Retired power battery rapid sorting method based on interval frequency electrochemical impedance spectroscopy

Abstract

The application discloses a kind of based on interval frequency electrochemical impedance spectroscopy's quick sorting method of decommissioned power battery, comprising:1 carries out appearance preliminary selection to the decommissioned battery of disassembly monomer;2 after all batteries are charged and discharged to a same voltage, the open-circuit voltage after being measured for the same time is used as a sorting index;3 in fixed frequency range, the impedance spectrum of each battery is measured by impedance measurement device;4 extract the time domain and frequency domain hybrid features based on impedance spectrum, after removing redundant features using mutual information method based on critical weight, obtain sorting index;5 using two-step method combined with Bayesian Gaussian mixture model, the battery containing sorting index is clustered and sorted, and the battery meeting consistency condition is obtained, and the battery not meeting condition enters scrap metal recycling link.The application can avoid capacity test of large quantities of decommissioned batteries and improve the consistency of group battery, so as to improve the efficiency of decommissioned battery sorting and the safety of group battery.

Description

Technical Field

[0001] This invention belongs to the field of power battery cascade utilization, specifically a rapid sorting method for retired power batteries based on interval frequency electrochemical impedance spectroscopy. Background Technology

[0002] In recent years, with the explosive growth in the production and sales of electric vehicles, power batteries, which have advantages such as high energy density, low self-discharge rate, long lifespan, and almost zero memory effect, have been widely used as energy storage devices for electric vehicles. However, these batteries become unsuitable for electric vehicles after their usable capacity drops below 80% or they fail, entering a period of large-scale retirement. The large number of discarded batteries, if not properly disposed of, will cause significant environmental damage. Therefore, the safe and environmentally friendly disposal of these retired batteries is of paramount importance. Research shows that retired power batteries still have high residual value. Besides being dismantled and having their various rare and precious metals, such as lithium, nickel, and cobalt, recovered, they can also be screened and reused in electric bicycles, touring vehicles, low-speed electric vehicles, smart grids, and wind and solar energy storage devices. Because retired power batteries have different historical usage records and varying performance parameters, directly assembling them can lead to premature aging, thermal runaway, and explosion risks. Therefore, retired batteries must be classified and screened to improve the consistency of battery pack assembly.

[0003] Currently, common sorting methods for retired power batteries can be categorized into capacity-based testing and impedance-based measurement. Capacity testing is the most direct method for estimating the state of health (SOH) of a battery, obtained by directly measuring the battery's fully charged capacity. However, capacity testing is time-consuming and has a high cumulative error. Ohmic impedance extraction typically involves establishing an equivalent circuit model and then identifying parameters to determine its relationship with SOH. The challenge lies in establishing a reasonable and effective equivalent model and designing suitable parameter identification experiments. Both methods are not only time-consuming and labor-intensive, but also result in inconsistent performance among batteries sorted based solely on current performance indicators due to variations in historical usage environments. Summary of the Invention

[0004] The present invention addresses the shortcomings of the existing technology by proposing a rapid sorting method for retired power batteries based on interval frequency electrochemical impedance spectroscopy. This method aims to significantly shorten testing time and improve sorting efficiency by using electrochemical impedance characteristics as a sorting index, while ensuring good consistency in the sorting of retired power batteries.

[0005] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0006] The present invention provides a rapid sorting method for decommissioned power batteries based on interval frequency electrochemical impedance spectroscopy, characterized by the following steps:

[0007] Step 1: Initial selection of battery appearance:

[0008] Collect appearance images of M0 batteries of the same type and input them into an image recognition model for defect detection. Select M1 batteries with intact appearance from the detection results.

