Method and system for scoring and screening virtual high capacity of retired battery

Abstract

The invention discloses a decommissioned battery virtual high capacity scoring and screening method and system, relates to the technical field of battery detection and evaluation, and aims to solve the problem that decommissioned battery virtual high capacity identification is difficult, and the method comprises the following steps: S1, performing discharge operation on a battery and standing, and monitoring an open-circuit voltage in a standing process in real time to obtain a voltage recovery curve; s2, extracting a multi-modal dynamic feature from the voltage recovery curve, wherein the multi-modal dynamic feature comprises a voltage recovery time constant and a curve fitting residual error; s3, performing weighted fusion on the multi-modal dynamic features and outputting a standardized score value to obtain a health score; and S4, comparing the health score with a preset threshold value, and outputting a virtual high-capacity risk grade screening result.

Description

Technical Field

[0001] The present invention relates to the technical field of battery detection and evaluation, and in particular to a method and system for screening inflated capacity scores of retired batteries. Background Art

[0002] With the rapid adoption of new energy vehicles, a large number of power batteries are entering retirement. Retired power cells typically retain a certain amount of residual capacity. Accurately assessing their true condition, sorting, and reassembling them for use in low-rate applications such as energy storage, can significantly extend their lifecycle and achieve resource recycling. However, quickly and accurately identifying the true capacity of retired cells remains a core technical challenge that urgently needs to be addressed in the current cascade utilization process. Currently, common cell screening methods rely on single-point parameters such as static voltage, voltage plateau, internal resistance, and residual capacity. While simple to implement, these methods fail to fully reflect the true health of the cells, particularly with significant blind spots in static recovery characteristics. In practical applications, it is common to observe that some retired cells experience a voltage rebound after being left for several hours or days, with their surface state parameters (such as open-circuit voltage and internal resistance) returning to normal ranges. This can lead to the system misidentifying them as "healthy" and placing them in the usable group. However, these cells often exhibit severe polarization, slow response, and reduced reversible capacity. Their voltage recovery is not accompanied by an increase in actual capacity, resulting in the so-called "falsely high capacity" phenomenon. Once this type of battery cell enters the cascade energy storage system, it is easy to cause SOC estimation deviation, performance imbalance within the cluster, premature exit of the battery cell from the cycle, and even cause thermal management risks, seriously affecting the stability and economy of the system.

[0003] Some existing research and patents have attempted to determine the health status of retired battery cells from perspectives such as voltage recovery behavior, electrochemical impedance spectroscopy, and curve feature extraction. For example, patent publication number CN118746769A uses the voltage recovery amplitude after standing still as a screening criterion, assessing the degree of aging by the ratio of the internal resistance before and after standing. Patent publication number CN111398833A performs cluster identification based on the shape of the recovery curve. These technologies have improved the accuracy of battery cell sorting to a certain extent, but they often focus on a single parameter or feature dimension and are not sufficient to fully reflect the true electrochemical performance of the battery cell.

[0004] Existing methods have the following limitations: First, they lack a coupled analysis mechanism between voltage and internal resistance, making it difficult to identify hidden degraded cells where "voltage recovers but capacity does not." Second, they ignore the inconsistency between recovery amplitude and rate, failing to incorporate reaction kinetics metrics such as the recovery time constant. Third, they lack a scalable, universal scoring system, making it difficult to adapt to engineered online screening scenarios. Fourth, they lack quantitative indicators for the "falsely high capacity" phenomenon, resulting in a high rate of misjudgment in screening results. Overall, a multimodal analysis system that integrates voltage recovery, internal resistance changes, and recovery dynamics has not yet been established. There is an urgent need for quantifiable, scorable, and deployable identification methods to improve the accuracy and consistency of screening retired cells for cascade utilization. Summary of the Invention

[0005] In order to solve the problem of difficulty in identifying the inflated capacity of retired battery cells, the present invention proposes a scoring and screening method and system for the inflated capacity of retired batteries. By quantifying the kinetic speed of voltage recovery during the static period, it can identify inflated capacity batteries with normal apparent capacity but hidden aging, thereby improving the accuracy of the initial screening of retired batteries.

