Method and system for evaluating repair state of failure repair lithium battery based on dynamic parameters

Abstract

The invention provides a method for evaluating the repair state of a failure repair lithium battery based on dynamic parameters. The method comprises the following steps: S1, collecting historical battery operation data, and constructing a training sample set by taking capacity attenuation information as a state monitoring label; s2, taking the training sample set as input, and constructing an initial double-exponential function repair state model based on fitting parameters; s3, based on a Monte Carlo parameter perturbation strategy, carrying out self-adaptive iteration updating of the double-exponential function repair state model; and S4, obtaining a repair state predicted value of the corresponding lithium battery by using the double-exponential function repair state model after the self-adaptive iteration update, and calculating the expected service life of the target lithium battery. The invention further provides a system for evaluating the repair state of the failure repair lithium battery based on the dynamic parameters, real-time, accurate and dynamic evaluation of the repair state and the expected service life of the target failure repair lithium battery is achieved, and the problem that error accumulation and parameter drift are prone to occurring during long-term prediction of a traditional physical model is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of lithium battery operating status evaluation, and in particular to a method and system for evaluating the repair status of a lithium battery after failure repair. Background Art

[0002] Lithium batteries, known for their high energy density, long cycle life, lack of memory effect, and environmentally friendly energy storage devices, have recently gained widespread application in electric vehicles, portable electronic devices, energy storage systems, and other fields. However, as lithium batteries cycle continuously in real-world use, their internal chemical reactions and physical changes inevitably lead to capacity fading, increased internal resistance, and other performance degradation issues, impacting the battery's overall performance and safety. The limited lifespan of batteries and the wave of battery retirements present dual challenges. Efficiently recycling spent batteries while minimizing environmental impact is a key component of achieving energy transition and sustainable development. Failed battery reconditioning involves structurally regenerating lithium-ion batteries with degraded electrochemical performance rather than disassembling them for recycling. This technology directly restores electrode material activity by precisely adding lithium to fill lattice vacancies, eliminating surface rock salt phases and lattice defects, and reconstructing lithium-ion transport pathways. Unlike traditional pyrometallurgical or hydrometallurgical recycling methods, this direct reconditioning process preserves the material's underlying structure, avoiding the energy-intensive decomposition and reconstruction steps, significantly reducing energy consumption and pollution. Key technological breakthroughs include salt-depleted solid-state sintering, the conversion of polycrystalline materials to single crystals, and the repair of grain boundary cracks. The repaired battery capacity recovery rate exceeds 95%, significantly extending its cycle life while reducing processing costs by over 70%, achieving a deep synergy between resource recycling and economic benefits. However, accurately and in real time assessing the state of repair (SOR) and predicting the expected service life (RUL) of lithium batteries have become key technical challenges that need to be addressed in the Battery Management System (BMS).

[0003] Currently, methods for predicting the operating status of lithium batteries primarily include physical model-based approaches and data-driven approaches. Physical model-based approaches simulate the internal reactions of the battery by establishing an equivalent circuit model (ECM) or electrochemical model to calculate the battery's operating status. However, traditional physical models rely on fixed parameters to describe the internal battery state. Over long-term use, battery aging leads to nonlinear parameter changes (such as SEI film thickening and active material loss), which fixed parameter models cannot track in real time. Due to the nonlinear characteristics and complex degradation mechanisms within the battery, traditional physical models are prone to error accumulation and parameter drift in long-term predictions. On the other hand, data-driven approaches, such as linear regression, support vector machines (SVM), and neural networks, use historical data to train models to predict the battery's operating status. Although these methods can capture nonlinear characteristics in battery data, their prediction accuracy still needs to be improved in the face of long-term dependencies and data noise. This is particularly true for evaluating the repair status of lithium batteries repaired with failed repair materials, where traditional methods are often difficult to apply.

