Retired power battery rapid sorting method and sorting system based on intelligent decision

Abstract

The invention discloses an intelligent decision-making fast sorting method and sorting system for retired power batteries, and relates to the technical field of battery recycling. The method comprises the following specific steps: measuring the open-circuit voltage of a battery monomer, dynamically adjusting a screening threshold according to a temperature compensation algorithm, comparing the open-circuit voltage value with a preset range subjected to temperature correction, and screening out a first batch of candidate battery monomers; applying a multi-order pulse test working condition to the first batch of candidate battery monomers to obtain dynamic response parameters, and screening out a second batch of candidate battery monomers according to the dynamic response parameters; carrying out extended frequency band impedance spectrum test on the second batch of candidate battery monomers to obtain an impedance characteristic matrix; and inputting the open-circuit voltage, the dynamic response parameter and the impedance characteristic matrix into a fuzzy neural network decision model, outputting a sorting decision result, and finally grading the single batteries. According to the method, the consistency and the safety of the cascade utilization battery pack are remarkably improved, and meanwhile, the efficiency and the sorting precision are improved on the basis of reducing the sorting cost.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of battery recycling, in particular to a quick sorting method and system for retired power battery based on intelligent decision-making. BACKGROUND

[0002] In the field of retired power battery sorting, the traditional technology mainly experienced two development stages of manual screening and single-parameter automatic detection, both of which have significant limitations. The early way of relying on manual experience for appearance inspection and voltage measurement is not only inefficient, but also has a high misjudgment rate and serious safety hazards. With the advancement of technology, although automatic devices such as internal resistance tester and static capacity testing equipment have appeared, there are still problems such as long testing period, large equipment investment, and single parameter. Although the current mainstream multi-parameter sorting technology integrates voltage, capacity and impedance testing, it still faces the problem of data fragmentation due to independent detection in stages, and generally ignores dynamic working condition simulation and temperature compensation, resulting in defects such as group difference > 15%, missing risk warning of thermal runaway. The publication number CN111580005B proposes a quick sorting scheme, but this technology shortens the testing time, but lacks practical working condition simulation capability, and the grading results have significant deviations from the actual performance of the battery. The industry is facing core pain points such as the difficulty of balancing testing accuracy and efficiency, large differences in the life of grouped batteries, and the difficulty of identifying safety hazards in advance, and there is an urgent need to develop new sorting technology that is fast, multi-dimensional and intelligent. SUMMARY

[0003] The purpose of the present application is to provide a quick sorting method and system for retired power battery based on intelligent decision-making to solve the problems raised in the background technology, significantly improve the consistency and safety of the graded battery group, and improve the efficiency and sorting accuracy on the basis of reducing the sorting cost.

[0004] To achieve the above-mentioned purpose, the present application provides the following solutions: on the one hand, a quick sorting method for retired power battery based on intelligent decision-making is provided, and the specific steps include the following: Measure the open-circuit voltage of the battery monomer and dynamically adjust the screening threshold according to the temperature compensation algorithm, compare the open-circuit voltage value with the preset range after temperature correction, and screen out the first batch of candidate battery monomers; Apply a multi-stage pulse test working condition to the first batch of candidate battery monomers, obtain dynamic response parameters, and screen out the second batch of candidate battery monomers according to the dynamic response parameters; Perform extended frequency range impedance spectrum test on the second batch of candidate battery monomers to obtain an impedance characteristic matrix; Input the open-circuit voltage, dynamic response parameters and impedance characteristic matrix into a fuzzy neural network decision-making model to output a sorting decision result; According to the sorting decision result, the battery monomer is finally graded.

[0005] Preferably, the multi-stage pulse test working condition comprises constant current charging, pulse discharging and standing recovery stage, and the voltage change characteristics and thermal behavior parameters in the charging and discharging process are recorded; the dynamic response parameters include voltage change rate, relaxation time constant and irreversible heat.

[0006] Preferably, the acquired dynamic response parameters are compared with the set threshold value, and the second batch of candidate battery monomers is screened out, and the battery monomers that do not meet the standard directly enter the disassembly and recycling process.

