Two-wheel electric vehicle battery off-line capacity attenuation predictive sorting method and system

Abstract

The present application relates to the technical field of battery testing, in particular to a two-wheeled electric vehicle battery offline capacity attenuation predictive sorting method and system, after adjusting the to-be-sorted batteries to the same state of charge, sequentially performing multiple rate pulse charge-discharge test and voltage relaxation test, collecting pulse voltage and relaxation open-circuit voltage data; calculating polarization internal resistance growth coefficient, large current polarization eigenvalue and ion diffusion relaxation eigenvalue according to the collected data to form a dynamic characteristic parameter set; inputting the parameter set into a pre-constructed capacity attenuation prediction model to output the predicted annual capacity attenuation rate of each battery; setting multiple sorting grades according to the predicted attenuation rate, rejecting unqualified products higher than the threshold; and grouping after fine-tuning and matching of batteries of the same grade according to initial capacity and initial internal resistance. The present application realizes the leap from static parameter sorting to dynamic attenuation behavior prediction sorting, significantly improves the capacity attenuation consistency of the grouped battery pack, and prolongs the service life of the battery pack.

Description

Technical Field

[0001] This invention relates to the field of battery testing technology, and in particular to a method and system for predictive sorting of battery capacity degradation after production for two-wheeled electric vehicles. Background Technology

[0002] Currently, lithium-ion battery packs are widely used as the power source for two-wheeled electric vehicles. In the battery production line, traditional sorting methods typically rely on static parameters such as initial capacity, initial open-circuit voltage, and AC internal resistance of individual battery cells for consistency screening. This method assumes that batteries with similar initial performance parameters will exhibit similar behavior in subsequent use. However, the operating conditions of two-wheeled electric vehicles are complex; frequent charging and discharging, different current rates, and changes in ambient temperature all contribute to battery capacity degradation. Extensive practical experience has shown that battery packs sorted solely based on initial static parameters exhibit significant differences in capacity degradation rates among individual cells shortly after being put into use. This inconsistency triggers the "weakest link" effect, where the usable capacity of the entire battery pack is limited by the cell with the fastest capacity degradation, resulting in an overall battery pack lifespan far below the design life, prematurely leading to severe range reduction, abnormal charging times, and even thermal runaway. Therefore, existing sorting methods fail to address the problem of predicting the consistency of capacity degradation during long-term battery use. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for predictive sorting of battery capacity degradation after production for two-wheeled electric vehicles, ensuring that battery packs assembled from batteries in the same sorting group can maintain consistent capacity degradation during subsequent use, thereby extending the effective service life of the battery pack.

[0004] To achieve the above objectives, in a first aspect, the present invention provides a method for predictive sorting of battery capacity degradation for two-wheeled electric vehicles, comprising the following steps: Each battery cell to be sorted is adjusted to the same preset state of charge. The battery cells in the preset state of charge are subjected to multi-rate pulse charge-discharge test and voltage relaxation test in sequence. The voltage value at the moment the pulse discharge ends and the open circuit voltage change data over time during the relaxation process are collected. The dynamic characteristic parameter set of each battery cell is calculated based on the collected data. The dynamic characteristic parameter set includes at least the polarization internal resistance growth coefficient, the high current polarization characteristic value, and the ion diffusion relaxation characteristic value. The dynamic feature parameter set is input into a pre-built capacity decay prediction model to obtain the predicted annual capacity decay rate of each battery cell. Multiple sorting levels are set according to the predicted annual capacity decay rate. The battery cells are classified into the corresponding levels, and battery cells with a predicted annual capacity decay rate higher than the preset rejection threshold are rejected. For individual battery cells within the same sorting level, fine-tuning and matching are performed based on their corresponding initial capacity and initial internal resistance to form a battery pack.

[0005] This includes adjusting each individual battery cell to the same preset state of charge, including: The battery cell is discharged at a rate of 0.5 times the discharge cutoff voltage, then charged at a constant current rate of 0.5 times the nominal capacity to 50%, and then switched to constant voltage charging until the charging current drops to a rate of 0.05 times the current, so that the battery cell reaches a state of 50% charge.

[0006] The multi-rate pulse charge-discharge test specifically includes: The battery cells are subjected to constant current pulse discharge at the first preset rate, the second preset rate and the third preset rate in sequence. The duration of each pulse discharge is ten seconds, and there is a rest period of twenty seconds between each two pulse discharges. The first preset multiplier is 0.5x, the second preset multiplier is 1.5x, and the third preset multiplier is 2.5x. At the moment each pulse discharge ends, the terminal voltage and current values ​​of the battery cells are simultaneously acquired and recorded.

[0007] The voltage relaxation test specifically includes: After the multi-rate pulse charge-discharge test is completed, immediately disconnect the charge-discharge circuit to keep the battery cells in an open circuit state. The open-circuit voltage is continuously recorded from the start of the settling period, with a total recording time of sixty seconds. In the first thirty seconds, a voltage point is recorded every 0.5 seconds, and in the last thirty seconds, a voltage point is recorded every second. The voltage rise rate between the second and fifth seconds after the start of rest was extracted from the recorded open-circuit voltage data and used as the ion diffusion relaxation characteristic value.

[0008] The calculation method for the dynamic feature parameter set is as follows: The first DC internal resistance is calculated based on the voltage and current at the moment the first multiplier pulse ends; the second DC internal resistance is calculated based on the voltage and current at the moment the second multiplier pulse ends; and the third DC internal resistance is calculated based on the voltage and current at the moment the third multiplier pulse ends. The difference between the second DC internal resistance and the first DC internal resistance is used as the polarization internal resistance growth coefficient. The difference between the third DC internal resistance and the second DC internal resistance is taken as the large current polarization characteristic value. The voltage rise rate between the second and fifth seconds of the voltage relaxation test was used as the characteristic value of ion diffusion relaxation.

[0009] The methods for constructing the capacity decay prediction model include: Sample batteries are extracted from the production batch, and the dynamic characteristic parameter set of each sample battery is extracted according to the methods of adjusting the state of charge, multi-rate pulse test and voltage relaxation test. The sample batteries were subjected to accelerated cycle aging tests, and the annual capacity decay rate of each sample battery was recorded when the initial capacity decayed to 80% of the initial capacity. Using the dynamic characteristic parameter set of the sample battery as input and the corresponding annual capacity decay rate as the output label, the capacity decay prediction model is trained using an ensemble regression algorithm.

[0010] The sorting levels include: the first level corresponds to a predicted annual capacity decay rate of less than 5%, the second level corresponds to a predicted annual capacity decay rate between 5% and 7%, and the third level corresponds to a predicted annual capacity decay rate between 7% and 9%. The preset rejection threshold is nine percent. Battery cells with a predicted annual capacity decay rate higher than nine percent are judged as degraded products and rejected.

