A battery SOC estimation method and system based on unscented Kalman filtering

Abstract

The present invention discloses a battery SOC estimation method and system based on unscented Kalman filtering, which belongs to the technical field of battery SOC estimation. It includes: establishing a state space equation of a lithium battery based on a second-order RC equivalent circuit model; conducting a discharge-stationary experiment and fitting to obtain an OCV-SOC relationship curve; calculating a voltage estimate based on the state space equation, and using the recursive least squares method to identify the terminal voltage of the model; sampling the open circuit voltage and current of the lithium battery to obtain the initial value of the open circuit voltage; using an unscented Kalman filtering algorithm to continuously optimize the state space equation, complete the continuous prediction and correction of the OCV-SOC relationship curve, and achieve the continuous convergence of the state of charge SOC estimate to the actual state of charge SOC value. The present invention successfully solves the problem of accurate estimation of SOC values ​​in lithium-ion battery applications by solving the nonlinear transfer problem of mean and covariance in the card.

Description

Technical Field

[0001] The present invention belongs to the technical field of battery SOC estimation, and in particular relates to a battery SOC estimation method and system based on unscented Kalman filtering. Background Art

[0002] During the complete operation cycle of the lithium-ion battery pack, the battery management system (BMS) is crucial to monitor and adjust the core parameter SOC, which will directly affect the effect and safety of emergency power output. Therefore, it is crucial to monitor the changes in SOC in a timely manner and ensure the working performance of the lithium-ion battery based on the monitoring results.

[0003] Since the SOC estimation technology in BMS is not yet fully mature, the existing safety hazards have seriously restricted the development of lithium-ion battery packs. For this type of battery, reliable BMS management requires accurate SOC values. When we know the SOC value, not only can reliable energy management and safety control be carried out, but also premature damage to the lithium-ion battery pack can be avoided, thereby extending its service life. Therefore, accurate estimation of the SOC value is crucial to ensure the working performance of the lithium-ion battery pack as well as energy and safety management.

[0004] With the continuous improvement and development of BMS technology, the methods and advantages and disadvantages of estimating battery SOC are as follows: the open circuit voltage method is simple and easy, but because obtaining the open circuit voltage of the battery requires the battery to be in a static state for a long time, it is not suitable for online calculation of the battery SOC; in contrast, the ampere-hour integration method is simple to calculate and easy to implement, but it is difficult to determine the initial value of the battery SOC using the ampere-hour integration method, and long-term integration calculations will produce cumulative errors; the neural network method is applicable to various batteries and has good SOC estimation capabilities for nonlinear systems such as batteries, but the estimation results will fall into local minimum points; the Kalman filter method calculates the battery SOC by establishing the Kalman filter equation, and estimates the battery SOC with high accuracy, but is limited by the battery model, the system design is difficult, and the amount of calculation is large. Summary of the invention

[0005] In order to solve the technical problems existing in the above background technology, the present invention provides a battery SOC estimation method based on unscented Kalman filtering, comprising the following steps:

[0006] Establishing a second-order RC equivalent circuit model, and establishing a state space equation of a lithium battery based on the second-order RC equivalent circuit model;

[0007] A discharge-stationary experiment is conducted to determine the relationship between the open circuit voltage OCV and the state of charge SOC of the lithium battery at different discharge stages, and the OCV-SOC relationship curve is obtained by fitting;

[0008] The voltage estimation value is calculated based on the state space equation, and the terminal voltage of the model is identified by using the recursive least square method to obtain the terminal voltage identification curve to verify the accuracy of the model; when the lithium battery is discharged at a constant current, the open circuit voltage and current of the lithium battery are sampled to obtain the initial value of the open circuit voltage;

[0009] The unscented Kalman filter algorithm is used to continuously optimize the state-space equation, complete the continuous prediction and correction of the OCV-SOC relationship curve, and achieve the continuous convergence of the state-of-charge SOC estimated value to the state-of-charge SOC actual value.