[0009] Step 2: Battery voltage stabilization:

[0010] Under constant temperature conditions, M1 intact batteries are discharged to their lower voltage limit U using a charging and discharging device. d After standing for a period of time, the terminal voltage {U} of each battery was measured. 0,m |m=1,2,…,M1} and used as sorting features, where U 0,m This represents the terminal voltage of the m-th battery with intact appearance;

[0011] Step 3: Obtain the electrochemical impedance spectroscopy (EIS) spectrum of the battery:

[0012] Step 3.1: Obtain the electrochemical impedance spectroscopy of new batteries of the same type, and use it to determine the frequency range of the impedance spectrum frequencies to be measured for M1 batteries. The frequency range is the imaginary part value - Z. im The frequency range from ≥0 to the last trough of the Nyquist curve;

[0013] Step 3.2: Under constant temperature environment, place the M1 batteries after they have been left to stand into the impedance measuring device, and apply small-amplitude sinusoidal voltage excitation signals of different frequencies in the frequency range to the m-th battery respectively, measure the corresponding different response current signals of the m-th battery, and then perform FTT transformation on the small-amplitude sinusoidal voltage excitation signal of the same frequency and its corresponding response current signal to calculate the impedance value of the m-th battery at the same frequency, thereby obtaining the impedance value of the m-th battery at different frequencies.

[0014] Step 4: Extract features based on impedance information:

[0015] Step 4.1: Using the real part of the impedance value of the m-th battery at different frequencies as the independent variable and the negative of the imaginary part as the dependent variable, fit the 5th order polynomial function of the real and imaginary parts of the m-th battery.

[0016] Plot the Nyquist curve of the m-th battery based on the 5th order polynomial function, and extract the classification features related to the performance and aging degree of the m-th battery from the Nyquist curve to obtain the classification features of all M1 batteries.

[0017] Step 4.2: Convert the frequency domain information of the impedance value of the m-th battery at different frequencies into time domain features using the DRT method, and plot the DRT curve of the m-th battery. Then extract the classification features related to the performance and aging degree of the m-th battery from the DRT curve to obtain the classification features of all M1 batteries.

[0018] Step 4.3: Use the mutual information method based on critical weights to remove redundant features from all classification features of M1 batteries, and obtain the remaining classification features as sorting features;

[0019] Step 5: Two-step battery sorting:

[0020] Step 5.1: After normalizing the sorting features obtained in Step 2 and Step 4, the M1 batteries are processed using the Bayesian-Gaussian mixture clustering algorithm based on the normalized sorting features to obtain several battery sets, and the battery set C1 with the best health status is selected.

[0021] Step 5.2: Based on the number of inflection points in the electrochemical impedance spectroscopy and the number of peaks in the DRT curve, a second screening is performed on the battery set C1. Batteries containing only 2 EIS inflection points and fewer than 4 DRT peaks are selected from C1 as the final selected batteries.

[0022] The rapid sorting method for decommissioned power batteries based on interval frequency electrochemical impedance spectroscopy described in this invention is also characterized in that the mutual information method based on critical weights in step 4.3 is performed as follows:

[0023] Step a: Denote the set of all classification features of all M1 batteries as the initial feature set S0, and denote any i-th classification feature vector of the M1 batteries in S0 as f. i Let f be any j-th classification feature vector of M1 batteries in S0. j ;

[0024] Step b: Calculate f using equation (3) i and f j Redundancy value R(f) i ,f j ), and determine R(f) i ,f j If f is greater than a given threshold, then f will be set to f. i NS number of filters i and f j NS number of filters j Add 1 to each item; otherwise, proceed to step c.