[0006] A further object of the present invention is to improve the robustness of false high risk identification through dynamic weight fusion.

[0007] In order to achieve the above objectives, the present invention adopts the following technical solution: a method for screening the inflated capacity of retired batteries, comprising: S1, discharge the battery and let it rest, monitor the open circuit voltage in real time during the resting process, and obtain a voltage recovery curve; S2, extracting multimodal dynamic features from the voltage recovery curve, the multimodal dynamic features including a voltage recovery time constant and a curve fitting residual; S3, performing weighted fusion on the multimodal dynamic features and outputting a standardized score value to obtain a health score; S4: Outputting a screening result of an inflated capacity risk level based on a comparison between the health score and a preset threshold.

[0008] In this technical solution, by fitting the open-circuit voltage-time curve during the static process, the voltage recovery time constant is obtained to quantify the speed of voltage recovery and the relaxation process, which can significantly improve the recognition accuracy and initial screening consistency.

[0009] Preferably, the method for extracting the curve fitting residual characteristics in step S2 is: calculating the overall deviation value between the measured voltage data and the fitted voltage curve during the static process, or extracting the dynamic morphological parameters of the voltage recovery curve, the overall deviation value quantifies the degree of deviation between the measured data and the fitting model, and the dynamic morphological parameters describe the changing trend characteristics of the recovery curve.

[0010] Preferably, step S3 includes: weighted fusion of the voltage recovery amplitude, the internal resistance change rate, the voltage recovery time constant and the curve fitting residual; the contribution weight of the voltage recovery amplitude to the health score is positively correlated, and the contribution weight of the voltage recovery time constant, the internal resistance change rate and the curve fitting residual characteristics to the health score is negatively correlated, and the weight coefficient is preset through expert experience or dynamically optimized by a machine learning model.

[0011] Preferably, in step S2, the voltage recovery time constant is inversely proportional to the open circuit voltage recovery speed, and a larger value indicates a more serious battery polarization phenomenon.

[0012] Preferably, step S2 comprises: fitting a voltage recovery curve by means of an exponential decay model, wherein the voltage recovery time constant is solved by means of an optimization algorithm, wherein the optimization algorithm aims to minimize the residual between the measured voltage data and the fitted voltage data.

[0013] Preferably, the optimization goal of the machine learning model is to minimize the error between the score and the actual capacity attenuation of the battery.

[0014] Preferably, the weighted fusion calculation outputs a standardized score value, and the standardized score value is mapped to an interval of 0 to 1, where 0 represents the highest risk of inflated capacity and 1 represents the highest reliability of real capacity.

[0015] Preferably, the preset thresholds include a health threshold and an elimination threshold, and the hierarchical decision of step S4 includes at least three levels of judgment: if the score is higher than the health threshold, it is judged as the "available" level and enters the high-priority tiered utilization group; if the score is between the health threshold and the elimination threshold, it is judged as the "warning" level and assigned to the low-priority tiered utilization group; if the score is lower than the elimination threshold, it is judged as the "elimination" level and is scrapped or disassembled for recycling.

[0016] Preferably, the health threshold and elimination threshold are set according to the initial capacity of the retired battery: when the initial capacity is ≥80% of the rated capacity, the health threshold is 0.8 and the elimination threshold is 0.6; when the initial capacity is <80% of the rated capacity, the health threshold is 0.7 and the elimination threshold is 0.5.

[0017] The present invention also adopts the following technical solution: a system for identifying the falsely high capacity of retired power battery cells, comprising: A cell testing device that discharges the cell to be tested to a preset low state of charge or cut-off voltage; The data acquisition module collects the open circuit voltage and AC internal resistance of the battery cell in real time during the static phase after discharge; Feature extraction module, extracts multimodal dynamic features from the collected data; A scoring module performs weighted fusion on the multimodal dynamic features and outputs a standardized scoring value; The determination module compares the health score with a preset threshold and outputs a sorting instruction for the false high capacity risk level.