[0010] In a further embodiment, the state space equation is as follows:

[0011]

[0012] It also includes an observation equation established based on the observation established by the state equation, and the observation equation is expressed as follows:

[0013] y(k)=[―1 ―1 0]x(k)+[―R0]I(k)+v(k);

[0014] Where x(k) is the system input value of convergence number k, and the state variable x = [U1 U2 SOC] T , y(k) is the output value of the system with convergence number k, Δt is the sampling time, R0 is the ohmic internal resistance of the lithium battery, R1 and R2 are the polarization resistances of the lithium battery, C1 and C2 are the polarization capacitances of the lithium battery, U1 is the voltage across R1, U2 is the voltage across R2, ω(k) is the system process noise with convergence number k, I(k) is the sampling current with convergence number k, v(k) is the observation noise with convergence number k, Q N is the capacitance value to be calculated after discretization, and k is the number of convergences.

[0015] In a further embodiment, the expression of the OCV-SOC relationship curve is as follows:

[0016] U OC =F(SOC)=31.26SOC 6 ―98.19SOC 5 +120.1SOC 4 ―70.56SOC 3 +19.07SOC 2 ―13.09SOC+3.15.

[0017] In a further embodiment, the identification process of the terminal voltage identification of the model using the recursive least squares method is as follows:

[0018] The following formula is used to calculate the battery parameter identification results:

[0019] In the formula, is the observed parameter matrix of the battery; θ is the parameter matrix, expressed as θ=[R0 R1 R2 C1 C2]; where R0 is the ohmic internal resistance of the lithium battery, R1 and R2 are the polarization resistances of the lithium battery, and C1 and C2 are the polarization capacitances of the lithium battery; e(k) is the error value.

[0020] In a further embodiment, the process of continuously optimizing the state space equation using the unscented Kalman filter algorithm is as follows:

[0021] Use unscented transformation to find the sampling points. In the state space equation, let the dimension of the state variable x be n, then;

[0022]

[0023] In the formula, is the Sigma point, is the optimal value of the state variable that converges k times, i represents the number of sampling points, and λ is the scaling factor, which is calculated as follows: λ=α 2 (n+k)―n, α is 10 ―4 An integer between 0 and 1;

[0024] Use the following formula to calculate the mean weight of Sigma points Covariance weights Initial mean and the initial covariance

[0025] Where, β≥0;

[0026] Define the state space equation as y = f(x), then substitute the Sigma point into y = f(x), and the estimated value of the system state variable that converges k+1 times is as follows:

[0027] y i =f(x i ), i = 0 to 2n;

[0028] In the formula, Represents the mean value of the state variable from the kth convergence to the k+1th convergence at sampling point i;

[0029] In the formula, Represents the optimal value of the state variable of sampling point i from convergence k to convergence k+1.

[0030] In a further embodiment, the parameter matrix is ​​optimized using the following method:

[0031] Calculate the gain value K(k+1) of the current convergence number k+1:

[0032]

[0033] Where P(k) converges to the error covariance matrix with k times, is the observation data matrix of the current convergence number k+1;

[0034] Calculate the parameter error e(k+1) of the current convergence number k+1 of the equivalent circuit model:

[0035]

[0036] In the formula, y (k+1) is the output value of the system at the current convergence number k+1, θ(k) represents the parameter matrix of the convergence number k, Represents the observation value of the current convergence number k+1;

[0037] The battery model parameter matrix θ(k+1) of the current convergence number k+1 is obtained using the following formula:

[0038] θ(k+1)=θ(k)+K(k+1)e(k+1).

