / ΔT, where ΔK_r is the change in the internal resistance compensation coefficient and ΔT is the temperature change. Cells with larger temperature-sensitive correction factors have unstable equilibration effects when the temperature fluctuates and need to be avoided during path selection. Cells with smaller temperature-sensitive correction factors have less equilibration effects affected by temperature and are more suitable as nodes in the transfer path.

[0059] The candidate transfer groups are ranked by efficiency based on transfer loss rate and temperature-sensitive correction factor to obtain the optimal transfer combination. The candidate transfer groups contain multiple feasible energy transfer channels, and the transfer efficiency of different channels varies. Therefore, it is necessary to comprehensively consider losses and temperature stability for optimal selection to ensure stable and efficient execution of the equalization operation under various operating conditions. For each channel in the candidate transfer group, the transfer loss rate of that channel and the temperature-sensitive correction factors of the source and target cells are read, and the comprehensive efficiency index E_eff=(1-η_loss) / (1+F_t_src+F_t_tgt), where η_loss is the transfer loss rate, F_t_src is the temperature-sensitive correction factor of the source cell, and F_t_tgt is the temperature-sensitive correction factor of the target cell. When the battery pack operates in high-temperature environments during summer, although the transfer channel from cell 3 to cell 7 in the alternative transfer group has a lower transfer loss rate, cell 3 is installed close to heat-generating components, resulting in a higher temperature-sensitive correction factor. When the ambient temperature rises, the actual transfer efficiency of this channel will decrease significantly, leading to a lower overall efficiency. In contrast, although the transfer loss rate of the channel from cell 5 to cell 9 in the alternative transfer group is slightly higher, the cells at both ends are installed in well-ventilated areas, resulting in a lower temperature-sensitive correction factor. Even with fluctuations in ambient temperature, the transfer efficiency of this channel remains stable, leading to a higher overall efficiency. The latter channel has a higher overall efficiency and should be prioritized. All channels in the alternative transfer group are ranked in descending order of overall efficiency. The top-ranked channels are extracted to form the preferred transfer combination, which balances low loss and high temperature stability.

[0060] An energy transfer path is established based on the optimal transfer combination. The optimal transfer combination selects the transfer channels with the highest efficiency and best temperature stability. These discrete channels need to be connected into a complete transfer path to achieve energy transport across multiple cells. The distribution of source and target cells in each channel of the optimal transfer combination is analyzed to identify tandemable channel sequences, where the target cell of the previous channel is exactly the source cell of the next channel. Multiple short channels are connected end-to-end to form a longer path with wider coverage. The battery pack contains 24 cells. The optimal transfer combination has three high-efficiency channels: from cell 1 to cell 4, from cell 4 to cell 7, and from cell 7 to cell 10. Adjacent channels have the same cell numbering. Connecting these three channels in series forms a complete energy transfer path from cell 1 to cell 10. Energy can be transferred step-by-step along this path from the high-energy cell 1 to the low-energy cell 10. The total path length is 9 cell spacings, and the total path loss is approximately 12% (the sum of the losses of the three channels). Check the series-connected paths for loops or dead ends, and eliminate abnormal branches that cannot effectively transfer energy to ensure that each energy transfer path is a unidirectional connected structure. Organize all valid transfer paths into an energy transfer path set by the starting unit number. Each energy transfer path records the sequence of units it passes through, the transfer loss rate of each segment, and the total path loss.

[0061] A basic balancing list is developed by combining the energy transfer path and the charge balance correlation table. For each energy transfer path, the records of the starting and ending cells in the charge balance correlation table are queried. The balancing priority and critical point time of the cells at both ends are read. For a certain energy transfer path of the battery pack, the starting cell is cell 3, which has a priority of 2 in the charge balance correlation table and a critical point of 6 weeks later. The ending cell is cell 7, which has a priority of 1 in the charge balance correlation table and a critical point of 2 weeks later. The higher priority and closer critical point of the ending cell indicate that the path should be started within 2 weeks to ensure that the ending cell receives timely energy replenishment. The recommended transfer amount is calculated based on the difference in the state of charge of the cells at both ends of the energy transfer path. The state of charge of cell 3 in the battery pack is 85%, while that of cell 7 is 62%, a difference of 23%. Considering the total loss of 12% in the energy transfer path, the recommended transfer amount is set to reduce the difference in the state of charge of the two ends to within 5%, i.e., transfer about 10% of the nominal capacity. The transfer amount does not exceed the carrying capacity of the path to avoid overcurrent. The path number, starting unit, ending unit, recommended transfer amount, execution priority, and suggested start time are organized into a basic balancing list. Each record in the basic balancing list corresponds to a specific balancing operation task with complete execution parameters. The basic balancing list provides a list of directly executable tasks for balancing scheduling.

