A method for sorting retired power battery considering unbalanced distribution of battery aging state

Abstract

The application discloses a kind of considering the retirement of power battery sorting method of battery aging state distribution imbalance, comprising;1. using voltage measuring equipment to measure and record the end voltage of retired battery;2. current batch of retired power battery data is randomly extracted as test battery;3 based on design step, test battery is executed test experiment and data is collected and stored;4. data cleaning and pretreatment are carried out to test data, and then the adaptive feature engineering designed is executed;5. using the output of last step, the pre-training of remaining energy estimation model embedded with feature distribution smoothing module is carried out.6. repeat steps 3-4 to the remaining battery, and deploy the pre-training model to execute remaining energy estimation.The application can overcome the error caused by aging distribution imbalance, so as to reduce the time cost and energy consumption cost of retired battery sorting work under the premise of ensuring sorting accuracy.

Description

Technical Field

[0001] This invention belongs to the field of retired battery cascade utilization, specifically a method for sorting retired power batteries that takes into account the uneven distribution of battery aging states. Background Technology

[0002] Power batteries are crucial energy storage components for new energy vehicles. They typically consist of numerous individual cells connected in series and parallel to form modules, which are then connected in series again to create battery packs that provide power to the vehicle. When retired after prolonged on-board use, the majority of retired power batteries are dismantled and enter the retirement process because their usable capacity has decreased to 80% of their original capacity. A significant portion, however, consists of excessively aged power batteries and healthy batteries retired due to significant inconsistencies with other batteries in the vehicle's battery pack. This imbalance in aging conditions directly affects the accuracy of subsequent residual energy estimation models, leading to sorting errors and inconvenience in secondary utilization. Therefore, it is necessary to accurately estimate the residual energy of these batteries with different aging levels before they enter the secondary utilization environment to achieve accurate sorting.

[0003] Currently, the sorting process for retired batteries generally involves the following steps: 1) Manually inspecting the structure and appearance of retired batteries to remove those that are obviously damaged or deformed and cannot be reused; 2) Measuring the performance parameters of retired batteries using testing equipment, and establishing corresponding databases to store the test data for different battery models and types. Simultaneously, batteries that are severely aged and cannot meet the requirements of any subsequent reuse scenarios (such as SOH below 40%) are included in the dismantling and recycling process; 3) Sorting the batteries in each database according to specific index parameters, grouping batteries with similar performance parameters together to ensure consistency in subsequent assembly work.

[0004] Most current methods for sorting retired batteries require discharging the remaining charge and then performing a full charge-discharge test. The batteries are then categorized based on their usable capacity and other performance parameters, and stored in their respective categories before being supplied according to the needs of different reuse scenarios. However, existing sorting methods share a common blind spot: they lack consideration for the uneven distribution of battery aging states within the same batch. This imbalance leads to unbalanced test data, affecting the effectiveness of the sorting method and consequently reducing the performance of retired batteries in reuse scenarios. Data-driven methods for sorting retired batteries need to further consider the impact of this imbalance; otherwise, the model will be influenced by the characteristics of the majority of aging states during learning, resulting in excessive estimation errors for the minority classes. Summary of the Invention

[0005] This invention aims to address the blind spots and shortcomings of current retired battery sorting technologies summarized above. By proposing a sorting method for retired power batteries that considers the uneven distribution of battery aging states, it seeks to overcome the impact of the imbalance phenomenon commonly found in industrial scenarios of retired battery sorting on the sorting results, avoid full charge and full discharge tests on large batches of retired batteries, and reduce redundant test processes, thereby reducing the energy consumption and time cost of retired battery sorting.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] The present invention provides a method for sorting retired power batteries that considers the uneven distribution of battery aging states, characterized by comprising the following steps:

[0008] Step 1: Visually screen a batch of retired power batteries to obtain the remaining retired batteries after screening;

[0009] Step 2: Randomly select from the remaining retired batteries to obtain a total of Test batteries;

[0010] Step 3: Collect test data for the test battery;

[0011] Step 3.1: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require After the test battery has been left to stand for a period of time, the initial terminal voltage of the test battery is measured. ,in, Representing the The terminal voltage of the test battery;

[0012] Step 3.2: Construct a charge / discharge strategy and use it for... Each test battery was tested to obtain a battery test dataset. ,in, Indicates the first Dynamic characteristic data of each test battery Indicates the first Static characteristic data of each test battery;

[0013] Step 4: [Regarding...] After preprocessing, the standardized feature sequences are obtained. ;in, Representing the first The standardized feature sequence of a test battery;

[0014] Step 5: The input is processed in the LASSO model to obtain the first... Preliminary characteristic sequence of a test battery;