[0025] R(f i ,f j )=2I(fi ,f j ) / (H(f i )+H(f i ,f j (3)

[0026] In equation (3), I(f) i ,f j ) represents f i and f j The mutual information value, where H represents entropy;

[0027] Step c: Calculate f using equation (4) i Objective weight WR i , if WR i >WR j Then NS i Add 1, if WR i <WR j Then NS j Add 1, if WR i =WR j Then proceed directly to step d, where WR j f j Objective weights;

[0028]

[0029] In equation (4), W i f i The weight, K i f i The characteristic variation index is calculated by equation (5); R i f i The characteristic conflict index is calculated by equation (6);

[0030]

[0031]

[0032] In equations (5) and (6), f im f is the value of the i-th classification feature in the m-th battery. tm Let |S0| be the value of the t-th classification feature in the m-th battery, and |S0| be the number of classification features in the initial feature set S0. Let be the mean of the i-th classification feature vector of M1 batteries. Let r be the mean of the t-th classification feature vector of M1 batteries. it f is the i-th classification feature vector of M1 batteries i With the t-th classification feature vector f t Similarity;

[0033] Step d: Determine whether pairwise comparisons have been completed among all classification features in S0. If so, sort them in ascending order according to the number of times each classification feature has been filtered, and select the top few classification features as the remaining classification features after redundancy removal to form the sorting feature set. Otherwise, return to step b and execute sequentially.

[0034] The present invention provides an electronic device, including a memory and a processor, wherein the memory is used to store a program that supports the processor in executing the rapid sorting method for retired power batteries, and the processor is configured to execute the program stored in the memory.

[0035] The present invention discloses a computer-readable storage medium on which a computer program is stored, wherein the computer program, when executed by a processor, performs the steps of the rapid sorting method for retired power batteries.

[0036] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0037] 1. The method of this invention eliminates the need for time-consuming capacity testing of batteries and the need for complex model identification of impedance parameters, saving significant time and improving sorting efficiency. According to the national standard GB / T 31484-2015, capacity testing includes five 1C CC-CV charging cycles and a 1C discharging cycle, taking the average discharge capacity as the capacity. This process also requires multiple long periods of rest, typically lasting tens of hours. In contrast, the method of this invention only requires charging and discharging to the same cutoff voltage, followed by a 30-minute rest period to measure the open-circuit voltage, and then another 3-5 minutes to measure the electrochemical impedance at a specific frequency range, thus greatly improving sorting efficiency.

[0038] 2. The present invention uses a mutual information method based on critical weights to remove redundant features, which not only reduces the dimensionality of the feature set, but also allows the selected features to better reflect the aging status of the battery. As a sorting index, it can improve the consistency performance of recombined batteries.

[0039] 3. The electrochemical impedance spectroscopy (EIS) used in this invention is a non-invasive technique. It involves applying a sinusoidal interference signal to the battery within a specific frequency range and measuring the response signal to calculate the impedance spectrum. Impedance spectroscopy can explain the historical aging state of the battery at a mechanistic level, providing information about different aging mechanisms. Therefore, applying EIS to battery sorting in this invention can significantly improve the consistency of recombined batteries. Furthermore, the measurement-based battery sorting process can reduce energy waste caused by battery cycle charging and discharging, and greatly shorten sorting time, saving sorting costs.

[0040] 4. The Bayesian-Gaussian mixture clustering algorithm used in this invention to analyze aging characteristic indicators is a soft clustering method based on a statistical model. By iteratively optimizing the model parameters, it can find the best fit. It has advantages such as eliminating the need for manual searching for the best category of clusters and automatically eliminating unnecessary clusters. Attached Figure Description

[0041] Figure 1 This is a flowchart of the battery sorting process based on interval frequency electrochemical impedance spectroscopy of the present invention;

[0042] Figure 2 Example graph showing the frequency range of electrochemical impedance spectroscopy measurements;

[0043] Figure 3 Impedance fitting and DRT curve plot;

[0044] Figure 4 This is a flowchart of the feature selection method of the present invention;

[0045] Figure 5 A scatter plot of cluster feature distribution;

[0046] Figure 6 This is a diagram of the Bayesian-Gaussian mixture model structure. Detailed Implementation

[0047] In this embodiment, a rapid sorting method for retired power batteries based on interval frequency electrochemical impedance spectroscopy is presented. The impedance measurement system includes an image recognition initial sorting device, a battery charge / discharge detection system, an electrochemical impedance measurement system, a constant temperature control device, and data analysis and operation equipment. Compared with traditional electrochemical workstation measurements, this impedance measurement system can measure a large number of individual cells at once. Furthermore, the system can be configured with different frequency ranges and excitation signal waveforms for measurement, offering greater flexibility and applicability, suitable for simultaneous measurement of large batches of batteries.