[0011] Fine-tuning the matching specifically includes: Within the same sorting level, the individual battery cells are sorted from largest to smallest according to their initial capacity. Several adjacent batteries are grouped into a coarse group, so that the range of initial capacity within the group is no more than 1.5 percent of the nominal capacity. Provided that the capacity range meets the requirements, calculate the initial internal resistance range of each battery cell in each coarse group. If the internal resistance range is greater than 10% of the average internal resistance in the group, then swap batteries between adjacent groups until the internal resistance range meets the requirements. If the requirements cannot be met through exchange, a weighted comprehensive scoring method is used for rematching, where the initial capacity has a weight of 70% and the initial internal resistance has a weight of 30%.

[0012] The total duration of the multi-rate pulse charge-discharge test and the voltage relaxation test is controlled within two minutes.

[0013] In a second aspect, the present invention provides a predictive sorting system for the capacity degradation of batteries for two-wheeled electric vehicles, applied to a predictive sorting method for the capacity degradation of batteries for two-wheeled electric vehicles as provided in the first aspect, comprising: The state of charge adjustment module is used to adjust each battery cell to be sorted to the same preset state of charge. The test data acquisition module is used to sequentially perform multi-rate pulse charge-discharge tests and voltage relaxation tests on the battery cells in the preset state of charge, and to acquire the voltage value at the moment the pulse discharge ends and the open-circuit voltage change data over time during the relaxation process. The feature parameter calculation module is used to calculate the dynamic feature parameter set of each battery cell based on the collected data. The dynamic feature parameter set includes at least the polarization internal resistance growth coefficient, the high current polarization feature value, and the ion diffusion relaxation feature value. The capacity degradation prediction module has a pre-built capacity degradation prediction model built in, which is used to receive the dynamic feature parameter set and output the predicted annual capacity degradation rate of each battery cell. The sorting execution module is used to set multiple sorting levels according to the predicted annual capacity decay rate, classify the battery cells into the corresponding levels, and remove battery cells whose predicted annual capacity decay rate is higher than the preset rejection threshold. The fine-tuning matching module is used to further fine-tune the matching of individual battery cells within the same sorting grade based on their initial capacity and initial internal resistance, and output the matching results of the battery pack.

[0014] This invention discloses a method and system for predictive sorting of battery capacity decay for two-wheeled electric vehicles. After adjusting the batteries to be sorted to the same state of charge, multi-rate pulse charge-discharge tests and voltage relaxation tests are performed sequentially, collecting pulse voltage and relaxation open-circuit voltage data. Based on the collected data, the polarization internal resistance growth coefficient, high-current polarization characteristic value, and ion diffusion relaxation characteristic value are calculated to form a dynamic characteristic parameter set. This parameter set is input into a pre-constructed capacity decay prediction model, which outputs the predicted annual capacity decay rate for each battery. Multiple sorting levels are set according to the predicted decay rate, and unqualified products exceeding the threshold are eliminated. Batteries in the same level are then fine-tuned and matched according to their initial capacity and initial internal resistance before being grouped together. This invention achieves a leap from static parameter sorting to dynamic decay behavior prediction sorting, significantly improving the consistency of capacity decay in the grouped battery packs, effectively suppressing the "weakest link" effect, extending the battery pack's lifespan, and without increasing production cycle time. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0016] Figure 1 This is a schematic diagram of the steps in a method for predicting capacity decay of batteries for two-wheeled electric vehicles according to the first embodiment of the present invention.

[0017] Figure 2 This is a simplified flowchart illustrating a method for predictive sorting of battery capacity decay in two-wheeled electric vehicles provided by the present invention.

[0018] Figure 3 This is a complete flowchart of the predictive sorting method for capacity decay of two-wheeled electric vehicle batteries provided by the present invention.

[0019] Figure 4This is a schematic diagram of the structural principle of the two-wheeled electric vehicle battery capacity decay prediction sorting system according to the second embodiment of the present invention.

[0020] Figure 5 This is a schematic diagram of the electronic device of the present invention.

[0021] In the diagram: 101-State of Charge Adjustment Module, 102-Test Data Acquisition Module, 103-Characteristic Parameter Calculation Module, 104-Capacity Attenuation Prediction Module, 105-Sorting Execution Module, 106-Fine-tuning Matching Module. Detailed Implementation

[0022] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0023] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0024] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0025] The first embodiment of this application is as follows: Please see Figures 1-3 This invention provides a method for predictive sorting of battery capacity degradation for two-wheeled electric vehicles, comprising the following steps: S1. Adjust each battery cell to be sorted to the same preset state of charge. Perform multi-rate pulse charge-discharge test and voltage relaxation test on the battery cells in the preset state of charge in sequence. Collect the voltage value at the moment the pulse discharge ends and the open circuit voltage change data over time during the relaxation process.

[0026] Specifically, to eliminate the impact of differences in state of charge (SOC) on subsequent test parameters, all individual cells must be standardized to the same SOC reference point. This invention selects a reference SOC of 50%. The adjustment process is performed in the following order: Pre-discharge treatment: First, using a standard battery charge / discharge test cabinet on the production line, the batteries to be sorted are discharged at a constant current rate of 0.5 times the battery manufacturer's stated current. During the discharge process, the terminal voltage of each battery cell is monitored in real time until it drops to the discharge cutoff voltage specified by the battery supplier (e.g., for ternary lithium batteries, the cutoff voltage is 2.8 volts per cell). The purpose of this step is to essentially deplete the residual charge in the battery and establish a clear zero-charge reference point.

[0027] Constant current and constant voltage charging to the target state of charge: After discharging, the battery is immediately charged at a constant current rate of 0.5 times the rated current. Simultaneously, the amount of charge added is calculated in real time using the ampere-hour integration method. Constant current charging ends when the added charge reaches 50% of the battery's nominal capacity. To further improve the accuracy of the state of charge, a constant voltage charging stage is then initiated: The charging voltage is set to the reference value of the battery's open-circuit voltage at this state of charge (e.g., for a battery with a nominal voltage of 3.6 volts, the open-circuit voltage corresponding to 50% state of charge is approximately 3.65 to 3.70 volts), and this voltage is maintained during charging until the charging current drops to 0.05 times the rated current. At this point, the battery is considered to have accurately reached 50% state of charge and has a low degree of internal polarization.

[0028] Quick Confirmation and Calibration: To verify the adjustment results, immediately measure the battery's open-circuit voltage after adjustment. If the voltage falls within the allowable fluctuation range of 50% of the corresponding state of charge (e.g., ±0.01 volts), the adjustment is considered successful; otherwise, repeat the charging steps described above once. The total time for this adjustment process is usually controlled within 40 seconds.

[0029] Parameter range and setting basis: Reference state of charge: 50% is chosen because this state of charge is in the middle region where the battery open-circuit voltage-state of charge curve is flattest, making it insensitive to small voltage errors. At the same time, the internal electrochemical reaction of the battery is relatively stable, which is conducive to obtaining repeatable pulse response and relaxation characteristics.

[0030] Charge / discharge rate (0.5 times): 0.5 times is a safety rate commonly used in the routine formation and testing of batteries for two-wheeled electric vehicles. It can ensure that the capacity adjustment is completed quickly without causing obvious side reactions or temperature rise.