[0015] Step Six: For the first After randomly combining any two features from the initial feature sequence of the test battery, grey relational analysis is performed to select features with high correlation coefficients to form the first feature. The optimal feature sequence of a test battery ,in, Representing the The k-th optimal feature of a test battery The number of optimal features;

[0016] Step 7: Construct a BiLSTM-based time-series regression prediction model, including: a feature extraction module, a feature smoothing module, a residual energy estimation module, and a temperature-based correction module, and use it to obtain the optimal feature sequence for n test batteries. The process is performed to obtain an estimate of the remaining energy of the retired batteries. ,in, This represents the estimated remaining energy of the m-th test battery;

[0017] Step 8: Obtain the corrected residual energy sequence using equation (8) :

[0018] (8)

[0019] In equation (8), This is the coefficient representing the effect of temperature on capacity. For reference temperature; To test the ambient temperature; This represents the corrected residual energy value of the m-th test cell;

[0020] Step 9: Construct the comprehensive loss function based on equation (9) :

[0021] (9)

[0022] In equation (9), There are two balance coefficients. The loss function represents the residual energy estimation. The loss function representing the feature smoothing process;

[0023] Step 10: Train the time series regression prediction model using the Adam optimizer and calculate the comprehensive loss function. To update the model parameters until The process continues until convergence, thus obtaining the optimal residual energy estimation model for retired batteries. This model is used to estimate the residual energy of remaining retired batteries, and based on the obtained residual energy and the calculated average dynamic internal resistance... To achieve battery sorting.

[0024] The method for sorting retired power batteries that considers the uneven distribution of battery aging states, as described in this invention, is also characterized in that step 3.2 includes the following steps:

[0025] Step 3.2.1: For Low-current hybrid pulse power characteristic tests were conducted on each test battery to obtain dynamic characteristic data that characterizes the battery's dynamic properties. ;in, Indicates pulse current, and , This represents the pulse current at the i-th multiplier. This represents the voltage change during the pulse test, and , This represents the voltage change of the m-th test battery in the i-th pulse current. Indicates the recovery voltage, and , This represents the recovery voltage of the m-th test battery. This indicates the time required to reach a stable recovery voltage, and , This represents the time it takes for the m-th test battery to reach a stable recovery voltage. This indicates the final voltage at which the test is completed, and ; This represents the final voltage of the m-th test battery;

[0026] Step 3.2.2 For Partial charging tests were performed on each test battery to obtain static charging data that characterizes the battery's aging state. ;in, This indicates the initial state of charge reached during discharge. This represents the set state of charge termination. This represents the voltage data during the charging process, and , This represents the charging voltage data of the m-th test battery; This represents the charging capacity data during the charging process, and , This represents the charging capacity data of the m-th test battery; This represents the charging energy data during the charging process, and , This represents the charging energy data of the m-th test battery; This represents the terminal voltage after partial charging is complete, and , This represents the final voltage of the m-th test battery;

[0027] 3. A method for sorting retired power batteries considering the uneven distribution of battery aging states according to claim 2, characterized in that step 3.2.1 is to sort the first... Each test battery was charged to its initial state of charge using a constant current, and then left to stand for a period of time; then, it was charged according to a set pulse current. After standing, the first The test battery was subjected to a pulse discharge test, and the results were recorded. Voltage changes of individual test batteries during pulse processing ,in, The pulse current representing the i-th multiplier. Represents the pulse current at the i-th multiplier. Next The voltage change of each test battery, z represents the number of rate options;

[0028] After the pulse process ends, for the first Each test battery was left to stand until its voltage stabilized, and the corresponding recovery voltage was recorded. Resting time and open circuit voltage .

[0029] Furthermore, step 3.2.2 is to... Each test battery was discharged to its initial state of charge. And record the initial voltage at this time. According to the set discharge current ratio, the first After the first test battery underwent initial discharge, the second... Each test battery was subjected to a constant current. Charge to the set state of charge. And record the first charge during the charging process. The voltage of each test battery Charging capacity and charging energy After charging is complete, the first... The test battery was left to stand for a period of time, and the reading was recorded. The final open-circuit voltage of each test battery after partial charging. .