[0004] Patent publication number CN119355561A discloses a lithium battery health monitoring and estimation method and system. This system extracts charge and discharge cycle data from lithium batteries during aging experiments, assigns state monitoring labels, performs time discretization and spatial embedding, and calculates the battery degradation coefficient. This method addresses the inaccuracy of lithium battery health estimation in existing technologies. However, this method relies on a machine learning model with fixed parameters and lacks an online update mechanism. This can lead to parameter drift over time due to battery aging. Summary of the Invention

[0005] In response to the technical problems that traditional physical models are prone to error accumulation and parameter drift, as well as limited prediction accuracy during long-term prediction, the present invention proposes a method and system for evaluating the repair status of failed repaired lithium batteries based on dynamic parameters. Through the four core technologies of multi-dimensional feature collaborative analysis, adaptive modeling, dynamic optimization mechanism and post-failure repair evaluation, real-time and accurate dynamic evaluation of the repair status (SOR) and expected service life (RUL) of the target failed repaired lithium batteries is achieved, solving the problems of error accumulation and parameter drift that traditional physical models are prone to during long-term prediction.

[0006] In order to achieve the above object, the technical solution of the present invention is achieved as follows:

[0007] A method for evaluating the repair status of a failed repaired lithium battery based on dynamic parameters, comprising the following steps:

[0008] S1: Collect historical battery operation data and use capacity decay information as the condition monitoring label to construct a training sample set;

[0009] S2: Taking the training sample set as input, an initial double exponential function repair state model based on the fitting parameters is constructed;

[0010] S3: Adaptive iterative update of the double exponential function repair state model based on the Monte Carlo parameter perturbation strategy;

[0011] S4: Apply the adaptive iteratively updated double exponential function repair state model to obtain the repair state prediction value of the target lithium battery and calculate the expected service life of the target lithium battery.

[0012] Furthermore, the method for collecting historical battery operation data in step S1 is as follows: perform a charge-discharge cycle test, record the charge-discharge cycle data each time a charge-discharge cycle is completed, and record the cycle data including the discharge voltage V under multiple charge-discharge cycle tests. dis,N , dynamic resistance R N The cumulative number of cycles N is sorted according to the time series and discretized.

[0013] Furthermore, the discretization processing method is: dividing the continuous charge and discharge cycle data into several discrete time periods, and statistically analyzing the corresponding capacity decay information in each time period as state monitoring label data to obtain a training sample set; the capacity decay information is the ratio of the actual capacity to the rated capacity of the commercial lithium battery.

[0014] Furthermore, the double exponential function repair state model is:

[0015]

[0016] Among them, S N Indicates the repair status value, V dis,N Indicates the discharge voltage of the lithium battery in the Nth cycle, R N represents the dynamic resistance of the lithium battery after the Nth cycle, and β1, β2, β3, and β4 are the model parameters to be fitted.

[0017] Furthermore, the method for constructing an initial double exponential function repair state model based on fitting parameters is: using the first M charge and discharge cycle data in the training sample set as fitting data, using the least squares method to fit the initial parameters of the double exponential function repair state model, and applying the fitted initial parameters to the double exponential function repair state model to obtain the initial double exponential function repair state model.

[0018] Furthermore, the adaptive iterative update method is:

[0019] S3.1: Take the data after the first M charge-discharge cycles in the training sample set as input, and trigger the update with the number of charge-discharge cycles K as the update period;

[0020] S3.2: Perform Monte Carlo parameter perturbation to update model parameters based on the current double exponential function repair state model parameters;

[0021] S3.3: Based on the model parameters under Monte Carlo parameter perturbation, perform sensitivity analysis to adjust the update period and Monte Carlo parameter perturbation amplitude of the next update;

[0022] S3.4: Repeat S3.2 to S3.3 until the training is completed using the charge and discharge cycle data to obtain a trained double exponential function repair state model.

[0023] Furthermore, the trigger update method described in step S3.1 is: after every K charge and discharge cycles, collect the discharge voltage V of the latest cycle dis,N , dynamic resistance R N and condition monitoring labels, calculating the mean square error between the prediction results of the current double exponential function repair state model and the condition monitoring labels;

[0024] The implementation method of step S3.2 is: take the current model parameter β i As the mean, several groups of normally distributed samples are randomly generated according to the disturbance amplitude of no more than 5%. Each set of disturbance parameters is applied to the double exponential function repair state model, and the discharge voltage V dis,N , dynamic resistance R N The proposed method uses the state monitoring labels as input data and calculates the mean square error between the prediction results of the double exponential function repair state model and the state monitoring labels for each set of disturbance parameters. The disturbance parameter combinations with the top fixed percentage of the mean square error are retained and averaged to obtain the new model coefficients.