[0031] Constant voltage charging cutoff current (0.05 times): This value ensures that the battery has basically reached electrochemical equilibrium, avoiding residual polarization from affecting subsequent tests.

[0032] After completing the state of charge adjustment, the individual battery cells were immediately subjected to multi-rate pulse charge-discharge tests. The purpose of this test was to measure the battery's DC internal resistance and polarization characteristics at different current rates. The test procedure is as follows: First-rate pulse discharge: The battery is first subjected to a constant-current pulse discharge at a first preset rate. The first preset rate is set to 0.5 times. The duration of the pulse discharge is ten seconds. Before the discharge begins, the initial open-circuit voltage and temperature of the battery are recorded; at the moment the discharge pulse ends (i.e., at the end of the tenth second), the battery's terminal voltage and discharge current values ​​are simultaneously acquired and recorded (the current value is known and is the current setting value). Subsequently, the discharge is stopped, and the battery is allowed to rest for twenty seconds to allow the voltage to partially recover. First preset rate (0.5 times): This rate is close to the average discharge rate of the battery during normal use and is used to measure the reference DC internal resistance as a baseline for subsequent comparisons.

[0033] Second-rate pulse discharge: After the above resting period, immediately perform a second constant-current pulse discharge at the second preset rate. The second preset rate is set to 1.5 times. The pulse duration is also ten seconds. At the moment the pulse ends, the battery terminal voltage and current values ​​are collected and recorded again. After the discharge is completed, allow it to rest for another twenty seconds. Second preset rate (1.5 times): This rate simulates medium-load conditions such as acceleration and hill climbing in a two-wheeled electric vehicle. By comparing the difference in DC internal resistance between 1.5 times and 0.5 times, the degree of increase in polarization internal resistance of the battery at the medium rate can be reflected.

[0034] Third-rate pulse discharge: Finally, a third constant-current pulse discharge is performed at the third preset rate. The third preset rate is set to 2.5 times. The pulse duration is ten seconds. The terminal voltage and current values ​​are collected at the moment the pulse ends. This completes the multi-rate pulse test. The entire test process (including the rest interval) takes approximately (10+20+10+20+10) = seventy seconds. Third preset rate (2.5 times): This rate is close to the typical upper limit of short-term high-current output (such as rapid acceleration) in two-wheeled electric vehicles. The internal resistance at 2.5 times and the difference from the previous step can significantly amplify the battery's internal charge transfer impedance and ohmic polarization defects, making it very sensitive to predicting long-term degradation. Pulse duration (ten seconds): Ten seconds is a commonly used short pulse duration in industry, sufficient to fully establish polarization without causing a significant temperature rise. Too long a duration will reduce production line cycle time, while too short a duration will result in incomplete polarization. Resting time (20 seconds): Twenty seconds allows the battery voltage to recover to a relatively stable level, avoiding interference from residual polarization from the previous pulse in the next measurement. Experiments have verified that this time is sufficient for most depolarization in lithium-ion batteries.

[0035] The voltage values ​​at the instant the pulse ends are denoted as V1 (0.5C), V2 (1.5C), and V3 (2.5C), and the corresponding current values ​​are denoted as I1, I2, and I3. In the subsequent second step, these data will be used to calculate the DC internal resistance at different rates: DC internal resistance = (open-circuit voltage - pulse end voltage) / pulse current. Note that the open-circuit voltage is the measured voltage at the instant before each pulse begins (which can be approximately equal to the stable voltage after the state-of-charge adjustment is completed, or the voltage before the first pulse).

[0036] After completing the multi-rate pulse test, varying degrees of polarization have accumulated inside the battery. At this point, a voltage relaxation test is immediately performed to capture the characteristics of ion diffusion and redistribution within the battery.

[0037] The testing process is as follows: Entering a static state: After the third-rate pulse discharge ends and data acquisition is completed, immediately disconnect all charging and discharging circuits to put the battery in an open-circuit state. At the same time, start the high-precision voltage acquisition system (sampling frequency not less than once per second).

[0038] Continuous recording of open-circuit voltage: Starting from the moment of rest (recorded as zero seconds), continuously record the open-circuit voltage value of the battery. The recording duration is set to sixty seconds. During the recording process, record one voltage point every 0.5 seconds in the first thirty seconds, and record one voltage point every second in the next thirty seconds, to balance data accuracy and storage capacity.

[0039] Extracting key features: From the recorded open-circuit voltage versus time curves, focus on the voltage rise rate during the initial settling period (e.g., the first five to ten seconds). Specifically, divide the voltage difference between the second and fifth seconds after the start of settling by the time difference to obtain the voltage rise rate per unit time. This rate is the ion diffusion relaxation characteristic value required for the subsequent second step. The reason for choosing the second to fifth seconds is that the voltage change in the first second is significantly affected by ohmic polarization recovery, while from the second second onwards, it primarily reflects the diffusion process of ions in the solid phase.

[0040] Total recording time (60 seconds): 60 seconds is sufficient to capture the main characteristics of diffusion relaxation without significantly increasing the overall test time. In a two-minute total budget on the production line, 70 seconds are used for pulse testing, 60 seconds for relaxation testing, plus 40 seconds for adjusting the state of charge, totaling 170 seconds (approximately 2 minutes and 50 seconds). If strict control within two minutes is required, relaxation recording can be appropriately shortened to 40 seconds or the state of charge adjustment steps can be optimized. However, this solution prioritizes data integrity, and the actual production cycle time can be addressed through parallel workstations. In the preferred embodiment, adjusting the state of charge can be achieved using a rapid charge-discharge method (e.g., directly discharging to the voltage point corresponding to 50% state of charge, combined with short-term constant voltage), compressing the adjustment time to within 20 seconds.

[0041] Recording frequency: The high frequency (0.5 seconds / point) for the first 30 seconds is to accurately capture the initial rapid rise phase, while the low frequency for the following 30 seconds is used to observe the slow changing trend. However, for the prediction model, the initial rise rate is sufficient. Feature extraction period (second to fifth second): This period avoids the abrupt change region of ohmic polarization recovery (0-1 second) and also avoids the later voltage change being too gradual and losing sensitivity. Experiments show that the voltage rise rate during this period is strongly correlated with the lithium-ion diffusion coefficient of the battery negative electrode, and thus directly related to the long-term capacity decay rate.

[0042] S2. Calculate the dynamic characteristic parameter set for each battery cell based on the collected data. The dynamic characteristic parameter set includes at least the polarization internal resistance growth coefficient, the high current polarization characteristic value, and the ion diffusion relaxation characteristic value.

[0043] Specifically, in the first step, each battery cell obtained the following raw data: The voltage value at the instant the first rate (0.5 rate) pulse discharge ends is denoted as Voltage 1, and the corresponding current value (known as the 0.5 rate current, denoted as Current 1).