[0039] In a further embodiment, the following steps are also included:

[0040] Predict the error covariance matrix that converges k+1 times:

[0041]

[0042] In the formula, Q k+1 is the system noise covariance matrix that converges k+1 times;

[0043] Substitute the predicted value point set of the state quantity that converges k+1 times into the observation equation, and use the observation equation to obtain the observed predicted value that converges k+1 times Perform a state transfer in the observation equation, and then weight it to get the mean of the converged k+1 measurements

[0044] Calculate the variance matrix P of the measurements that converge k+1 times xy，k+1 :

[0045]

[0046] Calculate the covariance matrix P of the state quantity and measurement value that converges k+1 times yy，k+1 :

[0047]

[0048] Calculate the Kalman filter gain K k+1 :K k+1 =P xy，k+1 / P yy，k+1 ;

[0049] Update system status variable values

[0050] Error variance matrix update P x,k+1|k+1 : In the formula, is the optimal value of the state variable when the variable x converges k+1 times, y (k+1) is the output value of the convergence k+1 times, is the optimal output value of the system with convergence number k+1.

[0051] In a further embodiment, the calculation formula of the error covariance matrix P(k+1) of the current convergence number k+1 is as follows:

[0052] Where I is the unit matrix, P(k) is the error covariance matrix with convergence times k, K(k+1) is the gain value of the current convergence times k+1, is the observation data matrix of the current convergence number k+1.

[0053] A battery SOC estimation system based on unscented Kalman filtering is used to implement the battery SOC estimation method as described above, comprising:

[0054] The first module is configured to establish a second-order RC equivalent circuit model, and establish a state space equation of the lithium battery based on the second-order RC equivalent circuit model;

[0055] The second module is configured to conduct a discharge-stationary experiment to determine the relationship between the open circuit voltage OCV and the state of charge SOC of the lithium battery at different discharge stages, and to obtain an OCV-SOC relationship curve by fitting;

[0056] The third module is configured to calculate a voltage estimate based on the state space equation, identify the terminal voltage of the model using a recursive least squares method, and obtain a terminal voltage identification curve to verify the accuracy of the model; when the lithium battery is discharged at a constant current, the open circuit voltage and current of the lithium battery are sampled to obtain an initial value of the open circuit voltage;

[0057] The fourth module is configured to use the unscented Kalman filter algorithm to continuously optimize the state-space equation, complete the continuous prediction and correction of the OCV-SOC relationship curve, and realize the continuous convergence of the state-of-charge SOC estimated value to the state-of-charge SOC actual value.

[0058] The beneficial effects of the present invention are as follows: by using traceless transformation to process nonlinear functional relationships in the Kalman filter algorithm, the nonlinear transmission problem of mean and covariance is solved, and the difficult problem of accurate estimation of SOC value in lithium-ion battery applications is successfully solved. This method realizes the effective iterative calculation of SOC value of lithium-ion battery, while improving the estimation accuracy and stability, laying a foundation for SOC estimation of lithium-ion battery packs and real-time monitoring of working status. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 This is a second-order RC equivalent circuit model diagram in the present invention.

[0060] Figure 2 This is a diagram of the OCV and SOC fitting results in the present invention.

[0061] Figure 3 This is a diagram of the terminal voltage identification result of the present invention.

[0062] Figure 4 This is a diagram of the SOC estimation value in the present invention.

[0063] Figure 5 This is the SOC estimation error diagram in the present invention.

[0064] Figure 6 This is the flow chart of the unscented Kalman filter in the present invention. DETAILED DESCRIPTION

[0065] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings. The following drawings are used to illustrate the present invention, but are not used to limit the scope of the present invention.

[0066] like Figure 6 As shown, a battery SOC estimation method based on unscented Kalman filtering includes the following steps:

[0067] As attached Figure 1 As shown, a second-order RC equivalent circuit model is established, and a state space equation of a lithium battery is established based on the second-order RC equivalent circuit model;

[0068] A discharge-stationary experiment is conducted to determine the relationship between the open circuit voltage OCV and the state of charge SOC of the lithium battery at different discharge stages, and the OCV-SOC relationship curve is obtained by fitting;

[0069] The voltage estimation value is calculated based on the state space equation, and the terminal voltage of the model is identified by using the recursive least square method to obtain the terminal voltage identification curve to verify the accuracy of the model; when the lithium battery is discharged at a constant current, the open circuit voltage and current of the lithium battery are sampled to obtain the initial value of the open circuit voltage;

[0070] The unscented Kalman filter algorithm is used to continuously optimize the state-space equation, complete the continuous prediction and correction of the OCV-SOC relationship curve, and achieve the continuous convergence of the state-of-charge SOC estimated value to the state-of-charge SOC actual value.

