Lithium battery pack online feeding and sorting dynamic compensation system based on digital twinning

Abstract

The application discloses a lithium battery PACK online feeding sorting dynamic compensation system based on digital twinning, belongs to the field of lithium battery PACK manufacturing, and comprises a battery cell deviation feature acquisition module, an assembly position compensation feature construction module, a dynamic steady-state adaptation degree calculation module and a feeding compensation decision module; by constructing a dynamic matching relationship between a battery cell deviation demand value and an assembly position equivalent compensation capacity value, and determining a target assembly position and a feeding sequence of a battery cell to be fed based on a dynamic steady-state adaptation degree, the initial difference of the battery cell can be jointly absorbed by the balancing capacity, the thermal boundary capacity and the connection path feature of the assembly position in the PACK operation process, so that the problem that the existing lithium battery PACK online feeding sorting can only be graded according to the static consistency before assembly and it is difficult to ensure the operation consistency after assembly is solved.

Description

Technical Field

[0001] This invention relates to the field of lithium battery PACK manufacturing technology, and more specifically, to a dynamic compensation system for material sorting on a lithium battery PACK production line based on digital twins. Background Technology

[0002] In the lithium battery pack assembly process, cell sorting is a crucial step affecting pack consistency and operational stability. Existing online sorting methods typically use initial static parameters such as open-circuit voltage, AC internal resistance, and capacity as the primary criteria, grouping cells with similar parameters into the same category before proceeding to the module or pack assembly process. While this method can reduce initial differences in cell conditions to some extent, it implicitly assumes that the closer the cells' parameters are before assembly, the better their operational consistency after assembly.

[0003] However, in actual PACK structures, different assembly positions are not entirely equivalent. Due to differences in the distance between each assembly position and the heat dissipation boundary, the bus connection path, the soldering contact state, the current stress, and the BMS balancing branch capability, even cells with similar parameters before assembly may exhibit different voltage drift, temperature rise response, branch impedance changes, and balancing recovery speeds during operation when assigned to the same PACK. In other words, the differences in the state of the assembled cells are not only determined by the cell itself but also influenced by the thermal boundary, connection path, soldering contact, and balancing capability of the assembly position.

[0004] While existing technologies include cell sorting solutions based on digital detection or digital twin models, most still focus on detecting, modeling, and classifying the current state of the cells. Their sorting objectives are mainly to reduce the initial parameter differences before cell assembly. They have not fully considered the absorption capacity of different assembly positions of the target PACK for the initial deviation of the cells, nor have they matched the cell deviation with the compensation capacity of the assembly position before determining the loading position.

[0005] Therefore, the existing sorting method is prone to the following problems: some cells with similar parameters before assembly will have their operating deviation amplified after entering a position with weak thermal boundary, high connection path impedance or insufficient equalization capability; while some cells with slight differences before assembly may maintain better convergence during operation if they are assigned to a position with strong compensation capability. However, the existing sorting method is difficult to identify and utilize this position compensation relationship. Summary of the Invention

[0006] To address the problems mentioned in the background section, the present invention provides the following technical solution: A dynamic compensation system for material sorting on a lithium battery PACK line based on digital twins includes a cell deviation feature acquisition module, an assembly position compensation feature construction module, a dynamic steady-state adaptability calculation module, and a material loading compensation decision module. The cell deviation feature acquisition module is used to acquire the voltage deviation, internal resistance deviation, capacity deviation and dynamic response deviation of the cell to be loaded, and generate the cell deviation requirement value of the cell to be loaded based on the above deviations. The assembly position compensation feature construction module is used to obtain the equalization capability, thermal boundary capability, connection path impedance and current stress weight corresponding to each assembly position in the target PACK based on the digital twin model of the target PACK, and generate the assembly position equivalent compensation capability value of each assembly position according to the obtained parameters. The dynamic steady-state adaptability calculation module is used to match the cell deviation requirement value with the equivalent compensation capability value of each assembly position, predict the degree of convergence of the running deviation after the cell to be loaded is assigned to each assembly position within a preset running time window, and generate the dynamic steady-state adaptability between the cell to be loaded and each assembly position. The loading compensation decision module is used to determine the target assembly position and loading sequence of the battery cell to be loaded based on the dynamic steady-state adaptability, so that the initial deviation of the battery cell to be loaded is jointly compensated by the equalization capability, thermal boundary capability and connection path impedance of the corresponding assembly position during the operation of the target PACK. Furthermore, the cell deviation feature acquisition module includes a reference feature determination unit, a detection feature extraction unit, and a deviation requirement value generation unit; The reference feature determination unit is used to generate batch reference features based on the voltage distribution, internal resistance distribution, capacity distribution and dynamic response distribution of the qualified cells in the current assembly batch of the target PACK. The detection feature extraction unit is used to obtain the open circuit voltage, internal resistance, capacity and dynamic response parameters of the battery cell to be loaded. The dynamic response parameters include at least two of the following: voltage recovery slope obtained from pulse charge-discharge test, temperature rise slope obtained from short-time current excitation test, self-discharge rate obtained from the open circuit voltage difference before and after resting, and capacity decay slope obtained from historical cycle data fitting. The deviation requirement value generation unit is used to calculate the normalized deviation of the open circuit voltage, internal resistance, capacity and dynamic response parameters of the cell to be loaded with the batch reference characteristics, and to fuse the normalized deviation according to the weight corresponding to the target PACK operating condition to generate the cell deviation requirement value.

[0007] Furthermore, the assembly position compensation feature construction module includes a position feature extraction unit, a compensation capability calculation unit, and a position capability correction unit; The location feature extraction unit is used to extract the upper limit of equalization current, allowable equalization time, heat dissipation boundary parameters, connection path impedance, predicted welding contact impedance and current stress weight corresponding to each assembly position in the target PACK based on the digital twin model of the target PACK. The compensation capability calculation unit is used to generate an active compensation capability value for the corresponding assembly position based on the upper limit of the equalization current, the allowable equalization time and the heat dissipation boundary parameters, and to generate a passive compensation constraint value for the corresponding assembly position based on the connection path impedance, the predicted welding contact impedance and the current stress weight. The position capability correction unit is used to fuse the active compensation capability value and the passive compensation constraint value to generate the assembly position equivalent compensation capability value for each assembly position, so that the degree of compensation for the initial deviation of the battery cell to be loaded at different assembly positions is quantified.

[0008] Furthermore, the dynamic steady-state fitness calculation module includes a difference mapping unit, a running deviation prediction unit, and a fitness generation unit; The difference mapping unit is used to match the cell deviation requirement value with the equivalent compensation capability value of each assembly position to generate the difference compensation relationship between the cell to be loaded and each assembly position. The operation deviation prediction unit is used to predict, based on the difference compensation relationship, the cumulative amount of voltage deviation, cumulative amount of temperature deviation, cumulative amount of equivalent branch impedance deviation and cumulative amount of attenuation trend deviation formed within a preset operation time window after the battery cell to be loaded is assigned to the corresponding assembly position. The fit generation unit is used to generate a dynamic steady-state fit between the battery cell to be loaded and the corresponding assembly position based on the cumulative voltage deviation, cumulative temperature deviation, cumulative equivalent branch impedance deviation, and cumulative attenuation trend deviation. The dynamic steady-state fit is used to characterize the degree to which the operating state of the battery cell to be loaded converges to the steady-state distribution of the target PACK after it enters the corresponding assembly position.