[0062] Step S150: Use the basic balance list to carry out balance priority allocation to form a balance order table. Determine the balance start threshold based on the balance order table and the balance critical position. Perform step verification on the balance order table according to the balance start threshold and output the balance execution command.

[0063] In some embodiments, the step of using the basic equilibrium list to allocate equilibrium priorities and form an equilibrium order table includes: sorting the basic equilibrium list according to the degree of imbalance to establish a degree ranking table; assessing the heat accumulation risk of continuous equilibrium in the same region based on the degree ranking table to generate heat constraint markers; performing adaptive segmentation on the degree ranking table based on the heat constraint markers to generate priority splitting points; and determining the equilibrium order table based on the priority splitting points.

[0064] A ranking table is established by sorting the basic balancing list according to the degree of imbalance. The basic balancing list records the recommended transfer amount information for each balancing task. The larger the transfer amount, the greater the difference in state of charge between the individual cells involved in the task and the average level of the group, and the more severe the imbalance. If tasks with severe imbalance are not handled in time, the target cells may reach the discharge cutoff state first, affecting the usable capacity of the entire group. The recommended transfer amount of each task in the basic balancing list is extracted as an imbalance degree index. The larger the imbalance degree index, the higher the priority of the task. The basic balancing list of the battery pack contains 12 task records. Among them, the recommended transfer amount of task 3 is 15% of the nominal capacity, corresponding to the imbalance degree index 15. The recommended transfer amount of task 7 is only 2% of the nominal capacity, corresponding to the imbalance degree index 2. The imbalance degree of task 3 is much higher than that of task 7, and it should receive a higher execution priority. All tasks in the basic balance list are sorted in descending order of imbalance degree index to form a degree ranking table. The degree ranking table places the most severely imbalanced tasks at the top and the less severely imbalanced tasks at the bottom. Each record in the degree ranking table includes task number, involved entities, imbalance degree index, and recommended transfer amount, providing an orderly task sequence for subsequent assessment of heat accumulation risk and division of priority intervals.

[0065] A thermal constraint marker is generated by assessing the risk of continuous heat accumulation in the same area based on a severity ranking table. The severity ranking table arranges all tasks to be executed according to their degree of imbalance. If multiple adjacent tasks involve cells in the same physical area of ​​the battery pack, executing these tasks consecutively will lead to continuous heat accumulation in that area, posing a risk of exceeding temperature limits. Analyzing the physical location relationships of cells involved in adjacent tasks in the severity ranking table identifies consecutive task sequences located in the same module or adjacent slots. When the battery pack adopts a compact layout, the module spacing is only 5mm, resulting in a short heat conduction path, making it easy for heat generated by continuous balancing to accumulate in localized areas. The five tasks ranked 3rd to 7th in the severity ranking table all involve cells within module 2. If these five tasks are executed sequentially according to the severity ranking table, each task generates approximately 3W of balancing heat dissipation. Executing these five tasks consecutively will accumulate 15W of heat within module 2. Based on the module's thermal resistance, the temperature rise could reach 20°C, and the internal temperature of the module could climb from 30°C to 50°C, exceeding the safety threshold of 45°C. For continuous task sequences with heat accumulation risk, the heat accumulation value generated by continuous equilibrium is calculated and compared with the safety threshold. For sequences that exceed the threshold, a heat constraint mark is generated. The heat constraint mark records the constrained task number range of number 3 to 7, the estimated heat accumulation value of 120% of the safety threshold, and the suggested cooling interval of 15 minutes. The heat constraint mark provides a constraint basis for subsequent segmentation.