[0044] The voltage value at the moment the second-rate (1.5-rate) pulse discharge ends is denoted as voltage two, and the corresponding current value (1.5-rate current, denoted as current two).

[0045] The voltage value at the moment the third-rate (2.5-rate) pulse discharge ends is denoted as voltage three, and the corresponding current value (2.5-rate current, denoted as current three).

[0046] In the voltage relaxation test, the open-circuit voltage value and the corresponding time point are recorded at regular intervals from the second to the fifth second after the start of the rest period.

[0047] Based on the above data, this step calculates four key feature parameters in sequence and combines them into a feature vector for subsequent capacity decay prediction.

[0048] First, it is necessary to obtain the DC internal resistance value of each battery cell at three different discharge rates. The basic logic of the calculation is that the DC internal resistance reflects the sum of the ohmic resistance and partial polarization resistance inside the battery at the moment of discharge. The specific calculation method is as follows: For the first rate pulse, before the discharge begins, the initial open-circuit voltage of the battery is known (this voltage was measured and recorded at the end of the first step of state-of-charge adjustment, and is denoted as the initial open-circuit voltage). At the instant the pulse ends, the voltage drops to voltage one, and the current is current one. Therefore, the DC internal resistance at the first rate is equal to the difference between the initial open-circuit voltage and voltage one, divided by current one. This calculated result is called the first DC internal resistance. It mainly represents the battery's overall internal resistance at a relatively small current, including ohmic internal resistance and part of the activation polarization.

[0049] Similarly, for the second-rate pulse, the battery has already experienced one pulse and one rest period before the second pulse begins. Although the battery voltage recovers somewhat after a 20-second rest period, it may not have fully returned to its initial open-circuit voltage. To more accurately calculate the DC internal resistance at the second rate, the voltage at the instant before the pulse begins is used as the reference voltage. This reference voltage can be directly obtained from the data log (i.e., the open-circuit voltage measured one second before the second pulse discharge, denoted as pulse-before voltage two). Therefore, the DC internal resistance at the second rate is equal to the difference between pulse-before voltage two and voltage two, divided by current two. The calculated result is called the second DC internal resistance.

[0050] Similarly, for the third-rate pulse, take the open-circuit voltage (pulse-before-pulse voltage three) at the instant before the third pulse begins, subtract voltage three, and divide by current three to obtain the third DC internal resistance.

[0051] The three DC internal resistance values ​​mentioned above each reflect the voltage response characteristics of the battery under different current loads. Generally speaking, the larger the current, the larger the measured DC internal resistance, because the polarization phenomenon is more significant.

[0052] Knowing only the three DC internal resistance values ​​is insufficient; it is also necessary to extract characteristics that reflect the increasing trend of internal resistance as the current increases. Therefore, the polarization internal resistance growth coefficient is calculated using the following method: Subtracting the first DC internal resistance from the second DC internal resistance yields a difference. This difference represents the increase in internal resistance when the current increases from 0.5 times to 1.5 times. As the current increases, both electrochemical polarization and concentration polarization within the battery intensify. The magnitude of this difference directly reflects the battery's sensitivity to moderate current loads. This difference is named the polarization resistance growth factor.

[0053] The larger this coefficient, the faster the battery's internal resistance increases with increasing current, indicating a higher charge transfer impedance or insufficient ion diffusion capability within the battery. Subsequent experiments showed that this coefficient is positively correlated with the battery's capacity decay rate during long-term cycling.

[0054] To further capture the battery polarization behavior under higher currents (approaching the peak operating conditions of two-wheeled electric vehicles), the high-current polarization characteristic values ​​are calculated using the following method: Subtracting the second DC internal resistance from the third DC internal resistance yields the second difference. This difference represents the additional increase in internal resistance when the current increases from a factor of 1.5 to 2.5. This is named the high-current polarization characteristic value.

[0055] Unlike the polarization resistance growth coefficient, the high-current polarization characteristic value more sensitively reflects the polarization limit of the battery under high-rate discharge. If this value is abnormally large, it indicates that there may be defects inside the battery such as obstructed lithium-ion diffusion, unstable electrode structure, or poor electrolyte wetting. These defects will accelerate capacity decay with the increase of cycle number.

[0056] Next, ion diffusion relaxation characteristic values ​​were extracted from the voltage relaxation test data. This characteristic reflects the rate at which internal ions redistribute and the voltage recovers after the battery stops discharging, and is directly related to the diffusion coefficient of lithium ions in the solid phase of the electrode material.

[0057] The specific extraction method is as follows: In the voltage relaxation test, continuous open-circuit voltage values ​​were recorded after the initial rest period. The time period from the second to the fifth second after the initial rest period was selected. First, the voltage value recorded at the second second was recorded as the relaxation initiation voltage; then, the voltage value recorded at the fifth second was recorded as the relaxation termination voltage. Next, the difference between the relaxation termination voltage and the relaxation initiation voltage was calculated, and then divided by the time difference (i.e., three seconds) to obtain the voltage rise rate per unit time. This rate is the characteristic value of ion diffusion relaxation.

[0058] The second to fifth seconds were chosen because during the first second after settling, the voltage rise primarily stems from the rapid elimination of ohmic polarization. During this phase, the voltage change is more closely related to internal resistance than to ion diffusion. From the second second onwards, ohmic polarization essentially disappears, and the continued voltage rise is mainly driven by ion diffusion in the solid phase. Therefore, the average rise rate between the second and fifth seconds more accurately reflects the ion diffusion capability. The weaker the diffusion capability, the slower the voltage rise, and the smaller this characteristic value.

[0059] After calculating the above four feature parameters, they are combined in a fixed order to form a dynamic feature parameter set, also known as a feature vector. This feature vector specifically contains the following four elements: Polarization internal resistance growth factor (obtained by subtracting the first DC internal resistance from the second DC internal resistance); High-current polarization characteristic value (obtained by subtracting the second DC internal resistance from the third DC internal resistance); Ion diffusion relaxation characteristic value (obtained from the average voltage rise rate from the second to the fifth second of the relaxation test); To increase the robustness of the prediction, the third DC internal resistance (i.e., the DC internal resistance at 2.5 times the rate) can also be selectively included in the eigenvector as a reference for the absolute value of the total internal resistance under high current.

[0060] Therefore, the final eigenvector can be expressed as: (polarization internal resistance growth coefficient, high-current polarization eigenvalue, ion diffusion relaxation eigenvalue, third DC internal resistance). The first three are derived features after difference or slope processing, and have stronger physical meaning and predictive ability.

[0061] S3. Input the set of dynamic characteristic parameters into the pre-built capacity decay prediction model to obtain the predicted annual capacity decay rate of each battery cell.

[0062] Specifically, the capacity decay prediction model used in this invention is a nonlinear regression model based on ensemble learning. The model's structure can be figuratively understood as a "black box mapper," receiving four feature values ​​as input and producing a continuous value at output—the predicted annual capacity decay rate. Internally, the model consists of multiple weak learners (e.g., decision trees). Each weak learner provides a preliminary decay rate estimate based on the feature vector, and these preliminary estimates are then merged into the final output through weighted averaging or voting. This ensemble structure has strong anti-overfitting capabilities and can automatically capture the complex nonlinear relationship between features and the decay rate without requiring manually defined empirical formulas.