[0048] like Figure 1 As shown, this rapid sorting method for retired power batteries involves first performing a voltage test on batteries initially selected by appearance, then measuring the open-circuit voltage after allowing them to stand. Next, the electrochemical impedance spectroscopy (EIS) information of the batteries is collected, and DRT analysis is performed on the EIS spectra to obtain aging characteristics. These characteristics are then used as model input for a two-step sorting process based on Bayesian-Gaussian mixture clustering, thereby achieving the goal of recombining batteries with good consistency for secondary use. Specifically, the method proceeds as follows:

[0049] Step 1: Initial selection of battery appearance:

[0050] Step 1.1: Disassemble the battery pack and use a high-definition camera to capture images of the appearance of M0 batteries of the same type;

[0051] Step 1.2: Using a Convolutional Neural Network (CNN) model, classify the M0 batteries based on the acquired images, classifying them into categories such as "intact," "damaged," and "bulging." The function definition of the CNN classification model is shown in equation (1):

[0052]

[0053] In equation (1), input represents the input image, f is the functional relationship implemented by the classification algorithm, and prob classn The probability of different categories is represented. M1 batteries with an "intact" appearance are selected to proceed to the next step, while batteries that do not meet the appearance requirements are discarded and sent to the dismantling and recycling stage.

[0054] Step 2: Battery voltage stabilization:

[0055] Under a constant temperature environment of 25℃, after fully charging the M1 batteries selected in step one, discharge them to a uniform lower voltage limit U. d After standing for 30 minutes, measure the terminal voltage {U} of each battery. 0,m |m=1,2,…,M1} and used as sorting features, where U 0,m This represents the terminal voltage of the m-th intact battery. At this point, due to polarization, U... 0m >U d Battery polarization can be classified into three types: electrochemical polarization, concentration polarization, and ohmic polarization. These three types of polarization reflect the resistance to the electrochemical reactions inside the battery. Due to the different internal aging processes of different batteries, their electrochemical environments also differ, resulting in variations in U0 for each battery. This means that the terminal voltage U0 can reflect the degree of polarization of the battery and can serve as an important characteristic for battery sorting.

[0056] Step 3: Measurement of the battery's electrochemical impedance spectroscopy:

[0057] Step 3.1, Determination of the frequency range for electrochemical impedance spectroscopy measurement: Obtain the electrochemical impedance spectra of new batteries of the same type, and use them to determine the frequency range of the impedance spectra to be measured for M1 batteries. The frequency range is the imaginary part value - Z. im The frequency range from ≥0 to the last trough of the Nyquist curve, such as... Figure 2 As shown.

[0058] Step 3.2, Electrochemical impedance spectroscopy measurement method:

[0059] Under constant temperature conditions, M1 batteries after being left to stand are placed in an impedance measuring device. Small-amplitude sinusoidal voltage excitation signals of different frequencies within the frequency range are applied to the m-th battery, and the corresponding different response current signals of the m-th battery are measured. Then, the small-amplitude sinusoidal voltage excitation signal of the same frequency and its corresponding response current signal are subjected to FTT transformation to calculate the impedance value of the m-th battery at the same frequency, thereby obtaining the impedance value of the m-th battery at different frequencies.

