A method for estimating the cycle life of lithium-ion power batteries based on SOC

Abstract

This invention discloses a method for estimating the cycle life of lithium-ion power batteries based on SOC (State of Charge). The method includes: acquiring impedance data of the battery under different states of charge and cycle counts using electrochemical impedance spectroscopy (EIS) to obtain an initial dataset of capacitance characteristics including double-layer capacitance and Faraday capacitance; statistically analyzing the correlation between capacitance characteristics and state of charge within each cycle count segment in the independent variation pattern to obtain the decay trends of the two types of capacitance with the number of cycles; updating the three-dimensional relationship model based on the life estimation results, iteratively optimizing the model parameters using real-time impedance spectroscopy data to determine the capacitance characteristic change pattern reflecting the latest aging mechanism; obtaining the optimized three-dimensional relationship model, and repeating the steps of separating double-layer capacitance and Faraday capacitance for newly acquired impedance spectroscopy data to obtain dynamically adjusted decay trend parameters.

Description

Technical Field

[0001] This invention belongs to the field of battery life prediction technology, and particularly relates to a method for estimating the cycle life of lithium-ion power batteries based on SOC. Background Technology

[0002] As a core component in the new energy field, the lifespan of lithium-ion power batteries directly determines the reliability and economy of key applications such as electric vehicles and energy storage systems. Research on battery lifespan is not only a cornerstone for promoting energy transition but also a crucial link in achieving sustainable development. However, current methods for estimating battery lifespan often rely on single parameters, such as capacity decay or increased internal resistance. These methods often fail to accurately reflect the true aging state of batteries under complex usage scenarios, especially in capturing the dynamic changes at different states of charge (SOC). Existing solutions typically ignore the diversity of internal electrochemical processes within batteries, limiting the accuracy and universality of lifespan predictions.

[0003] In this field, the core challenge lies in comprehensively characterizing the aging mechanism of batteries during cycling, particularly the deep-seated mechanism of the impact of State of Charge (SOC) on lifespan, which remains incompletely understood. Among these challenges, capacitance characteristics, as a crucial indicator of battery health, require further investigation into their correlation with SOC and cycle count. Different types of capacitors, such as double-layer capacitors and Faraday capacitors, contribute significantly differently to the aging process, but existing research lacks a systematic differentiation and quantitative analysis of these capacitance parameters. Furthermore, monitoring capacitance degradation trends and their coupling with SOC usage history is difficult to predict accurately due to a lack of effective modeling methods. These unresolved technical factors result in the dual challenges of insufficient accuracy and excessive model complexity in practical applications of lifespan estimation. Summary of the Invention

[0004] This invention proposes a method for estimating the cycle life of lithium-ion power batteries based on SOC (State of Charge) to solve the problems existing in the prior art.

[0005] To achieve the above objectives, this invention provides a method for estimating the cycle life of lithium-ion power batteries based on SOC, comprising the following steps:

[0006] Impedance data of the battery under different states of charge and number of cycles were collected to obtain an initial dataset of capacitance characteristics including double-layer capacitance and Faraday capacitance.

[0007] Based on the initial dataset, the contributions of double-layer capacitance and Faraday capacitance are separated. The impedance spectrum data are fitted using an equivalent circuit model to determine the independent variation law of the two capacitance characteristics under different charging states.

[0008] Statistical analysis was performed on the correlation between capacitance characteristics and state of charge within each cycle number segment in the independent variation law, and the decay trend of the two types of capacitance with the number of cycles was obtained.

[0009] A three-dimensional relationship model between capacitance characteristics, cycle number, and state of charge was established using decay trend data.

[0010] Capacitive characteristic decay values ​​under specific states of charge and number of cycles are extracted from the three-dimensional relational model. Weighting is performed based on the differences between double-layer capacitance and Faraday capacitance to obtain comprehensive decay trend parameters.

[0011] If the overall degradation trend parameter exceeds the preset threshold, the remaining battery life is predicted by the support vector regression algorithm to obtain the life estimation result based on the capacitance characteristics under the current cycle number.

[0012] The three-dimensional relationship model is updated based on the lifetime estimation results, and the model parameters are iteratively optimized using real-time impedance spectroscopy data.

[0013] Obtain the optimized three-dimensional relationship model, and obtain the dynamically adjusted attenuation trend parameters based on the newly acquired impedance spectrum data;

[0014] By rerunning the support vector regression algorithm with dynamically adjusted degradation trend parameters, an updated predicted battery life is obtained.

[0015] Preferably, obtaining the initial dataset of capacitance characteristics includes:

[0016] Impedance data of the battery under different states of charge and number of cycles were collected using electrochemical impedance spectroscopy to obtain an initial dataset.

[0017] Capacitive characteristic data, including double-layer capacitance and Faraday capacitance, are extracted from the initial dataset to obtain a subset of capacitance characteristic data.

[0018] Preferably, determining the independent variation law of the two capacitance characteristics under different charging states includes:

[0019] The equivalent circuit model is loaded using impedance spectrum data to obtain the fitted parameter set;

[0020] The contributions of double-layer capacitance and Faraday capacitance are extracted from the fitted parameter set to obtain the separated capacitance dataset.

[0021] Using the separated capacitor dataset, the characteristic parameters of the double-layer capacitor under different charging states are determined;

[0022] The distribution of Faraday capacitance is calculated using characteristic parameters, and its independent variation curves under different charging states are obtained.

[0023] If the independent variation curve deviates from the preset threshold, the optimized capacitance characteristic data can be obtained by adjusting the parameters of the equivalent circuit model.

[0024] Based on the optimized capacitance characteristic data, the dynamic response of the two types of capacitors under changes in state of charge is determined, and the final independent variation law is obtained.

[0025] Preferably, the degradation trend of the two types of capacitance with the number of cycles includes:

[0026] By loading the dataset corresponding to the number of cycles, the capacitance characteristic data within each cycle segment is obtained. The distribution characteristics of the state of charge are extracted from the obtained capacitance characteristic data to obtain the correlation parameters. The correlation parameters are processed by statistical analysis to determine the decay trend of the double-layer capacitance. By calculating the dynamic response of the Faraday capacitance, its variation law with the number of cycles is obtained.

[0027] If the change pattern deviates from the preset threshold, the parameters of the statistical analysis are adjusted to obtain the optimized attenuation trend. Based on the optimized attenuation trend, the independent distribution characteristics of the two capacitors under the charging state are determined. A complete dataset is generated through the independent distribution characteristics to obtain the final attenuation trend.

[0028] Preferably, the step of establishing a three-dimensional relationship model between capacitance characteristics, cycle number, and state of charge using decay trend data includes:

[0029] A preliminary mapping relationship between capacitance characteristics and cycle number is generated by using decay trend data. The mapping relationship is then processed using a multinomial regression algorithm to obtain preliminary dynamic characteristics.

[0030] The distribution parameters of the state of charge are extracted from the preliminary features. The data basis of the three-dimensional relationship is determined based on the distribution parameters. The parameters of the polynomial regression are adjusted based on the three-dimensional relationship data to obtain the optimized dynamic change features. The intermediate expression of the aging mechanism is calculated through the optimized dynamic change features to obtain the mathematical expression of the three-dimensional relationship model.

[0031] Preferably, the obtained comprehensive attenuation trend parameters include:

[0032] The decay values ​​corresponding to the state of charge and the number of cycles are extracted by a three-dimensional relationship model. The difference between the electric double layer capacitance and the Faraday capacitance is quantified by weighted processing to obtain the comprehensive decay parameters.