[0009] Furthermore, the feeding compensation decision module includes a location filtering unit, a feeding sequence generation unit, and an abnormal diversion unit; The position filtering unit is used to sort the dynamic steady-state fit between the battery cell to be loaded and each assembly position, and to determine the assembly position whose dynamic steady-state fit meets the preset fit conditions as the candidate assembly position. The loading sequence generation unit is used to select the assembly position with the highest dynamic steady-state adaptability from the candidate assembly positions as the target assembly position, and generate the loading sequence of the cells to be loaded according to the assembly cycle of the target assembly position in the target PACK. The abnormal current diversion unit is used to import the battery cell to be loaded into a delayed buffer path, a re-inspection path, or a state shaping path when there is no candidate assembly position that meets the preset adaptation conditions. The state shaping path is used to perform micro-charging, micro-discharging, or temperature homogenization processing on the battery cell to be loaded, so that the battery cell to be loaded can regain the battery cell deviation requirement value that can participate in the assembly position matching.

[0010] Furthermore, it also includes a terminal test residual attribution update module; The end-test residual attribution update module is used to obtain the voltage discrete residual, temperature rise residual, branch current residual and equalization response residual in the end test of the target PACK after the target PACK is assembled. The voltage discrete residual, temperature rise residual, branch current residual, and equilibrium response residual are then correlated with the cell deviation requirement, assembly position equivalent compensation capability, and dynamic steady-state adaptability of the corresponding assembly position in the target PACK to generate cell body attribution, connection contact attribution, thermal boundary attribution, and equilibrium response attribution. The end-test residual attribution update module is also used to update the weights of the cell deviation requirement value, the weights of the assembly position equivalent compensation capability value, and the dynamic steady-state adaptability calculation weights in the subsequent cell sorting process based on the cell body attribution value, connection contact attribution value, thermal boundary attribution value, and equalization response attribution value.

[0011] In summary, the present invention has the following beneficial effects: By constructing a dynamic matching relationship between the cell deviation requirement value and the equivalent compensation capability value of the assembly position, and determining the target assembly position and loading sequence of the cells to be loaded based on the dynamic steady-state adaptability, the initial differences of the cells can be absorbed by the balancing capability, thermal boundary capability and connection path characteristics of the assembly position during PACK operation. This solves the problem that the current lithium battery PACK line sorting is based only on the static consistency before assembly and it is difficult to guarantee the consistency of operation after assembly. Detailed Implementation

[0012] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] Example 1

[0014] The present invention will be described in further detail below.

[0015] This invention provides a technical solution: a dynamic compensation system for material sorting on a lithium battery PACK line based on digital twins, including a cell deviation feature acquisition module, an assembly position compensation feature construction module, a dynamic steady-state adaptability calculation module, and a material loading compensation decision module; The cell deviation feature acquisition module is used to acquire the voltage deviation, internal resistance deviation, capacity deviation and dynamic response deviation of the cell to be loaded, and generate the cell deviation requirement value of the cell to be loaded based on the above deviations. The assembly position compensation feature construction module is used to obtain the equalization capability, thermal boundary capability, connection path impedance and current stress weight corresponding to each assembly position in the target PACK based on the digital twin model of the target PACK, and generate the assembly position equivalent compensation capability value of each assembly position according to the obtained parameters. The dynamic steady-state adaptability calculation module is used to match the cell deviation requirement value with the equivalent compensation capability value of each assembly position, predict the degree of convergence of the running deviation after the cell to be loaded is assigned to each assembly position within a preset running time window, and generate the dynamic steady-state adaptability between the cell to be loaded and each assembly position. The feeding compensation decision module is used to determine the target assembly position and feeding sequence of the battery cell to be fed according to the dynamic steady-state adaptability, so that the initial deviation of the battery cell to be fed is jointly compensated by the equalization capability, thermal boundary capability and connection path impedance of the corresponding assembly position during the operation of the target PACK. The cell deviation feature acquisition module includes a reference feature determination unit, a detection feature extraction unit, and a deviation requirement value generation unit; The reference feature determination unit is used to generate batch reference features based on the voltage distribution, internal resistance distribution, capacity distribution and dynamic response distribution of the qualified cells in the current assembly batch of the target PACK. The detection feature extraction unit is used to obtain the open circuit voltage, internal resistance, capacity and dynamic response parameters of the battery cell to be loaded. The dynamic response parameters include at least two of the following: voltage recovery slope obtained from pulse charge-discharge test, temperature rise slope obtained from short-time current excitation test, self-discharge rate obtained from the open circuit voltage difference before and after resting, and capacity decay slope obtained from historical cycle data fitting. The deviation requirement value generation unit is used to calculate the normalized deviation of the open circuit voltage, internal resistance, capacity and dynamic response parameters of the cell to be loaded with the batch reference characteristics, and to fuse the normalized deviation according to the weight corresponding to the target PACK operating condition to generate the cell deviation requirement value. The assembly position compensation feature construction module includes a position feature extraction unit, a compensation capability calculation unit, and a position capability correction unit; The location feature extraction unit is used to extract the upper limit of equalization current, allowable equalization time, heat dissipation boundary parameters, connection path impedance, predicted welding contact impedance and current stress weight corresponding to each assembly position in the target PACK based on the digital twin model of the target PACK. The compensation capability calculation unit is used to generate an active compensation capability value for the corresponding assembly position based on the upper limit of the equalization current, the allowable equalization time and the heat dissipation boundary parameters, and to generate a passive compensation constraint value for the corresponding assembly position based on the connection path impedance, the predicted welding contact impedance and the current stress weight. The position capability correction unit is used to fuse the active compensation capability value and the passive compensation constraint value to generate the assembly position equivalent compensation capability value for each assembly position, so that the degree of compensation for the initial deviation of the battery cell to be loaded at different assembly positions is quantified. The dynamic steady-state fitness calculation module includes a difference mapping unit, a running deviation prediction unit, and a fitness generation unit. The difference mapping unit is used to match the cell deviation requirement value with the equivalent compensation capability value of each assembly position to generate the difference compensation relationship between the cell to be loaded and each assembly position. The operation deviation prediction unit is used to predict, based on the difference compensation relationship, the cumulative amount of voltage deviation, cumulative amount of temperature deviation, cumulative amount of equivalent branch impedance deviation and cumulative amount of attenuation trend deviation formed within a preset operation time window after the battery cell to be loaded is assigned to the corresponding assembly position. The fit generation unit is used to generate a dynamic steady-state fit between the battery cell to be loaded and the corresponding assembly position based on the cumulative voltage deviation, cumulative temperature deviation, cumulative equivalent branch impedance deviation, and cumulative attenuation trend deviation. The dynamic steady-state fit is used to characterize the degree to which the operating state of the battery cell to be loaded converges to the steady-state distribution of the target PACK after it enters the corresponding assembly position.

[0016] In this embodiment: Before the battery cells enter the assembly station, the compatibility between the battery cells to be loaded and different assembly positions in the target PACK is determined, and the target assembly position and loading sequence of the battery cells to be loaded are determined based on the determination result. In this embodiment, the target PACK includes multiple battery cell assembly positions. Due to differences in distance from the heat dissipation boundary, bus connection path, local welding contact state, current stress, and BMS balancing branch capability, the absorption capacity of the battery cells for initial deviations during PACK operation varies. Therefore, this embodiment does not simply sort the battery cells according to the similarity of their initial voltage, internal resistance, and capacity, but rather dynamically compensates and allocates them based on the convergence degree of operational deviations after the battery cells enter different assembly positions.

[0017] In specific implementation, the system includes a cell deviation characteristic acquisition module, an assembly position compensation characteristic construction module, a dynamic steady-state adaptability calculation module, and a material loading compensation decision module. After the cells to be loaded enter the online sorting station via the barcode scanning unit, the cell deviation characteristic acquisition module reads the batch number, formation cabinet number, capacity grading channel number, historical cycle record, and current detection station number of the cells to be loaded, and collects the open-circuit voltage, AC internal resistance, DC internal resistance, and capacity verification value of the cells to be loaded under a 25℃ environment. The open-circuit voltage is collected at 1s intervals, and a stable average value is taken after 10 consecutive collections; the AC internal resistance uses the 1kHz AC impedance test value; the DC internal resistance is obtained based on the voltage change under short-time pulse current; the capacity verification value uses the available capacity in the capacity grading record and is normalized in combination with the capacity distribution of the current assembly batch.