[0018] The beneficial effects of the present invention are: 1) Efficiently identify inflated capacity battery cells, improve the consistency of initial screening and sorting of battery cells, and reduce misjudgment and manual intervention; 2) Improved system safety and performance stability, reducing the risk of thermal runaway; 3) The method is highly versatile and applicable to cells of different types and aging levels; 4) Improve resource utilization and enhance the economic value of retired battery cells. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 The present invention is a flow chart of a method for identifying artificially high capacity of retired power battery cells.

[0020] Figure 2 This is a flow chart of a method for identifying artificially high capacity of retired power battery cells according to embodiment 1 of the present invention. DETAILED DESCRIPTION

[0021] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific implementation method described herein is only an optimal embodiment of the present invention, which is only used to explain the present invention and does not limit the scope of protection of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0022] Example 1 This embodiment provides a method for screening retired batteries with falsely high capacity scores. A certain type of retired electric vehicle lithium iron phosphate battery cell is selected as a test sample. However, this method is also applicable to screening scenarios of other types of lithium-ion batteries or sodium-ion batteries. It includes several steps. Figure 1 Flowchart with step-by-step instructions.

[0023] Step S1: discharge pre-treatment, rest and data acquisition stage.

[0024] The retired cells to be tested are placed in a constant temperature test environment, and the retired cells to be tested are pre-discharged. The polarization state is activated through the constant current discharge mode, and the state of charge (SOC) is reduced to the preset cut-off condition.

[0025] Specifically, a discharge current of 0.5C to 1C rate is used (C is the rated capacity of the battery cell) and the discharge is continued until the preset cut-off condition is reached.

[0026] In this embodiment, the test environment is 25±2°C, and the preset cutoff condition can be a preset low SOC point, such as a cutoff voltage of 20% SOC or lower. Specifically, it can be set according to the battery cell type, for example, 2.5V for lithium iron phosphate batteries and 3.0V for ternary material batteries.

[0027] The core purpose of this step is to fully expose the potential degradation defects inside the battery cell and lay the foundation for subsequent recovery characteristics analysis.

[0028] The resting process is started immediately after the discharge is completed. At this time, the battery cell status is monitored in real time through a high-precision data acquisition device. The high-precision data acquisition device can be a battery testing system or a dedicated data recorder.

[0029] It should be emphasized that the acquisition frequency needs to be set within the range of 10 to 100 times per second to ensure the accuracy of capturing dynamic behaviors.

[0030] The core parameters recorded include: the initial value of the open circuit voltage (marked as V0) and the initial value of the AC internal resistance (marked as R0) at the beginning of the static state, as well as the subsequent real-time open circuit voltage V t and real-time AC internal resistance R t .

[0031] The duration of the static state is usually set to 2 to 4 hours, or until the voltage change tends to be stable, for example, the fluctuation is less than 5mV for 30 consecutive minutes.

[0032] Understandably, temperature stability must be ensured during this phase. If the ambient temperature fluctuates by more than ±3°C, retesting or enabling the temperature compensation algorithm may be necessary.

[0033] At this stage, obviously aged batteries can be screened out based on the real-time open circuit voltage and real-time AC internal resistance during the static process and classified into the "elimination" level.

[0034] Step S2: multimodal dynamic feature extraction stage.

[0035] Based on the time series data collected in step S1, the following four key features are calculated by the data processing unit.

[0036] First, the voltage recovery amplitude characteristics are extracted.

[0037] The calculation method is the difference between the open-circuit voltage at the end of the static period and the starting voltage. This parameter indicates the apparent recovery ability of the battery cell. A larger value generally indicates a more significant voltage recovery.

[0038] Secondly, the internal resistance change rate feature is extracted.