[0033] The time series distribution of attenuation values ​​is obtained from the comprehensive attenuation parameters. Linear interpolation is performed on the changing trend of the number of cycles to determine the continuous expression of the attenuation values. Based on the continuous expression, the dynamic distribution characteristics of the state of charge are extracted. The features are classified by the random forest algorithm to determine the contribution ratio of the double-layer capacitance and the Faraday capacitance.

[0034] If the contribution ratio exceeds the preset threshold, the overall attenuation parameter is recalculated by adjusting the weighting parameters to obtain the corrected overall attenuation trend.

[0035] Preferably, obtaining the lifetime estimation result based on capacitance characteristics at the current cycle number includes:

[0036] By processing the comprehensive degradation trend parameters through the support vector regression algorithm, a preliminary prediction of the remaining battery life is obtained. The preliminary prediction is then corrected for time series using the number of iterations to determine the corrected life estimation distribution.

[0037] The capacitance characteristic variation features are extracted from the corrected lifetime estimation distribution to obtain the dynamic distribution data of the features. If the dynamic distribution data exceeds the preset threshold, the features are fitted twice by a regression algorithm to obtain the corrected prediction results.

[0038] Based on the revised prediction results, the correlation between the number of cycles and capacitance characteristics is analyzed, and the significance of the correlation is determined. The comprehensive attenuation trend parameter is then weighted and adjusted based on the significance to obtain the adjusted trend parameter distribution.

[0039] Extract the lifetime estimate under the current state from the adjusted trend parameter distribution to determine the final prediction output.

[0040] Preferably, determining the capacitance characteristic change pattern reflecting the latest aging mechanism includes:

[0041] The three-dimensional relationship model is updated by real-time impedance spectrum data, and the model parameters are adjusted by iterative optimization method to obtain the feature distribution reflecting the current state. The correlation data between capacitance characteristics and change law is extracted from the feature distribution to obtain the preliminary distribution of dynamic law.

[0042] If the initial distribution of the dynamic pattern exceeds the preset threshold, the related data is fitted by the support vector regression algorithm to determine the corrected distribution data. Based on the corrected distribution data, the changing trend between real-time impedance and aging characteristics is analyzed to obtain the adjustment value of the trend parameter. The parameters of the three-dimensional relationship model are updated a second time by the adjustment value to obtain the updated model parameter distribution.

[0043] Dynamic data reflecting capacitance characteristics are extracted from the updated model parameter distribution to determine the evolution direction of aging characteristics. The judgment results are used to perform time series correction on the characteristic distribution to determine the final evolution trend distribution.

[0044] Preferably, the dynamically adjusted attenuation trend parameters include:

[0045] The three-dimensional model is updated using newly acquired impedance spectrum data. The double-layer capacitance and Faraday capacitance are separated to obtain a preliminary capacitance characteristic distribution. The attenuation trend is extracted from the preliminary capacitance characteristic distribution.

[0046] The adjusted trend parameters are determined by a dynamic adjustment method. The accuracy of real-time processing is verified by impedance spectroscopy data to obtain the verified parameter distribution. If the verified parameter distribution exceeds the preset threshold, the attenuation trend is fitted by the support vector regression algorithm to obtain the corrected trend data.

[0047] The 3D model is updated based on the corrected trend data, the optimized model parameter distribution is obtained, the dynamically adjusted attenuation trend is extracted from the optimized model parameter distribution, the direction of change of capacitance characteristics is determined, the time series of data collection is corrected by the direction of change, and the final trend parameter distribution is determined.

[0048] Preferably, obtaining the updated predicted battery remaining life includes:

[0049] By running a support vector regression algorithm on the decay trend data, a preliminary remaining lifetime prediction value is obtained. Based on the preliminary remaining lifetime prediction value, aging mechanism characteristics are extracted, the aging characteristic distribution is determined, and the change in state of charge is analyzed through the aging characteristic distribution to obtain the preliminary distribution of coupling relationships.

[0050] If the initial distribution of coupling relationships exceeds the preset threshold, the trend of change is corrected by the support vector regression algorithm to obtain the adjusted distribution data. The decay trend is updated based on the adjusted distribution data to determine the direction of change of the aging mechanism. The time series of the state of charge is corrected by the direction of change to determine the final distribution of coupling relationships.

[0051] Update the remaining lifetime prediction based on the final coupling relationship distribution to obtain the optimized prediction result.

[0052] Compared with the prior art, the present invention has the following advantages and technical effects:

[0053] This invention discloses a method for estimating the cycle life of lithium-ion power batteries based on State of Charge (SOC). By collecting impedance data under different states of charge and cycle counts, the method separates the double-layer capacitance and Faraday capacitance, establishing a three-dimensional relationship model between capacitance characteristics, cycle count, and state of charge. This method utilizes a polynomial regression algorithm to quantify dynamic change characteristics, determine the mathematical expression of the aging mechanism, and predict the remaining battery life using a support vector regression algorithm. Furthermore, this invention can update model parameters in real time and dynamically adjust the degradation trend, thereby accurately reflecting the latest aging mechanism. This method can effectively assess battery performance degradation, achieve accurate life prediction, and provide an important basis for optimizing battery management systems and extending battery life. Attached Figure Description

[0054] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0055] Figure 1 This is a flowchart of a method according to an embodiment of the present invention;

[0056] Figure 2 This is a schematic diagram of the method according to an embodiment of the present invention. Detailed Implementation

[0057] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0058] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0059] Example 1

[0060] like Figure 1-2 As shown in the figure, this embodiment provides a method for estimating the cycle life of a lithium-ion power battery based on SOC, including the following steps:

[0061] Impedance data of the battery under different states of charge and number of cycles were collected to obtain an initial dataset of capacitance characteristics including double-layer capacitance and Faraday capacitance.

[0062] Based on the initial dataset, the contributions of double-layer capacitance and Faraday capacitance are separated. The impedance spectrum data are fitted using an equivalent circuit model to determine the independent variation law of the two capacitance characteristics under different charging states.

[0063] Statistical analysis was performed on the correlation between capacitance characteristics and state of charge within each cycle number segment in the independent variation law, and the decay trend of the two types of capacitance with the number of cycles was obtained.

[0064] A three-dimensional relationship model between capacitance characteristics, cycle number, and state of charge was established using decay trend data.

[0065] Capacitive characteristic decay values ​​under specific states of charge and number of cycles are extracted from the three-dimensional relational model. Weighting is performed based on the differences between double-layer capacitance and Faraday capacitance to obtain comprehensive decay trend parameters.

[0066] If the overall degradation trend parameter exceeds the preset threshold, the remaining battery life is predicted by the support vector regression algorithm to obtain the life estimation result based on the capacitance characteristics under the current cycle number.

[0067] The three-dimensional relationship model is updated based on the lifetime estimation results, and the model parameters are iteratively optimized using real-time impedance spectroscopy data.

[0068] Obtain the optimized three-dimensional relationship model, and obtain the dynamically adjusted attenuation trend parameters based on the newly acquired impedance spectrum data;

[0069] By rerunning the support vector regression algorithm with dynamically adjusted degradation trend parameters, an updated predicted battery life is obtained.

[0070] Specifically, the following steps are included:

[0071] S101. Acquire impedance data of the battery under different states of charge and cycle numbers using electrochemical impedance spectroscopy to obtain an initial dataset. Extract capacitance characteristic data, including double-layer capacitance and Faraday capacitance, from the initial dataset to obtain a subset of capacitance characteristics.