[0030] Furthermore, step four includes:

[0031] Step 4.1: For After data cleaning, a clean dataset is obtained. ;

[0032] Step 4.2: Based on Build a basic feature library ,in, Indicates the first A set of dynamic characteristics of a test battery. Indicates the first A set of static features of a test battery;

[0033] Step 4.3: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require The feature sequence consisting of all dynamic and static characteristics of each test battery and its capacity label. After concatenation and normalization preprocessing, the standardized feature sequence is obtained. ,in, The capacity label represents the m-th test battery;

[0034] Furthermore, step 4.2 includes the following steps:

[0035] Step 4.2.1: For Feature extraction was performed on the low-current hybrid pulse power characteristic test data to obtain a dynamic feature set. ;

[0036] Step 4.2.2: For The voltage, energy, and capacity data obtained from partial charging tests are averaged, and the maximum and minimum extreme values ​​are taken. These values ​​are then used to calculate variance, kurtosis, skewness, quantiles, and curve slope, thereby obtaining a static feature set. ,in, Indicates the first Voltage-related characteristics of each test battery Indicates the first Capacity-related characteristics of each test battery Indicates the first Energy-related characteristics of the test battery Indicates the first The characteristics of the partial charging IC curve of the test battery.

[0037] Furthermore, step 4.2.1 includes the following steps:

[0038] Step 4.2.1.1: According to the first... Voltage changes of each test battery With pulse current Dynamic internal resistance obtained from doing business Using equations (1) and (2), we can obtain the first... Average dynamic internal resistance of each test battery and the rate of change of dynamic internal resistance And accordingly as the first The first dynamic characteristics of the test battery Second dynamic characteristics :

[0039] (1)

[0040] (2)

[0041] In equations (1) and (2), Let represent the dynamic internal resistance calculated under the i-th pulse current, and , This represents the dynamic internal resistance of the m-th test battery under the current pulse current. This indicates the maximum internal resistance. This indicates the minimum internal resistance.

[0042] Step 4.2.1.2: Based on the recovery voltage Recovery time and initial terminal voltage Using equations (3) and (4), we can obtain the first... Difference in recovery voltage among individual test batteries and pulse recovery rate And accordingly as the first The third dynamic characteristic of the test battery and the fourth dynamic feature Thus, a dynamic feature set of n test batteries is obtained. ;

[0043] (3)

[0044] (4)

[0045] In equations (3) and (4), This represents the recovery voltage of the m-th test battery; This represents the recovery time of the m-th test battery; This represents the initial terminal voltage of the m-th test battery.

[0046] Furthermore, step seven includes the following steps:

[0047] Step 7.1: The feature extraction module utilizes a BiLSTM layer to... Process and output the first... Depth features of each test battery ,in, Indicates the first The first test battery One deep feature, The total number of depth features;

[0048] Step 7.2: The feature smoothing module applies a feature distribution smoothing strategy to... The process is performed to obtain a smoothed feature sequence. ;

[0049] Step 7.3: The residual energy estimation module consists of a fully connected layer and a regression estimation layer; wherein, the fully connected layer... Perform feature mapping to generate the first The mapping characteristics of the test battery; the regression estimation layer is based on the regression function on the first test battery. The mapping features of the test battery are processed to generate the first test battery. Residual energy estimate of each test battery .

[0050] Furthermore, step 7.2 includes the following steps:

[0051] Step 7.2.1: Obtain the updated feature mean using equations (5) and (6). and the updated feature variance :

[0052] (5)

[0053] (6)

[0054] In equations (5) and (6), Indicates the smoothing coefficient; and They represent the first The feature mean and feature variance of each test battery before the update;

[0055] Step 7.2.2: Use equation (7) to... Smoothing is performed to obtain the smoothed feature sequence. ;

[0056] (7)

[0057] In equation (7), To prevent tiny constants with a denominator of zero; Indicates the first The first test battery after smoothing One deep feature;

[0058] Furthermore, in step 9, the loss function for residual energy estimation is obtained using equations (10) and (11) respectively. loss function for feature smoothing process ;

[0059] (10)

[0060] (11).

[0061] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0062] 1. The present invention collects test data by designing a test process specifically for testing the dynamic characteristics of batteries and conducting partial charge-discharge experiments. It uses only partial data to estimate residual energy and perform subsequent sorting work, avoiding the first step of emptying and full charge-discharge testing of retired batteries. This significantly reduces redundant test steps in the sorting process and reduces the energy consumption and sorting time for retired battery manufacturers.

[0063] 2. This invention designs an adaptive feature engineering approach. Based on local test data, it includes all dynamic and static feature information. Static features are processed through eight types of statistical methods and IC curve-related features to jointly construct a complete feature library and optimization pipeline containing battery dynamic performance and aging information. The optimal feature combination is automatically selected for different types of batteries, thereby improving the generalization ability of the estimation model and its adaptability to various retirement sorting scenarios for subsequent residual energy estimation of retired batteries.

[0064] 3. The residual energy estimation model in the method of this invention is based on a BiLSTM neural network. An FDS layer (Feature Distribution Smooth) is embedded to smooth the feature distribution during the training process of the neural network model, thereby overcoming the impact of imbalance and improving the residual energy estimation accuracy and sorting accuracy based on residual energy after the deployment of the subsequent pre-trained residual energy estimation model.