[0066] For example, the step of adaptively segmenting the degree ranking table based on the thermal constraint markers to generate priority split points includes: extracting differential features from the thermal constraint markers to obtain a differential marker sequence; adaptively segmenting the degree ranking table based on the differential marker sequence to generate candidate boundary positions; performing boundary stability analysis on the candidate boundary positions to screen effective boundaries; and summarizing the effective boundaries to form priority split points.

[0067] A graded feature extraction process is performed on the thermal constraint markers to obtain the graded marker sequence. The thermal constraint markers record the cumulative heat value and suggested cooling interval for each constrained task sequence. Different constraint sequences exhibit varying levels of risk, requiring graded feature extraction to determine the priority of segmentation. Segmentation points with larger risk grade differences should be prioritized to effectively isolate high-risk and low-risk tasks. The difference in cumulative heat value between adjacent constraint records in the thermal constraint markers is calculated as the graded value. The sign of the graded value reflects the direction of risk change, and the absolute value reflects the magnitude of risk change. The battery pack's thermal constraint markers contain three constraint records: the first corresponds to module 2 with a cumulative heat value of 120% of the safety threshold; the second corresponds to module 4 with a cumulative heat value of 135% of the safety threshold; and the third corresponds to module 6 with a cumulative heat value of 110% of the safety threshold. A graded difference of +15% between the first and second records indicates an increasing risk from module 2 to module 4, while a graded difference of -25% between the second and third records indicates a significant decrease in risk from module 4 to module 6. Arrange the difference values ​​of each adjacent constraint record in order to form a difference mark sequence. The difference mark sequence reveals the changing trend of risk distribution in the thermal constraint mark. Positions with larger absolute difference values ​​in the difference mark sequence, such as -25%, are often candidate points for priority segmentation. Segmentation at these positions can effectively isolate task groups with significant differences in risk characteristics.

[0068] Candidate cut-off points are generated by adaptively segmenting the severity ranking table based on the differential marker sequence. The differential marker sequence identifies locations in the thermal constraint markers where risk changes drastically. These locations correspond to potential cut-off points in the severity ranking table. Breaking the task sequence at these cut-off points can block the transmission of heat accumulation and insert cooling intervals. Locations in the differential marker sequence where the absolute value of the differential exceeds the 20% threshold are scanned and mapped to their corresponding indices in the severity ranking table. During mapping, the constraint sequence numbers in the thermal constraint markers need to be converted to task indices in the severity ranking table. The battery pack's differential marker sequence shows a significant differential of -25% exceeding the 20% threshold at position 2. This position corresponds to the end of the constraint sequence for module 4, i.e., after task 9 in the severity ranking table. This indicates a significant risk discontinuity between task 9 and task 10, making it a suitable cut-off point. Segmenting at this point can isolate tasks involving the high-risk module 4 from those involving the low-risk module 6. In addition to the split points triggered by the graded marker sequence, natural breaks in the imbalance index of the severity ranking table also need to be checked. The imbalance index of task 12 in the battery pack is 6%, while that of task 13 is 2%, a difference of 4%, which is more than 1.5 times the overall standard deviation of 2.5%. A split point is added at this position to distinguish tasks with different urgency levels. The index positions of all potential split points are summarized to form a candidate split point list. Each element in the candidate split point list records the index of the split point, the triggering reason, and the split strength index.

[0069] Stability analysis was conducted on candidate boundary points to screen for effective boundaries. Some candidate boundary points may be unstable and triggered by edge conditions. These boundary points may fail when operating conditions change slightly, leading to frequent fluctuations in priority allocation and scheduling chaos. Stability analysis is needed to screen these points and retain truly reliable ones. For each candidate boundary point, three tasks before and after it were selected to form a local window. The fluctuation of the imbalance index and heat accumulation value within the window was analyzed. The fluctuation amplitude reflects the uncertainty level of the data in that area. The imbalance level at position 12 in the battery pack candidate boundary point is 4%, but the imbalance level of the three tasks before and after this boundary point exhibits a random fluctuation of ±3%. The 4% difference and the 3% fluctuation amplitude are close, resulting in a stability index of only 1.3 for this boundary point, slightly higher than the threshold of 1.0. It is considered marginally stable and should be used with caution. However, the imbalance level at position 9 in the candidate boundary point is 8%, and the fluctuation amplitude of the tasks before and after it is only ±1.5%. The stability index reaches 5.3, far exceeding the threshold, and it is considered an effective boundary. The ratio of the difference between each candidate boundary position and the local fluctuation amplitude is calculated as a stability index. Boundary points with a stability index greater than the set threshold of 2.0 are determined to be effective boundaries with sufficient distinguishability. Boundary points with a stability index lower than the threshold are eliminated. After screening, the battery pack candidate boundary positions No. 9 and No. 17 are retained as two effective boundaries.

