Adaptive Joint Estimation Method for State of Charge and Battery Capacity of Lithium-Ion Batteries

Abstract

The present invention relates to the technical field of lithium-ion batteries, and particularly to an adaptive joint estimation method for the state of charge and battery capacity of a lithium-ion battery, mainly solving the technical problem that the existing monitoring methods for the state of charge and battery capacity of lithium-ion batteries cannot accurately reflect the true state of the battery. The method includes: S1. Collect data and establish a first-order RC equivalent circuit model; S2. Introduce a constant-current pseudo-steady-state resistance R<subgt;CCQS< / subgt> to form a target circuit model with the constant-current pseudo-steady-state resistance R<subgt;CCQS< / subgt>; S3. Establish a state equation and an observation equation; S4. Construct a non-Gaussian Lévy noise sequence; S5. Use a particle filter algorithm with variational Bayesian to update the particle state, and jointly estimate the state of charge and battery capacity of the lithium-ion battery in a Lévy noise environment. The method of the present invention has strong model adaptability, and the particle filter updates the weights of these states by simulating a large number of possible battery states and based on actual measurements, so that the state estimation can more truly reflect the actual situation of the battery.

Description

Technical Field

[0001] The present invention relates to the technical field of lithium - ion batteries, and particularly to an adaptive joint estimation method for the state of charge and battery capacity of a lithium - ion battery. Background Art

[0002] With the growing global demand for sustainable energy, electric vehicles, renewable energy systems, and smart grid technologies have developed rapidly. As the most core energy storage component, the performance and safety of the battery have a decisive impact on the reliability and efficiency of the entire system. Among them, lithium - ion batteries are widely used in various commercial and industrial fields due to their high energy density and long life. However, lithium - ion batteries will gradually age and degrade during use, which will affect their storage capacity and output performance. Therefore, accurately monitoring the state of charge and battery capacity of lithium - ion batteries has become crucial.

[0003] In the prior art, the monitoring methods for the state of charge and battery capacity of lithium - ion batteries are generally developed based on the Gaussian noise assumption. In actual applications, the charge - discharge process of the battery is affected by various complex factors, such as the influence of the sensor by the complex industrial environment, the change of battery temperature, the aging process, and the non - linear characteristics of the electrochemical reaction. These factors often cause the measurement noise of the battery state to deviate from the Gaussian distribution. Therefore, when the traditional state - estimation algorithms still assume that the noise is Gaussian distributed in actual applications, they may not accurately reflect the true state of the battery. This gap indicates that it is necessary to further research and develop advanced estimation technologies that can adapt to non - Gaussian noise environments. Summary of the Invention

[0004] To overcome the technical defect that the existing monitoring methods for the state of charge and battery capacity of lithium - ion batteries cannot accurately reflect the true state of the battery, the present invention provides an adaptive joint estimation method for the state of charge and battery capacity of a lithium - ion battery.

[0005] The adaptive joint estimation method for the state of charge and battery capacity of a lithium - ion battery provided by the present invention successively includes the following steps:

[0006] S1. Collect the current data, voltage data, and time - interval data of the lithium - ion battery in the charge - discharge state, and establish a first - order RC equivalent circuit model;

[0007] S2. Introduce a constant - current pseudo - steady - state resistance R CCQS , to form a target circuit model with the constant - current pseudo - steady - state resistance R CCQS . The expression of the target circuit model is:

[0008]

[0009] Where Up represents the polarization voltage, R p represents the polarization resistance, R0 represents the internal resistance, R sum represents the total resistance, represents the current flowing through R p , SOC represents the state of charge, Δt represents the time interval, U out represents the terminal voltage, τ P represents the time constant, OCV represents the open-circuit voltage, C represents the battery capacity;

[0010] S3. Establish a state equation and an observation equation through the expression of the target circuit model;

[0011] S4. Construct a non-Gaussian Lévy noise sequence to simulate the interference of the real environment and sensor errors on the battery system;

[0012] S5. Through the state equation and the observation equation, use the particle filter algorithm with variational Bayesian to update the particle state, and jointly estimate the state of charge and battery capacity of the lithium-ion battery in a Lévy noise environment.