[0065] 4. The test steps in this invention use charge and discharge data obtained from low-current hybrid pulse power characteristic (L-HPPC) testing and partial charge testing for feature extraction, model training, and subsequent residual energy estimation. This ensures the uniformity of the test environment and data attributes for each test battery, while avoiding redundant test steps caused by performing full charge and full discharge tests on all batteries.

[0066] 5. The method used in this invention is based on two indicators: residual energy estimation using a data-driven approach and average dynamic internal resistance calculated from dynamic characteristics, to assist in sorting. Compared with the current method of obtaining residual energy through full charge and full discharge, this solution is more suitable for use in factory testing processes, reduces the complexity of manual screening work, and significantly improves sorting accuracy and efficiency. It is universally applicable to sorting scenarios for various types of batteries. Attached Figure Description

[0067] Figure 1 This is a structural diagram of the sorting equipment involved in the present invention;

[0068] Figure 2 This is a flowchart of the entire sorting method of the present invention;

[0069] Figure 3 This is a flowchart of the residual energy estimation process used in this invention;

[0070] Figure 4 This is a flowchart illustrating the feature distribution smoothing method. Detailed Implementation

[0071] In this embodiment, a sorting system structure is used in a method for sorting retired power batteries that considers the uneven distribution of battery aging states, such as... Figure 1 As shown, it includes: voltage measurement equipment, charge / discharge test cabinet, data acquisition equipment, and storage and processing equipment. The functions of each device are as follows:

[0072] Voltage measuring equipment is used to quickly measure the current voltage value of retired batteries with different residual energy without needing to perform charge and discharge to obtain the curve;

[0073] Charge and discharge test cabinets are the most commonly used testing equipment for a large number of individual batteries. They have the characteristics of multi-channel high power and can be configured by a host computer to send corresponding steps to each channel of the battery to perform charge and discharge tests.

[0074] The data acquisition equipment is responsible for collecting battery test data, such as current, voltage, and energy, as well as external physical parameters, such as test temperature and battery size changes.

[0075] Data storage and processing devices are typically integrated into the same host computer. The storage device stores the collected battery information, while the processing device is responsible for data processing and training and application of the remaining energy estimation model. The flowchart of the entire sorting method is as follows: Figure 2 As shown, the specific steps include:

[0076] Step 1: Visually screen a batch of retired power batteries to obtain the remaining retired batteries after screening; the current state of charge of the retired batteries is not considered, and there is no need to discharge the remaining energy of the batteries.

[0077] Step Two: Randomly select test batteries; the total number obtained through random sampling is... Retired batteries were used as test batteries.

[0078] Step 3: Collect test data for the test battery;

[0079] Step 3.1: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require After the test battery has been left to stand for a period of time, the initial terminal voltage of the test battery is measured. ,in, Representing the The terminal voltage of the test battery;

[0080] Step 3.2: Construct a charge / discharge strategy and use it for... Each test battery was tested to obtain a battery test dataset. ,in, Indicates the first Dynamic characteristic data of each test battery Indicates the first Static characteristic data of each test battery;

[0081] Step 3.2.1: For Low-current hybrid pulse power characteristic tests were conducted on each test battery to obtain dynamic characteristic data that characterizes the battery's dynamic properties. ;in, Indicates pulse current, and , This represents the pulse current at the i-th multiplier. This represents the voltage change during the pulse test, and , This represents the voltage change of the m-th test battery in the i-th pulse current. Indicates the recovery voltage, and , This represents the recovery voltage of the m-th test battery. This indicates the time required to reach a stable recovery voltage, and , This represents the time it takes for the m-th test battery to reach a stable recovery voltage. This indicates the final voltage at which the test is completed, and ; This represents the final voltage of the m-th test battery.

[0082] No. Each test battery was charged to its initial state of charge using a constant current, and then left to stand for a period of time; then, it was charged according to a set pulse current. After standing, the first The test battery was subjected to a pulse discharge test, and the results were recorded. Voltage changes of individual test batteries during pulse processing ,in, The pulse current representing the i-th multiplier. Represents the pulse current at the i-th multiplier. Next The voltage change of each test battery, z represents the number of rate options;

[0083] After the pulse process ends, for the first Each test battery was left to stand until its voltage stabilized, and the corresponding recovery voltage was recorded. Resting time and open circuit voltage .