[0063] The model's input layer has four fixed nodes, corresponding to the four feature parameters mentioned above. The model's output layer has one node, with output values ​​ranging from 3% to 15%, covering the common annual degradation range of two-wheeled electric vehicle batteries. The model's intermediate layers (i.e., the number and connection method of weak learners) are automatically determined during training, requiring no manual design by the user.

[0064] Building this model requires the following five sub-steps: Sub-step 1: Extract sample batteries from the production batch.

[0065] Before mass production begins, a certain number of battery cells are randomly selected as samples from the same production batch (same material system, same process parameters, same batch number) to be sorted. The sample size is typically between one hundred and three hundred cells, depending on the scale of the production line and the acceptable modeling cost. The sample should cover normal cells as well as cells with potential boundary performance (e.g., individuals with high or low internal resistance) to ensure the model's generalization ability.

[0066] Sub-step 2: Perform accelerated cycle aging tests on the sample batteries and record the actual capacity decay data.

[0067] All sampled batteries were placed in an accelerated aging test apparatus. Test conditions should simulate typical usage of a two-wheeled electric vehicle, such as constant current charging to full capacity at a 1x current rate, resting for ten seconds, then constant current discharging at a 1x current rate to the cutoff voltage, and repeating this cycle. To accelerate aging, the ambient temperature could be set to 45 degrees Celsius, and a slightly higher charge / discharge rate (e.g., 1.5x) could be used. During the aging test, every 100 cycles, the test was paused, and the actual remaining capacity of each sample battery at room temperature (the capacity discharged at a standard 0.5x rate to the cutoff voltage) was measured. This process was repeated until the remaining capacity of each sample battery decayed to 80% of its initial capacity. The total number of cycles experienced by each sample battery from the start to 80% decay was recorded, or more directly, the annual capacity decay rate of each sample battery was calculated. The annual capacity decay rate was calculated as follows: the initial capacity minus the capacity difference at 80% decay, divided by the initial capacity, and then divided by the number of years converted from the total number of test cycles (e.g., 365 cycles per day based on one full charge / discharge cycle per day, corresponding to one year). For cases where the annual decay rate has not yet fully decayed to 80% but testing time is limited, extrapolation can be used to estimate the annual decay rate, but it is preferable to actually measure 80%.

[0068] Sub-step 3: Obtain the dynamic feature parameter set for each sample battery.

[0069] Before starting the accelerated aging test, for each sample battery, a rapid offline inspection procedure was strictly performed according to the methods described in steps one and two: the state of charge was adjusted to 50%, multi-rate pulse testing and voltage relaxation testing were conducted, and the polarization internal resistance growth coefficient, high-current polarization characteristic value, ion diffusion relaxation characteristic value, and third DC internal resistance of each sample battery were calculated. This yielded the input feature vector for each sample battery. Note that the acquisition of these feature vectors must be completed before the aging test to ensure that they represent the initial state of the battery.

[0070] Sub-step four: Train the regression model.

[0071] Using the feature vectors of all sample batteries obtained in sub-step three as input and the actual annual capacity decay rate of each sample battery obtained in sub-step two as the output label, a training dataset is formed. Then, an ensemble regression algorithm (such as random forest regression or gradient boosting regression tree), suitable for small sample sizes and strong at modeling nonlinear relationships, is used to train this dataset. The training process essentially seeks an optimal mapping from a four-dimensional feature space to a one-dimensional decay rate value. Specifically, the algorithm automatically learns from the training data by randomly generating multiple decision trees, each splitting based on a subset of samples and features, and finally averaging the results of all trees as the output. After training, the model's internal parameters (i.e., the node splitting conditions and leaf node values ​​for each tree) are fixed. To verify the model's prediction accuracy, the training dataset is typically randomly divided into two parts: 80% for training and 20% for validation, ensuring that the model's prediction error (e.g., mean absolute error) on unseen samples is less than 0.5 percentage points.

[0072] Sub-step five: Model solidification and deployment.

[0073] The trained model parameters are saved as a file or embedded into the production line's inspection and control software. This model can be used continuously for the same production line and the same battery model until significant changes occur in the battery's material system or manufacturing process. When changes occur, samples need to be re-extracted and sub-steps one through four above need to be re-executed to update the model.

[0074] For each battery entering the offline inspection station, its predicted annual capacity degradation rate is obtained according to the following process: The first step is feature extraction. After processing in the first and second steps, the battery's polarization internal resistance growth coefficient, high-current polarization characteristic value, ion diffusion relaxation characteristic value, and third DC internal resistance are calculated in real time, forming the battery's feature vector.

[0075] The second step is model inference. The feature vector is input into the capacity degradation prediction model already embedded in the production testing software. Internally, the model automatically performs the following operations: the feature vector is simultaneously fed into each decision tree within the model. Each tree, based on the splitting rules learned during training, guides the feature vector layer by layer to a leaf node. This leaf node stores a value, representing the tree's "vote" for the battery's annual capacity degradation rate. Then, the model arithmetically averages the votes from all decision trees to obtain the final output value. The entire inference process is completed in milliseconds.

[0076] The third step is to output the prediction results. The model outputs a value with a percentage sign, such as "6.3 percent", which represents the expected annual capacity degradation rate of the battery under normal use conditions (one full charge and discharge cycle per day, average ambient temperature of 25 degrees Celsius). This value is displayed in real time on the workstation screen and recorded in the battery's barcode or QR code information for use in the fourth step of sorting.

[0077] S4. Set multiple sorting levels according to the predicted annual capacity decay rate, classify the battery cells into the corresponding levels, and remove battery cells whose predicted annual capacity decay rate is higher than the preset rejection threshold.

[0078] Specifically, the number and boundary values ​​of sorting levels are not arbitrarily determined, but are scientifically set based on the following three criteria: Basis 1: Typical lifespan targets for battery packs in two-wheeled electric vehicles.

[0079] Two-wheeled electric vehicle users typically expect battery packs to function normally for two to three years, with the maximum range decreasing by no more than 20% of the initial value during this period. Converting this to annual capacity degradation, if the annual degradation rate is controlled within 7%, the total degradation over three years will be approximately 21%, generally meeting user expectations. If the annual degradation rate exceeds 9%, the total degradation over three years could reach over 27%, significantly worsening the user experience and increasing the likelihood of insufficient capacity complaints during the warranty period. Therefore, using 9% as a rejection threshold is reasonable.

[0080] Basis 2: The allowable difference in degradation rate between individual cells within the battery pack.