[0007] Preferably, the formula of the temperature compensation algorithm is: ; Wherein, T is the ambient temperature.

[0008] Preferably, the extended frequency range impedance spectrum test adopts a discrete frequency point measurement method, and the test frequency range covers 0.01Hz to 10kHz, specifically including: measuring high-frequency ohmic impedance at 10kHz, 2kHz and 500Hz frequency points ; Measuring medium-frequency charge transfer impedance at 100Hz, 50Hz and 10Hz frequency points ; Measuring low-frequency diffusion impedance at 1Hz, 0.1Hz and 0.01Hz frequency points .

[0009] Preferably, the fuzzy neural network decision model comprises a preprocessing layer, a fuzzy inference layer and a BP neural network optimization layer; the preprocessing layer is used for normalizing the open circuit voltage, the dynamic response parameters and the impedance characteristic matrix; the fuzzy inference layer is used for fuzzy processing the preprocessed multi-source detection data through a Gaussian membership function to obtain fuzzy output; the BP neural network optimization layer is used for receiving the fuzzy output, and after transformation through an activation function, the sorting decision result is generated by applying a Softmax function and a Sigmoid function in the output layer.

[0010] Preferably, the preprocessing layer adopts Min-Max normalization processing, and the formula is: ; Wherein, represents the minimum value of a specific detection parameter in the data set, represents the maximum value of the same specific detection parameter in the data set.

[0011] In another aspect, a retired power battery quick sorting system for intelligent decision-making is provided, comprising a voltage detection module, a dynamic working condition simulation power supply module, an extended frequency band impedance analysis module, a multi-parameter fusion decision-making module, and an automatic sorting execution module. The voltage detection module is configured to measure the open-circuit voltage of the battery monomer and dynamically adjust the screening threshold according to a temperature compensation algorithm, compare the open-circuit voltage value with the preset range after temperature correction, and screen out the first batch of candidate battery monomers. The dynamic working condition simulation power supply module is configured to apply a multi-stage pulse test working condition to the first batch of candidate battery monomers, obtain dynamic response parameters, and screen out the second batch of candidate battery monomers according to the dynamic response parameters. The extended frequency band impedance analysis module is configured to perform extended frequency band impedance spectrum testing on the second batch of candidate battery monomers to obtain an impedance characteristic matrix. The multi-parameter fusion decision-making module is configured to input the open-circuit voltage, the dynamic response parameters, and the impedance characteristic matrix into a fuzzy neural network decision-making model, and output a sorting decision result. The automatic sorting execution module is configured to perform final grading on the battery monomer according to the sorting decision result.

[0012] According to the specific embodiments of the present application, the following technical effects are disclosed: (1) The static characteristics, dynamic response, and thermal behavior are innovatively integrated, multi-dimensional indicators are realized, comprehensive performance evaluation is realized, the evaluation dimension is enriched, the accuracy of the evaluation is improved, the dynamic testing method based on multi-parameter fusion makes the sorting precision reach the group difference ≤5%, which is more than 3 times higher than the prior art, and effectively prolongs the cycle life of the battery pack; (2) The advantages of fuzzy reasoning and BP neural network are combined, the model significantly improves the sorting precision and thermal runaway early warning capability, significantly improves the safety of grade utilization, and supports real-time decision-making; (3) Through the innovative sorting system, the comprehensive performance evaluation of the single battery can be completed within 15 minutes, the sorting period of the traditional 3-7 days is shortened by more than 98%, and the economic feasibility of grade utilization is greatly improved. BRIEF DESCRIPTION OF DRAWINGS

[0013] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0014] Figure 1 The method flowchart of the present application; Figure 2 A multi-stage pulse test working condition design diagram for the application; Figure 3 An impedance characteristic matrix construction schematic diagram for the application; Figure 4 A fuzzy neural network model structure diagram for the application; Figure 5 A system structure diagram for the application. DETAILED DESCRIPTION

[0015] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the application.