[0025] The implementation method of step S3.3 is: calculate the gradient contribution of each model parameter to the mean square error under the Monte Carlo parameter perturbation, and calculate the sensitivity of each model parameter according to the sensitivity calculation formula, and convert the model parameter sensitivity S i The perturbation amplitude of the model parameters exceeding the sensitivity threshold is doubled, and the next update period is shortened to half the number of charge and discharge cycles.

[0026] Furthermore, the sensitivity calculation formula is:

[0027]

[0028] Where K is the number of samples of the Monte Carlo parameter perturbation normal distribution, and k is the index of the sample of the Monte Carlo parameter perturbation normal distribution;

[0029] In step S4, the repair state prediction value S of the corresponding lithium battery is obtained by using the trained double exponential function repair state model. current After that, key characteristic data of the repaired lithium battery is collected, including capacity recovery rate C rec and the internal resistance change rate ΔR, the repair state prediction value S current Compared with the measured capacity recovery rate C rec Perform residual verification ∈=|S current -C rec / 100|, if the residual is greater than the validation threshold for multiple consecutive times, the Monte Carlo parameter re-update is triggered to suppress the prediction drift until the residual is less than or equal to the validation threshold.

[0030] Furthermore, after the residual verification is completed, the expected service life is calculated based on the cumulative number of cycles of the corresponding charge and discharge cycles and the predicted value of the repair status of the corresponding charge and discharge cycles:

[0031]

[0032] Among them, N current is the cumulative number of charge and discharge cycles;

[0033] The expected service life is corrected using the internal resistance change rate ΔR. When |ΔR|>15%, the life prediction value is discounted according to the discrete reduction coefficient table, and the optimized remaining service life L is finally output. final :

[0034] L final =L×f(ΔR);

[0035] Where, f(ΔR) is the reduction factor;

[0036] In step S4, the repair state prediction value S corresponding to the charge and discharge cycle is used. current The repaired lithium batteries are divided into different repair levels:

[0037] Level Ⅰ: S current ≥80%, suitable for electric vehicle power systems;

[0038] Level II: 60% ≤ S current ≤80%, suitable for energy storage systems or medium power scenarios;

[0039] Grade III: S current <60, suitable for low-speed electric vehicles or backup power supplies.

[0040] A system for evaluating the repair status of failed repaired lithium batteries based on dynamic parameters.

[0041] It includes a data acquisition and feature engineering module, a double exponential model initialization module, a Monte Carlo parameter optimization module, and a repair status estimation and remaining life prediction module connected in sequence;

[0042] Data acquisition and feature engineering module: used to collect historical battery operation data and construct training sample sets using capacity decay information as status monitoring labels;

[0043] Double exponential model initialization module: takes the training sample set as input and constructs the initial double exponential function repair state model based on the fitting parameters;

[0044] Monte Carlo parameter optimization module: used for adaptive iterative update of the double exponential function repair state model based on the Monte Carlo parameter perturbation strategy;

[0045] Repair status estimation and remaining life prediction module: Apply the adaptive iteratively updated double exponential function repair status model to obtain the repair status prediction value of the target lithium battery and calculate the expected service life of the target lithium battery.

[0046] Beneficial effects of the present invention:

[0047] Through the synchronous collection and fusion of multi-dimensional data, a comprehensive quantitative description of the lithium battery degradation process is achieved; the repair status model constructed with a double exponential function fully reflects the synergistic degradation effect between the discharge voltage and dynamic resistance of the battery during the degradation process, thereby improving the accuracy of the model prediction; the introduction of Monte Carlo parameter perturbation strategy and parameter sensitivity analysis realizes real-time adaptive updating of model parameters, so that the model can always capture the latest degradation dynamics of the battery and solve the problems of error accumulation and parameter drift; for the dedicated evaluation process designed for repaired batteries, through the collection of key parameters and the prediction of expected service life, it can accurately quantify the remaining use value and provide reliable technical support for safety management of repaired batteries. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0049] Figure 1 Flow chart of the method of the present invention.

[0050] Figure 2 Schematic diagram of the system of the present invention. DETAILED DESCRIPTION

[0051] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without creative work are within the scope of protection of the present invention.

[0052] A method for evaluating the repair status of a failed repaired lithium battery based on dynamic parameters, such as Figure 1 As shown, the steps include:

[0053] S1: Collect historical battery operation data and construct a training sample set using capacity decay information as the status monitoring label.