[0060] Step 3.3, Principle of obtaining electrochemical impedance spectroscopy:

[0061] Step 3.3.1: Configure a bipolar power supply to output different frequencies ω to the battery. l A small-amplitude sinusoidal AC voltage excitation signal V (l=1,2,…,k) l (l=1,2,...,k), where k is the number of frequencies to be measured, and the value of k is determined by taking ten points for every ten frequency ranges;

[0062] Step 3.3.2: Simultaneously acquire the excitation voltage signal V of the bipolar power supply using the data acquisition card. l (l=1,2,...,k) and the response current signal I of the battery under test l (l = 1, 2, ..., k);

[0063] Step 3.3.3, the measured excitation voltage signal V l With response current signal I l After FTT transformation, f(V) is obtained l ) and f(I l Then, the AC impedance Z at the l-th frequency can be calculated using equation (2). l :

[0064]

[0065] In equation (2), Let j be the real part of the impedance at the l-th frequency, and j be a complex number. Let be the imaginary part of the impedance at the l-th frequency. The x-axis is... Plotting the Nyquist curve on the ordinate is equivalent to an electrochemical impedance spectroscopy (EIS).

[0066] Step 4: Extract features based on impedance information:

[0067] Step 4.1: Fit the Nyquist curve and extract aging features:

[0068] Using the real part of the impedance value of the m-th battery at different frequencies as the independent variable and the imaginary part as the dependent variable, the 5th-order polynomial function of the real and imaginary parts of the m-th battery is fitted using the least squares fitting function polyfit(X,Y,N) provided in MATLAB. An example of the fitting is shown below. Figure 3 As shown in part a of the diagram.

[0069] The Nyquist curve of the m-th battery is plotted based on the 5th-order polynomial function. Classification features related to the performance and aging degree of the m-th battery are extracted from the Nyquist curve, including: the position of the curve inflection point, the peak and valley values ​​and their positions, the derivative of the curve in a specific frequency range, the intercept of the curve with the horizontal axis, and the imaginary part of the initial and final frequencies, thereby obtaining the classification features of all M1 batteries.

[0070] Step 4.2: Convert the frequency domain information of the impedance value of the m-th battery at different frequencies into time domain characteristics using the DRT method, and plot the DRT curve of the m-th battery, as shown below. Figure 3 As shown in section b, classification features related to the performance and aging degree of the m-th battery are extracted from the DRT curve, including the peak value and position of each peak, thus obtaining the classification features of all M1 batteries.

[0071] Step 4.3: Use the Mutual Information Based on Critical Weights (MIBCW) method to remove redundant features from all classification features S0 of all M1 batteries, obtaining a feature set containing four features. As a sorting feature, the feature selection process is as follows: Figure 4 As shown, proceed as follows:

[0072] Step a: Denote the set of all classification features of all M1 batteries as the initial feature set S0, and denote any i-th classification feature vector of the M1 batteries in S0 as f. i Let f be any j-th classification feature vector of M1 batteries in S0. j .

[0073] Step b: Calculate f using equation (3) i and f j Redundancy value R(f) i ,f j ), and determine R(f) i ,f j If f is greater than a given threshold, then f will be set to f. i NS number of filters i and f j NS number of filters j Add 1 to each item; otherwise, proceed to step c.

[0074] R(f i ,f j )=2I(f i ,f j ) / (H(f i )+H(f i ,f j (3)

[0075] In equation (3), I(f) i ,f j ) represents f i and f j The mutual information value, H represents entropy.

[0076] Step c: Calculate f using equation (4) i Objective weight WR i f j Objective weight WR j The calculation method is the same; if WR i >WR j Then NS i Add 1, if WR i <WR j Then NS j Add 1, if WR i =WR j If so, proceed directly to step d;

[0077]

[0078] In equation (4), W i f i The weight, K i f i The characteristic variation index is calculated by equation (5); R i f i The characteristic conflict index is calculated by equation (6);

[0079]

[0080]

[0081] In equations (5) and (6), f im f is the value of the i-th classification feature in the m-th battery. tm Let |S0| be the value of the t-th classification feature in the m-th battery, and |S0| be the number of classification features in the initial feature set S0. Let be the mean of the i-th classification feature vector of M1 batteries. Let r be the mean of the t-th classification feature vector of M1 batteries. itf is the i-th classification feature vector of M1 batteries i With the t-th classification feature vector f t The similarity.