[0016] The purpose of the application is to provide a smart decision-making method for quickly sorting retired power batteries, as shown in Figure 1 The specific steps include the following: S1, measure the open-circuit voltage of the battery monomer and dynamically adjust the screening threshold according to the temperature compensation algorithm, compare the open-circuit voltage value with the preset range corrected by temperature, and screen out the first batch of candidate battery monomers; S2, apply a multi-stage pulse test working condition to the first batch of candidate battery monomers, obtain dynamic response parameters, and screen out the second batch of candidate battery monomers according to the dynamic response parameters; S3, perform an extended frequency range impedance spectrum test on the second batch of candidate battery monomers to obtain an impedance characteristic matrix; S4, input the open-circuit voltage, dynamic response parameters and impedance characteristic matrix into a fuzzy neural network decision-making model, and output a sorting decision result; S5, perform final grading on the battery monomer according to the sorting decision result.

[0017] Further, in S1, the battery surface temperature is monitored in real time by an infrared thermal imager, and the screening threshold of the open-circuit voltage is dynamically adjusted according to the temperature compensation algorithm. The specific formula is: Wherein, T is the ambient temperature. Compare the open-circuit voltage value with the dynamically adjusted threshold range to screen out the first batch of candidate battery monomers. For the battery monomer whose open-circuit voltage exceeds the range, it is directly determined as unqualified and does not enter the subsequent test.

[0018] Further, in order to solve the problem of long traditional sorting period, the application completes single battery evaluation within 15 minutes through the following technical solutions:​ Hierarchical screening mechanism: first, unqualified batteries are quickly removed by open-circuit voltage, then secondary screening is combined with dynamic response parameters, and finally impedance spectrum test is performed on candidate batteries, which greatly reduces invalid detection.

[0019] Differentiated test scheme: in view of the different needs of power batteries and energy storage batteries, as a specific implementation and optimization strategy of multi-stage pulse test, pulse test method (fast evaluation) and multi-point test method (high-precision evaluation) are adopted, which takes into account the efficiency and consistency requirements.

[0020] Safety protection strategy: real-time monitoring of voltage drop (>10%) or abnormal temperature rise (>1℃ / s), automatic termination of test and marking of risk batteries to ensure the safety of the sorting process.

[0021] Specifically, in view of the different characteristics of power batteries and energy storage batteries, differentiated direct current internal resistance test schemes are adopted. For power battery sorting, the standard pulse test method applies a 10A constant current discharge pulse for 3 seconds, and then applies a 10A charging pulse in the opposite direction for 3 seconds, and the whole process is completed within 8 to 12 seconds. The test system collects voltage change data during the pulse period in real time, and takes the average value of ΔV / ΔI in the discharge and charge phases as the final internal resistance value. This method is suitable for rapid sorting in production lines.

[0022] For energy storage batteries that pay more attention to consistency, high-precision multi-point test method is used for more detailed evaluation. This scheme applies 5A, 10A and 15A multi-current step loading, each current step maintains 5 seconds to make the battery reach steady state, the system records the steady-state voltage value at each current, and then uses least squares method for linear regression analysis to calculate the internal resistance. The whole test process takes 20 to 30 seconds. This method effectively separates the ohmic polarization and concentration polarization by measuring multiple current points, and improves the test accuracy to ±0.5%, which is especially suitable for evaluating energy storage batteries with strict group consistency requirements.

[0023] Both test methods integrate safety protection strategies, which immediately terminate the test when voltage drop exceeds 10% or temperature rise rate is greater than 1℃ / s. Test data is automatically uploaded to the MES system, which, together with open-circuit voltage, impedance spectrum and other parameters, forms the basic data set for battery health state evaluation. Through comparison and verification, the pulse test method for power batteries and the multi-point test method for energy storage batteries can achieve the best balance between test efficiency and accuracy in their respective application scenarios. The system automatically switches the test mode according to the battery type identification without manual intervention, while retaining a manual review interface for special verification in disputed cases.