[0013] Further, in step S3, the expressions of the state equation and the observation equation are:

[0014]

[0015] where ω k , ν k represent the measurement noise and the observation noise respectively, and I k represents the current measured at the current moment.

[0016] Further, the step S5 includes:

[0017] S51. Based on the state and input of the previous moment, use the state equation to generate the predicted value of this moment;

[0018] S52. Use the predicted value and the observation equation to generate the estimated value of this moment and use the observation equation to generate the observed value of this moment

[0019] S53. Use variational Bayesian, the estimated value and the observed value to approximate the Student-t distribution, which is used as the likelihood function to calculate the weights of the particles, resample according to the weights of the particles, and estimate the latest SOC of the lithium-ion battery based on the resampled particle set;

[0020] S54. Compare the estimated SOC reduction with the actually consumed charge quantity, and adjust and update the estimated value of the battery capacity.

[0021] Further, the step S53 includes:

[0022] S531. Set initial hyperparameters a k , b k , c k and d k corresponding to the scale parameter and degrees of freedom of the Student-t distribution, and introduce a forgetting factor ρ for representing the degree of noise fluctuation, with the expression as follows:

[0023]

[0024] wherein, the superscript "-" represents the predicted value of the hyperparameter;

[0025] S532. According to and the generated difference, perform iterative updates on a k , b k , c k and d k , and the update formula is as follows:

[0026]

[0027] Calculate the variational distributions q(Λ k ), q(u k ) and q(v k ) of the posterior distributions of the parameters Λ k ), q(u k ) and q(v k ), and the calculation formulas are as follows:

[0028]

[0029] wherein, the superscript " * " represents the updated value of the hyperparameter, u k and v k are used to describe the characteristics of the measurement noise and jointly act to characterize the scale and shape of the noise, while Λ k is used to control the degrees of freedom of the Student-t distribution;

[0030]

[0031] Approximate a Student - t distribution with the parameters updated by the above variational Bayesian method. This Student - t distribution serves as the likelihood function for calculating the weights in particle filtering. Calculate the weights of each particle through the observation equation, perform resampling based on the weights of the particles, and estimate the latest state of charge (SOC) of the battery based on the resampled particle set. The calculation formula is as follows:

[0032]

[0033] Wherein, and represent the particle weights at the previous moment and this moment respectively, is the Student - t distribution approximated by the parameters iteratively updated by the above variational Bayesian method.

[0034] Furthermore, the step S54 includes:

[0035] S541. Capture the total battery discharge by comparing the reduction in SOC at this moment with that at the previous moment;

[0036] S542. Calculate the total battery discharge within a given time period according to the average value of the current I t-1 and I t and the time interval Δt. The calculation formula is as follows:

[0037]

[0038] Wherein, Ah t and Ah t-1 are the total battery discharges at this moment and the previous moment respectively, Δt is the time interval, I t-1 and I t are the current values at this moment and the previous moment respectively;

[0039] S543. Fit a straight line with the total battery discharge as the dependent variable and the SOC value at this moment as the independent variable, and calculate the battery capacity at this moment through the slope of the fitted straight line and the battery capacity at the previous moment. The calculation formula is as follows:

[0040] Capacity k =Capacity k-1 ×α + Slope×β;

[0041] Wherein, α is the attenuation factor, β is the regression influence factor, Capacity k and Capacity k-1 are the capacity values at this moment and the previous moment respectively, and Slope represents the slope of the fitted straight line.