[0018] To avoid missorting of boundary cells due to relying solely on static parameters, the cell deviation characteristic acquisition module further acquires the dynamic response offset of the cells to be loaded. Specifically, a 2-second pulse charge-discharge test is applied to the cells to be loaded, and a voltage recovery curve is acquired within 30 seconds after the pulse ends. The voltage recovery slope is obtained based on the trend of change in the voltage recovery curve from 5 seconds to 30 seconds. A short-term current excitation of 5 seconds is applied to the cells to be loaded, and the temperature change of the cell shell surface is acquired through an infrared thermometer. The temperature rise slope is obtained based on the temperature change during the excitation. The self-discharge rate is obtained by correlating the open-circuit voltage difference of the cells to be loaded before and after resting with the resting time. The capacity decay slope is obtained based on the capacity change trend of the cells to be loaded in historical cycle data. The voltage recovery slope, temperature rise slope, self-discharge rate, and capacity decay slope are used to characterize the polarization recovery capability, thermal response capability, voltage holding capability, and decay trend of the cells to be loaded during operation.

[0019] When generating the required values ​​for battery cell deviations, the system first generates batch baseline characteristics based on the open-circuit voltage distribution, internal resistance distribution, capacity distribution, and dynamic response distribution of the qualified battery cells in the current target PACK assembly batch. The batch baseline characteristics are determined using the mean or median value after removing outliers. In this embodiment, the batch baseline characteristics include baseline open-circuit voltage, baseline internal resistance, baseline capacity, baseline voltage recovery slope, baseline temperature rise slope, baseline self-discharge rate, and baseline capacity decay slope. Subsequently, the various detection values ​​of the battery cells to be loaded are compared with the corresponding batch baseline characteristics, and normalized according to the allowable fluctuation range of each parameter, forming normalized deviation items for open-circuit voltage, internal resistance, capacity, voltage recovery slope, temperature rise slope, self-discharge rate, and capacity decay slope.

[0020] After obtaining the normalized deviation items, the system assigns different evaluation weights to different deviation items based on the target PACK's operating conditions, and then integrates these deviation items into a cell deviation requirement value. Under assembly conditions prioritizing safety, the evaluation weights for internal resistance deviation, temperature rise rate deviation, and self-discharge rate deviation are higher than those for capacity deviation; under assembly conditions prioritizing range consistency, the evaluation weights for capacity deviation and capacity decay rate deviation are increased. Therefore, the cell deviation requirement value does not simply represent the difference between the current value of the cell to be supplied and the batch baseline value, but rather represents the comprehensive deviation burden that the cell to be supplied needs to absorb after entering the target PACK, through the balancing capability of the assembly location, thermal boundary capability, and connection path characteristics.

[0021] The assembly position compensation feature construction module generates equivalent compensation capability values ​​for each assembly position based on the digital twin model of the target PACK. The digital twin model of the target PACK is established based on the structural data, BMS equalization parameters, thermal boundary parameters, connection path parameters, and welding process parameters of the target PACK. The structural data includes the spatial coordinates of each assembly position, the relationship between adjacent cells, the orientation of the electrode tabs, and the bus connection relationship; the BMS equalization parameters include the upper limit of the equalization current, the allowable equalization duration, and the equalization control response time of each equalization branch; the thermal boundary parameters include the distance from each assembly position to the heat sink, the length of the local heat conduction path, the proximity of the cooling channels, and the temperature rise suppression coefficient; the connection path parameters include the bus length, bus width, number of connection nodes, and path impedance estimation value corresponding to each assembly position; the welding process parameters include the welding energy window, the electrode tab clamping force range, the predicted value of the weld area, and the predicted value of the welding contact impedance.

[0022] For each assembly location in the target PACK, the system extracts the corresponding upper limit of balancing current, allowable balancing time, heat dissipation boundary parameters, connection path impedance, predicted welding contact impedance, and current stress weight. The upper limit of balancing current and allowable balancing time characterize the ability of this location to pull back voltage deviations through BMS balancing; the heat dissipation boundary parameters characterize the ability of this location to suppress temperature rise deviations; the connection path impedance and predicted welding contact impedance characterize the impact of this location on the equivalent branch impedance; and the current stress weight characterizes the degree to which this location withstands current surges under the operating conditions of the target PACK. The system normalizes and comprehensively evaluates the above location characteristics to obtain the equivalent compensation capability value of each assembly location. For locations with better heat dissipation and stronger balancing capabilities, the equivalent compensation capability value of the assembly location is higher; for locations with higher connection path impedance, larger predicted welding contact impedance, or higher current stress, the equivalent compensation capability value of the assembly location is adjusted accordingly.

[0023] The dynamic steady-state adaptability calculation module matches the cell deviation requirement value of the battery cell to be loaded with the equivalent compensation capability value of each assembly position, and predicts the convergence degree of the operating deviation of the battery cell to be loaded within a preset operating time window after entering different assembly positions. Specifically, the system establishes the difference compensation relationship between the battery cell to be loaded and each assembly position based on the target PACK digital twin model. This difference compensation relationship is used to characterize the operating deviation trend of the initial deviation of the battery cell to be loaded in different assembly positions after the combined effects of equalization capability, thermal boundary capability, connection path impedance, and welding contact impedance. For the same battery cell to be loaded, if its temperature rise slope is high, the position with stronger heat dissipation boundary capability can obtain a lower predicted temperature deviation value; if its internal resistance is slightly low, it can be matched to a position with slightly higher connection path impedance but better thermal boundary, so that the internal resistance of the battery cell and the connection path impedance together form an equivalent branch impedance that is closer to the balanced distribution of the target PACK.

[0024] In this embodiment, the preset operating time window is the first 30 minutes of the simulated charge-discharge test. The system samples every 10 seconds to predict the voltage deviation trend, temperature deviation trend, equivalent branch impedance deviation trend, and attenuation trend deviation of the battery cells to be loaded after they are assigned to their respective assembly positions, and performs cumulative evaluation for each. The cumulative voltage deviation result is used to characterize the degree of deviation of the battery cell from the steady-state voltage distribution of the target PACK; the cumulative temperature deviation result is used to characterize the degree of deviation of the battery cell from the steady-state temperature distribution of the target PACK; the cumulative equivalent branch impedance deviation result is used to characterize the degree of branch deviation after the combined effect of the battery cell's body impedance and the connection path impedance; the cumulative attenuation trend deviation result is used to characterize the degree of deviation between the future attenuation trend of the battery cell and the overall attenuation trend of the target PACK. The system generates a dynamic steady-state fit degree based on the above cumulative evaluation results. The smaller the cumulative deviation, the higher the dynamic steady-state fit degree, indicating that the battery cells to be loaded are more likely to converge to the steady-state distribution of the target PACK after being assigned to their corresponding assembly positions.

