Method for quickly classifying complementary energy of retired power batteries

Abstract

The invention provides a method for quickly classifying the complementary energy of a decommissioned power battery, which comprises the following steps of: measuring the residual capacity of the decommissioned power battery by adopting a high-rate constant-current discharge method, and generating a voltage-capacity curve; feature parameters are extracted based on the voltage-capacity curve, feature data sets are generated, and then the feature data sets are screened; optimizing the screened characteristic parameters, and inputting the optimized characteristic parameters into an ACO-SVR model for training and verification to obtain a decommissioned power battery complementary energy evaluation model; and according to an output result of the decommissioned power battery complementary energy evaluation model, the complementary energy of the batteries is divided into different grades, and different secondary utilization or recovery processing suggestions are provided for the batteries with different complementary energy grades. According to the method, the problems of low test precision and long time consumption of the retired power battery are effectively solved, and grade classification is carried out on the evaluated complementary energy of the retired power battery.

Description

Technical Field

[0001] The present invention belongs to the field of battery energy management, and in particular relates to a method for quickly classifying the remaining energy of retired power batteries. Background Art

[0002] With the popularity of electric vehicles and energy storage devices, the demand for lithium batteries has increased significantly. However, lithium batteries will experience capacity and energy decay during long-term use, eventually leading to battery retirement. Such retired power batteries still retain a certain amount of residual energy and have great potential for recycling.

[0003] The degradation of retired power batteries is mainly reflected in the following aspects: first, the capacity degradation of the battery, and second, the energy degradation of the battery. Capacity degradation is usually due to the loss of active lithium and the aging of electrode materials during the battery cycle, which leads to a reduction in the available capacity of the battery. Energy degradation is closely related to the voltage drop inside the battery and the increase in the internal impedance of the battery. These changes will directly affect the discharge capacity of the battery. However, the retirement management of power batteries requires accurate assessment of the remaining energy of the battery to ensure its reliability and safety after retirement. Traditional residual energy classification methods usually rely on complex testing equipment and a long testing process. Summary of the invention

[0004] In view of the problems existing in the existing methods for assessing and classifying the remaining energy of retired batteries, the present invention proposes a method for quickly classifying the remaining energy of retired power batteries, which aims to improve the timeliness and accuracy of the assessment and provide a scientific basis for the secondary utilization and recycling of retired batteries.

[0005] The technical solution of the present invention is as follows: A method for quickly classifying the remaining energy of retired power batteries, comprising: The remaining capacity of retired batteries is measured using a high-rate constant current discharge method to generate a voltage-capacity curve; Extracting characteristic parameters based on the voltage-capacity curve, generating a characteristic data set, and then screening the characteristic data set; The screened characteristic parameters are optimized based on adaptive weights and ant colony optimization algorithms, and the optimized characteristic parameters are input into the ACO-SVR model for training and verification to obtain a retired power battery remaining energy assessment model; According to the output results of the retired power battery residual energy assessment model, the battery's residual energy is divided into different levels, and different secondary utilization or recycling treatment suggestions are proposed for batteries with different residual energy levels.

[0006] Furthermore, the high-rate constant current discharge method is used to measure the remaining capacity of the retired battery, and the voltage-capacity curve is generated as follows: Discharge the battery at the specified discharge rate until the set cut-off voltage is reached. Calculate the amount of charge released during the discharge process and compare it with the nominal capacity of the battery to obtain the remaining capacity percentage of the battery. Record the battery voltage and discharge capacity data in real time to generate a voltage-capacity curve.

[0007] Furthermore, the extracting of characteristic parameters based on the voltage-capacity curve, generating a characteristic data set, and then screening the characteristic data set is specifically as follows: Extracting capacity loss characteristics, active lithium loss characteristics, and impedance characteristics based on the voltage-capacity curve to form a feature data set; The capacity loss characteristics, active lithium loss characteristics, and impedance characteristics are calculated respectively; Finally, the Pearson correlation coefficient and hypothesis test were used for analysis and screening.

[0008] Furthermore, the formulas for calculating the capacity loss characteristics, active lithium loss characteristics, and impedance characteristics are as follows: , in, is the capacity loss characteristic of the battery, is the initial capacity, is the capacity after cycling, is the voltage platform change, is the initial voltage platform voltage, Indicates the voltage value after the voltage platform changes. represents the total impedance, is the series resistance, is the charge transfer resistance, is the electrolyte resistance.