[0042] The technical solution provided by the present invention has the following advantages compared with the prior art:

[0043] The self - adaptive joint estimation method for the state of charge and battery capacity of the lithium - ion battery provided by the present invention uses the concepts of the constant - current pseudo - steady - state resistance R CCQS and the total resistance R sum so that the model can more accurately simulate the resistance change of the battery under different working conditions. This method not only improves the reflection accuracy of the model for the internal state of the battery, but also enhances the adaptability of the model to the change of the battery resistance with factors such as external load and battery self - aging; moreover, this method uses a particle algorithm with variational Bayes to deal with non - Gaussian noise. Compared with the traditional filtering technology based on Gaussian assumptions, the particle filter with variational Bayes can provide a more flexible way to represent and update the state distribution of the lithium battery; furthermore, the particle filter makes the state estimation more truly reflect the actual situation of the battery by simulating a large number of possible battery states and updating the weights of these states according to actual measurements. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The drawings herein are incorporated into the specification and form a part of this specification, showing embodiments in accordance with the present invention and, together with the specification, are used to explain the principles of the present invention.

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0046] Figure 1 represents the flowchart of the self - adaptive joint estimation method in the embodiments of the present invention;

[0047] Figure 2 represents the schematic diagram of the first - order RC equivalent circuit model;

[0048] Figure 3 represents the schematic diagram of the target circuit model in the embodiments of the present invention;

[0049] Figure 4 is the schematic diagram of current and voltage under the WLTP (World Light Vehicle Test Procedure) working condition;

[0050] Figure 5 is the schematic diagram of current and voltage under the DST (Dynamic Stress Test) working condition;

[0051] Figure 6 is the comparison schematic diagram of the SOC estimation result and the true SOC value under the WLTP working condition in Gaussian noise;

[0052] Figure 7 It is a schematic diagram of the error of the SOC estimation result under the WLTP working condition in Gaussian noise;

[0053] Figure 8 It is a schematic diagram comparing the estimated battery capacity result with the true capacity value under the WLTP working condition in Gaussian noise;

[0054] Figure 9 It is a schematic diagram of the error of the battery capacity estimation result under the WLTP working condition in Gaussian noise;

[0055] Figure 10 It is a schematic diagram comparing the SOC estimation result with the true SOC value under the DST working condition in Gaussian noise;

[0056] Figure 11 It is a schematic diagram of the error of the SOC estimation result under the DST working condition in Gaussian noise;

[0057] Figure 12 It is a schematic diagram comparing the estimated battery capacity result with the true capacity value under the DST working condition in Gaussian noise;

[0058] Figure 13 It is a schematic diagram of the error of the battery capacity estimation result under the DST working condition in Gaussian noise;

[0059] Figure 14 It is a schematic diagram comparing the SOC estimation result with the true SOC value under the WLTP working condition in Lévy noise;

[0060] Figure 15 It is a schematic diagram of the error of the SOC estimation result under the WLTP working condition in Lévy noise;

[0061] Figure 16 It is a schematic diagram comparing the estimated battery capacity result with the true capacity value under the WLTP working condition in Lévy noise;

[0062] Figure 17 It is a schematic diagram of the error of the battery capacity estimation result under the WLTP working condition in Lévy noise;

[0063] Figure 18 It is a schematic diagram comparing the SOC estimation result with the true SOC value under the DST working condition in Lévy noise;

[0064] Figure 19 It is a schematic diagram of the error of the SOC estimation result under the DST working condition in Lévy noise;

[0065] Figure 20 It is a schematic diagram comparing the estimated battery capacity result with the true capacity value under the DST working condition in Lévy noise;

[0066] Figure 21 Schematic diagram of the error of the battery capacity estimation result under the DST condition in Lévy noise Detailed implementation manners

[0067] In order to more clearly understand the above objects, features and advantages of the present invention, the solutions of the present invention will be further described below. It should be noted that, without conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other

[0068] Many specific details are set forth in the following description in order to fully understand the present invention, but the present invention can also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only a part of the embodiments of the present invention, rather than all the embodiments