[0024] Further, in S2, a multi-stage pulse test condition is applied to the first batch of candidate battery monomers screened out, which includes constant current charging, pulse discharging and standing recovery stages, such asFigure 2 as shown, specifically comprising the following steps: S21, constant current charging at 0.5C rate for 120 seconds, recording the charging starting voltage and ending voltage , calculating the voltage change value: ; S22, after charging, rest for 60 seconds, record the voltage recovery curve, extract the relaxation time constant τ; S23, apply a test sequence containing three-1C pulse discharges, each pulse lasts for 5 seconds, interval 10 seconds rest period, record the voltage response and temperature rise rate during discharging , and calculate the irreversible heat ; S24, according to the dynamic response parameters ( , τ, , ), screening out the second batch of candidate battery monomers.

[0025] In the process of rapid sorting of retired power batteries, the temperature rise rate ( ) is one of the key parameters for evaluating the thermal safety and state of health of the battery. Under normal circumstances, the temperature rise rate of lithium ion battery in pulse discharge test should be controlled at a low level, which is specifically manifested as the temperature rise rate is usually not more than 0.5℃ / s under the working condition of-1C pulse discharge for 5 seconds. When the temperature rise rate exceeds 0.8℃ / s, the system will determine it as an abnormality and trigger the early warning mechanism. This threshold is set based on the characteristics of battery materials and actual test data. Excessive temperature rise rate is often closely related to internal short circuit, separator defect or electrolyte decomposition and other safety hazards in the battery. In order to further improve the accuracy of judgment, the system will dynamically adjust the threshold in combination with the environmental temperature and the state of charge (SOC) of the battery. For example, in high temperature environment (>40℃) or high SOC (>80%) conditions, the safety threshold of temperature rise rate will be adjusted down to 0.6℃ / s or 0.7℃ / s accordingly, in order to avoid false judgment caused by extreme conditions. In addition, the system will also compare the current temperature rise rate with the average value of the same batch of batteries. If it exceeds the average value by more than 50%, even if it does not reach the absolute threshold, it will also be marked as a potential risk battery and enter the reinspection process. Through this multi-dimensional and dynamic judgment method, the system can ensure the sorting efficiency while effectively identifying the batteries with thermal runaway risk, ensuring the safety of grade utilization. The relaxation time constant (τ) is an important indicator for evaluating battery polarization characteristics, reflecting the speed of voltage recovery after charging and discharging. During the sorting process, after the batteries undergo constant current charging and resting recovery stages, the system records the voltage recovery curve and extracts the relaxation time constant. Normal batteries typically have a short τ value, indicating a rapid internal polarization reaction and quick recovery to steady-state voltage. If the τ value is abnormally prolonged (e.g., exceeding 120 seconds), it indicates severe polarization, possibly due to electrode material aging, decreased electrolytic material performance, or increased interfacial impedance. Such batteries are prone to accelerated capacity decay or performance instability during subsequent use and therefore require close monitoring. The system dynamically sets a judgment threshold based on the distribution range of τ values ​​and historical data. For example, if the average τ value of a batch of batteries is 60 seconds, but the τ value of individual batteries exceeds 90 seconds, it will be judged as abnormal. Simultaneously, the system also combines temperature rise rate and impedance data for comprehensive analysis to eliminate the influence of test conditions or external interference on the τ value. By accurately determining the relaxation time constant, the system can effectively identify severely polarized batteries, preventing them from entering demanding application scenarios, thereby improving the overall performance and lifespan of battery packs used in cascade applications. Irreversible heat ( The temperature rise rate (TFR) is a key parameter for measuring battery energy loss and thermal behavior, directly reflecting the efficiency loss during charging and discharging. In the sorting test, the system calculates the irreversible heat generated by the battery by recording the temperature rise curve and current-voltage data during the pulse discharge phase. Normal batteries have a low TFR, typically less than 5 J / g, indicating high energy conversion efficiency and low internal loss. If the TFR exceeds 10 J / g, it indicates significant irreversible reactions inside the battery, such as increased side reactions, increased internal resistance, or localized overheating. These batteries are not only inefficient but also have a high risk of thermal runaway, and are therefore directly classified as Grade C, entering the dismantling and recycling process. To improve the accuracy of the judgment, the system cross-validates the TFR by combining temperature rise rate and impedance characteristics. For example, if a battery has a high TFR accompanied by an excessive temperature rise rate or a sudden increase in impedance in the high-frequency region, it can further confirm a serious internal defect. In addition, the system dynamically adjusts the judgment threshold based on historical battery data and batch characteristics to adapt to batteries with different aging levels and material systems. Through precise analysis of irreversible heat, the system can effectively identify batteries with low energy efficiency and high safety risks, providing reliable data support for secondary utilization.