[0039] The initial AC internal resistance value is obtained at the beginning of the static stage, and the ending AC internal resistance value is obtained at the end of the static stage. The internal resistance recovery characteristics are characterized by calculating the relative change degree between the ending AC internal resistance value and the initial AC internal resistance value, that is, the relative change rate of the AC internal resistance at the end of the static stage compared to the initial internal resistance is calculated. The larger the change degree value, the more serious the internal degradation of the battery.

[0040] Specifically, it is expressed as the ratio of the difference between the AC internal resistance at the end of the static period and the initial internal resistance divided by the initial internal resistance. This indicator reflects the efficiency of internal resistance recovery during the polarization elimination process, and abnormal changes often indicate hidden degradation.

[0041] It is worth noting that, in the present invention, the voltage recovery time constant feature is also extracted.

[0042] A function fitting is performed on the curve of the open circuit voltage changing with time, preferably using an exponential decay model, such as a single exponential or double exponential form. The fitting parameters are solved by an optimization algorithm to minimize the deviation between the measured data and the fitting curve.

[0043] The obtained time constant parameter characterizes the speed of voltage recovery rate. The larger the value, the slower the polarization elimination.

[0044] It should be noted that this parameter is the core indicator for identifying slow-recovery inflated capacity.

[0045] In some other embodiments, computational efficiency can be improved by piecewise exponential approximation. For example, fitting with a single exponential model in the early stage of static state and switching to a double exponential model in the later stage can significantly improve computational efficiency.

[0046] Specifically, in this embodiment, the static process is divided into an instantaneous recovery period (0-30 seconds), a main recovery period (30 seconds-30 minutes) and a long-term relaxation period (>30 minutes) according to the polarization type. A single exponential model is used for fitting in the instantaneous recovery period, and a double exponential model is used for fitting in the main recovery period and the long-term relaxation period, thereby improving processing efficiency while ensuring accuracy.

[0047] Finally, the morphological features of the recovery curve are extracted.

[0048] The recovery curve morphology features include two sub-features: the first is the fitting residual index, which quantifies the overall deviation between the measured voltage data and the fitted curve, such as the root mean square error; the second is dynamic morphology parameters, such as the voltage rise slope during the initial static phase or the location of the curve inflection point. These parameters can help identify abnormal recovery patterns.

[0049] For the fitting residual index, specifically, the difference between the measured voltage data point and the corresponding position of the fitting curve is calculated, and the overall deviation intensity is quantified by the root mean square error.

[0050] For example, when the root mean square error exceeds 5 millivolts, it indicates that there is a systematic structural abnormality in the battery, such as internal micro-short circuit or electrolyte drying up.

[0051] Dynamic morphological parameters focus on capturing local abnormal characteristics of the curve.

[0052] For example, a smooth exponential decay represents a healthy cell, a double-platform decay corresponds to a lithium plating failure, and a step-like decline indicates the shedding of active materials.

[0053] By calculating the second-order derivative of the voltage curve in real time, the peak value at the inflection point can be extracted. This value reflects the steepness of the voltage recovery process. When the inflection point peak exceeds 0.15 millivolts per square second, it indicates a high risk of lithium plating in the battery's negative electrode.

[0054] Another key parameter is the segmented slope change rate, calculated by dividing the curve into ten equal parts and calculating the total slope fluctuation in each segment. A parameter exceeding 0.08 millivolts per square second indicates active material shedding. Although such cells may have normal short-term capacity, they are susceptible to sudden capacity drops under vibration.

[0055] These two types of features form a collaborative diagnostic mechanism. For example, when the inflection point peak exceeds the standard alone, the lithium plating repair process is triggered. When the slope change rate exceeds the standard, it is allocated to the low-speed electric vehicle power pack. If both parameters exceed the standard at the same time, it is judged as composite aging and directly dismantled.

[0056] The present invention converts microscopic aging characteristics into quantifiable indicators, which can solve the problem of traditional methods missing detection of batteries with normal voltage recovery amplitude but internal deterioration, and provides key technical guarantees for the hierarchical utilization of retired batteries.