[0054] To ensure the accuracy of the data, the present invention conducts a systematic aging cycle experiment on brand-new commercial lithium batteries provided by the manufacturer, and uses the rated capacity of the commercial lithium battery as the benchmark capacity for calculating the capacity decay information during the aging cycle experiment. At the same time, the rated capacity of the commercial lithium battery is also used as the benchmark capacity for the capacity decay information of the battery after failure repair. Subsequently, the battery after the aging cycle experiment is repaired and then the repaired lithium battery is subjected to charge and discharge cycles to simulate the charge and discharge cycle process of the repaired lithium battery in actual battery use for evaluation in the subsequent step S4.

[0055] Furthermore, a charge and discharge test procedure was constructed in accordance with GB / T31486 and ISO 12405-4 standards. The charge and discharge test procedure strictly follows GB / T31486 and ISO12405-4 standards, requiring the test temperature to be maintained in the room temperature range of 25℃±2℃, setting the 1 / 3C current specification according to the rated capacity, and executing a cycle process of constant current charging to the cut-off voltage after standing for 10 minutes, and constant current discharging to the cut-off voltage after standing for another 10 minutes. A charge and discharge cycle test was performed 1000 times, and the charge and discharge cycle data was recorded each time a charge and discharge cycle was completed. During each charge and discharge cycle, constant current charging to 4.4V and constant current discharging to 2.7V were used as standard test conditions. The voltage and current were recorded at a high sampling rate of 100Hz-1000Hz. In this embodiment, multi-channel synchronous data acquisition was performed at 500Hz to obtain the discharge voltage curve of the lithium battery during the discharge process of each charge and discharge cycle. At the end of the discharge process, a small disturbance signal was applied to obtain the dynamic resistance data of the lithium battery in each charge and discharge cycle, and the actual charge and discharge capacity of the lithium battery in each charge and discharge cycle was calculated using the coulomb counting method. At the same time, the cumulative number of cycles in each charge and discharge cycle was recorded.

[0056] Furthermore, the ratio of the actual capacity of each charge and discharge cycle to the rated capacity of a commercial lithium battery is used to represent capacity decay information and serve as a status monitoring label. That is, the capacity decay information is calculated using the rated capacity of a commercial lithium battery as the baseline capacity. For example, if the actual capacity of a lithium battery in a certain charge and discharge cycle is measured at 95% of the rated capacity, the battery status corresponding to that cycle is considered to be 0.95. Each charge and discharge cycle data is assigned a label reflecting its battery status, which can intuitively reflect the degradation of the battery during the aging process.

[0057] Furthermore, the cycle data under multiple charge and discharge cycle tests including the discharge voltage V dis,N , dynamic resistance R N The cumulative number of cycles N is sorted according to the time series and discretized to divide the continuous charge and discharge cycle data into several discrete time periods. The corresponding state monitoring label data is counted in each time period to obtain the training sample set.

[0058] Specifically, in this embodiment, in Matlab, firstly, the discharge voltage V recorded by time stamp is converted to dis,N , dynamic resistance R N The data of 1000 charge and discharge cycles are integrated into a time series using the cycle number N. The retime function is used to perform discretized statistics on the data according to the set window, and the ratio of the actual capacity mean in the segment to the rated capacity of the commercial lithium battery is extracted as the state monitoring label to form a training sample set.

[0059] S2: Taking the training sample set as input, an initial double exponential function repair state model based on the fitting parameters is constructed.

[0060] Specifically, in this embodiment, a double exponential function is used as the repair state model to define the relationship between the lithium battery repair state and the discharge voltage and dynamic resistance. The expression of the double exponential function repair state model is:

[0061]

[0062] Among them, V dis,N Indicates the discharge voltage of the lithium battery in the Nth cycle, R N represents the dynamic resistance of the lithium battery after the Nth cycle, and β1, β2, β3, and β4 are the model parameters to be fitted. The prediction results using the double exponential function are more accurate than those using a single variable, with a smaller mean absolute error.

[0063] Specifically, the method for constructing an initial double exponential function repair state model based on fitting parameters is: using the first 200 charge and discharge cycle data in the training sample set as fitting data, using the least squares method to fit the initial parameters of the double exponential function repair state model, and applying the fitted initial parameters to the double exponential function repair state model to obtain the initial double exponential function repair state model.