[0069] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings

[0070] Referring to Figure 1 , this embodiment provides an adaptive joint estimation method for the state of charge and battery capacity of a lithium-ion battery, including steps S1 to S5

[0071] S1. Collect the current data, voltage data and time interval data of the lithium-ion battery in the charge and discharge states, and establish a first-order RC equivalent circuit model

[0072] Specifically, referring to Figure 2 , the expression of the first-order RC equivalent circuit model is

[0073]

[0074] wherein, U p represents the polarization voltage, R p represents the polarization resistance, R0 represents the internal resistance, I represents the current, SOC represents the state of charge, dt represents the time interval, U out represents the terminal voltage, τ P represents the time constant, OCV represents the open-circuit voltage, and C represents the battery capacity

[0075] z S2. Referring to Figure 3 , introduce a constant-current pseudo-steady-state resistance R CCQS into the first-order RC equivalent circuit model to form a target circuit model with the constant-current pseudo-steady-state resistance R CCQS . The expression of the target circuit model is

[0076]

[0077] wherein, U p represents the polarization voltage, R p represents the polarization resistance, R0 represents the internal resistance, Rsum represents the total resistance, represents the current flowing through R p , SOC represents the state of charge, Δt represents the time interval, U out represents the terminal voltage, τ P represents the time constant, OCV represents the open circuit voltage, and C represents the battery capacity.

[0078] It is easy to understand that a lithium battery consists of a positive electrode, a separator, a negative electrode, an electrolyte, and a battery case. The positive electrode uses nickel-cobalt-manganese ternary oxide, and the negative electrode uses graphite. After introducing the constant current pseudo-steady state resistance R CCQS , the resistance inside the battery at the initial stage of battery charging is mainly composed of the ionic migration resistance of the electrolyte and the internal resistance R0 of the electrode material. At this time, the resistance inside the battery is relatively high because the flow paths of lithium ions and electrons are not fully open. As the charging progresses, lithium ions inside the battery gradually fill the voids in the graphite, and the concentration distribution of lithium ions in the electrolyte tends to be uniform, and the resistance inside the battery gradually decreases and stabilizes at a relatively low value, which is the constant current pseudo-steady state resistance R CCQS . At this time, the rising speed of the battery voltage slows down, and the charging efficiency of the battery increases. At the initial stage of discharge, due to the large concentration gradient of lithium ions in the electrolyte, the internal resistance is relatively high, and the current flow is blocked. As the discharge continues, lithium ions are released from the graphite and pass through the electrolyte to reach the nickel-cobalt-manganese ternary oxide, and the lithium ion concentration gradually tends to be uniform, and the internal resistance decreases and stabilizes, indicating that a new dynamic equilibrium state is reached inside the battery. At this time, the decreasing speed of the discharge voltage of the battery slows down, and the energy release is more stable.

[0079] Specifically, the constant current pseudo-steady state resistance R CCQS is used to represent the equivalent resistance of the lithium-ion battery when it reaches an approximate steady state during long-term constant current charging or discharging. In this solution, the constant current pseudo-steady state resistance R CCQS is added to the first-order RC equivalent circuit model, and the polarization resistance R p in the traditional first-order RC equivalent circuit model and the internal resistance R0 are combined to form the total resistance R sum . Under the constant current pseudo-steady state, the total resistance R sum will gradually converge towards the constant current pseudo-steady state resistance R CCQS , and the expression is as follows:

[0080]

[0081] S3. Establish the state equation and the observation equation through the expression of the target circuit model.

[0082] It should be noted that when the battery is charged and discharged at a constant current I to the constant current pseudo-steady state, the polarization impedance R p C pIt will degenerate into the polarization resistance R p , and together with the internal resistance R0, they form the constant-current pseudo-steady-state resistance R CCQS .

[0083] Specifically, the expressions of the state equation and the observation equation are as follows:

[0084]

[0085] where ω k , ν k represent the measurement noise and the observation noise respectively, and I k represents the current measured at the current moment.