[0070] Priority dividing points are formed by summarizing effective boundaries. The number of effective boundaries may decrease after stability screening. It's necessary to check if the remaining boundaries are sufficient to divide the priority ranking table into reasonable intervals. Too few intervals will lead to too many tasks within a batch, increasing the risk of heat accumulation; too large interval spans will result in significant differences in the urgency of tasks within the same batch, affecting scheduling efficiency. The number and distribution of effective boundaries are statistically analyzed. If too few effective boundaries lead to excessively large interval spans, auxiliary boundaries are added within those intervals. The location of the auxiliary boundaries is selected where the imbalance index changes most gradually within the interval to ensure stability. For the battery pack, there are only two effective boundaries: positions 9 and 17. Between positions 9 and 17, there are eight tasks, and the imbalance index of these eight tasks gradually changes from 11% to 6%, a large range that is not conducive to refined scheduling. In the middle of this interval, position 13 has the lowest rate of change in the imbalance index, only 0.3% / task. An auxiliary boundary is added here to further divide this interval into two sub-intervals. The effective and supplementary auxiliary boundary points are arranged in ascending order of index position to form the final priority division point set. The priority division points of the battery pack include three positions: No. 9, No. 13, and No. 17. After the priority division points are determined, the priority ranking table is divided into four priority intervals with clear boundaries. The degree of task imbalance in each interval is similar and the risk of heat accumulation is controllable.

[0071] The balancing order table is determined based on priority split points. These points divide the severity ranking table into multiple priority intervals. Tasks within each interval have similar levels of imbalance and no thermal accumulation conflicts, allowing them to be scheduled as a single balancing batch. Cooling intervals are inserted between different batches to ensure thermal safety. The severity ranking table is further divided into sub-tables based on the location of the priority split points. Each sub-table is assigned a priority number; sub-tables with lower priority numbers require priority execution, and batches with lower priority numbers enjoy higher scheduling weight during resource balancing. The battery pack's severity ranking table is divided into four intervals by three priority split points. Interval 1 contains tasks 1 through 9 (the nine most severely imbalanced tasks) and is assigned priority 1. Interval 2 contains tasks 10 through 13 and is assigned priority 2. Interval 3 contains tasks 14 through 17 and is assigned priority 3. Interval 4 contains tasks 18 and onwards and is assigned priority 4. Tasks within each interval are arranged in descending order of imbalance to determine the execution order within a batch. The task lists and priority numbers of each interval are organized into a balance order table. Each record in the balance order table includes the task number, priority, imbalance index, and execution order within the batch. The balance order table supports batch execution of the balance operation according to priority. Tasks within the same batch can be executed in parallel or in rapid rotation to improve balance efficiency. A 15-minute cooling interval is inserted between different batches to ensure that the module temperature drops to a safe range before starting the next batch.

[0072] In some embodiments, determining the equilibrium start threshold based on the equilibrium order table and the equilibrium critical position includes: parsing the imbalance intensity range of each priority based on the equilibrium order table; extracting priority distribution features from the equilibrium order table to generate a charge sensitivity coefficient; constructing an adaptive boundary matrix based on the imbalance intensity range and the charge sensitivity coefficient; and determining the equilibrium start threshold by boundary calibration based on the adaptive boundary matrix and the equilibrium critical position.