[0064] S3: Adaptive iterative update of the double exponential function repair state model based on the Monte Carlo parameter perturbation strategy.

[0065] The adaptive iterative update method is:

[0066] First, the data after the first 200 charge and discharge cycles in the training sample set is used as input, and the update is triggered with the number of charge and discharge cycles being 50: after every 50 charge and discharge cycles, the discharge voltage V of the latest cycle is collected. dis,N , dynamic resistance R N and condition monitoring labels, calculating the mean square error (MSE) between the prediction results of the current double exponential function repair state model and the condition monitoring labels;

[0067] Furthermore, the Monte Carlo parameter perturbation is performed to update the model parameters based on the current double exponential function repair state model parameters:

[0068] With the current model parameter β i As the mean, randomly generate several groups (for example, 1000 groups) of normally distributed samples according to the disturbance amplitude of no more than 5%. Each set of disturbance parameters is applied to the double exponential function repair state model, and the discharge voltage V of the current cycle of 50 charge and discharge cycles is used. dis,N , dynamic resistance R N and condition monitoring labels as input data, and the mean square error (MSE) between the prediction results of the double exponential function repaired state model and the condition monitoring labels is calculated for each set of disturbance parameters. The disturbance parameter combinations ranked in the top 10% by mean square error (MSE) are retained and averaged to obtain new model coefficients, thereby adapting the model to the latest degradation trend.

[0069] Furthermore, based on the model parameters under Monte Carlo parameter perturbation, sensitivity analysis is performed to adjust the update cycle and Monte Carlo parameter perturbation amplitude of the next update:

[0070] Calculate the gradient contribution of each model parameter to the mean square error (MSE) under the Monte Carlo parameter perturbation, and calculate the sensitivity of each model parameter according to the sensitivity calculation formula, and convert the model parameter sensitivity S iThe disturbance amplitude of model parameters exceeding 0.15 is increased from 5% to 10%, and the next update cycle is shortened to 25 cycles to track battery degradation changes at a higher frequency.

[0071] The sensitivity calculation formula is:

[0072]

[0073] Where K is the number of samples in the Monte Carlo parameter perturbation normal distribution, and k is the index of the sample in the Monte Carlo parameter perturbation normal distribution. Through this adaptive update mechanism, the model can maintain high prediction accuracy over long periods of time, ensuring the accuracy and reliability of the repair status assessment results.

[0074] Furthermore, the model only uses the discharge voltage V in the initial parameter fitting and adaptive update process. dis,N , dynamic resistance R N The three key parameters, namely, the number of cycles N, are used as input without introducing any other external features, thus ensuring that the implementation of the method is simple and has good versatility.

[0075] Further, when completed After a round of adaptive iterative updates, a trained double exponential function repair state model is obtained for subsequent repair state prediction of the repaired lithium battery.

[0076] S4: Apply the adaptive iteratively updated double exponential function repair state model to obtain the repair state prediction value of the target lithium battery and calculate the expected service life of the target lithium battery.

[0077] The failure of lithium batteries is mainly manifested in performance degradation caused by the loss of active lithium ions and degradation of the electrode interface. After the battery undergoes long-term charge and discharge cycles, the active materials of the positive and negative electrodes undergo irreversible phase changes, resulting in a significant attenuation of capacity and a significant decrease in the battery output power, which ultimately causes the battery to be unable to meet application requirements. When the battery fails, the present invention restores the capacity of the battery to more than 95% of the initial value through repair methods such as direct repair, and then uses the model of the present invention to predict the operating status and remaining life of the battery after failure repair. Failure repair is not the research objective of the present invention, and how to perform failure repair will not be described here.

[0078] First, after the lithium battery electrode material is repaired and reorganized, its internal structure and electrochemical properties undergo certain changes, and the traditional model may not be directly applicable. Therefore, the present invention collects key characteristic data of the repaired lithium battery, including the capacity recovery rate C rec and the internal resistance change rate ΔR.

[0079] The calculation formula for capacity recovery rate is:

[0080]

[0081] Among them, C repaired is the capacity after repair, C initial It is the benchmark capacity, i.e. the rated capacity of commercial lithium battery.

[0082] The calculation formula for the resistance change rate is:

[0083]

[0084] Among them, R repaired is the resistance after repair, R initial is the benchmark resistance, i.e. the rated resistance of commercial lithium batteries.