[0081] When assembling batteries, if the annual capacity degradation rates of individual cells within the same group differ significantly—for example, one at 5% and another at 8%—the capacity difference may exceed 3% after one year of use, resulting in a significant "weakest link" effect. To control the maximum capacity difference within a group within an acceptable range (e.g., no more than 5% over two years), the intra-group range of annual degradation rates needs to be controlled within 2%. Accordingly, the width of the sorting gradations can be set to 2%, for example, the first gradation being below 5%, the second gradation being 5% to 7%, and the third gradation being 7% to 9%. This way, the maximum range within the same gradation is 2%, meeting the design requirements.

[0082] Basis 3: The economic balance between the actual production capacity and the pass rate of the production line.

[0083] The finer the sorting levels, the fewer batteries are in each level after sorting, making matching during grouping more difficult and potentially leading to battery backlog. Simulations of production lines at different production volumes have shown that setting three to four effective sorting levels (plus one rejection level) is a cost-effective approach. Fewer than three levels result in insufficient improvement in intra-group consistency; more than four levels significantly increase the complexity of production line sorting and inventory management, while offering limited additional benefits.

[0084] Based on the above, the preferred gear setting of the present invention is as follows: First tier (premium): Predicted annual capacity degradation rate is less than 5%. These batteries degrade extremely slowly and are suitable for high-end vehicles or applications requiring high range retention.

[0085] The second tier (standard tier): The predicted annual capacity degradation rate is between 5% and 7%. This type of battery meets the normal usage requirements of most two-wheeled electric vehicles and is the main tier with the highest production volume.

[0086] The third tier (qualified): The predicted annual capacity degradation rate is between 7% and 9%. These batteries degrade slightly faster, but can still be used in low-intensity scenarios or as spare parts.

[0087] Elimination category (substandard products): Batteries with a predicted annual capacity degradation rate exceeding 9%. These batteries will degrade rapidly during future use, directly causing premature battery pack failure, and therefore should not be included in any assembly process.

[0088] For the boundary values ​​of the above gear levels, the intervals belonging to "below" and "between" can be divided using either left-closed and right-open or left-open and right-closed, but this must be consistent throughout the entire production process. For example, the second gear level is defined as greater than or equal to 5% and less than 7%.

[0089] The sorting process is completed in an automated offline sorting station, which is directly connected to the output of the prediction model in step three. The specific process is as follows: Sub-step 1: Obtain and record the prediction results.

[0090] Once a battery cell has completed the third step of model inference, its predicted annual capacity degradation rate is sent to the central control system of the production line. The control system then associates this value with the battery's unique identifier (such as a barcode or QR code) and stores it in a database.

[0091] Sub-step two: Gear selection.

[0092] The control system compares and judges the predicted annual capacity degradation rate based on preset threshold values. The comparison process is entirely executed automatically by a programmable logic controller or industrial computer. For example, if the predicted value of a battery is 6.3%, the control system determines that it belongs to the second threshold (5% to 7%); if the predicted value is 9.2%, it is determined to be in the eliminated threshold.

[0093] Sub-step 3: Physical sorting action.

[0094] The individual battery cells move sequentially on a conveyor belt, passing through a sorting actuator driven by a cylinder or servo motor. The control system sends commands to the actuator based on the gear selection result. There are two common implementation methods: Type 1 (Push Rod Sorting): Multiple collection slots are provided on both sides of the conveyor belt, each corresponding to a specific gear position. When a battery reaches the collection slot corresponding to its gear position, the control system triggers a solenoid valve, which drives a push rod to push the battery from the conveyor belt into the collection slot.

[0095] Form 2 (robotic gripper type): A multi-axis robotic arm follows the movement of the battery, grabs the battery according to the gear signal, and places it into the corresponding gear tray.

[0096] Regardless of the method used, the timing of the sorting action needs to be precisely controlled. Typically, before the battery enters the sorting station, it passes through a photoelectric sensor to detect its position. The system calculates the delay time based on the conveyor belt speed and the distance from the sensor to the sorting execution point, and issues a sorting command when the battery just reaches the execution point.

[0097] Sub-step four: Remove gear positions.

[0098] Batteries deemed unsuitable for rejection (predicted annual capacity degradation rate exceeding 9%) are marked as "degraded" by the control system and sent to a dedicated scrap or repair channel. These batteries will not be sent to subsequent assembly processes or stored with normal batteries. If necessary, rejected batteries can be disassembled and analyzed to provide feedback for improving upstream processes.

[0099] Sub-step 5: Recording and tracing the sorting results.

[0100] Each collection trough or tray at each sorting level has a corresponding barcode. The control system stores the identification code of each battery along with its sorting level, sorting time, and sorting station number into the production database. This allows for quick retrieval of all available batteries based on their sorting level during subsequent batching, and the predicted degradation value of each battery can be traced back to the original test data.

[0101] S5. For battery cells within the same sorting level, fine-tune and match them according to their initial capacity and initial internal resistance to form a battery pack.

[0102] Specifically, during the initial state-of-charge adjustment process and before the multi-rate pulse test, the system recorded the following two static parameters for each battery cell: Initial capacity: This parameter is the actual discharge capacity of the battery measured during the initial formation and capacity testing process, typically expressed in ampere-hours. This value is recorded under the battery's unique identifier and stored in association with the data from the first and second steps of this invention.

[0103] Initial internal resistance: This refers to the AC internal resistance measured using the AC impedance method when the battery is at 50% charge, typically in milliohms. This parameter is also obtained during the formation and capacity testing process, or the first DC internal resistance calculated using the first rate pulse in the first step of this invention can be used as a reference. To maintain consistency with industry practice, this invention preferentially uses the AC internal resistance value.

[0104] When entering the fifth step, the system reads the initial capacity and initial internal resistance values ​​of all battery cells within the same degradation rate range from the database.

[0105] The goal of fine-tuning matching is to group battery cells with the same predicted degradation rate (i.e., belonging to the same range) according to the similarity of their initial capacity and initial internal resistance, so that the difference in initial capacity among all cells within the same battery group does not exceed a threshold, and the difference in initial internal resistance also does not exceed a threshold. The specific rules are as follows: Rule 1: Capacity priority principle.

[0106] Since initial capacity directly affects the initial driving range of a battery pack, and capacity differences have the greatest impact on battery packs used in series, they are first sorted and grouped according to their initial capacity. The maximum allowable difference in capacity matching is set at 1.5 percent of the nominal capacity. For example, for a battery with a nominal capacity of 20 Ah, the maximum capacity difference within a group must not exceed 0.3 Ah. If the capacity distribution of batteries within the same category is wide, batteries with similar capacities are first grouped together by capacity sorting.

[0107] Rule 2: Internal resistance-assisted matching.

[0108] Assuming the capacity difference meets the requirements, the initial internal resistance is further considered. The maximum allowable difference in internal resistance matching is set at 10% of the average internal resistance. For example, if the average internal resistance of the batteries in the group is 20 milliohms, the maximum internal resistance difference must not exceed 2 milliohms. If the internal resistance difference exceeds this range, fine-tuning and replacement with candidate batteries of similar capacity are required until the internal resistance difference meets the standard.