[0073] The imbalance intensity range of each priority level is analyzed based on the balancing priority table. The balancing priority table divides the tasks to be executed into several batches according to priority. The imbalance degree of tasks within each priority batch falls within a specific range. It is necessary to clearly define the imbalance degree boundaries of each priority level so that the activation conditions can be dynamically adjusted to match the response sensitivity of the balancing system with the actual state of the battery pack. The maximum and minimum values ​​of the imbalance degree index for all tasks within each priority batch in the balancing priority table are statistically analyzed. The interval formed by the maximum and minimum values ​​is defined as the imbalance intensity range of that priority level. The battery pack's balancing priority table contains four priorities. Priority 1 contains nine tasks, with a maximum imbalance degree index of 15% and a minimum of 11%, thus its imbalance intensity range is 11% to 15%. Priority 2's imbalance intensity range is 8% to 10%, Priority 3's is 5% to 7%, and Priority 4's is 2% to 4%. The imbalance intensity ranges of each priority level do not overlap, forming a stepped distribution that facilitates clear definition of task assignment. The imbalance intensity ranges of each priority are organized into an interval list, which serves as the basic input for constructing the adaptive boundary matrix. The imbalance intensity range clarifies the entry threshold for each priority. When the imbalance degree of a task falls into a specific range, it belongs to the corresponding priority.

[0074] Priority distribution characteristics are extracted from the equilibrium order table to generate a charge sensitivity coefficient. The proportion of tasks in each priority batch in the equilibrium order table is statistically analyzed. An excessive number of high-priority batches indicates a severe overall imbalance in the battery bank, requiring improved response sensitivity in the equilibrium scheduling process to intervene early and prevent further deterioration. The ratio of the proportion of high-priority batches to the proportion of low-priority batches is calculated as a distribution skewness index. A larger distribution skewness index indicates a greater concentration of tasks in high-priority batches and a more severe imbalance in the battery bank. In the equilibrium order table of the battery bank, priorities 1 and 2 together contain 13 tasks, accounting for 65% of the total 20 tasks, while priorities 3 and 4 contain only 7 tasks, accounting for 35%. The distribution skewness index is 65% / 35% = 1.86, indicating that most tasks are concentrated in high-priority batches. The charge sensitivity coefficient S_q = B_skew × σ_imb is obtained by multiplying the distribution skew index by the standard deviation of the imbalance index in the equilibrium order table, where B_skew is the distribution skew index (1.86) and σ_imb is the standard deviation of the imbalance index (4.2%). The calculated charge sensitivity coefficient is 7.8, which is considered relatively high. A higher charge sensitivity coefficient indicates that the equilibrium start threshold needs to be lowered accordingly to accelerate the equilibrium response speed.

[0075] An adaptive boundary matrix is ​​constructed based on the imbalance intensity range and charge sensitivity coefficient. The basic framework of the adaptive boundary matrix is ​​built using priority number as the row index and boundary type as the column index. The boundary type includes two columns: lower boundary and upper boundary. The imbalance intensity range of each priority is filled into the corresponding row, with the lower boundary being the minimum value of the imbalance intensity range and the upper boundary being the maximum value, forming the initial static boundary configuration. The boundary values ​​are dynamically adjusted according to the charge sensitivity coefficient, enabling the balancing system to automatically adjust its response sensitivity based on the real-time status of the battery pack. A high charge sensitivity coefficient indicates severe imbalance, requiring a reduction in the lower boundary of each priority to grant more tasks higher priority and accelerate processing. A low charge sensitivity coefficient indicates slight imbalance, allowing for a slight increase in the lower boundary to avoid over-balancing. The battery pack's charge sensitivity coefficient is 7.8, which is relatively high. The lower boundary of each priority is lowered by 10%, with the lower boundary of priority 1 adjusted from 11% to 9.9%, elevating some tasks originally belonging to priority 2 to priority 1, allowing more tasks to be executed first and accelerating the balancing process. The corrected boundary values ​​are filled into the adaptive boundary matrix A=[P_id,L_low,L_up], where P_id is the priority number, L_low is the corrected lower boundary, and L_up is the corrected upper boundary. The adaptive boundary matrix provides a dynamically adjusted priority interval definition for subsequent boundary calibration.