[0086] S4. Construct a non-Gaussian Lévy noise sequence to simulate the interference of the real environment and sensor errors on the battery system.

[0087] It should be noted that constructing a non-Gaussian Lévy noise sequence to simulate the environmental impact of unreliable sensors or other complex industrial applications causes the measurement data to deviate from the normal value, forming a heavy-tailed distribution. At the same time, the battery model parameters change with time, temperature, battery aging, etc. If the model parameters are not adjusted in a timely manner, it may lead to a mismatch between the model and the actual situation, further enhancing the heavy-tailed characteristics of the measurement noise.

[0088] S5. Through the state equation and the observation equation, use the particle filter algorithm with variational Bayes to update the particle state, and jointly estimate the state of charge and battery capacity of the lithium-ion battery in a Lévy noise environment.

[0089] Specifically, step S5 includes:

[0090] S51. Based on the state and input of the previous moment, use the state equation to generate the predicted value of this moment. The formula is as follows:

[0091]

[0092] where represents the predicted value of this moment, and f represents the state equation;

[0093] S52. Use the predicted value and the observation equation to generate the estimated value of this moment and use the observation equation to generate the observed value of this moment The formula is as follows:

[0094]

[0095] where g represents the observation equation; [[ID=5%]]

[0096] S53. Use variational Bayes, the estimated value and the observed value to approximate the Student-t distribution, which is used as a likelihood function to calculate the weights of the particles. Resampling is performed according to the weights of the particles, and the latest SOC of the lithium-ion battery is estimated based on the resampled particle set;

[0097] S54. Compare the estimated SOC reduction with the actual consumed charge, and adjust and update the estimated value of the battery capacity.

[0098] More specifically, step S53 includes:

[0099] S531. Set the initial hyperparameters a k , b k , c k and d k corresponding to the scale parameter and degrees of freedom of the Student-t distribution, and introduce a forgetting factor ρ to represent the degree of noise fluctuation. The expression is as follows:

[0100]

[0101] where the superscript "-" represents the predicted value of the hyperparameter;

[0102] S532. According to and the generated difference, iteratively update a k , b k , c k and d k . The update formula is as follows:

[0103]

[0104] Calculate the variational distributions q(Λ k ), q(u k ) and q(v k ) of the posterior distributions of the parameters Λ k ), q(u k ) and q(v k ) using the updated hyperparameters. The calculation formulas are as follows:

[0105]

[0106] where the superscript " * " represents the updated value of the hyperparameter, u k and v k are used to describe the characteristics of the measurement noise and jointly act to characterize the scale and shape of the noise, while Λ k is used to control the degrees of freedom of the Student-t distribution;

[0107]

[0108] S533. Approximate a Student-t distribution using the parameters updated by the variational Bayesian method. This Student-t distribution is used as the likelihood function for the particle filter weight calculation. The weight of each particle is calculated using the observation equation. Resampling is performed based on the particle weight. The latest SOC of the battery is estimated based on the resampled particle set. The calculation formula is as follows:

[0109]

[0110] in, and Represented as the particle weights at the previous moment and this moment respectively, It is the Student-t distribution approximated by the parameters of the above variational Bayes update iteration.

[0111] More specifically, step S54 includes:

[0112] S541. Capture the total battery discharge by comparing the SOC reduction at this moment with the previous moment;

[0113] S542. According to the current I t-1 and I t The total battery discharge in a given time period is calculated using the average value of the time interval Δt and the following formula:

[0114]

[0115] Among them, Ah t and Ah t-1 are the total battery discharge at this moment and the previous moment respectively, Δt is the time interval, I t-1 and I t are the current values at this moment and the previous moment respectively;

[0116] S543. Using the total battery discharge as the dependent variable and the SOC value at this moment as the independent variable, a straight line is fitted. The battery capacity at this moment is calculated using the slope of the fitted line and the battery capacity at the previous moment. The calculation formula is as follows:

[0117] Capacity k =Capacity k-1 ×α+Slope×β;

[0118] Where α is the attenuation factor, β is the regression impact factor, Capacity k and Capacity k-1 are the capacity values at this moment and the previous moment respectively, and Slope is the slope of the fitting line.