[0084] Step 3.2.2 For Partial charging tests were performed on each test battery to obtain static charging data that characterizes the battery's aging state. ;in, This indicates the initial state of charge reached during discharge. This represents the set state of charge termination. This represents the voltage data during the charging process, and , This represents the charging voltage data of the m-th test battery; This represents the charging capacity data during the charging process, and , This represents the charging capacity data of the m-th test battery; This represents the charging energy data during the charging process, and , This represents the charging energy data of the m-th test battery; This represents the terminal voltage after partial charging is complete, and , This represents the final voltage of the m-th test battery.

[0085] The first Each test battery was discharged to its initial state of charge. And record the initial voltage at this time. According to the set discharge current ratio, the first After the first test battery underwent initial discharge, the second... Each test battery was subjected to a constant current. Charge to the set state of charge. And record the first charge during the charging process. The voltage of each test battery Charging capacity and charging energy After charging is complete, the first... The test battery was left to stand for a period of time, and the reading was recorded. The final open-circuit voltage of each test battery after partial charging. .

[0086] Step 3.3: Use data acquisition equipment to collect the charging and discharging data of retired batteries in each channel and the ambient temperature data during the test, and transmit them to the host computer for storage via protocol communication.

[0087] Step 4: Data preprocessing and feature engineering;

[0088] Step 4.1: For After data cleaning, a clean dataset is obtained. ;

[0089] Step 4.2: Based on Build a basic feature library ,in, Indicates the first A set of dynamic characteristics of a test battery. Indicates the first A set of static features of a test battery;

[0090] Step 4.2.1: For Feature extraction was performed on the low-current hybrid pulse power characteristic test data to obtain a dynamic feature set. ;

[0091] Step 4.2.1.1: According to the first... Voltage changes of each test battery With pulse current Dynamic internal resistance obtained from doing business Using equations (1) and (2), we can obtain the first... Average dynamic internal resistance of each test battery and the rate of change of dynamic internal resistance And accordingly as the first The first dynamic characteristics of the test battery Second dynamic characteristics :

[0092] (1)

[0093] (2)

[0094] In equations (1) and (2), Let represent the dynamic internal resistance calculated under the i-th pulse current, and , This represents the dynamic internal resistance of the m-th test battery under the current pulse current. This indicates the maximum internal resistance. This represents the minimum internal resistance.

[0095] Step 4.2.1.2: Based on the recovery voltage Recovery time and initial terminal voltage Using equations (3) and (4), we can obtain the first... Difference in recovery voltage among individual test batteries and pulse recovery rate And accordingly as the first The third dynamic characteristic of the test battery and the fourth dynamic feature Thus, a dynamic feature set of n test batteries is obtained. ;

[0096] (3)

[0097] (4)

[0098] In equations (3) and (4), This represents the recovery voltage of the m-th test battery; This represents the recovery time of the m-th test battery; This represents the initial terminal voltage of the m-th test battery.

[0099] Step 4.2.2: For The voltage, energy, and capacity data obtained from partial charging tests are averaged, and the maximum and minimum extreme values ​​are taken. These values ​​are then used to calculate variance, kurtosis, skewness, quantiles, and curve slope, thereby obtaining a static feature set. ,in, Indicates the first Voltage-related characteristics of each test battery Indicates the first Capacity-related characteristics of each test battery Indicates the first Energy-related characteristics of the test battery Indicates the first Relevant characteristics of partial charging IC curves of individual test batteries;

[0100] The above feature sequences are used together to construct a static feature library. The specific static statistical characteristics can be calculated using formulas (5)-(12):

[0101] Maximum value:

[0102] (5)

[0103] Minimum value:

[0104] (6)

[0105] Mean:

[0106] (7)

[0107] Kuroshi:

[0108] (8)

[0109] Skewness:

[0110] (9)

[0111] Quantiles:

[0112] (10)

[0113] variance:

[0114] (11)

[0115] Slope:

[0116] (12)

[0117] In the above statistical calculation formulas, It is the data sequence of the d-th sample; This represents the number of data points in the sample. Represents quantiles, Represents the lower bound of the quantile. It is frequency It is the group spacing.

[0118] The IC curve calculation and feature selection are shown in equations (13)-(14):

[0119] The IC curve is the ratio of capacitance change to terminal voltage change. For the overall coordinates, the corresponding voltage The graph represents the internal electrochemical reactions during battery aging, plotted on the horizontal axis.

[0120] (13)

[0121] In equation (13), and They represent the corresponding The battery capacity and its terminal voltage at any given time; It is in discrete form. During its constant current charging and discharging phase, the calculation formula can be further transformed into:

[0122] (14)

[0123] In equation (14), This refers to the current during the constant current phase. The sampling interval is... Sampling time.

[0124] Based on the calculated IC curve, three indicators are selected as relevant features of the IC curve: the peak value on the IC curve and its corresponding position, and the area under the IC curve.