[0025] To illustrate the above calculation process, this embodiment selects four battery cells to be loaded (C101, C102, C103, and C104) and four candidate assembly positions (P03, P06, P09, and P12) in the target PACK for fit calculation. The cell deviation requirement value of each battery cell to be loaded, the equivalent compensation capability value of each candidate assembly position, and the calculated dynamic steady-state fit are shown in the table below: Table 1: Calculation results of dynamic steady-state fit between battery cell and candidate assembly location

[0026] The data above shows that the cell deviation requirement of cell C101 is 0.42, and its dynamic steady-state fit in candidate assembly position P03 is 74.6, which is higher than its dynamic steady-state fit in P06, P09, and P12. Therefore, C101 is preferentially matched to P03. The cell deviation requirement of cell C102 is higher than that of C101, but due to the high equivalent compensation capability of assembly position P03, its dynamic steady-state fit in P03 still reaches 69.7, indicating that the initial deviation of this cell can be well absorbed at this position. The cell deviation requirement of cell C104 is 0.63, which is higher than the equivalent compensation capability of some candidate assembly positions. Therefore, its dynamic steady-state fit in P06 and P09 is significantly reduced. The system will not only allocate based on whether the capacity or voltage is close to the batch average, but will also consider P03 or P12 based on the convergence degree after operation. Therefore, this embodiment can avoid the problem of post-assembly state divergence caused by relying solely on static consistency indicators in traditional sorting.

[0027] To further verify the effectiveness of online material sorting based on dynamic steady-state adaptability in this embodiment, 240 lithium-ion cells from the same production batch that passed appearance inspection, voltage inspection, and capacity inspection were selected and grouped for sorting at 25°C. 120 cells were assembled into packs using the traditional static consistency sorting method, which primarily uses open-circuit voltage difference, AC internal resistance difference, and capacity difference as sorting criteria. The other 120 cells were assembled into packs using the dynamic steady-state adaptability method described in this embodiment. This method determines the target assembly position of the cells based on the cell deviation requirement value, the equivalent compensation capability value of the assembly position, and the dynamic steady-state adaptability.

[0028] During testing, each group assembled 10 identical test packs, each containing 12 cells. After assembly, each test pack was left to stand at 25°C for 2 hours, followed by 0.5C charging, 1C discharging, 10-minute short-time rate discharging, and BMS equalization response testing. Test items included end-open-circuit voltage dispersion, maximum temperature difference at the discharge end, equivalent branch impedance dispersion, equalization completion time, and capacity retention dispersion after 5 cycles. To reduce random errors, each test pack was tested three times, and the average value was used as the test result for the corresponding pack. Table 2: Consistency test results of PACK assembly after different sorting methods

[0029] The test results above show that although the traditional static sorting group controls the open-circuit voltage, internal resistance, and capacity difference before assembly, there are still significant voltage dispersion, temperature rise differences, and branch impedance differences after assembly. This is because the traditional static sorting group only considers the initial parameter similarity of the cells before loading, without considering the differences in thermal boundary capabilities, connection path impedance, welding contact conditions, and equalization capabilities between different assembly positions of the target PACK. Therefore, some cells with similar initial parameters may still experience amplified operational deviations after entering different assembly positions.

[0030] In contrast, the dynamic steady-state matching group matches the cell deviation requirement value with the equivalent compensation capability value of the assembly position before loading, and predicts the convergence degree of the operating deviation after the cell enters different assembly positions through the dynamic steady-state matching degree. Therefore, some cells whose initial parameters are not the closest can still obtain better matching results because they form a complementary relationship with the thermal boundary capability, connection path impedance and equalization capability of the target assembly position. The test results show that the dynamic steady-state matching group has lower dispersion of terminal open circuit voltage, maximum temperature difference at the discharge end, dispersion of equivalent branch impedance, dispersion of equalization completion time and capacity retention than the traditional static sorting group. This indicates that this embodiment can perform compensable pre-allocation of initial differences of cells before PACK assembly, thereby improving the operational consistency after PACK assembly.

[0031] Through the above implementation method, the open-circuit voltage, internal resistance, capacity, and dynamic response parameters of the battery cells to be loaded are converted into battery cell deviation requirement values. The balancing capability, thermal boundary capability, connection path impedance, and current stress weight of different assembly positions in the target PACK are converted into equivalent compensation capability values ​​of the assembly positions. The system then calculates the operational convergence relationship between the battery cells and the assembly positions based on the dynamic steady-state adaptability. Therefore, this embodiment can determine whether the initial deviation of the battery cells can be absorbed and compensated in a specific assembly position before PACK assembly, and determine the target assembly position and loading sequence accordingly, thereby improving the voltage consistency, thermal consistency, and branch impedance consistency after PACK assembly, and reducing the risk of local battery cell state divergence after assembly.

[0032] Example 2

[0033] The material feeding compensation decision module includes a location filtering unit, a material feeding sequence generation unit, and an abnormal diversion unit. The position filtering unit is used to sort the dynamic steady-state fit between the battery cell to be loaded and each assembly position, and to determine the assembly position whose dynamic steady-state fit meets the preset fit conditions as the candidate assembly position. The loading sequence generation unit is used to select the assembly position with the highest dynamic steady-state adaptability from the candidate assembly positions as the target assembly position, and generate the loading sequence of the cells to be loaded according to the assembly cycle of the target assembly position in the target PACK. The abnormal current diversion unit is used to import the battery cell to be loaded into a delayed buffer path, a re-inspection path, or a state shaping path when there is no candidate assembly position that meets the preset adaptation conditions. The state shaping path is used to perform micro-charging, micro-discharging, or temperature homogenization processing on the battery cell to be loaded, so that the battery cell to be loaded can regain the battery cell deviation requirement value that can participate in the assembly position matching.

[0034] In this embodiment: after obtaining the dynamic steady-state fit between the battery cell to be loaded and each candidate assembly position in Embodiment 1, the system further determines the target assembly position, loading sequence, and abnormal diversion path of the battery cell to be loaded. This embodiment focuses on explaining the specific execution process of the loading compensation decision module, enabling the system to convert the dynamic steady-state fit calculation results into executable control actions for the online loading equipment, buffer channel, re-inspection station, and state shaping station.

[0035] In practical implementation, the feeding compensation decision module includes a location screening unit, a feeding sequence generation unit, and an abnormal current diversion unit. The location screening unit receives the adaptation results output by the dynamic steady-state adaptation calculation module and establishes a candidate assembly location list for each cell to be fed. The candidate assembly location list records the candidate assembly location number, the equivalent compensation capability value of the assembly location, the dynamic steady-state adaptation between the cell to be fed and the candidate assembly location, the current assembly cycle window of the candidate assembly location, whether the candidate assembly location has been occupied by other cells to be fed, and the cumulative evaluation of the temperature deviation and the cumulative evaluation of the equivalent branch impedance deviation of the candidate assembly location.

[0036] In this embodiment, the selection of candidate assembly positions considers both fit and compensation capability conditions. The fit condition is that the dynamic steady-state fit between the cell to be loaded and the candidate assembly position is not lower than a preset fit threshold. The compensation capability condition is that the cell deviation requirement of the cell to be loaded does not exceed the equivalent compensation capability value of the candidate assembly position, or although the cell deviation requirement of the cell to be loaded is slightly higher than the equivalent compensation capability value of the assembly position, the excess does not exceed a preset boundary compensation margin. In this embodiment, the preset fit threshold is set to 65, and the preset boundary compensation margin is set to 0.05. Therefore, for some cells to be loaded that are in a boundary state but whose operational deviation still shows a convergence trend, the system does not directly eliminate them but allows them to enter the candidate assembly position sorting stage.

[0037] After screening candidate assembly positions, the position selection unit sorts multiple candidate assembly positions corresponding to the same cell to be loaded into order of dynamic steady-state fit from high to low. When the difference in dynamic steady-state fit between two candidate assembly positions is less than 2, the system further compares the cumulative evaluation of temperature deviation and the cumulative evaluation of equivalent branch impedance deviation, prioritizing the position with the lower cumulative evaluation of temperature deviation; if the cumulative evaluations of temperature deviation are still close, the position with the lower cumulative evaluation of equivalent branch impedance deviation is prioritized; if the above evaluations are still close, the position that entered the assembly cycle window earlier is selected. Through this method, the system can avoid allocating positions based solely on a single fit value, reducing the probability of localized thermal risk concentration and loading cycle conflicts.