[0072] Specifically, electrochemical impedance spectroscopy (EIS) is one of the core methods in battery performance analysis, used to collect impedance data of batteries under different states of charge and cycle numbers. EIS can reflect the electrochemical reaction characteristics inside the battery, such as charge transfer resistance, double-layer capacitance, and dynamic changes in diffusion processes.

[0073] For example, when testing lithium-ion batteries, the state of charge can be set to 20%, 50%, and 80%, corresponding to low, medium, and high charge ranges, respectively, while the cycle count can be set to 0, 100, and 500 cycles to simulate the process of a battery aging from new to old. Using AC excitation with a frequency range from 0.01Hz to 100kHz, the real and imaginary parts of the impedance are obtained to form an initial dataset. The advantage of this method is that it can comprehensively capture the performance degradation characteristics of the battery at different stages of use, providing data support for subsequent optimization. Extracting capacitance characteristic data, including double-layer capacitance and Faraday capacitance, from the initial dataset is an effective way to further focus on the battery's electrochemical behavior. Double-layer capacitance mainly reflects the charge storage capacity of the electrode surface, while Faraday capacitance is closely related to the redox reaction of the electrode material.

[0074] Specifically, Faraday capacitance information can be extracted from the straight lines in the low-frequency region of the impedance spectrum Nyquist plot, while the semicircles in the mid-frequency region correspond to the double-layer capacitance.

[0075] For example, in a certain experiment, the double-layer capacitance might be 50 μF / cm at the initial cycle. 2As the number of cycles increased to 500, the concentration decreased to 30 μF / cm. 2 This indicates a decrease in electrode surface activity. This extraction process helps quantify the impact of battery aging on capacitance characteristics, providing a basis for material improvement.

[0076] In this embodiment, after obtaining the subset of capacitance characteristics, it can be analyzed in conjunction with the actual battery operating scenario.

[0077] For example, in the case of power batteries for electric vehicles, assuming a 50% state of charge, the initial double-layer capacitance is high, indicating a stable electrode-electrolyte interface. However, after 300 cycles, the capacitance decreases by 20%, possibly due to thickening of the solid electrolyte interfacial film. This can be verified by comparing the changes in the shape of the impedance spectrum at different cycle numbers; for example, an increase in the semi-circle diameter reflects an increase in interfacial resistance.

[0078] Preferably, this analysis can also guide the battery management system to adjust its charging and discharging strategies, thereby extending its service life.

[0079] It should be noted that the extraction of double-layer capacitance and Faraday capacitance is not carried out in isolation, but rather they support each other to form a complete description of electrochemical characteristics.

[0080] In this embodiment, if testing a lithium-ion battery reveals a significant decrease in Faraday capacitance under high charge conditions, it may be related to the collapse of the cathode material structure. In this case, by combining the double-layer capacitance data, if the value remains stable, it can be inferred that the problem primarily lies in the active material, rather than the electrode interface. This multi-faceted analysis can improve diagnostic accuracy.

[0081] Understandably, this method is not only applicable to laboratory research, but can also provide technical support for quality control in industrial production.

[0082] For example, during the battery research and development stage, by analyzing the capacitance characteristics of electrodes with different formulations using the methods described above, materials with better cycle resistance can be selected.

[0083] In this embodiment, one electrode formulation, at 80% charge, showed a Faraday capacitance decrease of only 10% after 500 cycles, while another formulation showed a decrease of 25%, clearly demonstrating the former's superior performance. The benefit of this comparative analysis lies in accelerating the R&D process and reducing trial-and-error costs. In summary, acquiring and analyzing capacitance characteristic data through electrochemical impedance spectroscopy can reveal the intrinsic mechanisms of battery performance changes from multiple dimensions, providing a solid foundation for improving battery design and application efficiency.

[0084] S102. Based on the initial dataset, separate the contributions of the double-layer capacitance and the Faraday capacitance, use the equivalent circuit model to fit the impedance spectrum data, and determine the independent variation law of the two capacitance characteristics under different charging states.

[0085] An equivalent circuit model is loaded using impedance spectroscopy data to obtain a fitted parameter set. The contributions of the double-layer capacitance and Faraday capacitance are extracted from this set to obtain a separated capacitance dataset. Using this dataset, the characteristic parameters of the double-layer capacitance under different charging states are determined. The distribution of the Faraday capacitance is calculated using these parameters, yielding its independent variation curves under different charging states. If the independent variation curves deviate from a preset threshold, the parameters of the equivalent circuit model are adjusted to obtain optimized capacitance characteristic data. Based on the optimized capacitance characteristic data, the dynamic responses of the two types of capacitance under changing charging states are determined, resulting in the final independent variation law.

[0086] Specifically, the equivalent circuit model is loaded using impedance spectrum data to obtain the fitted parameter set.

[0087] Understandably, the core of this step lies in transforming complex impedance spectrum data into analyzable circuit element parameters. For example...

[0088] In this embodiment, an equivalent circuit model including resistors, capacitors, and inductors can be selected. The impedance spectrum data of the battery is input into the model, and the parameter values ​​of each component are obtained through software fitting. For example, the resistance may be 50 milliohms, and the capacitance may be 200 microfarads. The advantage of this method is that it can transform abstract spectral data into intuitive physical meaning, which is convenient for subsequent analysis. The contributions of double-layer capacitance and Faraday capacitance are extracted from the fitted parameter set to obtain the separated capacitance dataset.

[0089] Specifically, the contributions of the two types of capacitors can be distinguished based on the behavior of different components in the circuit model.

[0090] For example, double-layer capacitance is typically associated with rapid charging and discharging of the electrode surface, exhibiting a response at higher frequencies, while Faraday capacitance is associated with electrochemical reactions and is more evident in the low-frequency region.

[0091] In this embodiment, it is assumed that the capacitance measured in the high-frequency region is 150 microfarads, which can be classified as a double-layer capacitor, while the additional 50 microfarads in the low-frequency region is classified as a Faraday capacitor. This separation helps to more clearly understand the role of the two types of capacitors. Using the separated capacitance dataset, the characteristic parameters of the double-layer capacitor under different charging states are determined.

[0092] It should be noted that the characteristic parameters may include the slope of the capacitance value as a function of voltage or the response time.

[0093] For example, as the state of charge increases from 20% to 80%, the double-layer capacitance may increase from 140 μF to 160 μF, with a slope of 0.33 μF / %. This analysis reflects the dynamic characteristics of the capacitor during battery operation, providing a basis for optimized design. The distribution of Faraday capacitance is calculated using characteristic parameters, yielding its independent variation curves under different states of charge.

[0094] Preferably, a curve can be plotted by fitting parameters.

[0095] In this embodiment, the Faraday capacitance contribution is 40 microfarads when the state of charge is 30%, and increases to 60 microfarads at 70%, exhibiting a non-linear growth. This distribution reveals the variation of Faraday capacitance with the intensity of electrochemical reactions, which helps predict battery performance. If the independent variation curve deviates from a preset threshold, optimized capacitance characteristic data can be obtained by adjusting the parameters of the equivalent circuit model.

[0096] For example, assuming the preset threshold is a curve slope of less than 0.5, but the actual measured slope is 0.7, the inductor in the model can be increased or the resistance value adjusted, and the model refitted until the curve matches the expectation. This adjustment improves the accuracy of the data and ensures the reliability of the analysis results. Based on the optimized capacitance characteristic data, the dynamic response of the two types of capacitors under changing states of charge is determined, and the final independent variation law is obtained.

[0097] In this embodiment, it can be observed that the double-layer capacitance tends to stabilize near a 50% state of charge, while the Faraday capacitance increases significantly at high charge levels. This indicates that the two have different response mechanisms; the former relies more on surface effects, while the latter is related to reaction depth. Understanding this pattern helps improve the accuracy of battery management strategies.