[0082] Step d: Determine whether pairwise comparisons have been completed among all classification features in S0. If so, sort them in ascending order according to the number of times each classification feature has been filtered, and select the top few classification features as the remaining classification features after redundancy removal to form the sorting feature set. Otherwise, return to step b and execute sequentially. The feature set obtained by this method... It includes four classification features with relatively low residual redundancy: the derivative of the EIS curve in a specific frequency range, the intercept of the EIS curve with the horizontal axis, the maximum peak value of the DRT curve, and the position of the second-to-last peak of the DRT curve.

[0083] Step 5: Two-step battery sorting:

[0084] Step 5.1: Based on the sorting features obtained in Step 2 and the sorting feature set obtained in Step 4. We jointly construct a sorting feature set S = {s1, s2, s3, s4, s5} containing five features of M1 batteries, and normalize all features to a standard feature set F = {F1, F2, F3, F4, F5}. We define the i-th standard feature F... i ={F i1 ,F i2 ,...,F im}, i = 0, 1, ..., 5, using equation (7) to obtain the normalized value F of the i-th feature of the m-th sample. im :

[0085]

[0086] In equation (7), s im Let s be the initial feature value of the m-th sample. imin Let s be the minimum value of the i-th feature among all samples. imax This represents the maximum value of the i-th feature across all samples. A normalized feature scatter plot is shown below. Figure 5 As shown.

[0087] Step 5.2 Two-step battery sorting based on Bayesian Gaussian mixture model: Using the feature set F obtained in Step 5.1 as an example, and the classification category as the label, the Bayesian Gaussian mixture clustering algorithm is used to process M1 batteries, resulting in several battery sets of different categories. The battery set C1 with the best health condition is then selected. The structure of the Bayesian Gaussian mixture model is as follows: Figure 6 As shown, squares represent parameters, circles represent random variables, and arrows represent conditional dependencies.

[0088] Step 5.3, Secondary Sorting: Based on the number of inflection points in the electrochemical impedance spectroscopy and the number of peaks in the DRT curve, the batteries of category C1 are screened a second time. Batteries with only 2 EIS inflection points and less than 4 DRT peaks are selected for group application, and the remaining batteries are re-entered into the screening process along with the next batch of retired batteries.

[0089] 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.

[0090] 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 rapid sorting method for decommissioned power batteries based on interval frequency electrochemical impedance spectroscopy, characterized in that, Includes the following steps: Step 1: Initial selection of battery appearance: Collect appearance images of M0 batteries of the same type and input them into an image recognition model for defect detection. Select M1 batteries with intact appearance from the detection results. Step 2: Battery voltage stabilization: Under constant temperature conditions, M1 intact batteries are discharged to their lower voltage limit U using a charging and discharging device. d After standing for a period of time, the terminal voltage {U} of each battery was measured. 0,m |m=1,2,…,M1} and used as sorting features, where U 0,m This represents the terminal voltage of the m-th battery with intact appearance; Step 3: Obtain the electrochemical impedance spectroscopy (EIS) spectrum of the battery. Step 3.1: Obtain the electrochemical impedance spectroscopy of new batteries of the same type, and use it to determine the frequency range of the impedance spectrum frequencies to be measured for M1 batteries. The frequency range is the imaginary part value - Z. im The frequency range from ≥0 to the last trough of the Nyquist curve; Step 3.2: Under constant temperature environment, place the M1 batteries after they have been left to stand into the impedance measuring device, and apply small-amplitude sinusoidal voltage excitation signals of different frequencies in the frequency range to the m-th battery respectively, measure the corresponding different response current signals of the m-th battery, and then perform FTT transformation on the small-amplitude sinusoidal voltage excitation signal of the same frequency and its corresponding response current signal to calculate the impedance value of the m-th battery at the same frequency, thereby obtaining the impedance value of the m-th battery at different frequencies. Step 4: Extract features based on impedance information: Step 4.1: Using the real part of the impedance value of the m-th battery at different frequencies as the independent variable and the negative of the imaginary part as the dependent variable, fit the 5th order polynomial function of the real and imaginary parts of the m-th battery. Plot the Nyquist curve of the m-th battery based on the 5th order polynomial function, and extract the classification features related to the performance and aging degree of the m-th battery from the Nyquist curve to obtain the classification features of all M1 batteries. Step 4.2: Convert the frequency domain information of the impedance value of the m-th battery at different frequencies into time domain features using the DRT method, and plot the DRT curve of the m-th battery. Then extract the classification features related to the performance and aging degree of the m-th battery from the DRT curve to obtain the classification features of all M1 batteries. Step 4.3: Use the mutual information method based on critical weights to remove redundant features from all classification features of M1 batteries, and obtain the remaining classification features as sorting features; Step 5: Two-step battery sorting: Step 5.1: After normalizing the sorting features obtained in Step 2 and Step 4, the M1 batteries are processed using the Bayesian-Gaussian mixture clustering algorithm based on the normalized sorting features to obtain several battery sets, and the battery set C1 with the best health status is selected. Step 5.2: Based on the number of inflection points in the electrochemical impedance spectroscopy and the number of peaks in the DRT curve, a second screening is performed on the battery set C1. Batteries containing only 2 EIS inflection points and fewer than 4 DRT peaks are selected from C1 as the final selected batteries.