[0038] The loading sequence generation unit generates the loading sequence of the battery cells to be loaded based on the candidate assembly position screening results. In this embodiment, the loading sequence is not directly determined by the order in which the battery cells enter the detection station, but rather by a combination of factors including the assembly cycle time of the target assembly position, the estimated time for the battery cells to arrive at the assembly station, the occupancy status of the candidate assembly positions, the stability of the battery cells in the buffer path, and the dynamic steady-state adaptability. For battery cells with a determined target assembly position, the system generates a loading priority score. The higher the loading priority score, the more suitable the battery cell is to be loaded with priority within the current time window. The loading priority score is determined by a combination of dynamic steady-state adaptability, cycle time matching degree, and buffer stability degree, with dynamic steady-state adaptability having the main evaluation weight. The cycle time matching degree and buffer stability degree are used to prevent the loading sequence from mismatching with the actual assembly cycle time.

[0039] When multiple battery cells awaiting loading are simultaneously matched to the same target assembly position, the loading sequence generation unit prioritizes the battery cell with higher dynamic steady-state adaptability. For battery cells not selected, the system re-queries its candidate assembly position list. If a suboptimal candidate assembly position exists and its dynamic steady-state adaptability still meets the preset adaptability conditions, it is assigned to that suboptimal candidate assembly position. If no suboptimal candidate assembly position meets the conditions, it is handed over to the abnormal flow diversion unit for processing. Thus, the system can avoid loading conflicts caused by multiple battery cells vying for the same assembly position and reduce production line cycle time losses due to waiting.

[0040] The abnormal current shunting unit is used to shun the battery cells to be loaded when there are no candidate assembly positions that meet the preset adaptation conditions. In this embodiment, the abnormal current shunting path includes a delay buffer path, a re-inspection path, and a state shaping path. The delay buffer path is used to process battery cells to be loaded whose deviation requirement value is close to the boundary of the candidate assembly position and whose voltage recovery slope or temperature recovery state is still changing; the re-inspection path is used to process battery cells to be loaded whose detection results conflict or whose detection reliability is insufficient; the state shaping path is used to process battery cells to be loaded that still have a chance to meet the adaptation conditions again after a small amount of power state adjustment or temperature state adjustment.

[0041] For the battery cells to be loaded via the delayed buffer path, the system controls them to enter the temperature-controlled buffer channel. The buffer channel temperature is controlled at 25°C, with an allowable temperature fluctuation range of ±1°C. The delay time is determined based on the open-circuit voltage drift state and temperature recovery state of the battery cell to be loaded. In this embodiment, the delay time is set to 5 min to 20 min. After the delay, the system re-acquires the open-circuit voltage, surface temperature, and voltage recovery slope, and regenerates the battery cell deviation requirement value and its dynamic steady-state fit with each candidate assembly position. If the re-evaluated dynamic steady-state fit reaches the preset fit threshold, the battery cell to be loaded re-enters the loading sequence generation unit; if the conditions are still not met, it continues to enter the re-inspection path, state shaping path, or isolation channel.

[0042] For battery cells to be loaded into the re-inspection path, the system first determines the source of the anomaly. If the open-circuit voltage offset is inconsistent with the self-discharge rate, the internal resistance offset is inconsistent with the temperature rise slope, or the dispersion of the internal resistance value continuously output from the same testing station exceeds the preset station dispersion threshold, a detection conflict is identified for the battery cell to be loaded. The re-inspection path re-executes open-circuit voltage detection, AC internal resistance detection, short-time pulse DC internal resistance detection, and infrared temperature rise detection on the battery cell to be loaded, and compares the re-inspection results with the initial detection results. If the deviation between the re-inspection results and the initial detection results does not exceed the re-inspection allowable error, the weighted evaluation result of the two detection results is used to regenerate the battery cell deviation requirement value; if the deviation between the re-inspection results and the initial detection results exceeds the re-inspection allowable error, the battery cell to be loaded is imported into the isolation channel and marked as awaiting manual confirmation or re-evaluation in subsequent batches.

[0043] For the battery cells to be loaded into the state shaping path, the system selects micro-charging, micro-discharging, or temperature homogenization processing based on the type of deviation. If the open-circuit voltage of the battery cell to be loaded is low and the voltage recovery slope is normal, micro-charging is performed, with the micro-charging current set to 0.05C to 0.1C and the micro-charging time set to 30s to 180s. If the open-circuit voltage of the battery cell to be loaded is high and the self-discharge rate does not exceed the preset upper limit, micro-discharging is performed, with the micro-discharge current set to 0.05C to 0.1C and the micro-discharge time set to 30s to 180s. If the temperature rise slope of the battery cell to be loaded is high but the internal resistance offset does not exceed the preset threshold, temperature homogenization is performed, and the battery cell to be loaded is placed in the temperature homogenization channel, with the channel temperature controlled at 25℃ and the homogenization time set to 5min to 15min. After the state shaping is completed, the system re-collects the open circuit voltage, temperature rise slope and voltage recovery slope of the cell to be loaded, and re-evaluates its cell deviation requirement value and dynamic steady-state adaptability.

[0044] To illustrate the material feeding compensation decision-making process, this embodiment selects 6 battery cells to be fed into the same material feeding time window. The system generates the following decision results based on the dynamic steady-state adaptability, candidate assembly position status, and assembly cycle time.

[0045] Table 3: Decision Results of Compensation for Battery Cells to be Loaded

[0046] As shown in Table 3, the dynamic steady-state compatibility between cell C201 and P05 reaches 79.4, and P05 is in an available state. Therefore, the system directly assigns C201 to P05 and determines its priority for loading within the current time window based on its loading priority score. The dynamic steady-state compatibility between cell C202 and P08 reaches 71.6, meeting the preset compatibility conditions, and is therefore assigned to P08. The highest compatibility position for cell C203 is also P08, but P08 is already occupied by C202. After querying its candidate assembly position list, the system assigns it to the second-best assembly position P11. The dynamic steady-state compatibility between C203 and P11 still meets the preset compatibility conditions, so there is no need to enter the abnormal routing path.

[0047] Cell C204, awaiting delivery, has a maximum dynamic steady-state fit of 63.8, which is lower than the preset fit threshold of 65. However, its voltage recovery state is still changing, and the difference between the cell deviation requirement and the equivalent compensation capability of the candidate assembly position does not exceed the boundary compensation margin. Therefore, the system imports it into the delay buffer path. Cell C205, awaiting delivery, has a maximum dynamic steady-state fit of 66.4, but there is a conflict between its internal resistance offset and temperature rise slope. That is, the internal resistance offset is within the normal range, but the temperature rise slope is too high. The system judges that its test results are not reliable enough, so it is imported into the re-inspection path. Cell C206, awaiting delivery, has a high cell deviation requirement, and its maximum dynamic steady-state fit does not reach the preset fit threshold. However, its open-circuit voltage is too high, and its self-discharge rate does not exceed the upper limit. The system judges that its voltage deviation requirement can be reduced through micro-discharge processing, so it is imported into the state shaping path.

[0048] After processing the cells that entered the abnormal shunt path, the system re-evaluated their cell deviation requirements and dynamic steady-state adaptability. The processing results are shown in the table below.

[0049] Table 4. Re-matching results after abnormal traffic splitting.