[0085] Furthermore, the repaired lithium battery is charged and discharged in cycles to simulate the charge and discharge cycle process of the repaired lithium battery in actual battery use, and the discharge voltage and dynamic resistance of the repaired lithium battery are obtained and input into the trained double exponential function repair state model to obtain the repair state prediction value S corresponding to the charge and discharge cycle. current , by comparing the repair status prediction value S current Compared with the measured capacity recovery rate C rec and verify model reliability.

[0086] Specifically, the repair state prediction value S current Compared with the measured capacity recovery rate C rec Perform residual verification ∈=|S current -C rec / 100|, if ∈>0.02 for three consecutive times, the Monte Carlo parameter re-update is triggered to suppress the prediction drift until the residual ∈ is less than or equal to the threshold, and the residual verification ends.

[0087] Taking the measured data of a waste NMC ternary lithium battery after repair as an example, the measured discharge voltage is 2.68V, the dynamic resistance is 0.092Ω, and the capacity recovery rate C rec =90.2%. The predicted health level S is calculated based on the parameters of the double exponential model (β1=0.56, β2=-0.24, β3=0.29, β4=8.7). current =0.94, and the capacity recovery rate C rec Converted to decimal form 0.902. At this time, the residual ∈=|0.94-0.902|=0.038, which significantly exceeds the threshold of 0.02, triggering the model recalibration mechanism. The system then starts the Monte Carlo optimization process: a random perturbation with a standard deviation of 5% is applied to the parameters to generate 1000 sets of new parameter combinations, and the new parameter set that minimizes the residual is screened (such as β1 is optimized to 0.55, β2 to -0.26). After optimization, the residual dropped to 0.017 (below the threshold), and the model returned to a reliable state. This case verifies the core value of the residual formula - by quantifying the model prediction value S currentCompared with the physical measured index C rec Deviations are dynamically maintained to maintain prediction accuracy. In engineering applications, a calibration mechanism is triggered when the residual ∈>0.02 for three consecutive times. This effectively suppresses prediction drift caused by electrode material phase changes and provides a self-correcting health management solution for the battery management system (BMS). Furthermore, after residual verification is completed, the expected service life is calculated based on the cumulative number of charge and discharge cycles and the predicted repair status of the corresponding charge and discharge cycles:

[0088]

[0089] Among them, N current is the cumulative number of charge and discharge cycles.

[0090] Furthermore, the expected service life is corrected using the internal resistance change rate ΔR. When |ΔR|>15%, the life prediction value is discounted according to the discrete reduction coefficient table, and the optimized remaining service life L is finally output. final .

[0091] Discrete reduction factor table

[0092]

[0093]

[0094] Specifically, the life correction formula is:

[0095] L final =L×f(ΔR);

[0096] Take the current state of a ternary lithium battery: basic life L = 600 cycles, measured internal resistance change rate ΔR = 25%;

[0097] The reduction factor is obtained from the table:

[0098]

[0099] Corrected lifespan:

[0100] L final =600×0.775=465 cycles.

[0101] Furthermore, based on the repair state prediction value S corresponding to the charge and discharge cycle current The repaired lithium batteries are divided into different repair levels so that the batteries can be used in a gradient manner:

[0102] Level Ⅰ: S current ≥80%, suitable for electric vehicle power systems;

[0103] Level II: 60% ≤ S current≤80%, suitable for energy storage systems or medium power scenarios;

[0104] Grade III: S current <60, suitable for low-speed electric vehicles or backup power supplies.

[0105] Furthermore, the model parameters β at each major trigger update node are listed. i The value and corresponding mean square error (MSE) evolution trend are used to generate an evaluation report including battery repair status, remaining life prediction value and maintenance recommendations. In specific implementation, the specific implementation method of the lithium battery repair status evaluation report is as follows: every 50 charge and discharge cycles are completed, the system is updated, and the discharge voltage V is collected through high-precision voltage sampling (±1mV) and HPPC pulse method. dis,N With dynamic resistance R N Based on the data, the Monte Carlo parameter perturbation strategy is used to optimize and update the parameters β1~β4 of the double exponential function model; the remaining life is linearly extrapolated based on the capacity decay trajectory. Maintenance recommendations are generated based on fault diagnosis decisions. Finally, the cycle count, SOR, RUL, and maintenance code are fed back to the battery management system (BMS) via the CAN bus using the J1939 protocol extended frame (PGN 65200). Simultaneously, reliability verification is implemented. If the predicted-measured residual exceeds 2% for three consecutive times, model recalibration is triggered, and historical parameters and mean square error (MSE) curves are stored in EEPROM and fed back to the battery management system (BMS) to trace the battery degradation trajectory and evaluate model reliability.