[0125] Step 4.3: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require The feature sequence consisting of all dynamic and static characteristics of each test battery and its capacity label. To splice together, and , The capacity label represents the m-th test battery; after normalization preprocessing, the standardized feature sequence is obtained. ,in, Representing the first The standardized feature sequence of each test battery; The feature sequence, composed of all dynamic and static features of each test battery, is concatenated with the battery's capacity label, and then normalized before being preprocessed to obtain the standardized feature sequence. ,in, Representing the first The standardized feature sequence of a test battery;

[0126] Will The input is processed in the LASSO model to obtain the first... Preliminary characteristic sequence of a test battery;

[0127] For the After randomly combining any two features from the initial feature sequence of the test battery, grey relational analysis is performed to select features with high correlation coefficients to form the first feature. The optimal feature sequence of a test battery ,in, Representing the The k-th optimal feature of a test battery The number of optimal features.

[0128] The optimization algorithm used is Least Absolute Shrinkage and Selection Operator (LASSO). LASSO, based on commonly used multiple linear regression, compresses feature coefficients by adding a penalty coefficient, adjusting the regression coefficients of low-correlation features to zero, thereby achieving feature selection. Subsequently, grey relational analysis (GRA) is performed on the results of LASSO feature optimization. The feature combinations with high correlation are selected as the final input for residual energy estimation.

[0129] Step 5: Construct a BiLSTM-based time-series regression prediction model, including: a feature extraction module, a feature smoothing module, a residual energy estimation module, and a temperature-based correction module, and use it to obtain the optimal feature sequence for n test batteries. The process is performed to obtain an estimate of the remaining energy of the retired batteries. ,and , This represents the estimated remaining energy of the m-th test battery.

[0130] Step 5.1: The feature extraction module uses a single BiLSTM layer to... Process and output the first... Depth features of each test battery ,in, Indicates the first The first test battery One deep feature, The total number of depth features; such as Figure 3 As shown;

[0131] Step 5.2: As Figure 4 As shown, the feature smoothing module uses a feature distribution smoothing strategy to... The process is performed to obtain the smoothed feature sequence. :

[0132] Step 5.2.1: Obtain the updated feature mean using equations (15) and (16). and the updated feature variance :

[0133] (15)

[0134] (16)

[0135] In equations (15) and (16), Indicates the smoothing coefficient; This indicates the m-th test battery. One characteristic. Let represent the feature mean and feature variance of the m-th test battery before the update, respectively.

[0136] Step 5.2.2: Use equation (17) to... Features in the deep feature sequence Smoothing is performed to obtain the smoothed feature sequence. ;

[0137] (17)

[0138] In equation (17), To prevent tiny constants with a denominator of zero, Indicates the first The first test battery after smoothing One deep feature;

[0139] Step 5.3: The residual energy estimation module consists of a fully connected layer and a regression estimation layer; wherein, the fully connected layer... Perform feature mapping to generate the first The mapping characteristics of the test battery; the regression estimation layer is based on the regression function on the first test battery. The mapping features of the test battery are processed to generate the first test battery. Residual energy estimate of each test battery ;

[0140] Step 5.4: Obtain the corrected residual energy sequence using equation (18). ,and , Represents the corrected residual energy value of the m-th test cell:

[0141] (18)

[0142] In equation (18), The coefficient representing the effect of temperature on capacity can be set to 0.0075 in this example. ; For reference temperature; To test the ambient temperature.

[0143] Step 5.5: Construct the comprehensive loss function according to equation (19) :

[0144] (19)

[0145] In equation (19), There are two balance coefficients. The loss function represents the residual energy estimation. Let represent the loss function in the feature smoothing process, and we have:

[0146] (20)

[0147] (twenty one)

[0148] Step 5.6: Train the time series regression prediction model using the Adam optimizer and calculate the comprehensive loss function. To update the model parameters until The process continues until convergence, thus obtaining the optimal residual energy estimation model for retired batteries. This model is used to estimate the residual energy of remaining retired batteries, and based on the obtained residual energy and the calculated average dynamic internal resistance... To achieve battery sorting.

[0149] Step Six: Estimation and sorting of remaining retired power batteries;

[0150] For the remaining retired batteries, repeat steps 3-4 above, performing only low-current hybrid pulse power characteristic (L-HPPC) testing and partial charging testing during the testing phase. After feature optimization, deploy a pre-trained residual energy estimation model to estimate the residual energy of each retired battery, and sort the batteries based on the obtained residual energy and the corresponding average dynamic internal resistance.

[0151] In this embodiment, an electronic device includes a memory and a processor. The memory stores a program that supports the processor in executing the above-described method, and the processor is configured to execute the program stored in the memory.

[0152] In this embodiment, a computer-readable storage medium stores a computer program, which is executed by a processor to perform the steps of the above method.