[0026] Furthermore, in S3, extended-band impedance spectroscopy testing was performed on the second batch of candidate battery cells. The extended-band impedance spectroscopy test employed a discrete frequency point measurement method, covering a frequency range from 0.01Hz to 10kHz. Figure 3 As shown, it specifically includes: Measure the ohmic impedance in the high-frequency region at 10kHz, 2kHz, and 500Hz. ; Measure the intermediate frequency area charge transfer impedance in 100Hz, 50Hz, 10Hz frequency point ; Measure the low frequency area diffusion impedance in 1Hz, 0.1Hz, 0.01Hz frequency point .

[0027] The above measurement results are constructed into a 3x3 impedance characteristic matrix, at the same time, the system supports full frequency band scanning mode for supplementary analysis during reinspection.

[0028] Further, in S4, the open circuit voltage, dynamic response parameters and impedance characteristic matrix are input into the fuzzy neural network decision model. The fuzzy neural network model includes 14 input nodes, respectively corresponding to the normalized open circuit voltage, dynamic response parameters (V0, Vmax, Vmin, τ, τ, τ, τ, τ, τ, τ, τ, τ, τ, τ) , τ, , ) and impedance characteristic matrix (9 parameters). After the input data is processed by Min-Max normalization, it is input into the hidden layer for decision, and the three output nodes (respectively corresponding to the health degree score, sorting grade and risk probability) It should be noted that, as shown in Figure 4 , the fuzzy neural network decision model includes a preprocessing layer, a fuzzy reasoning layer and a BP neural network optimization layer; the preprocessing layer is used to normalize the open circuit voltage, dynamic response parameters and impedance characteristic matrix, and the preprocessing layer adopts Min-Max normalization processing, and the formula is: ; Wherein, represents the minimum value of a specific detection parameter in the data set, represents the maximum value of the same specific detection parameter in the data set.

[0029] The preprocessed multi-source detection data is input into the fuzzy reasoning layer, which uses fuzzy logic processing based on expert knowledge to deal with uncertain information. The input features such as open circuit voltage and dynamic response parameters are processed by Gaussian membership function, and each feature is mapped to three linguistic variables "low", "medium" and "high". Taking the voltage parameter as an example, when the detection value is 3.32V, the fuzzy propositions "voltage medium" (membership degree 0.7) and "voltage high" (membership degree 0.3) may be activated at the same time. This partial membership feature effectively solves the rigidity defect of traditional threshold determination.

[0030] The fuzzy inference engine loads a pre-configured rule base, which encodes domain expert experience in the form of "IF-THEN" rules. A typical rule is "IF charge transfer resistance Rct is 'high' and temperature rise ΔT is 'fast', THEN risk level is 'high'". The rule firing strength is calculated using the Mamdani min-max method, i.e., taking the minimum value of the membership degrees of the premise part as the rule weight. The outputs of all activated rules are combined by weighted summation to generate the preliminary decision, forming fuzzy output quantities such as health status, risk level, etc. A rule confidence factor (0.8-1.0) is specially set in this stage to distinguish the influence difference between core rules and auxiliary rules.

[0031] The BP neural network optimization layer receives the structured feature vector from the fuzzy inference layer, with the number of input layer nodes corresponding to the discretized segment number of fuzzy output quantities. The LeakyReLU activation function (a=0.01) is used in the hidden layer to alleviate the gradient vanishing problem, and the He normal distribution is used for weight initialization. During the forward propagation process, the system first calculates the weighted input of the hidden layer, and then applies the Softmax function (classification task) and Sigmoid function (regression task) in the output layer to generate the final prediction. For the battery sorting scene, the output layer of the network is designed as a multi-task structure: the SOH evaluation uses the mean square error loss, the sorting level uses the classification cross-entropy, and the risk prediction uses the Focal Loss to solve the class imbalance.