[0106] Example 2

[0107] A system for evaluating the repair status of a lithium battery after repair based on characteristic parameters, such as Figure 2 As shown, it includes a data acquisition and feature engineering module, a double exponential model initialization module, a Monte Carlo parameter optimization module, and a repair status estimation and remaining life prediction module connected in sequence.

[0108] Data acquisition and feature engineering module: used to collect historical battery operation data and construct a training sample set using capacity decay information as the status monitoring label.

[0109] Double exponential model initialization module: takes the training sample set as input and constructs an initial double exponential function repair state model based on the fitting parameters.

[0110] Monte Carlo parameter optimization module: used for adaptive iterative update of the double exponential function repair state model based on the Monte Carlo parameter perturbation strategy.

[0111] Repair status estimation and remaining life prediction module: Apply the adaptive iteratively updated double exponential function repair status model to obtain the repair status prediction value of the target lithium battery and calculate the expected service life of the target lithium battery.

[0112] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for evaluating the repair status of a failed repaired lithium battery based on dynamic parameters, characterized in that: The steps are: S1: Collect historical battery operation data and use capacity decay information as the condition monitoring label to construct a training sample set; S2: Taking the training sample set as input, an initial double exponential function repair state model based on the fitting parameters is constructed; S3: Adaptive iterative update of the double exponential function repair state model based on the Monte Carlo parameter perturbation strategy; S4: Apply the adaptive iteratively updated double exponential function repair state model to obtain the repair state prediction value of the target lithium battery and calculate the expected service life of the target lithium battery.

2. The method for evaluating the repair status of a failed repaired lithium battery based on dynamic parameters according to claim 1, characterized in that: The method for collecting historical battery operation data in step S1 is as follows: perform a charge-discharge cycle test, record the charge-discharge cycle data each time a charge-discharge cycle is completed, and record the cycle data including the discharge voltage V under multiple charge-discharge cycle tests. dis,N , dynamic resistance R N The cumulative number of cycles N is sorted according to the time series and discretized.

3. The method for evaluating the repair status of a failed repaired lithium battery based on dynamic parameters according to claim 2, characterized in that: The discretization processing method is as follows: the continuous charge and discharge cycle data is divided into several discrete time periods, and the corresponding capacity decay information in each time period is statistically analyzed as state monitoring label data to obtain a training sample set; the capacity decay information is the ratio of the actual capacity to the rated capacity of the commercial lithium battery.

4. The method for evaluating the repair status of a failed repaired lithium battery based on dynamic parameters according to any one of claims 1 to 3, characterized in that: The double exponential function repair state model is: Among them, S N Indicates the repair status value, V dis,N Indicates the discharge voltage of the lithium battery in the Nth cycle, R N represents the dynamic resistance of the lithium battery after the Nth cycle, and β1, β2, β3, and β4 are the model parameters to be fitted.

5. The method for evaluating the repair status of a failed repaired lithium battery based on dynamic parameters according to claim 4, characterized in that: The method for constructing an initial double exponential function repair state model based on fitting parameters is as follows: using the first M charge and discharge cycle data in the training sample set as fitting data, using the least squares method to fit the initial parameters of the double exponential function repair state model, and applying the fitted initial parameters to the double exponential function repair state model to obtain the initial double exponential function repair state model.

6. The method for evaluating the repair status of a failed repaired lithium battery based on dynamic parameters according to claim 2 or 3, characterized in that: The adaptive iterative update method is: S3.1: Take the data after the first M charge-discharge cycles in the training sample set as input, and trigger the update with the number of charge-discharge cycles K as the update period; S3.2: Perform Monte Carlo parameter perturbation to update model parameters based on the current double exponential function repair state model parameters; S3.3: Based on the model parameters under Monte Carlo parameter perturbation, perform sensitivity analysis to adjust the update period and Monte Carlo parameter perturbation amplitude of the next update; S3.4: Repeat S3.2 to S3.3 until the training is completed using the charge and discharge cycle data to obtain a trained double exponential function repair state model.