[0050] As shown in Table 4, after a 12-minute delay buffer, the open-circuit voltage drift of C204 stabilized, its cell deviation requirement decreased from 0.62 to 0.55, and its highest dynamic steady-state fit increased from 63.8 to 68.7, meeting the conditions for re-entering the material queue. After re-inspection, the system confirmed that the initial abnormal temperature rise of C205 mainly came from fluctuations in the detection contact state. The highest dynamic steady-state fit after re-evaluation still reached 65.9, so it re-entered the material queue. After 90 seconds of 0.08C micro-discharge, the open-circuit voltage deviation of C206 decreased, the cell deviation requirement decreased from 0.67 to 0.58, and the highest dynamic steady-state fit increased to 67.2, so it could re-participate in assembly position matching. After re-inspection, the internal resistance re-measurement deviation of C208 was found to be excessive. After processing, its highest dynamic steady-state fit decreased to 58.6, and the system introduced it into the isolation channel, not participating in the current target PACK assembly.

[0051] This embodiment further verifies the production line cycle time of the traditional fixed threshold sorting method and the material loading compensation decision method of this embodiment. 300 lithium-ion cells that have completed initial testing from the same batch were selected, and material loading was organized using both the traditional fixed threshold sorting method and the material loading compensation decision method of this embodiment. The traditional fixed threshold sorting method directly transfers cells to the processing channel when they do not meet fixed voltage, internal resistance, or capacity thresholds; the material loading compensation decision method of this embodiment further processes cells that do not meet the current assembly position adaptation conditions through candidate position re-matching, delay buffering, re-inspection, or state shaping paths. The test environment temperature was 25℃, and each method was continuously run for 3 assembly batches, with 100 cells in each batch. The material loading success rate, average waiting time due to sorting, abnormal isolation rate, and assembly cycle time fluctuation were recorded. Table 5: Verification results of production line feeding cycle time under different sorting methods

[0052] As shown in Table 5, the traditional fixed threshold sorting method typically transfers boundary cells directly to the processing channel, leading to a decrease in loading success rate, an increase in average waiting time, and significant fluctuations in assembly cycle time due to the lack of suitable cells at some assembly positions for short periods. In contrast, the loading compensation decision method in this embodiment filters candidate positions and generates loading order based on dynamic steady-state adaptability. For cells that do not meet the conditions for direct loading, further evaluation is conducted through delay buffering, re-inspection, and state shaping paths. This allows some boundary cells to regain the cell deviation requirement value and dynamic steady-state adaptability that meet the assembly position matching requirements, thereby improving the loading success rate, reducing the abnormal isolation rate, and minimizing assembly cycle time fluctuations.

[0053] Through the above implementation methods, this embodiment transforms the dynamic steady-state fit calculation results into practically executable material loading decisions. For battery cells that meet the preset fit conditions, the system generates a loading sequence based on the target assembly position and assembly cycle time. For battery cells that do not meet the preset fit conditions, the system imports a delayed buffer path, a re-inspection path, or a state shaping path based on the anomaly type, and re-evaluates the battery cell deviation requirement value and dynamic steady-state fit after processing. Thus, this embodiment not only avoids the direct rejection of boundary battery cells but also prevents incompatible battery cells from being forcibly assembled into positions with insufficient compensation capabilities, enabling the on-line material sorting process to have both consistency control capabilities and production line cycle time stability.

[0054] Example 3

[0055] It also includes a terminal test residual attribution update module; The end-test residual attribution update module is used to obtain the voltage discrete residual, temperature rise residual, branch current residual and equalization response residual in the end test of the target PACK after the target PACK is assembled. The voltage discrete residual, temperature rise residual, branch current residual, and equilibrium response residual are then correlated with the cell deviation requirement, assembly position equivalent compensation capability, and dynamic steady-state adaptability of the corresponding assembly position in the target PACK to generate cell body attribution, connection contact attribution, thermal boundary attribution, and equilibrium response attribution. The end-test residual attribution update module is also used to update the weight of the cell deviation requirement value, the weight of the assembly position equivalent compensation capability value, and the weight of the dynamic steady-state adaptability calculation in the subsequent cell sorting process based on the cell body attribution value, connection contact attribution value, thermal boundary attribution value, and equalization response attribution value. In this embodiment, after the target PACK has been assembled and undergone end-of-line testing, the voltage dispersion residual, temperature rise residual, branch current residual, and equalization response residual observed in the end-of-line testing are inversely correlated to the front-end material loading and sorting process. Based on this, the evaluation weights for cell deviation requirement value, assembly position equivalent compensation capability, and dynamic steady-state adaptability in the subsequent cell sorting process are adjusted. This embodiment, in conjunction with Embodiments 1 and 2, enables the system to not only perform dynamic steady-state adaptability calculations before assembly but also to continuously correct the front-end sorting rules based on the actual test results after assembly.

[0056] In practice, after the target PACK completes cell assembly, soldering, bus connection, sampling line connection, and BMS connection, it enters the final testing station. The final testing station sequentially performs static voltage testing, low-rate charge / discharge testing, short-time rate discharge testing, thermal response testing, and BMS equalization response testing. Static voltage testing obtains the open-circuit voltage at each assembly location; low-rate charge / discharge testing obtains the voltage change of each cell under stable operating conditions; short-time rate discharge testing obtains the current sharing status of each branch; thermal response testing obtains the temperature rise data at each assembly location; and the BMS equalization response testing obtains the equalization time required for each cell to recover from its initial discrete voltage state to a preset equalization range.

[0057] Before the final test, the dynamic steady-state fit calculation module predicted the voltage deviation trend, temperature rise trend, equivalent branch impedance deviation trend, and equalization response time at each assembly location based on the target PACK digital twin model. The final test residual attribution update module compared the measured values ​​of the final test with the predicted values ​​before assembly to obtain the voltage discrete residual, temperature rise residual, branch current residual, and equalization response residual corresponding to each assembly location. If the measured voltage deviation at a certain assembly location is significantly higher than the predicted voltage deviation, it indicates that there is a voltage discrete residual at that location; if the measured temperature rise is significantly higher than the predicted temperature rise, it indicates that there is a temperature rise residual at that location; if the measured branch current sharing deviation is higher than the predicted result, it indicates that there is a branch current residual at that location; if the measured equalization completion time is longer than the predicted equalization completion time, it indicates that there is an equalization response residual at that location.

[0058] In this embodiment, the residual judgment adopts a normalization processing method. Voltage discrete residuals are evaluated using 10mV as the benchmark, temperature rise residuals using 3℃ as the benchmark, branch current residuals using 5% as the benchmark, and equilibrium response residuals using 20min as the benchmark. The larger the normalized residual value, the more significant the deviation between the measured result and the predicted result before assembly. When a certain type of residual appears continuously at the same assembly location or in the same type of assembly area, the system determines that this type of residual has rule update value; when a certain type of residual only appears occasionally and does not exceed the preset residual threshold, the system only performs temporary marking and does not immediately change the subsequent sorting rules.

[0059] To determine the source of residuals, the end-test residual attribution update module further establishes attribution features. These attribution features include cell body attribution features, connection contact attribution features, thermal boundary attribution features, and equalization response attribution features. Cell body attribution features include the cell deviation requirement, open-circuit voltage drift, internal resistance offset, temperature rise slope, and capacity decay slope for the cell assembled at that location. Connection contact attribution features include the predicted welding contact impedance, solder joint area detection, welding energy deviation, and bus path impedance for that assembly location. Thermal boundary attribution features include the distance from the assembly location to the heat dissipation boundary, local temperature rise suppression coefficient, temperature rise status of adjacent cells, and proximity of cooling channels. Equalization response attribution features include the upper limit of equalization current, equalization response delay time, equalization duration, and number of equalization control triggers for the equalization branch corresponding to that assembly location.