[0032] The momentum stochastic gradient descent (μ=0.9) is used in the back propagation phase, and the learning rate is dynamically adjusted by the cosine annealing strategy (initial value 0.001). The batch normalization layer is inserted between the fully connected layers to effectively suppress the internal covariate shift. To prevent overfitting, in addition to L2 regularization, Dropout (ratio 0.3) is also implemented after the first hidden layer. The loss function integrates the weighted sum of the three prediction tasks, where the SOH evaluation weight is set to 0.5, and the sorting and risk each account for 0.3 and 0.2, reflecting the importance difference of different output indicators.

[0033] The interaction between fuzzy inference and neural network is realized through feature concatenation: the linguistic variable membership degree output by fuzzy inference (such as "health status good = 0.6") is used as a supplementary feature, together with the original detection data to form the input vector of the neural network. This architecture not only retains the advantages of fuzzy systems in handling uncertainty, but also leverages the feature learning capabilities of neural networks. In the model training phase, a two-stage optimization strategy is adopted: first, fix the fuzzy rule parameters to train the neural network weights, and then fine-tune the membership function parameters through back propagation. In actual deployment, the system running delay is controlled within 15ms, meeting the real-time requirements of the production line.

[0034] The intermediate process results of the retired power battery rapid sorting are provided to technicians for diagnostic analysis through a visual interface, such as a heat map showing the contribution of different rules to the final decision. The system is specially provided with a conflict detection mechanism. When the difference between the neural network output and the fuzzy rule conclusion exceeds a threshold value, an expert review process is automatically triggered to ensure the explainability of the decision-making process.

[0035] This hybrid intelligent architecture has been tested and verified to improve the classification accuracy by 12% compared to pure neural network models in battery sorting scenarios, while maintaining the transparency of the rule system. Its core advantage lies in the fuzzy reasoning layer handling parameter boundary uncertainty and the neural network learning complex nonlinear relationships, complementing each other to form a decision-making system with both expert knowledge and data-driven characteristics.

[0036] Through membership calculation and rule reasoning, the final decision result is output.

[0037] A-level battery: health score ≥ 80%, suitable for electric vehicle backup power.

[0038] B-level battery: health score 60%~80%, suitable for energy storage systems.

[0039] C-level battery: health score < 60% or risk probability > 5%, needs to be disassembled and recycled.

[0040] On the other hand, a retired power battery rapid sorting system with intelligent decision-making is provided, as shown in Figure 5 The system includes a voltage detection module, a dynamic working condition simulation power supply module, an extended frequency range impedance analysis module, a multi-parameter fusion decision-making module, and an automatic sorting execution module. Among them, The voltage detection module is used to measure the open-circuit voltage of the battery monomer and dynamically adjust the screening threshold according to the temperature compensation algorithm, compare the open-circuit voltage value with the temperature-corrected preset range, and screen out the first batch of candidate battery monomers; The dynamic working condition simulation power supply module is used to apply multi-stage pulse test working conditions to the first batch of candidate battery monomers, obtain dynamic response parameters, and screen out the second batch of candidate battery monomers according to the dynamic response parameters; The extended frequency range impedance analysis module is used to perform extended frequency range impedance spectrum testing on the second batch of candidate battery monomers to obtain an impedance feature matrix; The multi-parameter fusion decision-making module is used to input the open-circuit voltage, dynamic response parameters, and impedance feature matrix into a fuzzy neural network decision-making model to output a sorting decision result; The automatic sorting execution module is used to perform final grading on the battery monomers according to the sorting decision result.

[0041] The principles and implementation manners of the present application are described by using specific examples in the present application, and the above examples are only used to help understand the method of the present application and its core idea; meanwhile, for the general technical personnel in the art, the specific implementation manners and application ranges will be changed according to the idea of the present application. In conclusion, the content of the present specification should not be understood as the limitation of the present application.