[0109] Rule 3: Use a weighted comprehensive scoring method for optimal matching.

[0110] When there are many batteries in the same category (e.g., more than one hundred) that need to be assembled into multiple battery packs, relying solely on sorting and manual selection is inefficient. This invention employs a weighted comprehensive scoring method to achieve automated optimal matching. Specifically, the initial capacity and initial internal resistance of each battery are normalized, and then a comprehensive score is calculated with capacity accounting for 70% and internal resistance for 30%. Next, all batteries are sorted in ascending order of their comprehensive scores, and adjacent batteries (e.g., four, six, or eight, depending on the series-parallel structure of the battery pack) are directly grouped together. This method can ensure minimal capacity differences while maintaining consistency in internal resistance.

[0111] The following example illustrates the complete process of fine-tuning and matching. Suppose a production line needs to assemble a battery pack with four batteries connected in series. Currently, there are 400 battery cells waiting to be assembled within the standard range (predicted annual capacity degradation rate between 5% and 7%).

[0112] Sub-step 1: Data filtering and sorting.

[0113] The system extracts the initial capacity and initial internal resistance values ​​of these 400 batteries from the database. First, it removes batteries with an initial capacity lower than 95% of their nominal capacity (these batteries, although meeting the degradation rate requirements, have too low an initial capacity and are unsuitable for normal assembly; they can be downgraded to spare). This leaves 380 batteries. Then, they are sorted by initial capacity from largest to smallest; if the capacities are the same, they are sorted by internal resistance from smallest to largest.

[0114] Sub-step 2: Coarse grouping.

[0115] The 380 sorted batteries are then divided into groups of four, resulting in 95 coarse groups. Since the capacities of adjacent batteries are very close after sorting, the capacity range within each coarse group is typically less than 1.5%. If the capacity range of a coarse group exceeds a threshold, the system will flag that group and address it in subsequent fine-tuning.

[0116] Sub-step 3: Internal resistance detection and fine-tuning.

[0117] For each coarse group, the difference between the maximum and minimum internal resistance of the four batteries within the group is calculated. If the internal resistance range is less than or equal to 10% of the average internal resistance, the group is accepted and proceeds directly to the grouping process. If the internal resistance range exceeds the threshold, a fine-tuning procedure is triggered: the system searches for batteries with closer internal resistance values ​​near the current coarse group (e.g., the previous and next groups) for local swapping, while ensuring that the capacity range of the two groups still meets the requirements after the swap. For example, if the internal resistances of the first group are 18, 20, 22, and 25 milliohms, with a range of 7 milliohms, exceeding the allowable value of 2 milliohms, the system can find a battery with an internal resistance of 20 milliohms in the second group and swap it with the battery with an internal resistance of 25 milliohms in the first group. After the swap, the internal resistances of the first group become 18, 20, 22, and 20 (range of 4 milliohms), still exceeding the limit, so further swapping is required. If the internal resistance range requirement cannot be met after three attempts, the coarse group is broken up, and the four batteries are placed in the matching pool to be recombined with other unmatched batteries.

[0118] Sub-step four: Comprehensive score matching.

[0119] For production lines with high output, the above-mentioned fine-tuning can be time-consuming. In such cases, a comprehensive scoring method can be used: First, calculate the comprehensive score for each battery. A lower score indicates a smaller capacity or higher internal resistance, while a higher score indicates a larger capacity or lower internal resistance. After sorting all batteries according to their comprehensive scores, group every four consecutive batteries together. This minimizes the difference in comprehensive scores within each group, thus ensuring consistency in both capacity and internal resistance. This method eliminates the need for subsequent fine-tuning, is computationally simple, and is suitable for fully automated production lines.

[0120] Sub-step 5: Grouping and output.

[0121] For each battery pack that passes the fine-tuning and matching process, the system generates a unique group number and records the identification codes, initial capacity values, initial internal resistance values, predicted annual capacity degradation rates, and group consistency indicators (capacity range, internal resistance range) of the four batteries in that group into the database. Subsequently, these batteries are sent to an automated spot welding or laser welding station to be assembled into a complete battery pack.

[0122] The second embodiment of this application is as follows: Please see Figure 4 This invention provides a predictive sorting system for the capacity degradation of batteries used in two-wheeled electric vehicles, applied to a predictive sorting method for the capacity degradation of batteries used in two-wheeled electric vehicles as provided in the first embodiment, comprising: The state of charge adjustment module 101 is used to adjust each battery cell to be sorted to the same preset state of charge. The test data acquisition module 102 is used to sequentially perform multi-rate pulse charge-discharge test and voltage relaxation test on the battery cells in the preset state of charge, and to acquire the voltage value at the moment the pulse discharge ends and the open circuit voltage change data over time during the relaxation process. The feature parameter calculation module 103 is used to calculate the dynamic feature parameter set of each battery cell based on the collected data. The dynamic feature parameter set includes at least the polarization internal resistance growth coefficient, the high current polarization feature value, and the ion diffusion relaxation feature value. The capacity degradation prediction module 104 has a pre-built capacity degradation prediction model built in it, which is used to receive the dynamic characteristic parameter set and output the predicted annual capacity degradation rate of each battery cell. The sorting execution module 105 is used to set multiple sorting levels according to the predicted annual capacity decay rate, classify the battery cells into the corresponding levels, and remove battery cells whose predicted annual capacity decay rate is higher than the preset rejection threshold. The fine-tuning matching module 106 is used to further fine-tune the matching of individual battery cells within the same sorting grade based on their initial capacity and initial internal resistance, and output the matching results of the battery pack.

[0123] Regarding the system in the above embodiments, the specific ways in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0124] For the system embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0125] Accordingly, this application also provides an electronic device, including: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the above-described predictive sorting method for the capacity degradation of two-wheeled electric vehicle batteries. Figure 5 The diagram shown is a hardware structure diagram of any device with data processing capabilities within a two-wheeled electric vehicle battery capacity degradation prediction sorting system provided in an embodiment of the present invention. (Except for...) Figure 5In addition to the processor, memory, and network interface shown, any data processing device in the embodiment may also include other hardware depending on the actual function of the data processing device, which will not be described in detail here.

[0126] Accordingly, this application also provides a computer-readable storage medium storing computer instructions, which, when executed by a processor, implement the aforementioned predictive sorting method for capacity degradation of two-wheeled electric vehicle batteries. The computer-readable storage medium can be an internal storage unit of any data-processing device as described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium can also be an external storage device, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc., equipped on the device. Furthermore, the computer-readable storage medium can include both internal storage units of any data-processing device and external storage devices. The computer-readable storage medium is used to store the computer program and other programs and data required by the data-processing device, and can also be used to temporarily store data that has been output or will be output.