[0076] Boundary calibration based on the adaptive boundary matrix and equilibrium critical points determines the equilibrium start threshold. For each priority in the adaptive boundary matrix, the time distribution of the equilibrium critical points of the individual units involved in the corresponding task is read. The time difference between the earliest equilibrium critical point and the current time is calculated as an urgency index. The smaller the urgency index, the more urgently the task of that priority needs to be executed, and more lenient start conditions should be adopted to ensure timely triggering. In the battery pack adaptive boundary matrix, the earliest equilibrium critical point of the nine tasks corresponding to priority 1 is in two weeks, and the urgency index of two weeks indicates an emergency state. The earliest equilibrium critical point of the task corresponding to priority 2 is in two months, and the urgency index of eight weeks indicates a medium state. The urgency of priority 1 is much higher than that of priority 2, requiring a lower imbalance threshold and a larger time lead threshold. The urgency index is jointly encoded with the boundary values ​​in the adaptive boundary matrix to form the balancing start threshold T_th=[P_id,L_th,Δt_th] for each priority, where P_id is the priority number, L_th is the imbalance degree threshold taken as 80% of the lower boundary in the adaptive boundary matrix, and Δt_th is the time advance threshold taken as 50% of the urgency index. The balancing start threshold for battery pack priority 1 is triggered when the imbalance degree exceeds 7.9% or is less than 1 week away from the balancing critical position.

[0077] The balancing sequence table is validated stepwise according to the balancing start threshold, and the balancing execution command is output. Tasks in the balancing sequence table are checked from highest to lowest priority to see if they meet the trigger conditions for the balancing start threshold. The trigger conditions include two thresholds: the imbalance degree threshold and the time lead threshold. If either condition is met, the task is determined to require balancing operation. The battery pack is currently only one week away from the earliest balancing critical point of priority 1 task in the balancing sequence table, having already reached the time lead threshold of the balancing start threshold. All tasks of priority 1 meet the start conditions and should be allocated balancing resources immediately. However, the earliest balancing critical point of priority 2 task is far from being reached in eight weeks and will not be started for now. For tasks that meet the start conditions, the sequence of cells along the energy transfer path and the transfer loss rate of each segment are read. Combined with the recommended transfer amount, the balancing current and duration are calculated. For task 3 in priority 1 of the battery pack, the recommended transfer amount is 7.5Ah, the path carrying capacity is 2A, and the calculated duration is 4 hours. The task number, target cell sequence, balancing direction, balancing current, and duration are encapsulated into a balancing execution command C_exec=[T_id,N_seq,Dir,I_bal,Δt_dur], where T_id is the task number, N_seq is the target cell sequence, Dir is the balancing direction, I_bal is the balancing current, and Δt_dur is the duration. After the balancing execution command is sent to the balancing hardware module, the energy transfer operation is initiated, transferring energy from cells with high state of charge to cells with low state of charge to gradually eliminate the state of charge differences within the battery pack.

[0078] To implement the above-described method embodiments, a battery online balancing method is provided to achieve the corresponding functions and technical effects. See also... Figure 2 , Figure 2 This diagram illustrates a structural block diagram of an online battery balancing system 200 according to an embodiment of this application. For ease of explanation, only the parts relevant to this embodiment are shown. The online battery balancing system 200 provided in this embodiment includes:

[0079] Data acquisition module 201 is used to acquire voltage discrete characteristics, capacity deviation characteristics and outlier cell data of the battery pack under the operating state, and to perform consistency offset verification on the voltage discrete characteristics and the capacity deviation characteristics to generate a cell state matrix.

[0080] Mismatch analysis module 202 is used to perform inconsistent clustering analysis using the monomer state matrix to identify high-energy monomers and low-energy monomers, analyze the energy mismatch degree based on the positional distribution of the high-energy monomers and the low-energy monomers, and delineate a hierarchical equilibrium domain based on the energy mismatch degree.

[0081] Trend analysis module 203 is used to perform trend analysis on the outlier data to form a consistent degradation trajectory, extract internal resistance deviation characteristics from the consistent degradation trajectory to generate an internal resistance compensation coefficient, evaluate the degradation rate based on the consistent degradation trajectory and the internal resistance compensation coefficient to determine the equilibrium critical position, and establish a charge balance correlation table according to the equilibrium critical position and the hierarchical equilibrium domain.