[0009] Furthermore, the Pearson correlation coefficient is specifically: , in, and Represent the data points of characteristic parameters and battery remaining energy respectively, and They represent the average values ​​of characteristic parameters and battery remaining energy respectively. The value range of r is [-1, 1]. Positive values ​​indicate positive correlation, negative values ​​indicate negative correlation, and 0 indicates no linear correlation. n is the number of samples.

[0010] Furthermore, the hypothesis test includes t test and F test, specifically: , in, is the regression coefficient of a certain characteristic parameter, It represents the standard error of the regression coefficient. By comparing the t value with the critical value, we can judge whether the characteristic parameter has an impact on the remaining energy of the battery. n is the number of samples, k is the number of characteristic parameters, and by comparing the F value with the critical value, we can judge whether the result is significant.

[0011] Furthermore, the ACO-SVR model is specifically: , E represents the remaining battery energy. is the support vector coefficient, is a kernel function used to measure input features With support vector The similarity between them is optimized by optimizing the loss function to find the appropriate To minimize the error between the characteristic parameters and the remaining energy of the battery.

[0012] Furthermore, the battery residual energy is divided into different levels, and different secondary utilization or recycling treatment suggestions are proposed for batteries with different residual energy levels. Specifically, the battery residual energy is divided into different levels, and different secondary utilization or recycling treatment suggestions are proposed for batteries with different residual energy levels. The battery with a remaining energy of more than 70% is classified as high level, the battery with a remaining energy of between 50% and 70% is classified as medium level, and the battery with a remaining energy of less than 50% is classified as low level; It is recommended that high-grade batteries be used for secondary use; medium-grade batteries are recommended to be used in small electric equipment; and low-grade batteries are recommended to be recycled.

[0013] Compared with the prior art, the present invention has the following advantages: (1) The present invention can accurately predict the remaining energy of retired power batteries by extracting the characteristics of the voltage-capacity curve and combining the evaluation model of adaptive weight and ant colony optimization algorithm, thereby improving the timeliness and accuracy of the remaining energy evaluation of retired batteries.

[0014] (2) Based on the residual energy evaluation results, the present invention classifies retired batteries into different levels and proposes corresponding secondary utilization or recycling treatment suggestions to improve the reuse efficiency and resource utilization rate of retired batteries and optimize secondary utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The accompanying drawings generally illustrate various embodiments by way of example and not limitation, and together with the description and claims, serve to illustrate the embodiments of the invention. Where appropriate, the same reference numerals are used throughout the drawings to refer to the same or similar parts. Such embodiments are illustrative and are not intended to be exhaustive or exclusive embodiments of the present apparatus or method.

[0016] Figure 1 A schematic diagram of a specific process of the method of the present invention is shown; Figure 2The overall framework schematic diagram of the present invention is shown. DETAILED DESCRIPTION

[0017] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0018] like Figure 1-Figure 2 As shown, an embodiment of the present invention provides a method for quickly classifying the remaining energy of retired power batteries, comprising: S1, by extracting the voltage-capacity curve generated by the retired power battery during the high-rate constant-current discharge process; More specifically, they include: 100 retired power battery samples of different models and production dates were selected, including ternary lithium batteries, lithium iron phosphate batteries, etc. The experimental equipment includes constant current charging and discharging instruments, data acquisition systems and professional analysis software. By setting constant current discharge tests at different rates (1C, 2C, 5C), the battery is discharged from a full charge state to a cut-off voltage of 2.5V, and the voltage and capacity data are recorded every second during the discharge process.

[0019] The voltage-capacity data collected during the experiment are stored in the local database through the data acquisition system, and a voltage-capacity curve is generated.

[0020] S2, feature parameter extraction, including capacity loss feature, active lithium loss feature, impedance feature, and acquisition of feature data set; More specifically, they include: The following characteristic parameters are extracted from each voltage-capacity curve: capacity loss characteristic, active lithium loss characteristic, impedance characteristic, and the original data set is established. The data set is: , in, represents the battery voltage at the i-th time point, Indicates the corresponding capacity, represents the jth test cycle, represents the corresponding capacity loss value, Indicates the voltage platform area, Indicates the corresponding voltage average value.

[0021] Then the capacity loss characteristics, active lithium loss characteristics, and impedance characteristics are calculated respectively, and the calculation formula is: , in, is the initial capacity, is the capacity after cycling, is the initial voltage platform voltage, Indicates the voltage value after the voltage platform changes. represents the total impedance, is the series resistance, is the charge transfer resistance, is the electrolyte resistance.

[0022] Finally, we use the relevant statistical method, the Pearson correlation coefficient: , and They represent the data points of characteristic parameters (such as capacity loss, active lithium loss) and battery remaining energy, and They represent the average values ​​of characteristic parameters and battery remaining energy respectively. The value range of r is [-1, 1]. Positive values ​​indicate positive correlation, negative values ​​indicate negative correlation, and 0 indicates no linear correlation.