[0057] However, in this embodiment, the curve morphology characteristics are comprehensively considered with indicators such as the voltage recovery time constant to further improve the robustness of the grading assessment, as detailed in step S3.

[0058] Step S3: constructing the fusion scoring function.

[0059] The above four types of features are normalized to eliminate dimensional differences, and then the health score S is generated through weighted fusion.

[0060] Specifically, the normalization process uses linear scaling to map each eigenvalue to the range of 0 to 1.

[0061] Weight allocation can be done in two ways: one is to preset weight coefficients based on the experience of battery experts, such as giving time constant characteristics a higher weight to strengthen the influence of kinetic behavior; the other is to dynamically optimize weights through machine learning models.

[0062] Specifically, by collecting a large number of retired battery cell samples with known actual capacity and inflated capacity, i.e., labeled training data, the model is trained using machine learning algorithms such as linear regression, support vector machine regression, neural networks, or decision trees. The model automatically learns the optimal weight coefficients, enabling the health score to maximize the distinction between inflated capacity cells and healthy cells. For example, the weights can be optimized by minimizing scoring error or maximizing classification accuracy.

[0063] In this embodiment, the training data comes from historical battery cell samples with known inflated risk labels, and the optimization goal is to maximize the score differentiation between healthy batteries and inflated capacity batteries.

[0064] It should be noted that the design principle of the scoring function follows a specific directionality. The voltage recovery amplitude characteristic is positively correlated with the score, while the internal resistance change rate, time constant, and fitting residual are all negatively correlated with the score.

[0065] Step S4: hierarchical decision-making stage.

[0066] The calculated health score is compared with the preset threshold and a three-level judgment mechanism is implemented.

[0067] When the score is higher than the health threshold, the battery cell is judged to be "usable", indicating that its actual capacity is reliable and there is no obvious risk of inflated capacity, and it can be directly used for cascade utilization.

[0068] When the score is between the elimination threshold and the health threshold, it is judged as a "warning" or "pending" level, indicating that there is a potential risk of inflated capacity and further capacity verification testing or restrictive use is required.

[0069] When the score is lower than the elimination threshold, it is judged as the "elimination" level, indicating that the battery cell is severely polarized or the capacity is falsely restored, and should be scrapped.

[0070] It is understandable that the health threshold and the elimination threshold can be dynamically adjusted according to the initial capacity of the battery cell.

[0071] For example, cells with an initial capacity ≥ 80% of the rated capacity adopt stricter thresholds, while cells with low capacity have relaxed standards.

[0072] Preferably, when the initial capacity is ≥80% of the rated capacity, the healthy threshold is 0.8 and the elimination threshold is 0.6; when the initial capacity is <80% of the rated capacity, the healthy threshold is 0.7 and the elimination threshold is 0.5.

[0073] The method of the present invention can effectively avoid misjudging "voltage recovery type weak cells" with a surface voltage recovery but low actual capacity as healthy cells, thereby improving the accuracy of cell sorting.

[0074] In some other implementations, the data acquisition module can integrate a temperature sensor to synchronously record the cell surface temperature. If the temperature changes beyond a threshold, the data compensation algorithm is automatically triggered. This design enhances robustness under high and low temperature conditions.

[0075] It is important to emphasize that the function fitting for time constant extraction is not limited to exponential models. In actual deployment, other kinetic models, such as fractional derivative models, can be adapted based on the cell chemistry. The core objective is to quantify the recovery rate.

[0076] For machine learning weight optimization scenarios, the training data set should cover multiple types of typical degradation modes (such as lithium precipitation, SEI thickening, active material loss, etc.). Interaction terms can be introduced in the feature engineering stage to improve the sensitivity of identifying falsely high capacity.

[0077] When this method is applied to an online sorting system, the judgment module can directly link the mechanical sorting mechanism. For example, when the "eliminate" command is output, the pneumatic push rod automatically moves the battery cell to the scrap recycling channel, achieving full process automation.