[0082] Energy transfer module 204 is used to extract the transfer loss rate based on the energy mismatch, generate an energy transfer path based on the transfer loss rate and the internal resistance compensation coefficient, and formulate a basic balance list by combining the energy transfer path with the charge balance association table.

[0083] The equalization output module 205 is used to perform equalization priority allocation using the basic equalization list to form an equalization order table, determine the equalization start threshold based on the equalization order table and the equalization critical bit, and perform step verification on the equalization order table according to the equalization start threshold to output the equalization execution command.

[0084] The aforementioned online battery balancing system 200 can implement one of the online battery balancing methods described in the above-described method embodiments. The options described in the above method embodiments are also applicable to this embodiment and will not be detailed here. The remaining content of this application's embodiments can be referred to the content of the above method embodiments, and will not be repeated in this embodiment.

[0085] The above embodiments are not an exhaustive list based on the present invention, and there may be many other embodiments not listed. Any substitutions and improvements made without departing from the concept of the present invention are within the protection scope of the present invention.

Claims

1. An on-line battery equalization method, characterized by, The method comprises the following steps: obtaining voltage discrete features, capacity deviation features and outlier cell data in the running state of the battery pack, performing consistency offset checking on the voltage discrete features and the capacity deviation features to generate a cell state matrix, including: performing polarization voltage stripping on the voltage discrete features to obtain stable discrete components; forming a deviation classification spectrum based on the stable discrete components according to deviation amplitudes; performing time sequence correlation analysis on the deviation classification spectrum and the capacity deviation features to obtain an offset checking factor; and constructing a cell state matrix using the offset checking factor; performing inconsistent clustering analysis on the cell state matrix to identify high-energy cells and low-energy cells, analyzing energy mismatch degree according to the position distribution of the high-energy cells and the low-energy cells, and dividing hierarchical equalization domains based on the energy mismatch degree; performing trend analysis on the outlier cell data to form a consistent degradation trajectory, including: performing distributed clustering mining on the outlier cell data to establish a time sequence aggregation zone; identifying jump abnormalities and normal sections based on the time sequence aggregation zone to generate a jump marker set and a normal degradation section, respectively; performing gradient risk screening on the jump marker set and the normal degradation section to generate a segmented degradation trajectory; continuously extrapolating the segmented degradation trajectory to form a consistent degradation trajectory; extracting internal resistance deviation features from the consistent degradation trajectory to generate an internal resistance compensation coefficient, performing degradation rate evaluation based on the consistent degradation trajectory and the internal resistance compensation coefficient to determine an equalization critical position, and establishing a charge equalization correlation table according to the equalization critical position and the hierarchical equalization domains; extracting a transfer loss rate based on the energy mismatch degree, generating an energy transfer path according to the transfer loss rate and the internal resistance compensation coefficient, and formulating a basic equalization list combining the energy transfer path and the charge equalization correlation table; performing equalization priority allocation using the basic equalization list to form an equalization order table, determining an equalization start threshold based on the equalization order table and the equalization critical position, and performing ladder verification on the equalization order table according to the equalization start threshold to output an equalization execution command.

2. The method of claim 1, wherein, The energy mismatch degree is analyzed according to the position distribution of the high-energy cells and the low-energy cells, including: performing topological proximity analysis on the high-energy cells and the low-energy cells to locate adjacent positions and establish an adjacent position table; obtaining the charge and discharge current distribution ratio of the high-energy cells and the low-energy cells in the adjacent position table to generate a current offset degree; based on the current offset degree and the adjacent position table, filtering positions with overlapping influence domains to form an overlap identifier; determining the energy mismatch degree according to the overlap degree of the overlap identifier and the current offset degree.

3. The method of claim 1, wherein, The energy transfer path is generated according to the transfer loss rate and the internal resistance compensation coefficient, including: performing spatial gradient analysis based on the transfer loss rate to track transferable channels and generate an alternative transfer group; extracting temperature sensitive features from the internal resistance compensation coefficient to generate a temperature sensitive correction factor; performing efficiency sorting on the alternative transfer group according to the transfer loss rate and the temperature sensitive correction factor to obtain an optimal transfer combination; establishing an energy transfer path according to the optimal transfer combination.