[0098] S103. Statistical analysis was performed on the correlation between capacitance characteristics and state of charge within each cycle number segment in the independent variation law to obtain the decay trend of the two types of capacitance with the number of cycles.

[0099] By loading the dataset corresponding to the number of loops, the capacitance characteristic data within each loop segment is obtained. The distribution characteristics of the state of charge are extracted from the obtained capacitance characteristic data to obtain correlation parameters. The correlation parameters are processed by statistical analysis to determine the decay trend of the double-layer capacitor. The dynamic response of the Faraday capacitor is calculated to obtain its variation law with the number of loops. If the variation law deviates from the preset threshold, the parameters of statistical analysis are adjusted to obtain the optimized decay trend. Based on the optimized decay trend, the independent distribution characteristics of the two capacitors under the state of charge are determined. A complete dataset is generated by the independent distribution characteristics to determine the final decay trend.

[0100] Specifically, by loading the dataset corresponding to the number of loops, it can be understood as obtaining a set of capacitance data that evolves over time from experiments or simulations.

[0101] For example, in battery cycle testing, assuming there are 1000 cycles, the corresponding voltage and current changes are recorded for each cycle, and then the capacitance characteristic data is derived.

[0102] For example, this data might show that the capacitance value decreased by 5% at the 200th cycle and by 20% at the 800th cycle. This data provides a basis for subsequent analysis. The distribution characteristics of the state of charge are extracted from the acquired capacitance characteristic data.

[0103] Specifically, this can be achieved by statistically analyzing the charge accumulation within a voltage range.

[0104] In this embodiment, it is assumed that the state of charge is divided into three intervals: low, medium, and high, and the capacitance value distribution in each interval is statistically analyzed.

[0105] For example, double-layer capacitance dominates under low charge conditions, while the proportion of Faraday capacitance increases under high charge conditions. This distribution characteristic reflects the dynamic change of capacitance with charge state. Statistical analysis is used to process the correlation parameters.

[0106] It should be noted that this step aims to quantify the relationships between features.

[0107] For example, correlation coefficients can be used to analyze the relationship between the number of cycles and capacitance decay. Assuming the correlation coefficient is 0.9 at the 500th cycle, it indicates a significant decay trend. This method helps to extract patterns from the data and determine the decay trend of the double-layer capacitance.

[0108] Preferably, the results can be visually presented by fitting a curve.

[0109] For example, plotting the double-layer capacitance over the number of cycles reveals a relatively slow decay in the first 300 cycles, followed by an acceleration. This trend helps predict long-term performance. The dynamic response of the Faraday capacitance can be calculated.

[0110] In this embodiment, the peak value change can be observed in each cycle.

[0111] For example, the peak response is 50 mF at the 100th cycle and decreases to 30 mF at the 900th cycle. This variation reveals its evolution characteristics with cycling. If the variation deviates from the preset threshold, for example, the preset attenuation is no more than 15%, but it actually reaches 18%, then the parameters of the statistical analysis are adjusted.

[0112] In this embodiment, weighting factors can be added for reanalysis to ensure the results are closer to reality. This optimization improves the reliability of the data. The independent distribution characteristics of the two capacitors under the charging state are determined based on the optimized attenuation trend.

[0113] For example, double-layer capacitance remains stable under low charge conditions, while Faraday capacitance decays significantly under high charge conditions. This independence analysis helps to understand the mechanisms of their interaction. Generating a complete dataset through independent distribution characteristics can be understood as integrating all analytical results to form a comprehensive view.

[0114] For example, summarizing the decay data from 1000 cycles into a single table clearly shows the changes in the two types of capacitance. This complete dataset supports subsequent applications, including determining the final decay trend.

[0115] For example, the final result shows that the degradation rate of the electric double layer capacitor is approximately 2% per year, while that of the Faraday capacitor is 5%. This clear trend provides a basis for optimized design and helps to extend service life.

[0116] S104. Establish a three-dimensional relationship model between capacitance characteristics, cycle number and state of charge by using decay trend data, and use a polynomial regression algorithm to quantify the dynamic change characteristics in the three-dimensional relationship to determine the mathematical expression of the aging mechanism.

[0117] A preliminary mapping relationship between capacitance characteristics and cycle number is generated by using decay trend data. The mapping relationship is processed by a multinomial regression algorithm to obtain preliminary dynamic features. The distribution parameters of the state of charge are extracted from the preliminary features to determine the data basis of the three-dimensional relationship. The parameters of the multinomial regression are adjusted according to the three-dimensional relationship data to obtain optimized dynamic features. The intermediate expression of the aging mechanism is calculated through the optimized dynamic features to obtain a mathematical expression.

[0118] Specifically, when generating a preliminary mapping relationship between capacitance characteristics and cycle number using decay trend data.

[0119] Understandably, this process aims to establish a foundational model.

[0120] For example, we can assume that a capacitor's capacitance decays to 90% of its initial value after 1000 cycles, and to 70% after 5000 cycles. By recording these data points, the initial mapping relationship can be expressed in a simple two-dimensional coordinate form, intuitively reflecting the impact of the number of cycles on the capacitor's characteristics. Using a multinomial regression algorithm to process the mapping relationship is to further capture the trend of nonlinear changes.

[0121] In this embodiment, a quadratic or cubic polynomial can be selected to fit the data. For example, if the initial capacity is set to 100 microfarads, and the decay trend shows a flattening followed by acceleration, a smooth curve can be obtained through regression analysis. The advantage of this method is that it can more accurately reflect dynamic changes, rather than relying on a simple linear assumption. When extracting the distribution parameters of the state of charge from the initial features...

[0122] It should be noted that the state of charge is usually closely related to the actual operating state of the capacitor.

[0123] For example, suppose that in a certain cycle, the capacitor exhibits stable voltage when fully charged, but slight fluctuations when half-charged. By statistically analyzing the voltage distribution characteristics over multiple cycles, distribution parameters, such as the mean and variance, can be extracted to quantify the changes in the state of charge.

[0124] Specifically, if the mean gradually shifts with the number of cycles, it may indicate a decrease in the stability of the capacitor's internal material. This distribution parameter provides a data foundation for subsequently constructing a three-dimensional relationship, helping to more comprehensively understand the interaction effects between variables. When adjusting the parameters of the multinomial regression based on the three-dimensional relationship data...

[0125] Preferably, fine-tuning can be made based on the degree of deviation of the actual data.

[0126] For example, if it is found that the predicted value differs significantly from the actual decay value in the high cycle number range, the order of the polynomial can be increased or the weighting coefficients can be adjusted to make the curve closer to the true trend.

[0127] In this embodiment, if the original regression predicts a capacitance of 75 microfarads after 5000 cycles, while the actual measurement is 70 microfarads, the error is reduced by iteratively optimizing the parameters. This optimized dynamic change characteristic can more realistically reflect the law of capacitance evolution over time, laying the foundation for subsequent analysis. The intermediate expression of the aging mechanism is calculated using the optimized dynamic change characteristic.

[0128] Understandably, this expression aims to reveal the physical meaning behind the decay.

[0129] For example, in electric double-layer capacitors, aging may be related to the blockage of pores on the electrode surface, while in Faraday capacitors, it may be related to the loss of active material.

[0130] Specifically, we can assume that after 3000 cycles, the internal resistance of a capacitor increases by 20%, and deduce the mathematical relationship between the change in resistance and the capacitance decay through characteristic analysis. The advantage of this intermediate expression is that it not only provides a mathematical description of the data, but also provides clues for in-depth research on the aging mechanism.