[0127] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

[0128] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A method for predictive sorting of battery capacity decay in two-wheeled electric vehicles, characterized in that, Includes the following steps: Each battery cell to be sorted is adjusted to the same preset state of charge. The battery cells in the preset state of charge are subjected to multi-rate pulse charge-discharge test and voltage relaxation test in sequence. The voltage value at the moment the pulse discharge ends and the open circuit voltage change data over time during the relaxation process are collected. The dynamic characteristic parameter set of each battery cell is calculated based on the collected data. The dynamic characteristic parameter set includes at least the polarization internal resistance growth coefficient, the high current polarization characteristic value, and the ion diffusion relaxation characteristic value. The dynamic feature parameter set is input into a pre-built capacity decay prediction model to obtain the predicted annual capacity decay rate of each battery cell. Multiple sorting levels are set according to the predicted annual capacity decay rate. The battery cells are classified into the corresponding levels, and battery cells with a predicted annual capacity decay rate higher than the preset rejection threshold are rejected. For individual battery cells within the same sorting level, fine-tuning and matching are performed based on their corresponding initial capacity and initial internal resistance to form a battery pack.

2. The method for predictive sorting of battery capacity degradation for two-wheeled electric vehicles as described in claim 1, characterized in that, Each individual battery cell to be sorted is adjusted to the same preset state of charge, including: The battery cell is discharged at a rate of 0.5 times the discharge cutoff voltage, then charged at a constant current rate of 0.5 times the nominal capacity to 50%, and then switched to constant voltage charging until the charging current drops to a rate of 0.05 times the current, so that the battery cell reaches a state of 50% charge.

3. The method for predictive sorting of battery capacity degradation for two-wheeled electric vehicles as described in claim 1, characterized in that, The multi-rate pulse charge-discharge test specifically includes: The battery cells are subjected to constant current pulse discharge at the first preset rate, the second preset rate and the third preset rate in sequence. The duration of each pulse discharge is ten seconds, and there is a rest period of twenty seconds between each two pulse discharges. The first preset multiplier is 0.5x, the second preset multiplier is 1.5x, and the third preset multiplier is 2.5x. At the moment each pulse discharge ends, the terminal voltage and current values ​​of the battery cells are simultaneously acquired and recorded.

4. The method for predictive sorting of battery capacity degradation for two-wheeled electric vehicles as described in claim 1, characterized in that, Voltage relaxation testing specifically includes: After the multi-rate pulse charge-discharge test is completed, immediately disconnect the charge-discharge circuit to keep the battery cells in an open circuit state. The open-circuit voltage is continuously recorded from the start of the settling period, with a total recording time of sixty seconds. In the first thirty seconds, a voltage point is recorded every 0.5 seconds, and in the last thirty seconds, a voltage point is recorded every second. The voltage rise rate between the second and fifth seconds after the start of rest was extracted from the recorded open-circuit voltage data and used as the ion diffusion relaxation characteristic value.

5. The method for predictive sorting of battery capacity degradation for two-wheeled electric vehicles as described in claim 1, characterized in that, The method for calculating the dynamic feature parameter set is as follows: The first DC internal resistance is calculated based on the voltage and current at the moment the first multiplier pulse ends; the second DC internal resistance is calculated based on the voltage and current at the moment the second multiplier pulse ends; and the third DC internal resistance is calculated based on the voltage and current at the moment the third multiplier pulse ends. The difference between the second DC internal resistance and the first DC internal resistance is used as the polarization internal resistance growth coefficient. The difference between the third DC internal resistance and the second DC internal resistance is taken as the large current polarization characteristic value. The voltage rise rate between the second and fifth seconds of the voltage relaxation test was used as the characteristic value of ion diffusion relaxation.

6. The method for predictive sorting of battery capacity degradation for two-wheeled electric vehicles as described in claim 1, characterized in that, Methods for constructing capacity decay prediction models include: Sample batteries are extracted from the production batch, and the dynamic characteristic parameter set of each sample battery is extracted according to the methods of adjusting the state of charge, multi-rate pulse test and voltage relaxation test. The sample batteries were subjected to accelerated cycle aging tests, and the annual capacity decay rate of each sample battery was recorded when the initial capacity decayed to 80% of the initial capacity. Using the dynamic characteristic parameter set of the sample battery as input and the corresponding annual capacity decay rate as the output label, the capacity decay prediction model is trained using an ensemble regression algorithm.

7. The method for predictive sorting of battery capacity degradation for two-wheeled electric vehicles as described in claim 1, characterized in that, The sorting levels include: the first level corresponds to a predicted annual capacity decay rate of less than 5%, the second level corresponds to a predicted annual capacity decay rate between 5% and 7%, and the third level corresponds to a predicted annual capacity decay rate between 7% and 9%. The preset rejection threshold is nine percent. Battery cells with a predicted annual capacity decay rate higher than nine percent are judged as degraded products and rejected.

8. The method for predictive sorting of battery capacity degradation for two-wheeled electric vehicles as described in claim 1, characterized in that, Fine-tuning the match specifically includes: Within the same sorting level, the individual battery cells are sorted from largest to smallest according to their initial capacity. Several adjacent batteries are grouped into a coarse group, so that the range of initial capacity within the group is no more than 1.5 percent of the nominal capacity. Provided that the capacity range meets the requirements, calculate the initial internal resistance range of each battery cell in each coarse group. If the internal resistance range is greater than 10% of the average internal resistance in the group, then swap batteries between adjacent groups until the internal resistance range meets the requirements. If the requirements cannot be met through exchange, a weighted comprehensive scoring method is used for rematching, where the initial capacity has a weight of 70% and the initial internal resistance has a weight of 30%.

9. The method for predictive sorting of battery capacity degradation for two-wheeled electric vehicles as described in claim 1, characterized in that, The total duration of the multi-rate pulse charge-discharge test and the voltage relaxation test is controlled within two minutes.

10. A predictive sorting system for the capacity degradation of batteries for two-wheeled electric vehicles, applied to the predictive sorting method for the capacity degradation of batteries for two-wheeled electric vehicles as described in claim 1, characterized in that, include: The state of charge adjustment module is used to adjust each battery cell to be sorted to the same preset state of charge. The test data acquisition module is used to sequentially perform multi-rate pulse charge-discharge tests and voltage relaxation tests on the battery cells in the preset state of charge, and to acquire the voltage value at the moment the pulse discharge ends and the open-circuit voltage change data over time during the relaxation process. The feature parameter calculation module is used to calculate the dynamic feature parameter set of each battery cell based on the collected data. The dynamic feature parameter set includes at least the polarization internal resistance growth coefficient, the high current polarization feature value, and the ion diffusion relaxation feature value. The capacity degradation prediction module has a pre-built capacity degradation prediction model built in, which is used to receive the dynamic feature parameter set and output the predicted annual capacity degradation rate of each battery cell. The sorting execution module is used to set multiple sorting levels according to the predicted annual capacity decay rate, classify the battery cells into the corresponding levels, and remove battery cells whose predicted annual capacity decay rate is higher than the preset rejection threshold. The fine-tuning matching module is used to further fine-tune the matching of individual battery cells within the same sorting grade based on their initial capacity and initial internal resistance, and output the matching results of the battery pack.

Images