4. The method of claim 1, wherein, The equalization priority allocation based on the basic equalization list forms an equalization order table, including: The basic equalization list is sorted according to the imbalance degree to establish a degree order table; According to the degree order table, the thermal accumulation risk of continuous equalization in the same area is evaluated to generate a thermal constraint mark; Based on the thermal constraint mark, adaptive segmentation is carried out on the degree order table to generate a priority cutting point; According to the priority cutting point, an equalization order table is determined.

5. The method of claim 1, wherein, The equalization start threshold is determined based on the equalization order table and the equalization critical position, including: Relying on the equalization order table, the imbalance intensity range of each priority is analyzed; From the equalization order table, the priority distribution characteristics are extracted to generate a charge sensitivity coefficient; Based on the imbalance intensity range and the charge sensitivity coefficient, an adaptive boundary matrix is constructed; Based on the adaptive boundary matrix and the equalization critical position, the boundary is calibrated to determine the equalization start threshold.

6. The method of claim 1, wherein, The gradient risk screening of the jump mark set and the normal degradation section generates a segmented degradation trajectory, including: The distribution gradient of the jump mark set and the normal degradation section is detected to establish a difference coupling zone; In the difference coupling zone, the slope mutation position is identified to generate a mutation early warning point; The mutation early warning point is subjected to accelerated degradation risk assessment to generate a risk slope order table; According to the risk slope order table, high-risk slopes are screened to form a segmented degradation trajectory.

7. The method of claim 4, wherein, The adaptive segmentation of the degree order table based on the thermal constraint mark generates a priority cutting point, including: The differential characteristic extraction of the thermal constraint mark obtains a differential mark sequence; Based on the differential mark sequence, the degree order table is adaptively segmented to generate a candidate boundary position; The candidate boundary position is subjected to boundary stability analysis to screen an effective boundary; With the help of the effective boundary, the priority cutting point is summarized.

8. An on-line battery equalization system, characterized by, Including: The data acquisition module is used to acquire the voltage dispersion characteristics, capacity deviation characteristics and outlier cell data of the battery pack in the running state, and the consistency offset check is performed on the voltage dispersion characteristics and the capacity deviation characteristics to generate a cell state matrix, including: The polarization voltage of the voltage dispersion characteristics is stripped to obtain a steady-state dispersion component; Based on the steady-state dispersion component, the deviation classification spectrum is formed by classifying the deviation amplitude; The time correlation analysis of the deviation classification spectrum and the capacity deviation characteristics is performed to obtain an offset check factor; The offset check factor is used to construct a cell state matrix; The mismatch analysis module is used to identify high-energy cells and low-energy cells by means of the cell state matrix, analyze the energy mismatch degree according to the position distribution of the high-energy cells and the low-energy cells, and divide the hierarchical equalization domain based on the energy mismatch degree. The trend analysis module is used for carrying out trend analysis on the outlier monomer data to form a consistency degradation track, including: carrying out distributed cluster mining on the outlier monomer data to establish a time sequence aggregation zone; identifying a jump abnormality and a normal section based on the time sequence aggregation zone to generate a jump mark set and a normal degradation section respectively; carrying out gradient risk screening on the jump mark set and the normal degradation section to generate a segmented degradation track; continuously extrapolating by means of the segmented degradation track to form a consistency degradation track; extracting an internal resistance deviation feature from the consistency degradation track to generate an internal resistance compensation coefficient, carrying out degradation rate evaluation based on the consistency degradation track and the internal resistance compensation coefficient to determine an equilibrium critical position, and establishing a charge equalization association table in accordance with the equilibrium critical position and the layered equilibrium domain; The energy transfer module is used for extracting a transfer loss rate based on the energy mismatch degree, generating an energy transfer path according to the transfer loss rate and the internal resistance compensation coefficient, and formulating a basic equalization list in combination with the energy transfer path and the charge equalization association table; The equalization output module is used for carrying out equalization priority allocation by using the basic equalization list to form an equalization order table, determining an equalization start threshold based on the equalization order table and the equilibrium critical position, and performing ladder verification on the equalization order table in accordance with the equalization start threshold to output an equalization execution command.

Images