[0131] In this embodiment, by comparing the resistance increments under different states of charge, it can be verified whether aging is related to specific operating conditions, thereby providing a basis for optimizing capacitor design or usage strategies.

[0132] In this embodiment, the cascade analysis of these steps can also reveal potential performance bottlenecks.

[0133] For example, if the state-of-charge distribution parameters are found to be abnormally concentrated at high cycle counts, it may indicate that the capacitor is aging faster under certain conditions. This insight is of practical value for extending capacitor life or adjusting the operating environment.

[0134] Preferably, verification from multiple perspectives, such as combining the trends of voltage, resistance, and capacity decay, can jointly support the inference of the aging mechanism and ensure the reliability of the analysis results.

[0135] S105. Extract the capacitance characteristic decay value under specific charge state and cycle number from the three-dimensional relationship model, and perform weighted processing based on the difference between double-layer capacitance and Faraday capacitance to obtain comprehensive decay trend parameters.

[0136] The decay values ​​corresponding to the state of charge and the number of cycles are extracted by a three-dimensional relationship model. The difference between the electric double-layer capacitance and the Faraday capacitance is quantified by weighted processing to obtain the comprehensive decay parameter. The time series distribution of the decay value is obtained from the comprehensive decay parameter. Linear interpolation is performed on the changing trend of the number of cycles to determine the continuous expression of the decay value. The dynamic distribution characteristics of the state of charge are extracted based on the continuous expression. The features are classified by the random forest algorithm to determine the contribution ratio of the electric double-layer capacitance and the Faraday capacitance. If the contribution ratio exceeds the preset threshold, the comprehensive decay parameter is recalculated by adjusting the weighted processing parameters to obtain the corrected comprehensive decay trend.

[0137] Specifically, the core of extracting the decay values ​​corresponding to the state of charge and the number of cycles through a three-dimensional relationship model lies in separating key variables from multi-dimensional data. For example...

[0138] In this embodiment, capacitance test data can be grouped according to the number of cycles. Each group records the decay value from 100% to 20% of the state of charge. Assuming a battery's decay value is 5% at the 50th cycle and 8% at the 100th cycle, a preliminary outline of the decay curve can be constructed using these data points. This method helps to intuitively reflect the decay trend with increasing cycle number, laying the foundation for subsequent quantization. Weighted processing is used to quantify the difference between double-layer capacitance and Faraday capacitance.

[0139] It should be noted that double-layer capacitance mainly originates from physical adsorption, while Faraday capacitance is related to chemical reactions.

[0140] Specifically, weighting coefficients can be set, such as 0.6 for double-layer capacitance and 0.4 for Faraday capacitance. By decomposing the attenuation value, the comprehensive attenuation parameter can be calculated.

[0141] For example, in a test, the total attenuation value is 10%. After weighting, the double-layer capacitance contributes 6% and the Faraday capacitance contributes 4%, which clearly distinguishes the influence of the two mechanisms.

[0142] Preferably, adjusting the coefficients through multiple experiments can more accurately reflect the actual source of attenuation. The time series distribution of attenuation values ​​is obtained from the comprehensive attenuation parameters.

[0143] In this embodiment, the decay values ​​of each cycle can be arranged in chronological order to form a sequence, such as 5%, 5.2%, 5.5%, etc. Linear interpolation is then applied to the changing trend of the number of cycles, connecting these discrete points into a continuous curve.

[0144] For example, between the 50th and 51st cycles, the decay value smoothly transitions from 5% to 5.2%. This continuous representation facilitates the capture of subtle changes in decay, making it particularly suitable for long-term trend analysis. This continuous representation is also useful for extracting dynamic distribution characteristics of the state of charge.

[0145] It is understandable that the state of charge will exhibit a non-uniform distribution over time.

[0146] For example, suppose that in a certain loop, the state of charge rapidly decreases from 80% to 50%. Its dynamic characteristics are characterized by statistical distribution parameters such as mean and variance. Then, the features are classified using a random forest algorithm.

[0147] In this embodiment, features can be input into the model, and the contribution ratios of the double-layer capacitance and Faraday capacitance can be output, such as 55% and 45%. If the ratio exceeds a preset threshold, for example, the double-layer capacitance exceeds 60%, the weighting coefficients are adjusted and recalculated.

[0148] For example, by lowering the double-layer coefficient from 0.6 to 0.55 and re-decomposing the attenuation value, a corrected trend can be obtained, such as adjusting the attenuation from 10% to 9.5%. This iterative optimization method can improve the accuracy of the analysis.

[0149] In this embodiment, determining the contribution ratio is significant in revealing the capacitor aging mechanism.

[0150] For example, an increased proportion of double-layer capacitance may indicate physical structural degradation, while an increase in Faraday capacitance may be related to electrolyte decomposition. This classification provides data support for subsequent optimization designs.

[0151] Preferably, verifying the ratio change through multiple sets of experiments can provide a more comprehensive understanding of the inherent laws of attenuation.

[0152] S106. If the comprehensive degradation trend parameter exceeds the preset threshold, the remaining battery life is predicted by the support vector regression algorithm to obtain the life estimation result based on the capacitance characteristics under the current cycle number.

[0153] The overall degradation trend parameters are processed using a support vector regression algorithm to obtain a preliminary prediction of the battery's remaining lifespan. The preliminary prediction is then time-series corrected using the number of iterations to determine the corrected lifespan estimation distribution. Capacitance characteristic variation features are extracted from the corrected lifespan estimation distribution to obtain dynamic distribution data. If the dynamic distribution data exceeds a preset threshold, the features are refitted using a regression algorithm to obtain a revised prediction result. The correlation between the number of iterations and capacitance characteristics is analyzed based on the revised prediction result, and the significance of the correlation is determined. The overall degradation trend parameters are then weighted and adjusted based on the significance to obtain the adjusted trend parameter distribution. Finally, the lifespan estimation value under the current state is extracted from the adjusted trend parameter distribution to determine the final prediction output.

[0154] Specifically, when processing the comprehensive decay trend parameter using the support vector regression algorithm, it can be regarded as a mapping method to transform the multidimensional data of cycle number and capacitance characteristic decay into a preliminary prediction of the remaining lifetime.

[0155] In this embodiment, assuming a battery cycle life of 500 cycles, and considering that the overall degradation trend parameters show a 15% decrease in capacitance, support vector regression can be used to train a model based on historical data to predict a remaining lifespan of 2000 cycles. The core of this method lies in using kernel functions to capture nonlinear relationships, making it suitable for the complex changes in battery degradation. Time-series correction is then performed on the initial prediction using the cycle count.

[0156] Preferably, the concept of a time window can be introduced.

[0157] Specifically, the preliminary forecast values ​​are weighted and adjusted based on the decay trend of the most recent 100 cycles to reduce the impact of short-term fluctuations.

[0158] For example, if the initial prediction is 2000 cycles, but the last 50 cycles show accelerated decay, the corrected lifetime estimate might be adjusted to 1800 cycles. This correction method improves the dynamic adaptability of the prediction. When extracting the variation characteristics of capacitance properties from the corrected lifetime estimate distribution...

[0159] Understandably, the characteristics may include the slope or amplitude of the decay rate.

[0160] In this embodiment, if the estimated distribution shows that the lifetime gradually decreases from 1800 cycles, the extracted feature might be that the slope changes from 0.02 to 0.05, indicating accelerated decay. After obtaining the dynamic distribution data, if the slope exceeds the preset threshold of 0.04, a second fitting is required.