[0060] The end-test residual attribution update module correlates various test residuals with the aforementioned attribution features. If the voltage discrete residual highly corresponds to the open-circuit voltage drift, internal resistance offset, or capacity decay slope of the cell assembled at that location, then the residual is mainly attributed to the cell body deviation; if the branch current residual highly corresponds to the solder joint area detection value, welding energy deviation, or bus path impedance, then the residual is mainly attributed to the connection contact deviation; if the temperature rise residual highly corresponds to the local temperature rise suppression coefficient, the proximity relationship of the cooling channel, or the temperature rise state of the adjacent cell, then the residual is mainly attributed to the thermal boundary deviation; if the equalization response residual highly corresponds to the equalization current upper limit, equalization response delay time, or equalization control trigger number, then the residual is mainly attributed to the equalization response deviation.

[0061] To illustrate the process of collecting residuals during end-of-line testing, this embodiment selects the target PACK A01 after assembly for end-of-line testing. PACK A01 includes 12 assembly positions, and the following table lists the predicted results, measured results, and residual evaluation results for 6 typical assembly positions.

[0062] Table 6: Residual Acquisition Results of Target PACK Terminal Test

[0063] Table 6 shows that the voltage discrete residual, branch current residual, and equalization response residual at location P04 are all significantly high, but the temperature rise residual is not the highest. This indicates that the residuals at this location may be more due to the connection contact state and equalization response estimation bias, rather than a simple thermal boundary problem. The temperature rise residual at location P06 is high, but the voltage discrete residual and branch current residual are low, indicating that there may be an overestimation of the local thermal boundary capability at this location. All four types of residuals at location P10 are high, indicating that there may be a multi-factor superposition problem at this location, and the evaluation weights related to the connection path, thermal boundary, and equalization response need to be adjusted simultaneously.

[0064] The end-of-line test residual attribution update module further evaluates the attribution of the above assembly locations, obtaining the cell body attribution, connection contact attribution, thermal boundary attribution, and equalization response attribution. The attribution values ​​represent the degree of contribution of different factors to the test residual at that location. A higher attribution value indicates a more significant contribution of that factor to the current residual. The system determines the primary attribution type based on the magnitude of each attribution value and uses this primary attribution type as the basis for subsequent rule updates. Table 7: Attribution Results of Target PACK Terminal Test Residuals

[0065] Table 7 shows that the main attribution type for location P04 is connection contact and equalization response, indicating that the estimation of connection path impedance or weld contact impedance at this location was underestimated before assembly, while the estimate of the repairability of the equalization branch was overestimated. The main attribution type for location P06 is thermal boundary, indicating that the target PACK digital twin model overestimated the heat dissipation capacity at this location. The main attribution type for location P08 is equalization response, indicating that the actual response time of the corresponding equalization branch at this location is longer than the predicted result. Location P10 involves multiple factors, requiring a simultaneous downgrade of the thermal boundary capability evaluation, connection contact capability evaluation, and equalization response capability evaluation for this location.

[0066] Based on the above attribution results, the system updates the sorting rules for subsequent battery cells to be loaded. The updates include three categories.

[0067] Firstly, the evaluation weights for cell deviation requirements need to be updated. If the residual error in the final test is mainly attributed to the cell itself, it indicates that the front end is not sensitive enough to this type of cell deviation characteristic, and the evaluation weight of the corresponding characteristics in the cell deviation requirement needs to be increased. For example, when the voltage discrete residual and capacity retention discrete residual are mainly attributed to the cell itself, the evaluation weights of open-circuit voltage offset, internal resistance offset, and capacity decay slope should be increased; when the temperature rise residual is mainly related to the temperature rise slope of the cell itself, the evaluation weight of the temperature rise slope should be increased.

[0068] Secondly, the evaluation weights for the equivalent compensation capability of the assembly location are updated. If the end-test residual is mainly attributed to connection contact deviation, the estimated values ​​of the connection path impedance compensation capability and welding contact compensation capability of the corresponding assembly location are reduced; if the end-test residual is mainly attributed to thermal boundary deviation, the estimated value of the thermal boundary capability of the assembly location is reduced to prevent cells with high temperature rise slope from being assigned to that location; if the end-test residual is mainly attributed to equalization response deviation, the estimated value of the equalization capability of the assembly location is reduced to prevent cells with high voltage deviation requirements from being assigned to that location.

[0069] Third, the evaluation weights for dynamic steady-state fit are updated. If end-point testing shows that a certain type of operational deviation has a greater impact on final consistency in the actual PACK, the evaluation weight of this type of cumulative deviation in dynamic steady-state fit is increased. For example, when the temperature rise residual has a significant impact on end-point consistency, the influence of the cumulative temperature deviation evaluation is increased; when the branch current residual and impedance residual are significant, the influence of the cumulative equivalent branch impedance deviation evaluation is increased; when the equalization response residual is significant, the influence of the cumulative voltage deviation evaluation and the equalization response related evaluation is increased.

[0070] In this embodiment, the weight update adopts a gradual update method to avoid drastic changes in sorting rules caused by accidental test fluctuations of a single PACK. If an abnormal residual appears only in a single PACK at the same assembly location, the system temporarily corrects that location and retains a verification mark; if the same assembly location shows the same primary attribution type in three consecutive PACKs, the system solidifies the compensation capability correction result corresponding to that location into the next production batch. The magnitude of a single correction is controlled within 15% of the original evaluation weight or the original capability coefficient to ensure smooth changes in the production line sorting rules.

[0071] To illustrate the rule update process, the following table lists the results of partial evaluation weights and assembly position capability corrections after the final test of target PACK A01.

[0072] Table 8: Results of Sorting Rule Updates After End-Test Residual Attribution

[0073] As shown in Table 8, after identifying connection contact and equalization response deviations at position P04, the system lowers the reliability coefficient of the connection contact and the equalization capability correction coefficient at P04, subjecting subsequent battery cells to stricter connection contact and voltage deviation constraints when matching to P04. After identifying an overestimation of thermal boundary capability at position P06, the system lowers the thermal boundary capability correction coefficient at P06, preventing battery cells with high temperature rise slopes from being preferentially matched to P06. The system also increases the impact of cumulative temperature deviation evaluation and cumulative equivalent branch impedance deviation evaluation based on the overall residual change trend, making the dynamic steady-state fit closer to the actual PACK end-test results.

[0074] After the weights and correction coefficients are updated, when subsequent battery cells are included in the dynamic steady-state fit calculation, the system will automatically reduce the estimation of the partial compensation capabilities of P04, P06, and P08. For example, for battery cells with a high temperature rise slope, the updated system will no longer prioritize assigning them to P06; for battery cells with high voltage deviation requirements, the updated system will reduce their priority for matching to P04 or P08; for battery cells sensitive to connection paths, the system will increase the impact of the cumulative evaluation of equivalent branch impedance deviation on the dynamic steady-state fit, thereby avoiding assigning them to positions with high connection contact residuals.

[0075] To verify the effectiveness of the end-test residual attribution update mechanism, this embodiment selects two consecutive assembly batches for comparison. The first batch is assembled using the sorting rules in Embodiments 1 and 2, and the residual attribution update of this embodiment is performed after the end-test is completed. The second batch is sorted using the updated weights and correction coefficients under the same target PACK model, the same test environment, and the same assembly cycle. Each batch assembles 8 test PACKs, each test PACK containing 12 cells. The end-test environment is 25℃. After each PACK is left to stand for 2 hours, it undergoes 0.5C charging, 1C discharging, 10-minute short-time rate discharging, and BMS equalization response tests. The test results are as follows: Table 9: Comparison of PACK end-of-line test results before and after the sorting rule update

[0076] As shown in Table 9, after the first batch of end-point tests, the system identified overestimation of connection contact capability, thermal boundary capability, and equalization response capability at some assembly locations through residual attribution. Based on this, the relevant evaluation weights and correction coefficients in subsequent sorting processes were updated. After adopting the updated sorting rules for the second batch, the dispersion of end-point open-circuit voltage, maximum temperature difference at the discharge end, branch current deviation, equalization completion time, and equivalent branch impedance dispersion all decreased. This indicates that the residual attribution update mechanism can reverse the actual test deviations after assembly and apply them to the front-end material loading and sorting rules, thereby improving the consistency of subsequent PACK operations.