Claims

1. A method for rapid sorting of decommissioned power batteries using intelligent decision-making, characterized in that, The specific steps include the following: The open-circuit voltage of a single battery cell is measured and the screening threshold is dynamically adjusted according to the temperature compensation algorithm. The open-circuit voltage value is compared with the preset range after temperature correction to select the first batch of candidate battery cells. Multi-stage pulse test conditions were applied to the first batch of candidate battery cells to obtain dynamic response parameters; A second batch of candidate battery cells was selected based on the dynamic response parameters. The impedance characteristic matrix was obtained by performing extended frequency band impedance spectroscopy on the second batch of candidate battery cells. The open-circuit voltage, the dynamic response parameters, and the impedance characteristic matrix are input into the fuzzy neural network decision model, and the sorting decision results are output. The battery cells are then classified in the final stage based on the sorting decision results.

2. The intelligent decision-making method for rapid sorting of decommissioned power batteries according to claim 1, characterized in that, The multi-stage pulse test condition includes constant current charging, pulse discharging, and static recovery stages, recording the voltage change characteristics and thermal behavior parameters during the charging and discharging process; the dynamic response parameters include voltage change rate, relaxation time constant, and irreversible heat.

3. The intelligent decision-making method for rapid sorting of decommissioned power batteries according to claim 2, characterized in that, The obtained dynamic response parameters are compared with the set threshold to select the second batch of candidate battery cells. Battery cells that do not meet the standards are directly put into the dismantling and recycling process.

4. The intelligent decision-making method for rapid sorting of decommissioned power batteries according to claim 1, characterized in that, The formula for the temperature compensation algorithm is: ； in, T The ambient temperature.

5. The intelligent decision-making method for rapid sorting of decommissioned power batteries according to claim 1, characterized in that, The extended frequency band impedance spectrum test employs a discrete frequency point measurement method, covering a frequency range from 0.01Hz to 10kHz. Specifically, it includes measuring the ohmic impedance in the high-frequency region at 10kHz, 2kHz, and 500Hz. Measure the charge transfer impedance in the mid-frequency region at 100Hz, 50Hz, and 10Hz. The diffusion impedance in the low-frequency region was measured at 1 Hz, 0.1 Hz, and 0.01 Hz. .

6. The intelligent decision-making method for rapid sorting of decommissioned power batteries according to claim 1, characterized in that, The fuzzy neural network decision model includes a preprocessing layer, a fuzzy inference layer, and a BP neural network optimization layer. The preprocessing layer is used to normalize the open-circuit voltage, the dynamic response parameters, and the impedance feature matrix. The fuzzy inference layer is used to fuzzify the preprocessed multi-source detection data using a Gaussian membership function to obtain a fuzzy output. The BP neural network optimization layer receives the fuzzy output, transforms it using an activation function, and then applies the Softmax and Sigmoid functions to the output layer to generate the sorting decision result.

7. The intelligent decision-making method for rapid sorting of decommissioned power batteries according to claim 6, characterized in that, The preprocessing layer uses Min-Max normalization, with the following formula: ； in, This represents the minimum value of a specific detection parameter in the dataset. This represents the maximum value of the same specific detection parameter in the dataset.

8. A rapid sorting system for retired power batteries based on intelligent decision-making, characterized in that, It includes a voltage detection module, a dynamic operating condition simulation power supply module, an extended frequency band impedance analysis module, a multi-parameter fusion decision module, and an automatic sorting execution module; among which, The voltage detection module is used to measure the open-circuit voltage of a battery cell and dynamically adjust the screening threshold according to the temperature compensation algorithm. The open-circuit voltage value is compared with the preset range after temperature correction to screen out the first batch of candidate battery cells. The dynamic operating condition simulation power supply module is used to apply multi-stage pulse test conditions to the first batch of candidate battery cells, obtain dynamic response parameters, and select the second batch of candidate battery cells based on the dynamic response parameters. The extended frequency band impedance analysis module is used to perform extended frequency band impedance spectrum testing on the second batch of candidate battery cells to obtain the impedance characteristic matrix. The multi-parameter fusion decision module is used to input the open-circuit voltage, the dynamic response parameters, and the impedance feature matrix into the fuzzy neural network decision model and output the sorting decision result. The automatic sorting execution module is used to perform final grading of battery cells based on the sorting decision results.

Images