[0023] And hypothesis testing for analysis and screening: , in, is the regression coefficient of a certain characteristic parameter, Indicates the standard error of the regression coefficient. By comparing the t value with the critical value, it is determined whether the characteristic parameter has an impact on the remaining battery energy. n is the number of samples, k is the number of characteristic parameters, and by comparing the F value with the critical value, it is determined whether the result is significant.

[0024] The feature selection algorithm is then applied to the preliminary screening features for further screening to select the most representative features.

[0025] S3, introduces an evaluation model based on adaptive weight and ant colony optimization algorithm, which optimizes the screened capacity loss characteristics and other related parameters, and inputs the optimized characteristic parameters into the ACO-SVR model for training; More specifically, they include: The extracted characteristic parameters such as capacity loss, active lithium loss, impedance, etc. are input into the adaptive weight. In each round of iteration, the pheromone weight α and the heuristic function weight β are adaptively adjusted to dynamically adjust the pheromone update method: , is the self-use weight of path solution i, is the quality of path solution i, is the quality of the current optimal solution, is the adjustment factor, is the updated weight factor , is the updated weight factor , Current weight factor , is the current weight factor .

[0026] And introduce the ACO evaluation model for optimization: , is the pheromone concentration on path i to j, is the heuristic function (such as the empirical effect value of the path), α and β are the weights of pheromone and heuristic function respectively, is the path k in the set of currently available paths.

[0027] , in, For parameters The objective function is is the support vector coefficient, is the support vector corresponding coefficient, is the kernel function and n is the number of samples.

[0028] It is an objective function with support vector coefficient as parameter. Its purpose is to improve the performance of the model by optimizing the weight coefficient of support vector. Specifically, the positive term of the formula reflects the interaction between samples. Describe the sample , The negative term in the formula represents the penalty for each support vector weight during the optimization process to avoid overfitting caused by excessive weights. The process is actually to adjust To find and weigh the importance of support vectors, thereby highlighting key features and suppressing redundant or noisy features.

[0029] In the ACO-SVR framework, As a bridge, it provides the optimal parameter input for subsequent model training by optimizing support vector weights and feature selection.

[0030] optimization The support vector weight parameters obtained It is directly used as the input of the subsequent ACO-SVR model for accurate modeling of features and prediction of unknown samples.

[0031] Finally, the optimized feature parameters are input into the ACO-SVR model for training: , E represents the remaining battery energy. is the support vector coefficient, is a kernel function used to measure input features With support vector The similarity between them is optimized by optimizing the loss function to find the appropriate To minimize the error between the characteristic parameters and the remaining energy of the battery.

[0032] S4, verify the prediction accuracy and stability of the model through experimental data, and quantitatively evaluate the prediction accuracy of the model; More specifically, they include: The prediction accuracy and stability of the model were verified by K-fold cross validation method: , in, is the MSE of K-fold cross validation i The average value of is the root mean square error of the i-th validation, and K is the fold number. The sample data is divided into 10 groups equally, and 9 groups are used as training sets each time, and the remaining 1 group is used as the validation set, and repeated 10 times.

[0033] Then, the mean square error (MSE) and mean absolute error (MAE) are used to quantitatively evaluate the prediction performance of the model: , in, is the prediction residual energy of the ith sample, is the true residual energy, and n is the number of samples. The calculated MSE and MAE values ​​are 0.015 and 0.02, respectively, verifying that the model has high prediction accuracy and stability.

[0034] S5, residual energy level classification and application recommendations.

[0035] More specifically, they include: According to the model output results, the remaining energy of the battery is divided into three levels: high (greater than 70%), medium (50%-70%) and low (less than 50%).

[0036] For high-grade batteries, secondary utilization is recommended, such as in energy storage systems; for medium-grade batteries, use is recommended in small electric devices; and for low-grade batteries, recycling is recommended.

[0037] In the above embodiment, based on the voltage-capacity characteristic curve, a method for rapid evaluation and classification of the remaining energy of retired batteries is established. The remaining capacity of the retired power battery is measured by a high-rate constant current discharge method, and the voltage and discharge capacity data are recorded in real time to generate a voltage-capacity curve. By extracting the capacity loss characteristics, active lithium loss characteristics and impedance characteristics in the curve, a data set is constructed. Next, an evaluation model based on adaptive weight adjustment and ant colony optimization algorithm is used, and the extracted characteristic parameters are input into the ACO-SVR model for training, and the prediction accuracy of the model is verified. The MSE and MAE values ​​are 0.015 and 0.02, respectively, which can realize a rapid and accurate evaluation of the remaining energy of the battery, and classify the remaining energy level of the battery according to the evaluation results, and put forward corresponding secondary utilization or recycling treatment suggestions.