[0071] In a further embodiment, the state space equation of the lithium battery is established based on the second-order RC equivalent circuit model as follows:

[0072]

[0073] It also includes an observation equation established based on the observation established by the state equation, and the observation equation is expressed as follows:

[0074] y(k)=[―1 ―1 0]x(k)+[―R0]I(k)+v(k);

[0075] Where x(k) is the system input value of convergence number k, and the state variable x = [U1 U2 SOC] T , y(k) is the output value of the system with convergence number k, Δt is the sampling time, R0 is the ohmic internal resistance of the lithium battery, R1 and R2 are the polarization resistances of the lithium battery, C1 and C2 are the polarization capacitances of the lithium battery, U1 is the voltage across R1, U2 is the voltage across R2, ω(k) is the system process noise with convergence number k, I(k) is the sampling current with convergence number k, v(k) is the observation noise with convergence number k, Q N is the capacitance value to be calculated after discretization, and k is the number of convergences.

[0076] Based on the above description, different constant currents are used to operate the lithium battery during the discharge process, and the result of the battery's state of charge changing from 1 to 0 is obtained. At the same time, the time interval of each discharge capacity is set to 0.1S, and the period of the static cycle process is set to 60min. During the experiment, the battery is discharged with a 1C pulse, and the capacity of each discharge of the battery is 1Ah, so as to obtain the OCV-SCO data. The acquired experimental data is added to Matlab, and the fitting toolbox cftool is used to obtain the relationship curve between OCV and SOC. The recursive least squares method (RLS) is used to obtain the corresponding relationship between OCV and SOC. It is added to the Matlab simulation software, and the corresponding curve is obtained with the help of the fitting toolbox cftool, as shown in the figure. Figure 2 As shown, the expression of the OCV-SOC relationship curve is as follows:

[0077] U OC =F(SOC)=31.26SOC 6 ―98.19SOC 5 +120.1SOC 4―70.56SOC 3 +19.07SOC 2 ―13.09SOC+3.15.

[0078] The traditional least squares algorithm (LS) will recalculate the least squares solution of the entire data set each time the data is used for parameter estimation. This results in the LS method consuming a lot of computing resources and taking longer time for problems involving a large amount of data. Therefore, the identification process of the terminal voltage identification of the model using the recursive least squares method in this embodiment is as follows:

[0079] The following formula is used to calculate the battery parameter identification results:

[0080] In the formula, is the observed parameter matrix of the battery; θ is the parameter matrix, expressed as θ=[R0 R1 R2 C1 C2]; where R0 is the ohmic internal resistance of the lithium battery, R1 and R2 are the polarization resistances of the lithium battery, and C1 and C2 are the polarization capacitances of the lithium battery; e(k) is the error value.

[0081] The parameter matrix is ​​optimized using the following method:

[0082] Calculate the gain value K(k+1) of the current convergence number k+1:

[0083]

[0084] Where P(k) converges to the error covariance matrix with k times, is the observation data matrix of the current convergence number k+1;

[0085] Calculate the parameter error e(k+1) of the current convergence number k+1 of the equivalent circuit model:

[0086]

[0087] In the formula, y (k+1) is the output value of the system at the current convergence number k+1, θ(k) represents the parameter matrix of the convergence number k, Represents the observation value of the current convergence number k+1;

[0088] The battery model parameter matrix θ(k+1) of the current convergence number k+1 is obtained using the following formula:

[0089] θ(k+1)=θ(k)+K(k+1)e(k+1).