[0119] Figure 4 Shown is the WLTP driving cycle. Under the WLTP driving cycle, the battery will experience sharp increases and decreases in current due to the simulated energy changes during vehicle acceleration and deceleration, representing the charging (regenerative braking) and discharging (acceleration) cycles of the battery. The overall downward trend of the voltage curve shows the voltage drop of the battery after long-term operation, which is due to the gradual discharge and energy consumption of the battery.

[0120] Figure 5 Shown is the DST driving cycle. Under the DST driving cycle, the battery starts from a fully charged state and undergoes multi-stage discharging due to the simulated different discharging conditions in actual use. During the discharging stage, the discharging current of the battery gradually increases to simulate the discharging conditions under different loads. To simulate rapid charging under actual use conditions, the battery also undergoes intermittent charging.

[0121] Figures 6 to 13 It shows the estimation results of the state of charge and battery capacity of the battery under two different driving cycles, WLTP and DST, respectively, in a Gaussian environment, as well as the estimation errors after the estimation results.

[0122] Figures 14 to 21 It shows the comparison results of the estimation of the state of charge and battery capacity of the battery under two different driving cycles, WLTP and DST, respectively, using unscented Kalman filter and particle filter with variational Bayes in a Lévy noise environment, as well as the comparison results of the estimation errors.

[0123] Specifically, in a Gaussian environment, for the method involved in the present invention, under two different driving cycles, WLTP and DST, the RMSE of SOC estimation is 0.4597% and 0.6270% respectively; and the estimation errors of the capacity under both driving cycles are below 5%. In a Lévy noise environment, for the method involved in the present invention, under two different driving cycles, WLTP and DST, the RMSE of SOC estimation is 0.206% and 0.304% respectively; and the estimation errors of the capacity under both driving cycles are also below 5%. In contrast, under the unscented Kalman filter, under two different driving cycles, WLTP and DST, the RMSE of SOC estimation is 2.283% and 1.0688% respectively; and the estimation errors of the capacity under both driving cycles exceed 5%. Through the above analysis, it shows that in the presence of non-Gaussian noise interference, the particle filter with variational Bayes can significantly improve the estimation accuracy.

[0124] From the above Figures 6 to 21From the above analysis of the estimation results and the error, it can be seen that by using the adaptive joint estimation method of this embodiment, the state of charge and capacity of the battery can be accurately estimated regardless of different environments or working conditions. This method provides an effective tool for the battery management system (BMS), is particularly suitable for dynamic and complex battery operating environments, and provides important technical support for the optimization of battery performance in key application fields such as electric vehicles.

[0125] The above description is only the specific implementation manners of the present invention, enabling those skilled in the art to understand or implement the present invention. Although the foregoing embodiments have been described in detail, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments, and they should all be covered by the protection scope of the claims.

Claims

1. An adaptive joint estimation method for the state of charge and battery capacity of a lithium-ion battery, characterized in that The following steps are included in sequence: S1. Collect current data, voltage data, and time interval data of the lithium-ion battery during charge and discharge, and establish a first-order RC equivalent circuit model; S2. Introduce a constant-current pseudo-steady-state resistance R into the first-order RC equivalent circuit model CCQS , to form a target circuit model with the constant-current pseudo-steady-state resistance R CCQS . The expression of the target circuit model is as follows: Among them, U p represents the polarization voltage, R p represents the polarization resistance, R0 represents the internal resistance, and R sum represents the total resistance, represents the current flowing through R p , SOC represents the state of charge, Δt represents the time interval, and U out represents the terminal voltage, τ P represents the time constant, OCV represents the open circuit voltage, and C represents the battery capacity; S3. Establish state equations and observation equations by expressions of the target circuit model; S4. Construct a non-Gaussian Lévy noise sequence to simulate the interference of real environment and sensor error on the battery system; S5. Using the state equation and observation equation, a particle filter algorithm with variational Bayesian is used to update the particle state, and the state of charge and battery capacity of the lithium-ion battery are jointly estimated in a Lévy noise environment.