[0038] In summary, the present invention can realize the rapid evaluation of the remaining energy of retired power batteries by extracting the capacity loss, active lithium loss and impedance characteristics from the voltage-capacity curve, combining the evaluation model of adaptive weight and ant colony optimization algorithm, and greatly shortening the evaluation time. At the same time, the ACO-SVR model is trained based on the optimized characteristic parameters, making the prediction results more accurate and significantly improving the accuracy of the remaining energy evaluation of retired batteries.

[0039] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for quickly classifying the remaining energy of retired power batteries, characterized in that: include: The remaining capacity of retired batteries is measured using a high-rate constant current discharge method to generate a voltage-capacity curve; Extracting characteristic parameters based on the voltage-capacity curve, generating a characteristic data set, and then screening the characteristic data set; The screened characteristic parameters are optimized based on adaptive weights and ant colony optimization algorithms, and the optimized characteristic parameters are input into the ACO-SVR model for training and verification to obtain a retired power battery remaining energy assessment model; According to the output results of the retired power battery residual energy assessment model, the battery's residual energy is divided into different levels, and corresponding secondary utilization or recycling treatment suggestions are made for batteries of each residual energy level.

2. The method for rapid classification of residual energy of retired power batteries according to claim 1 is characterized in that: The high-rate constant current discharge method is used to measure the remaining capacity of the retired battery, and the voltage-capacity curve is generated as follows: Discharge the battery at the specified discharge rate until the set cut-off voltage is reached. Calculate the amount of charge released during the discharge process and compare it with the nominal capacity of the battery to obtain the remaining capacity percentage of the battery. Record the battery voltage and discharge capacity data in real time to generate a voltage-capacity curve.

3. The method for rapid classification of residual energy of retired power batteries according to claim 1 is characterized in that: The extracting characteristic parameters based on the voltage-capacity curve, generating a characteristic data set, and then screening the characteristic data set specifically includes: Extracting capacity loss characteristics, active lithium loss characteristics, and impedance characteristics based on the voltage-capacity curve to form a feature data set; The capacity loss characteristics, active lithium loss characteristics, and impedance characteristics are calculated respectively; Finally, the results were analyzed and screened through Pearson correlation coefficient and hypothesis test.

4. The method for rapid classification of residual energy of retired power batteries according to claim 3 is characterized in that: The formula for calculating the capacity loss characteristic, active lithium loss characteristic, and impedance characteristic is: ， in, is the capacity loss characteristic of the battery, is the initial capacity, is the capacity after cycling, is the voltage platform change, is the initial voltage platform voltage, Indicates the voltage value after the voltage platform changes. represents the total impedance, is the series resistance, is the charge transfer resistance, is the electrolyte resistance.

5. The method for rapid classification of residual energy of retired power batteries according to claim 3 is characterized in that: The Pearson correlation coefficient is specifically: ， in, and Represent the data points of characteristic parameters and battery remaining energy respectively, and They represent the average values ​​of characteristic parameters and battery remaining energy respectively. The value range of r is [-1, 1]. Positive values ​​indicate positive correlation, negative values ​​indicate negative correlation, and 0 indicates no linear correlation. n is the number of samples.

6. The method for rapid classification of residual energy of retired power batteries according to claim 3 is characterized in that: The hypothesis test includes t test and F test, specifically: ， in, is the regression coefficient of the characteristic parameter, represents the standard error of the regression coefficient, n is the number of samples, and k is the number of characteristic parameters.

7. The method for rapid classification of residual energy of retired power batteries according to claim 1, characterized in that: The ACO-SVR model is specifically: ， Where E represents the remaining energy of the battery. is the support vector coefficient, is a kernel function used to measure input features With support vector The similarities between.

8. The method for rapid classification of residual energy of retired power batteries according to claim 1, characterized in that: The battery residual energy is divided into different levels, and corresponding secondary utilization or recycling treatment suggestions are proposed for batteries of each residual energy level, specifically including: The battery with a remaining energy of more than 70% is classified as a high level, the battery with a remaining energy of between 50% and 70% is classified as a medium level, and the battery with a remaining energy of less than 50% is classified as a low level; It is recommended that high-grade batteries be used for secondary use; medium-grade batteries are recommended to be used in small electric equipment; and low-grade batteries are recommended to be recycled.

Images