Claims

1. A method for sorting retired power batteries considering the uneven distribution of battery aging states, characterized in that, It includes the following steps: Step 1: Visually screen a batch of retired power batteries to obtain the remaining retired batteries after screening; Step 2: Randomly select from the remaining retired batteries to obtain a total of Test batteries; Step 3: Collect test data for the test battery; Step 3.1: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require the full context.] After the test battery has been left to stand for a period of time, the initial terminal voltage of the test battery is measured. ,in, Representing the The terminal voltage of the test battery; Step 3.2: Construct a charge / discharge strategy and use it for... Each test battery was tested to obtain a battery test dataset. ,in, Indicates the first Dynamic characteristic data of each test battery Indicates the first Static characteristic data of each test battery; Step 4: [Regarding...] After preprocessing, the standardized feature sequences are obtained. ;in, Representing the first The standardized feature sequence of a test battery; Step 5: The input is processed in the LASSO model to obtain the first... Preliminary characteristic sequence of a test battery; Step Six: For the first After randomly combining any two features from the initial feature sequence of the test battery, grey relational analysis is performed to select features with high correlation coefficients to form the first feature. The optimal feature sequence of a test battery ,in, Representing the The k-th optimal feature of a test battery The number of optimal features; Step 7: Construct a BiLSTM-based time-series regression prediction model, including: a feature extraction module, a feature smoothing module, a residual energy estimation module, and a temperature-based correction module, and use it to obtain the optimal feature sequence for n test batteries. The process is performed to obtain an estimate of the remaining energy of the retired batteries. ,in, This represents the estimated remaining energy of the m-th test battery; Step 8: Obtain the corrected residual energy sequence using equation (8) : (8) In equation (8), This is the coefficient representing the effect of temperature on capacity. For reference temperature; To test the ambient temperature; This represents the corrected residual energy value of the m-th test cell; Step 9: Construct the comprehensive loss function based on equation (9) : (9) In equation (9), There are two balance coefficients. The loss function represents the residual energy estimation. The loss function representing the feature smoothing process; Step 10: Train the time series regression prediction model using the Adam optimizer and calculate the comprehensive loss function. To update the model parameters until The process continues until convergence, thus obtaining the optimal residual energy estimation model for retired batteries. This model is used to estimate the residual energy of remaining retired batteries, and based on the obtained residual energy and the calculated average dynamic internal resistance... To achieve battery sorting.

2. The method for sorting retired power batteries considering the uneven distribution of battery aging states according to claim 1, characterized in that, Step 3.2 includes the following steps: Step 3.2.1: For Low-current hybrid pulse power characteristic tests were conducted on each test battery to obtain dynamic characteristic data that characterizes the battery's dynamic properties. ;in, Indicates pulse current, and , This represents the pulse current at the i-th multiplier. This represents the voltage change during the pulse test, and , This represents the voltage change of the m-th test battery in the i-th pulse current. Indicates the recovery voltage, and , This represents the recovery voltage of the m-th test battery. This indicates the time required to reach a stable recovery voltage, and , This represents the time it takes for the m-th test battery to reach a stable recovery voltage. This indicates the final voltage at which the test is completed, and ; This represents the final voltage of the m-th test battery; Step 3.2.2 For Partial charging tests were performed on each test battery to obtain static charging data that characterizes the battery's aging state. ;in, This indicates the initial state of charge reached during discharge. This represents the set state of charge termination. This represents the voltage data during the charging process, and , This represents the charging voltage data of the m-th test battery; This represents the charging capacity data during the charging process, and , This represents the charging capacity data of the m-th test battery; This represents the charging energy data during the charging process, and , This represents the charging energy data of the m-th test battery; This represents the terminal voltage after partial charging is complete, and , This represents the final voltage of the m-th test battery.

3. The method for sorting retired power batteries considering the uneven distribution of battery aging states according to claim 2, characterized in that, Step 3.2.1 is to... Each test battery was charged to its initial state of charge using a constant current, and then left to stand for a period of time; then, it was charged according to a set pulse current. After standing, the first The test battery was subjected to a pulse discharge test, and the results were recorded. Voltage changes of individual test batteries during pulse processing ,in, The pulse current representing the i-th multiplier. Represents the pulse current at the i-th multiplier. Next The voltage change of each test battery, z represents the number of rate options; After the pulse process ends, for the first Each test battery was left to stand until its voltage stabilized, and the corresponding recovery voltage was recorded. Resting time and open circuit voltage .