[0161] For example, after refitting using a regression algorithm, the revised prediction result may be adjusted from 1800 iterations to 1700 iterations, improving prediction accuracy. This revised prediction result can then be used to analyze the correlation between the number of iterations and capacitance characteristics.

[0162] It should be noted that statistical tests can be used to determine the significance level.

[0163] In this embodiment, if the number of cycles increases from 500 to 1000 and the capacitance characteristic decreases from 15% to 25%, and the correlation coefficient calculation shows a high degree of significance, then the correlation is strong. This analysis helps to reveal the driving factors of degradation. When the comprehensive degradation trend parameter is weighted and adjusted by significance...

[0164] Specifically, weights can be dynamically allocated based on the correlation coefficient.

[0165] For example, if the contribution of double-layer capacitance is significant, the weight is adjusted from 0.4 to 0.6 to obtain an adjusted trend parameter distribution. This adjustment more accurately reflects the decay effect of different capacitor types. When extracting lifetime estimates from the adjusted trend parameter distribution...

[0166] Preferably, the final output can be determined by combining the current loop state.

[0167] In this embodiment, assuming the current cycle count is 700, the adjusted parameter distribution shows a stabilizing decay trend, and the final predicted output may be 1600 cycles. This method, through multi-level analysis, ensures that the prediction results are closer to reality, which helps to optimize battery management strategies.

[0168] S107. Update the three-dimensional relationship model based on the lifetime estimation results, use real-time impedance spectrum data to iteratively optimize the model parameters, and determine the law of change in capacitance characteristics that reflects the latest aging mechanism.

[0169] The three-dimensional relationship model is updated using real-time impedance spectroscopy data. The model parameters are adjusted using an iterative optimization method to obtain a feature distribution reflecting the current state. Correlation data between capacitance characteristics and their changing patterns are extracted from the feature distribution to obtain a preliminary distribution of dynamic patterns. If the preliminary distribution of dynamic patterns exceeds a preset threshold, the correlation data is fitted using a support vector regression algorithm to determine the corrected distribution data. The changing trend between real-time impedance and aging characteristics is analyzed based on the corrected distribution data to obtain the adjustment value of the trend parameter. The parameters of the three-dimensional relationship model are updated a second time using the adjustment value to obtain the updated model parameter distribution. Dynamic pattern data reflecting capacitance characteristics are extracted from the updated model parameter distribution to determine the evolution direction of aging characteristics. The judgment result is used to perform time series correction on the feature distribution to determine the final evolution trend distribution.

[0170] Specifically, updating the three-dimensional relationship model using real-time impedance spectrum data can be understood as using the dynamic impedance information during battery operation to optimize the model structure.

[0171] For example, during the charging and discharging process of a battery, the impedance spectrum data acquired in real time may show a frequency range of 1Hz to 1000Hz. This data is input into the initial three-dimensional relational model to reflect the aging state of the electrode material.

[0172] In this embodiment, the resistance and capacitance parameters in the model can be adjusted by comparing the differences between the old and new impedance spectra to make them more closely reflect the current battery state. When adjusting the model parameters using an iterative optimization method...

[0173] Specifically, the parameter values ​​can be gradually adjusted based on the idea of ​​gradient descent.

[0174] For example, suppose the initial model predicts a capacitance value of 500F, while real-time data shows an actual value closer to 480F. Iterative optimization will gradually narrow the gap between the prediction and the actual value through multiple iterations, ultimately yielding a feature distribution that reflects the current state. This distribution may show the range of capacitance value fluctuations at different iteration counts, for example, 480F to 490F after 100 iterations. The correlation data between capacitance characteristics and their changing patterns is then extracted from this feature distribution.

[0175] It should be noted that this step focuses on how the capacitance changes over time or the number of cycles.

[0176] In this embodiment, after 200 cycles, the capacitance decreased by 5%, while the impedance increased by 10%, thus extracting a preliminary correlation between the two. If the preliminary distribution of the dynamic pattern exceeds a preset threshold, such as a capacitance decrease rate exceeding 2%, it is considered abnormal, and the correlation data is fitted using a support vector regression algorithm.

[0177] Preferably, the algorithm trains the model based on historical data and outputs corrected distribution data, such as adjusting the abnormal decline rate to a smoother 1.8%. When analyzing the changing trend between real-time impedance and aging characteristics based on the corrected distribution data, it can be understood as focusing on the changes in the shape of the impedance spectrum curve.

[0178] For example, a significant increase in impedance at low frequencies may indicate electrolyte loss. By quantifying this trend, adjustment values ​​for trend parameters can be obtained, such as increasing the resistance parameter in the model from 0.1Ω to 0.15Ω. After updating the parameters of the three-dimensional relational model a second time using these adjustment values, the resulting model parameter distribution may show the dynamic patterns of capacitance characteristics over 300 cycles, such as a gradual flattening of the decreasing trend. This can be used to determine the direction of aging characteristics.

[0179] Specifically, by comparing the characteristic distributions of different cycling stages, it can be inferred whether the battery has entered an accelerated aging stage.

[0180] In this embodiment, if the capacitance decrease rate changes from 0.5% to 1% after 400 cycles, it indicates accelerated aging. The judgment result is used to perform time-series correction on the characteristic distribution.

[0181] For example, a time-weighted factor can be introduced, increasing the weight of recent data by 20%, to determine the final evolution trend distribution. This distribution might predict that the battery will have 450F of remaining capacitance after 500 cycles, which helps in planning maintenance strategies in advance.

[0182] S108. Obtain the optimized three-dimensional relationship model. Repeat the steps of separating the double-layer capacitance and Faraday capacitance for the newly acquired impedance spectrum data to obtain the dynamically adjusted attenuation trend parameters.

[0183] The 3D model is updated using newly acquired impedance spectroscopy data. The electric double-layer capacitance and Faraday capacitance are separated to obtain a preliminary capacitance characteristic distribution. The attenuation trend is extracted from the preliminary capacitance characteristic distribution, and a dynamic adjustment method is used to determine the adjusted trend parameters. For the adjusted trend parameters, the accuracy of real-time processing is verified using impedance spectroscopy data to obtain the verified parameter distribution. If the verified parameter distribution exceeds a preset threshold, the attenuation trend is fitted using a support vector regression algorithm to obtain corrected trend data. The 3D model is updated based on the corrected trend data to obtain the optimized model parameter distribution. The dynamically adjusted attenuation trend is extracted from the optimized model parameter distribution, and the direction of change of capacitance characteristics is determined. The time series of data acquisition is corrected based on the direction of change to determine the final trend parameter distribution.

[0184] Specifically, when updating the three-dimensional model using newly acquired impedance spectrum data.

[0185] Understandably, impedance spectroscopy data typically contains rich frequency information and can reflect the electrochemical properties of materials.

[0186] For example, in the scenario of lithium battery aging analysis, assuming that the impedance spectrum data collected comes from battery cycle testing, a three-dimensional model can be initially constructed using this data, with the horizontal axis representing frequency, the vertical axis representing impedance amplitude, and the depth axis representing phase angle.

[0187] In this embodiment, the separation of double-layer capacitance and Faraday capacitance can be achieved through an equivalent circuit model. For example, the double-layer capacitance can be considered as the response in the high-frequency range, while the Faraday capacitance is more prominent in the low-frequency region. For instance, if the high-frequency impedance is around 0.1 ohms and the low-frequency impedance rises to 1 ohm, it can be preliminarily determined that the double-layer capacitance attenuates less, while the Faraday capacitance contribution increases.