[0077] Through the above implementation method, this embodiment does not simply alarm or reject the PACK after an anomaly is found in the end-of-life test. Instead, it decomposes the end-of-life test residual into cell body attribution, connection contact attribution, thermal boundary attribution, and equilibrium response attribution. Based on the attribution results, it updates the evaluation weights in the cell deviation requirement value, assembly position equivalent compensation capability value, and dynamic steady-state adaptability. Therefore, the system can gradually correct the deviation between the target PACK digital twin model and the actual assembly result during continuous production. This allows the front-end material sorting rules to be continuously corrected according to the end-of-life test results, thus forming a closed-loop compensation mechanism between manufacturing-end material sorting and post-assembly testing, further improving the consistency and operational stability of the lithium battery PACK after assembly.

[0078] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0079] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the present invention. Various changes and modifications can be made to the present invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed.

Claims

1. A dynamic compensation system for material sorting on a lithium battery PACK line based on digital twins, characterized in that, It includes a cell deviation feature acquisition module, an assembly position compensation feature construction module, a dynamic steady-state adaptability calculation module, and a material feeding compensation decision module; The cell deviation feature acquisition module is used to acquire the voltage deviation, internal resistance deviation, capacity deviation and dynamic response deviation of the cell to be loaded, and generate the cell deviation requirement value of the cell to be loaded based on the above deviations. The assembly position compensation feature construction module is used to obtain the equalization capability, thermal boundary capability, connection path impedance and current stress weight corresponding to each assembly position in the target PACK based on the digital twin model of the target PACK, and generate the assembly position equivalent compensation capability value of each assembly position according to the above parameters. The dynamic steady-state adaptability calculation module is used to match the cell deviation requirement value with the equivalent compensation capability value of each assembly position, predict the degree of convergence of the running deviation after the cell to be loaded is assigned to each assembly position within a preset running time window, and generate the dynamic steady-state adaptability between the cell to be loaded and each assembly position. The feeding compensation decision module is used to determine the target assembly position and feeding sequence of the battery cell to be fed according to the dynamic steady-state adaptability, so that the initial deviation of the battery cell to be fed is jointly compensated by the equalization capability, thermal boundary capability and connection path impedance of the corresponding assembly position during the operation of the target PACK. The cell deviation feature acquisition module includes a reference feature determination unit, a detection feature extraction unit, and a deviation requirement value generation unit; The reference feature determination unit is used to generate batch reference features based on the voltage distribution, internal resistance distribution, capacity distribution and dynamic response distribution of the qualified cells in the current assembly batch of the target PACK. The detection feature extraction unit is used to obtain the open circuit voltage, internal resistance, capacity and dynamic response parameters of the battery cell to be loaded. The dynamic response parameters include at least two of the following: voltage recovery slope obtained from pulse charge-discharge test, temperature rise slope obtained from short-time current excitation test, self-discharge rate obtained from the open circuit voltage difference before and after resting, and capacity decay slope obtained from historical cycle data fitting. The deviation requirement value generation unit is used to calculate the normalized deviation of the open circuit voltage, internal resistance, capacity and dynamic response parameters of the cell to be loaded with the batch reference characteristics, and to fuse the normalized deviation according to the weight corresponding to the target PACK operating condition to generate the cell deviation requirement value. The assembly position compensation feature construction module includes a position feature extraction unit, a compensation capability calculation unit, and a position capability correction unit; The location feature extraction unit is used to extract the upper limit of equalization current, allowable equalization time, heat dissipation boundary parameters, connection path impedance, predicted welding contact impedance and current stress weight corresponding to each assembly position in the target PACK based on the digital twin model of the target PACK. The compensation capability calculation unit is used to generate an active compensation capability value for the corresponding assembly position based on the upper limit of the equalization current, the allowable equalization time and the heat dissipation boundary parameters, and to generate a passive compensation constraint value for the corresponding assembly position based on the connection path impedance, the predicted welding contact impedance and the current stress weight. The position capability correction unit is used to fuse the active compensation capability value and the passive compensation constraint value to generate the assembly position equivalent compensation capability value for each assembly position, so that the degree of compensation for the initial deviation of the battery cell to be loaded at different assembly positions is quantified.

2. The dynamic compensation system for material sorting on a lithium battery PACK line based on digital twins according to claim 1, characterized in that, The dynamic steady-state fitness calculation module includes a difference mapping unit, a running deviation prediction unit, and a fitness generation unit. The difference mapping unit is used to match the cell deviation requirement value with the equivalent compensation capability value of each assembly position to generate the difference compensation relationship between the cell to be loaded and each assembly position. The operation deviation prediction unit is used to predict, based on the difference compensation relationship, the cumulative amount of voltage deviation, cumulative amount of temperature deviation, cumulative amount of equivalent branch impedance deviation and cumulative amount of attenuation trend deviation formed within a preset operation time window after the battery cell to be loaded is assigned to the corresponding assembly position. The fit generation unit is used to generate a dynamic steady-state fit between the battery cell to be loaded and the corresponding assembly position based on the cumulative voltage deviation, cumulative temperature deviation, cumulative equivalent branch impedance deviation, and cumulative attenuation trend deviation. The dynamic steady-state fit is used to characterize the degree to which the operating state of the battery cell to be loaded converges to the steady-state distribution of the target PACK after it enters the corresponding assembly position.

3. The dynamic compensation system for material sorting on a lithium battery PACK line based on digital twins according to claim 1, characterized in that, The material feeding compensation decision module includes a location filtering unit, a material feeding sequence generation unit, and an abnormal diversion unit. The position filtering unit is used to sort the dynamic steady-state fit between the battery cell to be loaded and each assembly position, and to determine the assembly position whose dynamic steady-state fit meets the preset fit conditions as the candidate assembly position. The loading sequence generation unit is used to select the assembly position with the highest dynamic steady-state adaptability from the candidate assembly positions as the target assembly position, and generate the loading sequence of the cells to be loaded according to the assembly cycle of the target assembly position in the target PACK. The abnormal current diversion unit is used to import the battery cell to be loaded into a delayed buffer path, a re-inspection path, or a state shaping path when there is no candidate assembly position that meets the preset adaptation conditions. The state shaping path is used to perform micro-charging, micro-discharging, or temperature homogenization processing on the battery cell to be loaded, so that the battery cell to be loaded can regain the battery cell deviation requirement value that can participate in the assembly position matching.

4. The dynamic compensation system for material sorting on a lithium battery PACK line based on digital twins according to claim 1, characterized in that, It also includes a terminal test residual attribution update module; The end-test residual attribution update module is used to obtain the voltage discrete residual, temperature rise residual, branch current residual and equalization response residual in the end test of the target PACK after the target PACK is assembled. The voltage discrete residual, temperature rise residual, branch current residual, and equilibrium response residual are then correlated with the cell deviation requirement, assembly position equivalent compensation capability, and dynamic steady-state adaptability of the corresponding assembly position in the target PACK to generate cell body attribution, connection contact attribution, thermal boundary attribution, and equilibrium response attribution. The end-test residual attribution update module is also used to update the weights of the cell deviation requirement value, the weights of the assembly position equivalent compensation capability value, and the dynamic steady-state adaptability calculation weights in the subsequent cell sorting process based on the cell body attribution value, connection contact attribution value, thermal boundary attribution value, and equalization response attribution value.

Images