[0090] Among them, the calculation formula of the error covariance matrix P(k+1) of the current convergence number k+1 is as follows:

[0091] Where I is the unit matrix, P(k) is the error covariance matrix with convergence times k, K(k+1) is the gain value of the current convergence times k+1, is the observation data matrix of the current convergence number k+1.

[0092] Combined with the above method reference Figure 3 , a constant current pulse discharge experiment was carried out, and the actual terminal voltage measured was compared with the output terminal voltage obtained by model parameter identification. It can be clearly seen that the terminal voltage tracking effect of the second-order RC equivalent circuit model battery is good, which can characterize the terminal voltage of the battery when it is working. The curve of the actual terminal voltage measured in the experiment is basically consistent with the output terminal voltage curve obtained by model parameter identification; the absolute average error of the terminal voltage is only 0.0056V. This is because at the end of the discharge, the battery terminal voltage changes dramatically, and the tracking effect lags, resulting in an increase in the model error, but the overall error is still less than 0.05V.

[0093] The least squares algorithm used in this embodiment only needs to update the value of the previous parameter estimate each time new data is received, and there is no need to recalculate the least squares solution of the entire data set. In view of the uncertainty of system parameters, it can use known data according to the recursive formula to obtain the unknown parameter values. It has a certain correction ability itself, which makes the model parameter identification value more accurate, and can also solve the problem of model parameter uncertainty.

[0094] In another embodiment, the sampling points are found using unscented transformation, and the dimension of the state variable x in the state space equation is set to n, then;

[0095]

[0096] In the formula, is the Sigma point, is the optimal value of the state variable that converges k times, i represents the number of sampling points, and λ is the scaling factor, which is calculated as follows: λ=α 2 (n+k)―n, α is 10 ―4 An integer between 0 and 1;

[0097] Use the following formula to calculate the mean weight of Sigma points Covariance weights Initial mean and the initial covariance

[0098] Where, β≥0;

[0099] Define the state space equation as y = f(x), then substitute the Sigma point into y = f(x), and the estimated value of the system state variable that converges k+1 times is as follows:

[0100] y i =f(x i ), i = 0 to 2n;

[0101] In the formula, Represents the mean value of the state variable from the kth convergence to the k+1th convergence at sampling point i;

[0102] In the formula, Represents the optimal value of the state variable of sampling point i from convergence k to convergence k+1.

[0103] Predict the error covariance matrix that converges k+1 times:

[0104]

[0105] In the formula, Q k+1 is the system noise covariance matrix that converges k+1 times;

[0106] Substitute the predicted value point set of the state quantity that converges k+1 times into the observation equation, and use the observation equation to obtain the observed predicted value that converges k+1 times Perform a state transfer in the observation equation, and then weight it to get the mean of the converged k+1 measurements

[0107] Calculate the variance matrix P of the measurements that converge k+1 times xy，k+1 :

[0108]

[0109] Calculate the covariance matrix P of the state quantity and measurement value that converges k+1 times yy，k+1 :

[0110]

[0111] Calculate the Kalman filter gain K k+1 :K k+1 =P xy，k+1 / P yy，k+1 ;

[0112] Update system status variable values

[0113] Error variance matrix update P x,k+1|k+1 : In the formula, is the optimal value of the state variable when the variable x converges k+1 times, y (k+1) is the output value of the convergence k+1 times, is the optimal output value of the system with convergence number k+1.

[0114] In summary, combined with Figure 4 and Figure 5 Aiming at the goal of accurate SOC estimation of lithium-ion batteries, the present invention comprehensively considers the estimation accuracy, computational complexity and algorithm stability, and proposes a SOC estimation method for lithium-ion battery packs based on unscented Kalman filtering. Based on the full consideration of the working principle of lithium-ion batteries, the method combines the establishment of a second-order RC equivalent circuit model, realizes iterative calculation of SOC estimation of lithium-ion battery packs, and provides a basis for SOC estimation of lithium-ion batteries and real-time monitoring of working status.