[0188] Specifically, when extracting the attenuation trend from the initial capacitance characteristic distribution, the changes in impedance spectrum curves under different cycle numbers can be compared.

[0189] For example, the low-frequency impedance increases by 20% at the 50th cycle and by 50% at the 100th cycle, indicating that the attenuation trend is gradually intensifying.

[0190] Preferably, when using a dynamic adjustment method to determine trend parameters, a time weighting factor can be introduced, with recent data given a higher weight. For example, the data from the most recent 10 cycles has a weight of 0.8, while earlier data has a weight of 0.2, thus more accurately reflecting the current state.

[0191] It should be noted that when verifying the accuracy of real-time processing using impedance spectrum data, the impedance values ​​predicted by the model can be compared with the actual measured values. If the error is less than 5%, the parameter distribution is considered reliable.

[0192] In this embodiment, if the verified parameter distribution exceeds a preset threshold, for example, if the decay trend exceeds the expected 30%, then the data is fitted using a support vector regression algorithm.

[0193] For example, by inputting 100 sets of impedance values ​​corresponding to the number of cycles, the output is a smooth attenuation curve. The corrected trend data might show that the attenuation rate has adjusted from 0.5% / cycle to 0.4% / cycle. After updating the 3D model based on the corrected trend data, the optimized model parameter distribution might show that the proportion of double-layer capacitance has decreased from 60% to 50%, reflecting a change in aging characteristics. When extracting the dynamically adjusted attenuation trend, a slower rate of increase in low-frequency impedance can be observed, indicating that the capacitance characteristics are stabilizing.

[0194] For example, when determining the direction of change in capacitance characteristics, if the high-frequency impedance decreases while the low-frequency impedance increases, it can be inferred that the double-layer capacitance is weakening while the Faraday process is strengthening. When correcting the time series of data acquisition by changing the direction, the sampling frequency can be adjusted from every 10 cycles to every 5 cycles to capture more subtle trend changes.

[0195] In this embodiment, the final trend parameter distribution may show that the degradation rate stabilizes at 0.3% / cycle, which helps predict battery life and optimize usage strategies.

[0196] Understandably, this method can dynamically adapt to the aging process, improving the model's practicality.

[0197] S109. Rerun the support vector regression algorithm using the dynamically adjusted attenuation trend parameters to obtain the updated predicted battery remaining life and determine whether the coupling relationship between the aging mechanism and the state of charge has changed significantly.

[0198] By running a support vector regression algorithm on the decay trend data, a preliminary remaining lifetime prediction is obtained. Based on the preliminary remaining lifetime prediction, aging mechanism features are extracted, and the aging feature distribution is determined. The changes in the state of charge are analyzed through the aging feature distribution to obtain a preliminary distribution of coupling relationships. If the preliminary distribution of coupling relationships exceeds a preset threshold, the change trend is corrected using the support vector regression algorithm to obtain adjusted distribution data. The decay trend is updated based on the adjusted distribution data to determine the direction of change in the aging mechanism. The time series of the state of charge is corrected based on the direction of change to determine the final distribution of coupling relationships. The remaining lifetime prediction is updated based on the final distribution of coupling relationships to obtain the optimized prediction result.

[0199] Specifically, a support vector regression algorithm is run using decay trend data to obtain preliminary remaining lifespan predictions.

[0200] Understandably, support vector regression is used here to capture the non-linear characteristics in the decaying trend.

[0201] For example, in a battery aging scenario, suppose a battery's capacity decays exponentially with the number of cycles. The initial capacity is 1000mAh, and after 500 cycles it drops to 800mAh. An algorithm can predict its remaining lifespan to be approximately 300 cycles. Aging mechanism features are then extracted based on this preliminary remaining lifespan prediction.

[0202] Specifically, the aging mechanism may involve crack propagation in the electrode material or electrolyte decomposition.

[0203] In this embodiment, feature extraction can focus on the relationship between capacity decay rate and internal resistance increase. For example, when the internal resistance increases from 50mΩ to 80mΩ, it indicates that the solid electrolyte interface film thickens, and this feature can serve as an indicator of the aging mechanism. The distribution of aging characteristics is then determined.

[0204] For example, the distribution patterns of internal resistance and temperature can be observed by statistically analyzing data from multiple experiments.

[0205] For example, the internal resistance increases by 20 mΩ and 35 mΩ under 25℃ and 45℃ conditions, respectively, demonstrating the effect of temperature on the aging characteristic distribution. The changes in the state of charge are then analyzed using the aging characteristic distribution.

[0206] It should be noted that the state of charge is usually related to the open-circuit voltage of the battery.

[0207] In this embodiment, if the voltage of a battery drops from 3.7V to 3.65V at 50% charge, it may reflect a capacity decrease due to the loss of active material. After obtaining the preliminary distribution of coupling relationships, if it exceeds a preset threshold, such as a voltage change exceeding 0.1V, the trend is corrected using a support vector regression algorithm.

[0208] Preferably, environmental factors can be introduced during the correction, such as accelerated decay at 60% humidity, thus adjusting the predicted lifetime from 300 cycles to 280 cycles. The decay trend is then updated based on the adjusted distribution data.

[0209] For example, by combining temperature and cycle number, the decay curve changes from a single exponential form to a piecewise function form, reflecting the aging rate at different stages. This helps determine the direction of change in the aging mechanism.

[0210] In this embodiment, a slower rate of increase in internal resistance may indicate that the electrolyte decomposition is stabilizing. The time series of the state of charge can be corrected by changing the direction of change.

[0211] For example, correcting the state of charge in the original data after 100 cycles from 60% to 58% improves data consistency. This is also useful when determining the final coupling distribution.

[0212] Specifically, this can be verified through multiple sets of experiments. For example, at different charging rates, the difference in coupling distribution between 0.5C and 1C is reflected in a difference of 50 cycles in the predicted lifetime. The remaining lifetime prediction is then updated based on the final coupling distribution.

[0213] For example, after comprehensive adjustments, the estimated lifespan is 290 cycles, which is closer to reality than the initial prediction. This is when obtaining the optimized prediction results.

[0214] Preferably, the accuracy can be verified by comparing with measured data; for example, the error between the predicted value and the actual lifespan of 295 cycles is only 5 cycles. This method can effectively improve the reliability of the prediction and provide a more accurate reference for battery management.

[0215] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for estimating the cycle life of a lithium-ion power battery based on SOC, characterized in that, Includes the following steps: Impedance data of the battery under different states of charge and number of cycles were collected to obtain an initial dataset of capacitance characteristics including double-layer capacitance and Faraday capacitance. Based on the initial dataset, the contributions of double-layer capacitance and Faraday capacitance are separated. The impedance spectrum data are fitted using an equivalent circuit model to determine the independent variation law of the two capacitance characteristics under different charging states. Statistical analysis was performed on the correlation between capacitance characteristics and state of charge within each cycle number segment in the independent variation law, and the decay trend of the two types of capacitance with the number of cycles was obtained. A three-dimensional relationship model between capacitance characteristics, cycle number, and state of charge was established using decay trend data. Capacitive characteristic decay values ​​under specific states of charge and number of cycles are extracted from the three-dimensional relational model. Weighting is performed based on the differences between double-layer capacitance and Faraday capacitance to obtain comprehensive decay trend parameters. If the overall degradation trend parameter exceeds the preset threshold, the remaining battery life is predicted by the support vector regression algorithm to obtain the life estimation result based on the capacitance characteristics under the current cycle number. The three-dimensional relationship model is updated based on the lifetime estimation results, and the model parameters are iteratively optimized using real-time impedance spectroscopy data. Obtain the optimized three-dimensional relationship model, and obtain the dynamically adjusted attenuation trend parameters based on the newly acquired impedance spectrum data; By rerunning the support vector regression algorithm with dynamically adjusted degradation trend parameters, an updated predicted battery life is obtained.