2. The adaptive joint estimation method for the state of charge and battery capacity of a lithium-ion battery according to claim 1, characterized in that In step S3, the expressions of the state equation and observation equation are: where ω k , ν k represent measurement noise and observation noise respectively, and I k represents the current measured current at the current moment.

3. The adaptive joint estimation method for the state of charge and battery capacity of a lithium-ion battery according to claim 1, characterized in that The step S5 comprises: S51. Based on the state and input at the previous moment, the predicted value at this moment is generated using the state equation; S52. Generate the estimated value at this moment using the predicted value and the observation equation And generate the observed value at this moment using the observation equation S53. Using variational Bayes and an estimated value and an observed value to approximate the Student-t distribution, where the Student-t distribution is used as a likelihood function to calculate the weights of particles, resampling is performed according to the weights of the particles, and the latest SOC of the lithium-ion battery is estimated based on the resampled particle set; S54. Compare the estimated SOC reduction amount with the actual consumed charge amount, and adjust and update the estimated value of the battery capacity.

4. The adaptive joint estimation method for the state of charge and battery capacity of a lithium-ion battery according to claim 3, wherein The step S53 includes: Set initial hyperparameters a k , b k , c k and d k , and introduce a forgetting factor ρ to represent the degree of noise fluctuation. The expression is as follows: Among them, the superscript "-" indicates the predicted value of the hyperparameter; S532. According to and the generated difference, perform iterative updates on a k , b k , c k and d k The update formula is as follows: Calculate the parameter Λ using the updated hyperparameters k , u k and v k variational distributions q(Λ k ), q(u k ), and q(v k ) of the posterior distributions, and the calculation formulas are as follows: Among them, the superscript " * " represents the updated value of the hyperparameter, u k and v k are used to describe the characteristics of the measurement noise and jointly act on characterizing the scale and shape of the noise, while Λ k is used to control the degrees of freedom of the Student-t distribution; S533. Approximate a Student-t distribution using the parameters updated by the variational Bayesian method. This Student-t distribution is used as the likelihood function for the particle filter weight calculation. The weight of each particle is calculated using the observation equation. Resampling is performed based on the particle weight. The latest SOC of the battery is estimated based on the resampled particle set. The calculation formula is as follows: Among them, and represent the particle weights at the previous moment and this moment respectively, is a Student-t distribution approximated by the parameters of the above variational Bayesian update iteration.

5. The adaptive joint estimation method for the state of charge and battery capacity of a lithium-ion battery according to claim 3, characterized in that, The step S54 includes: S541. Capture the total battery discharge by comparing the SOC reduction at this moment with the previous moment; S542. Calculate the total amount of battery discharge within a given time period based on the average value of currents I t-1 and I t and the time interval Δt. The calculation formula is as follows: where Ah t and Ah t-1 are the total battery discharge amounts at this moment and the previous moment respectively, Δt is the time interval, I t-1 and I t are the current values at this moment and the previous moment respectively; S543. Using the total battery discharge as the dependent variable and the SOC value at this moment as the independent variable, a straight line is fitted. The battery capacity at this moment is calculated using the slope of the fitted line and the battery capacity at the previous moment. The calculation formula is as follows: Capacity k = Capacity k-1 ×α + Slope×β; where α is the attenuation factor, β is the regression influence factor, Capacity k and Capacity k-1 are the capacity values at this moment and the previous moment respectively, and Slope represents the slope of the fitted straight line.

Images