[0078] The present invention can serve as the core algorithm module of an intelligent sorting platform for retired power batteries and is widely used in scenarios such as energy storage, power peak regulation, and battery recycling.

[0079] Example 2 Based on the core technical framework of Example 1, this embodiment further introduces a deep quantization mechanism of polarization dynamics and multi-physics field collaborative analysis to significantly improve the accuracy of identifying "virtually high capacity" battery cells.

[0080] The following is an explanation of the NCM ternary cell screening scenario for decommissioning from a certain energy storage power station. The test environment is kept constant at 25±1°C.

[0081] Step S1, polarization activation and multi-dimensional data acquisition stage.

[0082] Place the battery cell to be tested on a constant temperature test platform and discharge it at a constant current rate of 1C to a cut-off voltage of 3.0V.

[0083] It should be emphasized that thickness deformation monitoring has been added in this stage. The change in battery cell thickness is recorded in real time through a laser displacement sensor. The sampling frequency is set to 10 times per second and stored synchronously with the voltage and internal resistance data.

[0084] After the discharge is completed, it immediately enters the static stage, which lasts for 3 hours.

[0085] Understandably, this embodiment focuses on rapid recovery behavior during the initial period of rest. Therefore, high-frequency sampling is performed at 100 times per second for the first five minutes of rest, and then reduced to 10 times per second. Simultaneously with this 100-times-per-second high-frequency sampling, an infrared thermal imager is activated to record the surface temperature distribution of the battery cells, focusing on capturing areas of localized temperature anomalies. This design effectively identifies micro-hotspots caused by lithium plating.

[0086] Step S2: multi-stage dynamic feature extraction stage.

[0087] Based on the collected time series data, the voltage recovery process is divided into three characteristic periods for differentiated processing.

[0088] First, during the transient recovery period (0-30 seconds), the initial voltage rise rate is extracted. This is calculated as the average slope of change over the first 10 voltage sampling points. The thickness recovery ratio during this period is also calculated to characterize the mechanical stress release characteristics. It should be noted that abnormal cells often exhibit delayed thickness recovery during this period.

[0089] Second, during the main recovery period (30 seconds to 30 minutes), the voltage curves were fitted using a novel kinetic model. This model describes ion diffusion behavior through a specific mathematical function, with key parameters including the diffusion anomaly index and the principal time constant.

[0090] The diffusion anomaly index reflects the degree of obstruction of lithium ion migration, with healthy cells close to 1 and deteriorated cells below 0.6; the main time constant quantifies the speed of charge transfer.

[0091] It is worth noting that the temperature uniformity index is extracted simultaneously in this stage. The calculation method is the difference between the highest and lowest temperatures on the surface. This parameter is indicative of local lithium deposition.

[0092] Finally, during the long relaxation period (>30 minutes), the final voltage recovery efficiency (VRE) was calculated, a parameter that characterizes the ability to eliminate deep polarization. The thickness stabilization time was also recorded, defined as the moment when the thickness fluctuation dropped to less than 0.1% of the initial value. Slow thickness stabilization is often accompanied by reversible capacity loss.

[0093] Step S3: Adaptive weight scoring system construction phase.

[0094] The seven extracted core features were normalized. Positive indicators (such as voltage rise rate) were calculated by comparing the measured value to a healthy baseline. Negative indicators (such as time constant) were calculated by comparing the healthy baseline to the measured value. The healthy baseline was derived from a database of new battery cells of the same model.

[0095] The scoring function uses a dynamic weighting mechanism. When significant temperature unevenness is detected, the weight of the kinetic parameters is automatically increased to 60%. If thickness recovery is abnormal, a veto rule is activated, and thickness stabilization time exceeding one hour is directly judged as "eliminated." In other cases, the standard weighting configuration is used.

[0096] In this embodiment, the criterion for determining significant temperature unevenness is that the temperature difference exceeds 5 degrees Celsius, and the criterion for determining abnormal thickness recovery is that the shrinkage rate is less than 0.3.

[0097] Step S4: Cloud-edge collaborative decision-making and sorting linkage stage.