2. The rapid sorting method for decommissioned power batteries based on interval frequency electrochemical impedance spectroscopy according to claim 1, characterized in that: The mutual information method based on critical weights in step 4.3 is performed as follows: Step a: Denote the set of all classification features of all M1 batteries as the initial feature set S0, and denote any i-th classification feature vector of the M1 batteries in S0 as f. i Let f be any j-th classification feature vector of M1 batteries in S0. j ; Step b: Calculate f using equation (3) i and f j Redundancy value R(f) i ,f j ), and determine R(f) i ,f j If f is greater than a given threshold, then f will be set to f. i NS number of filters i and f j NS number of filters j Add 1 to each item; otherwise, proceed to step c. R(f i ,f j )=2I(f i ,f j ) / (H(f i )+H(f i ,f j )) (3) In equation (3), I(f) i ,f j ) represents f i and f j The mutual information value, where H represents entropy; Step c: Calculate f using equation (4) i Objective weight WR i , if WR i >WR j Then NS i Add 1, if WR i <WR j Then NS j Add 1, if WR i =WR j Then proceed directly to step d, where WR j f j Objective weights; In equation (4), W i f i The weight, K i f i The characteristic variation index is calculated by equation (5); R i f i The characteristic conflict index is calculated by equation (6); In equations (5) and (6), f im f is the value of the i-th classification feature in the m-th battery. tm Let |S0| be the value of the t-th classification feature in the m-th battery, and |S0| be the number of classification features in the initial feature set S0. Let be the mean of the i-th classification feature vector of M1 batteries. Let r be the mean of the t-th classification feature vector of M1 batteries. it f is the i-th classification feature vector of M1 batteries i With the t-th classification feature vector f t Similarity; Step d: Determine whether pairwise comparisons have been completed among all classification features in S0. If so, sort them in ascending order according to the number of times each classification feature has been filtered, and select the top few classification features as the remaining classification features after redundancy removal to form the sorting feature set. Otherwise, return to step b and execute sequentially.

3. An electronic device, comprising a memory and a processor, characterized in that, The memory is used to store a program that supports the processor in executing the rapid sorting method for decommissioned power batteries according to claim 1 or 2, and the processor is configured to execute the program stored in the memory.

4. A computer-readable storage medium storing a computer program thereon, characterized in that, The computer program, when run by the processor, executes the steps of the rapid sorting method for retired power batteries as described in claim 1 or 2.

Images