[0115] Example 2

[0116] This embodiment discloses a battery SOC estimation system based on unscented Kalman filtering, which is used to implement the battery SOC estimation method described in Embodiment 1, including:

[0117] The first module is configured to establish a second-order RC equivalent circuit model, and establish a state space equation of the lithium battery based on the second-order RC equivalent circuit model;

[0118] The second module is configured to conduct a discharge-stationary experiment to determine the relationship between the open circuit voltage OCV and the state of charge SOC of the lithium battery at different discharge stages, and to obtain an OCV-SOC relationship curve by fitting;

[0119] The third module is configured to calculate a voltage estimate based on the state space equation, identify the terminal voltage of the model using a recursive least squares method, and obtain a terminal voltage identification curve to verify the accuracy of the model; when the lithium battery is discharged at a constant current, the open circuit voltage and current of the lithium battery are sampled to obtain an initial value of the open circuit voltage;

[0120] The fourth module is configured to use the unscented Kalman filter algorithm to continuously optimize the state-space equation, complete the continuous prediction and correction of the OCV-SOC relationship curve, and realize the continuous convergence of the state-of-charge SOC estimated value to the state-of-charge SOC actual value.

Claims

1. A battery SOC estimation method based on unscented Kalman filtering, characterized in that: The following steps are involved: Establishing a second-order RC equivalent circuit model, and establishing a state space equation of a lithium battery based on the second-order RC equivalent circuit model; A discharge-stationary experiment is conducted to determine the relationship between the open circuit voltage OCV and the state of charge SOC of the lithium battery at different discharge stages, and the OCV-SOC relationship curve is obtained by fitting; The voltage estimation value is calculated based on the state space equation, and the terminal voltage of the model is identified by using the recursive least square method to obtain the terminal voltage identification curve to verify the accuracy of the model; When the lithium battery is discharged at a constant current, the open circuit voltage and current of the lithium battery are sampled to obtain the initial value of the open circuit voltage; The unscented Kalman filter algorithm is used to continuously optimize the state space equation, complete the continuous prediction and correction of the OCV-SOC relationship curve, and realize the continuous convergence of the state of charge SOC estimated value to the state of charge SOC actual value; the expression of the OCV-SOC relationship curve is as follows: U OC =F(SOC)=31.26SOC 6 -98.19SOC 5 +120.1SOC 4 -70.56SOC 3 +19.07SOC 2 -13.09SOC+3.15; Using the unscented Kalman filter algorithm, the process of continuously optimizing the state space equation is as follows: Use unscented transformation to find the sampling points. In the state space equation, let the dimension of the state variable x be n, then; In the formula, is the Sigma point, is the optimal value of the state variable that converges k times, i represents the number of sampling points, and λ is the scaling factor, which is calculated as follows: λ=α 2 (n+k)-n, α is 10 -4 An integer between 0 and 1; Use the following formula to calculate the mean weight of Sigma points Covariance weights Initial mean and the initial covariance Define the state space equation as y = f(x), then substitute the Sigma point into y = f(x), and the estimated value of the system state variable that converges k+1 times is as follows: y i =f(x i ),i=0~2n; In the formula, Represents the mean value of the state variable from the kth convergence to the k+1th convergence at sampling point i; In the formula, Represents the optimal value of the state variable of sampling point i from convergence k to convergence k+1.

2. A battery SOC estimation method based on unscented Kalman filtering according to claim 1, characterized in that: The state space equation is as follows: It also includes an observation equation established based on the observation established by the state space equation, and the observation equation is expressed as follows: y(k)=[-1 -1 0]x(k)+[-R0]I(k)+v(k); Where x(k) is the system input value of convergence number k, and the state variable x = [U1U2SOC] T , y(k) is the output value of the system with convergence number k, Δt is the sampling time, R0 is the ohmic internal resistance of the lithium battery, R1 and R2 are the polarization resistances of the lithium battery, C1 and C2 are the polarization capacitances of the lithium battery, U1 is the voltage across R1, U2 is the voltage across R2, ω(k) is the system process noise with convergence number k, I(k) is the sampling current with convergence number k, v(k) is the observation noise with convergence number k, q N is the capacitance value to be calculated after discretization, and k is the number of convergences.