4. The method for sorting retired power batteries considering the uneven distribution of battery aging states according to claim 3, characterized in that, Step 3.2.2 is to... Each test battery was discharged to its initial state of charge. And record the initial voltage at this time. According to the set discharge current ratio, the first After the first test battery is initially discharged, the second... Each test battery was subjected to a constant current. Charge to the set state of charge. And record the first charge during the charging process. The voltage of each test battery Charging capacity and charging energy After charging is complete, the first... The test battery was left to stand for a period of time, and the reading was recorded. The final open-circuit voltage of each test battery after partial charging. .

5. A method for sorting retired power batteries considering the uneven distribution of battery aging states according to claim 4, characterized in that, Step four includes: Step 4.1: For After data cleaning, a clean dataset is obtained. ; Step 4.2: Based on Build a basic feature library ,in, Indicates the first A set of dynamic characteristics of a test battery. Indicates the first A set of static features of a test battery; Step 4.3: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require the full context.] The feature sequence consisting of all dynamic and static characteristics of each test battery and its capacity label. After concatenation and normalization preprocessing, the standardized feature sequence is obtained. ,in, The capacity label represents the m-th test battery.

6. The method for sorting retired power batteries considering the uneven distribution of battery aging states according to claim 5, characterized in that, Step 4.2 includes the following steps: Step 4.2.1: For Feature extraction was performed on the low-current hybrid pulse power characteristic test data to obtain a dynamic feature set. ; Step 4.2.2: For The voltage, energy, and capacity data obtained from partial charging tests are averaged, and the maximum and minimum extreme values ​​are taken. These values ​​are then used to calculate variance, kurtosis, skewness, quantiles, and curve slope, thereby obtaining a static feature set. ,in, Indicates the first Voltage-related characteristics of the test battery Indicates the first Capacity-related characteristics of each test battery Indicates the first Energy-related characteristics of the test battery Indicates the first The characteristics of the partial charging IC curve of the test battery.

7. A method for sorting retired power batteries considering the uneven distribution of battery aging states according to claim 6, characterized in that, Step 4.2.1 includes the following steps: Step 4.2.1.1: According to the first... Voltage changes of each test battery With pulse current Dynamic internal resistance obtained from doing business Using equations (1) and (2), we can obtain the first... Average dynamic internal resistance of each test battery and the rate of change of dynamic internal resistance And accordingly as the first The first dynamic characteristics of the test battery Second dynamic characteristics : (1) (2) In equations (1) and (2), Let represent the dynamic internal resistance calculated under the i-th pulse current, and , This represents the dynamic internal resistance of the m-th test battery under the current pulse current. This indicates the maximum internal resistance. This indicates the minimum internal resistance. Step 4.2.1.2: Based on the recovery voltage Recovery time and initial terminal voltage Using equations (3) and (4), we can obtain the first... Difference in recovery voltage among individual test batteries and pulse recovery rate And accordingly as the first The third dynamic characteristic of the test battery and the fourth dynamic feature Thus, a dynamic feature set of n test batteries is obtained. ; (3) (4) In equations (3) and (4), This represents the recovery voltage of the m-th test battery; This represents the recovery time of the m-th test battery; This represents the initial terminal voltage of the m-th test battery.

8. A method for sorting retired power batteries considering the uneven distribution of battery aging states according to claim 7, characterized in that, Step seven includes the following steps: Step 7.1: The feature extraction module utilizes a BiLSTM layer to... Process and output the first... Depth features of each test battery ,in, Indicates the first The first test battery One deep feature, The total number of depth features; Step 7.2: The feature smoothing module applies a feature distribution smoothing strategy to... The process is performed to obtain a smoothed feature sequence. ; Step 7.3: The residual energy estimation module consists of a fully connected layer and a regression estimation layer; wherein, the fully connected layer... Perform feature mapping to generate the first The mapping characteristics of the test battery; the regression estimation layer is based on the regression function on the first test battery. The mapping features of the test battery are processed to generate the first test battery. Residual energy estimate of each test battery .

9. A method for sorting retired power batteries considering the uneven distribution of battery aging states according to claim 8, characterized in that, Step 7.2 includes the following steps: Step 7.2.1: Obtain the updated feature mean using equations (5) and (6). and the updated feature variance : (5) (6) In equations (5) and (6), Indicates the smoothing coefficient; and They represent the first The feature mean and feature variance of each test battery before the update; Step 7.2.2: Use equation (7) to... Smoothing is performed to obtain the smoothed feature sequence. ; (7) In equation (7), To prevent tiny constants with a denominator of zero; Indicates the first The first test battery after smoothing One deep feature.

10. A method for sorting retired power batteries considering the uneven distribution of battery aging states according to claim 9, characterized in that, Step 9 involves using equations (10) and (11) to obtain the loss function for the residual energy estimation. loss function for feature smoothing process ; (10) (11)。

Images