[0098] The calculated health scores are transmitted to a cloud-based analytics platform, which continuously aggregates data from multiple production lines and dynamically optimizes thresholds.

[0099] The health threshold fluctuates between 0.75 and 0.85, and is calibrated in real time based on the screening pass rate of the same batch of battery cells in the same month. The elimination threshold is fixed at 0.55.

[0100] The judgment results are transmitted to the sorting terminal in real time. When the "eliminate" command is output, the pneumatic actuator moves the battery cell to the disassembly channel within 3 seconds, triggering an audible and visual alarm, recording the battery cell code and characteristic parameters, and establishing a degradation pattern traceability file.

[0101] It should be noted that this embodiment, by introducing multi-stage recovery analysis and temperature-deformation coordinated monitoring, achieves for the first time the accurate identification of concealed lithium-induced false high capacity.

[0102] Compared with the basic solution of Example 1, this embodiment can further reduce the missed detection rate of artificially high-capacity battery cells and significantly improve the safety margin of the cascade utilization battery pack.

[0103] The above implementation methods prove that this solution effectively solves the technical pain points in the sorting of retired batteries through multi-physical field coupling analysis and dynamic decision-making mechanism, and has significant industrial application value.

[0104] Example 3 This embodiment provides a system for identifying the falsely high capacity of retired power battery cells, the structure of which is as follows: Figure 2 As shown, including: The battery cell testing device is used to implement the preset discharge process and provide a stable testing environment.

[0105] The device typically includes a charging and discharging device, a temperature control module, and a battery cell fixture. The charging and discharging device can be a high-precision battery tester, and the temperature control module ensures that the test is performed at a constant temperature to avoid the impact of temperature on internal resistance and voltage recovery.

[0106] The battery cell testing device can discharge the battery cell with precise constant current or constant power to reach a preset low SOC point.

[0107] Data acquisition module: used to collect the open circuit voltage (V t ) and AC internal resistance (R t ) data and records the initial values (V0, R0) at the start of the static state. This module typically consists of a high-precision voltage collector (such as a digital multimeter or A / D converter) and an AC internal resistance meter (such as a milliohm meter or impedance analyzer) to ensure accurate and synchronized data collection.

[0108] Feature Extraction Module: Connected to the Data Acquisition Module, this module receives the collected voltage, internal resistance, and time data and automatically calculates the voltage recovery amplitude ΔV, internal resistance change rate ΔR / R0, voltage recovery time constant τ, and the recovery curve fitting residual or morphological parameters. This module can be implemented by a data processing unit such as a high-performance embedded processor, industrial control computer, or server, and has built-in corresponding data processing and fitting algorithms.

[0109] Scoring module: Connected to the feature extraction module, it embeds the capacity authenticity coupling scoring function S. It receives the characteristic parameters calculated by the feature extraction module and calculates the health score S of the battery cell based on preset weight coefficients or a trained machine learning model. This module can be an independent processing unit or a software module integrated into an industrial computer.

[0110] The judgment module, connected to the scoring module, compares the health score S output by the scoring module with a preset threshold and outputs a cell classification result of "usable / warning / rejected." The judgment module can display the result on the user interface or send the result to subsequent automated sorting equipment via a communication interface.

[0111] The system described in this invention can be preferably integrated into automated sorting equipment for the reuse of retired batteries. In such an integrated system, the output of the determination module can directly drive an automated robot or sorting mechanism to automatically sort the cells into different storage bins or subsequent processing lines, enabling online, high-throughput identification and elimination of cells with inflated capacity risks. For example, different sorting channels can be set to correspond to three categories of cells: "usable," "warning," and "eliminated."

[0112] Through the above-mentioned method and system, the present invention can efficiently and accurately identify the problem of inflated capacity of retired battery cells, avoiding the limitations of single parameter judgment in traditional methods, thereby significantly improving the overall efficiency and safety of the cascade utilization of retired battery cells.