3. The battery SOC estimation method based on unscented Kalman filtering according to claim 1 is characterized in that: The identification process of terminal voltage identification of the model using recursive least squares method is as follows: The following formula is used to calculate the battery parameter identification results: In the formula, is the observed parameter matrix of the battery; θ is the parameter matrix, expressed as θ=[R0 R1 R2 C1 C2]; where R0 is the ohmic internal resistance of the lithium battery, R1 and R2 are the polarization resistances of the lithium battery, and C1 and C2 are the polarization capacitances of the lithium battery; e(k) is the error value.

4. The battery SOC estimation method based on unscented Kalman filtering according to claim 3 is characterized in that: The parameter matrix is ​​optimized using the following method: Calculate the gain value K(k+1) of the current convergence number k+1: Where P(k) converges to the error covariance matrix with k times, is the observation data matrix of the current convergence number k+1; Calculate the parameter error e(k+1) of the current convergence number k+1 of the equivalent circuit model: In the formula, y (k+1) is the output value of the system at the current convergence number k+1, θ(k) represents the parameter matrix of the convergence number k, Represents the observation value of the current convergence number k+1; The battery model parameter matrix θ(k+1) of the current convergence number k+1 is obtained using the following formula: θ(k+1)=θ(k)+K(k+1)e(k+1).

5. The battery SOC estimation method based on unscented Kalman filtering according to claim 1 is characterized in that: The following steps are also included: Predict the error covariance matrix that converges k+1 times: In the formula, Q k+1 is the system noise covariance matrix that converges k+1 times; Substitute the predicted value point set of the state quantity that converges k+1 times into the observation equation, and use the observation equation to obtain the observed predicted value that converges k+1 times Perform a state transfer in the observation equation, and then weight it to get the mean of the converged k+1 measurements Calculate the variance matrix P of the measurements that converge k+1 times xy , k+1 : Calculate the covariance matrix P of the state quantity and measurement value that converges k+1 times yy , k+1 : Calculate the Kalman filter gain K k+1 :K k+1 =P xy , k+1 / P yy , k+1 ; Update system status variable values Error variance matrix update P x,k+1|k+1 : In the formula, is the optimal value of the state variable when the variable x converges k+1 times, y (k+1) is the output value of the convergence k+1 times, is the optimal output value of the system with convergence number k+1.

6. The battery SOC estimation method based on unscented Kalman filtering according to claim 4 is characterized in that: The calculation formula of the error covariance matrix P(k+1) of the current convergence number k+1 is as follows: Where I is the unit matrix, P(k) is the error covariance matrix with convergence times k, K(k+1) is the gain value of the current convergence times k+1, is the observation data matrix of the current convergence number k+1.

7. A battery SOC estimation system based on unscented Kalman filtering, used to implement the battery SOC estimation method according to any one of claims 1 to 6, characterized in that: include: The first module is configured to establish a second-order RC equivalent circuit model, and establish a state space equation of the lithium battery based on the second-order RC equivalent circuit model; The second module is configured to conduct a discharge-stationary experiment to determine the relationship between the open circuit voltage OCV and the state of charge SOC of the lithium battery at different discharge stages, and to obtain an OCV-SOC relationship curve by fitting; The third module is configured to calculate a voltage estimate based on the state space equation, identify the terminal voltage of the model using a recursive least squares method, and obtain a terminal voltage identification curve to verify the accuracy of the model; when the lithium battery is discharged at a constant current, the open circuit voltage and current of the lithium battery are sampled to obtain an initial value of the open circuit voltage; The fourth module is configured to use the unscented Kalman filter algorithm to continuously optimize the state-space equation, complete the continuous prediction and correction of the OCV-SOC relationship curve, and realize the continuous convergence of the state-of-charge SOC estimated value to the state-of-charge SOC actual value.

Images