2. The method according to claim 1, characterized in that, The initial dataset for obtaining capacitance characteristics includes: Impedance data of the battery under different states of charge and number of cycles were collected using electrochemical impedance spectroscopy to obtain an initial dataset. Capacitive characteristic data, including double-layer capacitance and Faraday capacitance, are extracted from the initial dataset to obtain a subset of capacitance characteristic data.

3. The method according to claim 1, characterized in that, The determination of the independent variation law of the two capacitance characteristics under different charging states includes: The equivalent circuit model is loaded using impedance spectrum data to obtain the fitted parameter set; The contributions of double-layer capacitance and Faraday capacitance are extracted from the fitted parameter set to obtain the separated capacitance dataset. Using the separated capacitor dataset, the characteristic parameters of the double-layer capacitor under different charging states are determined; The distribution of Faraday capacitance is calculated using characteristic parameters, and its independent variation curves under different charging states are obtained. If the independent variation curve deviates from the preset threshold, the optimized capacitance characteristic data can be obtained by adjusting the parameters of the equivalent circuit model. Based on the optimized capacitance characteristic data, the dynamic response of the two types of capacitors under changes in state of charge is determined, and the final independent variation law is obtained.

4. The method according to claim 1, characterized in that, The decay trends of the two types of capacitance with the number of cycles obtained include: By loading the dataset corresponding to the number of cycles, the capacitance characteristic data within each cycle segment is obtained. The distribution characteristics of the state of charge are extracted from the obtained capacitance characteristic data to obtain the correlation parameters. The correlation parameters are processed by statistical analysis to determine the decay trend of the double-layer capacitance. By calculating the dynamic response of the Faraday capacitance, its variation law with the number of cycles is obtained. If the change pattern deviates from the preset threshold, the parameters of the statistical analysis are adjusted to obtain the optimized attenuation trend. Based on the optimized attenuation trend, the independent distribution characteristics of the two capacitors under the charging state are determined. A complete dataset is generated through the independent distribution characteristics to obtain the final attenuation trend.

5. The method according to claim 1, characterized in that, The method for establishing a three-dimensional relationship model between capacitance characteristics, cycle number, and state of charge using decay trend data includes: A preliminary mapping relationship between capacitance characteristics and cycle number is generated by using decay trend data. The mapping relationship is then processed using a multinomial regression algorithm to obtain preliminary dynamic characteristics. The distribution parameters of the state of charge are extracted from the preliminary features. The data basis of the three-dimensional relationship is determined based on the distribution parameters. The parameters of the polynomial regression are adjusted based on the three-dimensional relationship data to obtain the optimized dynamic change features. The intermediate expression of the aging mechanism is calculated through the optimized dynamic change features to obtain the mathematical expression of the three-dimensional relationship model.

6. The method according to claim 1, characterized in that, The obtained comprehensive attenuation trend parameters include: The decay values ​​corresponding to the state of charge and the number of cycles are extracted by a three-dimensional relationship model. The difference between the electric double layer capacitance and the Faraday capacitance is quantified by weighted processing to obtain the comprehensive decay parameters. The time series distribution of attenuation values ​​is obtained from the comprehensive attenuation parameters. Linear interpolation is performed on the changing trend of the number of cycles to determine the continuous expression of the attenuation values. Based on the continuous expression, the dynamic distribution characteristics of the state of charge are extracted. The features are classified by the random forest algorithm to determine the contribution ratio of the double-layer capacitance and the Faraday capacitance. If the contribution ratio exceeds the preset threshold, the overall attenuation parameter is recalculated by adjusting the weighting parameters to obtain the corrected overall attenuation trend.

7. The method according to claim 1, characterized in that, The process of obtaining the lifetime estimation result based on capacitance characteristics at the current cycle number includes: By processing the comprehensive degradation trend parameters through the support vector regression algorithm, a preliminary prediction of the remaining battery life is obtained. The preliminary prediction is then corrected for time series using the number of iterations to determine the corrected life estimation distribution. The capacitance characteristic variation features are extracted from the corrected lifetime estimation distribution to obtain the dynamic distribution data of the features. If the dynamic distribution data exceeds the preset threshold, the features are fitted twice by a regression algorithm to obtain the corrected prediction results. Based on the revised prediction results, the correlation between the number of cycles and capacitance characteristics is analyzed, and the significance of the correlation is determined. The comprehensive attenuation trend parameter is then weighted and adjusted based on the significance to obtain the adjusted trend parameter distribution. Extract the lifetime estimate under the current state from the adjusted trend parameter distribution to determine the final prediction output.

8. The method according to claim 1, characterized in that, The iterative optimization of the model parameters includes: The three-dimensional relationship model is updated by real-time impedance spectrum data, and the model parameters are adjusted by iterative optimization method to obtain the feature distribution reflecting the current state. The correlation data between capacitance characteristics and change law is extracted from the feature distribution to obtain the preliminary distribution of dynamic law. If the initial distribution of the dynamic pattern exceeds the preset threshold, the related data is fitted by the support vector regression algorithm to determine the corrected distribution data. Based on the corrected distribution data, the changing trend between real-time impedance and aging characteristics is analyzed to obtain the adjustment value of the trend parameter. The parameters of the three-dimensional relationship model are updated a second time by the adjustment value to obtain the updated model parameter distribution. Dynamic data reflecting capacitance characteristics are extracted from the updated model parameter distribution to determine the evolution direction of aging characteristics. The judgment results are used to perform time series correction on the characteristic distribution to determine the final evolution trend distribution.

9. The method according to claim 1, characterized in that, The dynamically adjusted attenuation trend parameters include: The three-dimensional model is updated using newly acquired impedance spectrum data. The double-layer capacitance and Faraday capacitance are separated to obtain a preliminary capacitance characteristic distribution. The attenuation trend is extracted from the preliminary capacitance characteristic distribution. The adjusted trend parameters are determined by a dynamic adjustment method. The accuracy of real-time processing is verified by impedance spectroscopy data to obtain the verified parameter distribution. If the verified parameter distribution exceeds the preset threshold, the attenuation trend is fitted by the support vector regression algorithm to obtain the corrected trend data. The 3D model is updated based on the corrected trend data, the optimized model parameter distribution is obtained, the dynamically adjusted attenuation trend is extracted from the optimized model parameter distribution, the direction of change of capacitance characteristics is determined, the time series of data collection is corrected by the direction of change, and the final trend parameter distribution is determined.

10. The method according to claim 1, characterized in that, The obtained updated predicted battery remaining life includes: By running a support vector regression algorithm on the decay trend data, a preliminary remaining lifetime prediction value is obtained. Based on the preliminary remaining lifetime prediction value, aging mechanism characteristics are extracted, the aging characteristic distribution is determined, and the change in state of charge is analyzed through the aging characteristic distribution to obtain the preliminary distribution of coupling relationships. If the initial distribution of coupling relationships exceeds the preset threshold, the trend of change is corrected by the support vector regression algorithm to obtain the adjusted distribution data. The decay trend is updated based on the adjusted distribution data to determine the direction of change of the aging mechanism. The time series of the state of charge is corrected by the direction of change to determine the final distribution of coupling relationships. Update the remaining lifetime prediction based on the final coupling relationship distribution to obtain the optimized prediction result.

Images