Claims

1. A method for screening retired batteries with inflated capacity scores, characterized in that: The following steps are involved: S1, discharge the battery and let it rest, monitor the open circuit voltage in real time during the resting process, and obtain a voltage recovery curve; S2, extracting multimodal dynamic features from the voltage recovery curve, the multimodal dynamic features including a voltage recovery time constant and a curve fitting residual; S3, performing weighted fusion on the multimodal dynamic features and outputting a standardized score value to obtain a health score; S4: Outputting a screening result of an inflated capacity risk level based on a comparison between the health score and a preset threshold.

2. A method for screening retired battery falsely high capacity scores according to claim 1, characterized in that: The method for extracting the curve fitting residual characteristics in step S2 is: calculating the overall deviation value between the measured voltage data and the fitted voltage curve during the static process, or extracting the dynamic morphological parameters of the voltage recovery curve. The overall deviation value quantifies the degree of deviation between the measured data and the fitting model, and the dynamic morphological parameters describe the changing trend characteristics of the recovery curve.

3. The method for screening the inflated capacity of retired batteries according to claim 1, characterized in that: The step S3 includes: weighted fusion of the voltage recovery amplitude, the internal resistance change rate, the voltage recovery time constant and the curve fitting residual; the contribution weight of the voltage recovery amplitude to the health score is positively correlated, and the contribution weight of the voltage recovery time constant, the internal resistance change rate and the curve fitting residual characteristics to the health score is negatively correlated, and the weight coefficient is preset through expert experience or dynamically optimized by a machine learning model.

4. A method for screening retired battery falsely high capacity scores according to claim 1 or 3, characterized in that: In step S2, the voltage recovery time constant is inversely proportional to the open circuit voltage recovery speed.

5. The method for screening the inflated capacity of retired batteries according to claim 1, characterized in that: The step S2 includes fitting a voltage recovery curve, wherein the voltage recovery time constant is solved by an optimization algorithm, and the optimization algorithm aims to minimize the residual between the measured voltage data and the fitted voltage data.

6. A method for screening inflated capacity scores of retired batteries according to claim 3, characterized in that: The optimization goal of the machine learning model is to minimize the error between the score and the actual capacity attenuation of the battery.

7. A method for screening inflated capacity scores of retired batteries according to claim 1 or 3, characterized in that: The weighted fusion calculation outputs a standardized score value, which is mapped to a range of 0 to 1, where 0 represents the highest risk of inflated capacity and 1 represents the highest reliability of true capacity.

8. The method for screening the inflated capacity of retired batteries according to claim 1, characterized in that: The preset thresholds include a healthy threshold and an elimination threshold. The hierarchical decision in step S4 includes at least three levels of determination: if the score is higher than the healthy threshold, it is determined to be "available" and enters the high-priority tiered utilization group; If the score is between the healthy threshold and the elimination threshold, it is judged as "warning" level and assigned to the low-priority tiered utilization group; If the score is lower than the elimination threshold, it will be judged as "elimination" level and will be scrapped or dismantled for recycling.

9. A method for screening retired battery falsely high capacity scores according to claim 8, characterized in that: The health threshold and elimination threshold are set according to the initial capacity of the retired battery: when the initial capacity is ≥80% of the rated capacity, the health threshold is 0.8 and the elimination threshold is 0.6; when the initial capacity is <80% of the rated capacity, the health threshold is 0.7 and the elimination threshold is 0.

5.

10. A system for identifying inflated capacity of retired power battery cells, using the inflated capacity scoring and screening method for retired batteries according to any one of claims 1 to 9, characterized in that: include: A cell testing device that discharges the cell to be tested to a preset low state of charge or cut-off voltage; The data acquisition module collects the open circuit voltage and AC internal resistance of the battery cell in real time during the static phase after discharge; Feature extraction module, extracts multimodal dynamic features from the collected data; A scoring module performs weighted fusion on the multimodal dynamic features and outputs a standardized scoring value; The determination module compares the health score with a preset threshold and outputs a sorting instruction for the false high capacity risk level.

Images