7. The method for evaluating the repair status of a failed repaired lithium battery based on dynamic parameters according to claim 6, characterized in that: The trigger update method described in step S3.1 is: after every K charge and discharge cycles, collect the discharge voltage V of the latest cycle dis,N , dynamic resistance R N and condition monitoring labels, calculating the mean square error between the prediction results of the current double exponential function repair state model and the condition monitoring labels; The implementation method of step S3.2 is: take the current model parameter β i As the mean, several groups of normally distributed samples are randomly generated according to the disturbance amplitude of no more than 5%. Each set of disturbance parameters is applied to the double exponential function repair state model, and the discharge voltage V dis,N , dynamic resistance R N The proposed method uses the state monitoring labels as input data and calculates the mean square error between the prediction results of the double exponential function repair state model and the state monitoring labels for each set of disturbance parameters. The disturbance parameter combinations with the top fixed percentage of the mean square error are retained and averaged to obtain the new model coefficients. The implementation method of step S3.3 is: calculate the gradient contribution of each model parameter to the mean square error under the Monte Carlo parameter perturbation, and calculate the sensitivity of each model parameter according to the sensitivity calculation formula, and convert the model parameter sensitivity S i The perturbation amplitude of the model parameters exceeding the sensitivity threshold is doubled, and the next update period is shortened to half the number of charge and discharge cycles.

8. The method for evaluating the repair status of a failed repaired lithium battery based on dynamic parameters according to claim 7, characterized in that: The sensitivity calculation formula is: Where K is the number of samples of the Monte Carlo parameter perturbation normal distribution, and k is the sample index of the Monte Carlo parameter perturbation normal distribution. In step S4, the repair state prediction value S of the corresponding lithium battery is obtained by using the trained double exponential function repair state model. current After that, key characteristic data of the repaired lithium battery is collected, including capacity recovery rate C rec and the internal resistance change rate ΔR, the repair state prediction value S current Compared with the measured capacity recovery rate C rec Perform residual verification ∈=|S current -C rec / 100|, if the residual is greater than the validation threshold for multiple consecutive times, the Monte Carlo parameter re-update is triggered to suppress the prediction drift until the residual is less than or equal to the validation threshold.

9. The method for evaluating the repair status of a failed repaired lithium battery based on dynamic parameters according to claim 8, characterized in that: After the residual verification is completed, the expected service life is calculated based on the cumulative number of charge and discharge cycles and the predicted value of the repair status of the corresponding charge and discharge cycles: Among them, N current is the cumulative number of charge and discharge cycles; The expected service life is corrected using the internal resistance change rate ΔR. When |ΔR|>15%, the life prediction value is discounted according to the discrete reduction coefficient table, and the optimized remaining service life L is finally output. final : L final =L×f(ΔR) Where, f(ΔR) is the reduction factor; In step S4, the repair state prediction value S corresponding to the charge and discharge cycle is used. current The repaired lithium batteries are divided into different repair levels: Level Ⅰ: S current ≥80%, suitable for electric vehicle power systems; Level II: 60% ≤ S current ≤80%, suitable for energy storage systems or medium power scenarios; Grade III: S current <60, suitable for low-speed electric vehicles or backup power supplies.

10. A system for evaluating the repair status of a failed repairable lithium battery based on dynamic parameters, using the method for evaluating the repair status of a failed repairable lithium battery based on dynamic parameters according to any one of claims 1 to 9, characterized in that: It includes a data acquisition and feature engineering module, a double exponential model initialization module, a Monte Carlo parameter optimization module, and a repair status estimation and remaining life prediction module connected in sequence; Data acquisition and feature engineering module: used to collect historical battery operation data and construct training sample sets using capacity decay information as status monitoring labels; Double exponential model initialization module: takes the training sample set as input and constructs the initial double exponential function repair state model based on the fitting parameters; Monte Carlo parameter optimization module: used for adaptive iterative update of the double exponential function repair state model based on the Monte Carlo parameter perturbation strategy; Repair status estimation and remaining life prediction module: Apply the adaptive iteratively updated double exponential function repair status model to obtain the repair status prediction value of the target lithium battery and calculate the expected service life of the target lithium battery.

Images