Battery SOC estimation method based on NYSR-SVMR model

Abstract

The invention discloses a battery SOC estimation method based on an NYSR-SVMR model. The method comprises the following steps: S1, collecting voltage, current and battery SOC information of a battery under each working condition; s2, a training set D = (xi, yi) is selected from the collected information, i = 1, 2,..., n, xi is a set of measured values of current, voltage and temperature of the ith sample battery, yi is a battery SOC value of the ith sample battery, and n represents the number of samples; s3, carrying out normalization processing on the selected data; s4, initializing the model, and training the model; and S5, verifying the model. The estimation performance of the SOC of the battery can be effectively improved, and the estimation precision of the SOC of the battery is improved.

Description

Technical Field

[0001] The present invention relates to the field of lithium battery calculation technology, and in particular to a battery SOC estimation method based on a NYSR-SVMR model. Background Art

[0002] The battery's SOC (State-of-Charge) refers to the state of the battery's remaining power. The accuracy of calculating the battery's remaining power has always been a difficult problem in the industry. In particular, the battery will wear out after use, and the battery's capacity will decrease, resulting in an increasing error in the calculated SOC value.

[0003] The general methods of SOC estimation can be divided into three categories, namely parameter measurement method, model-based method and data-driven method. Parameter measurement methods include open circuit voltage method, coulomb counting method, impedance measurement method, etc. The parameter measurement method is simple and easy to implement, but it has disadvantages such as low practicality and poor estimation accuracy. Model-based methods rely on the understanding and mathematical modeling of the physical and chemical processes of the battery system, including Kalman filtering method, particle filtering method and sliding mode observer method. This type of method has high estimation accuracy, but the model establishment is relatively complex, and the accuracy is highly dependent on the accuracy of the model. Data-driven methods include neural network method and support vector machine method. Although this type of method has accurate estimation results, it has a large amount of calculation and requires high chip computing processing capabilities. Therefore, there is an urgent need for an estimation method that can improve the estimation performance of battery SOC. Summary of the invention

[0004] The present invention aims to solve the technical problems existing in the prior art, and in particular innovatively proposes a battery SOC estimation method based on the NYSR-SVMR model, which can effectively improve the estimation performance of the battery SOC and improve the estimation accuracy of the battery SOC.

[0005] In order to achieve the above object, the present invention provides a battery SOC estimation method based on the NYSR-SVMR model, comprising the following steps:

[0006] S1: Collect the battery voltage, current and battery SOC information under various working conditions;

[0007] S2: Select the training set D from the collected information = (x i ,y i ),i=1,2,···,n, where x i is the set of measured values ​​of current, voltage and temperature of the i-th sample battery, y i is the battery SOC value of the i-th sample, and n represents the number of samples;

[0008] S3: normalize the selected data;

[0009] S4: Initialize the model and train the model;

[0010] S5: Validate the model.

[0011] In the above scheme: step S4 also includes the following steps:

[0012] S4-1: Set the model as follows:

[0013]

[0014] Where y(t) is the estimated value of the battery SOC, x is the set of measured values ​​of the battery current, voltage and temperature; ω T is the coefficient, b is the deviation value;

[0015] S4-2: Accelerate the training speed of the support vector machine model through the Nystrom approximation algorithm to obtain NYSR-SVMR;

[0016] S4-3: Substitute the selected training set D into the set model and train the support vector machine model.

[0017] In the above scheme: step S4-3 also includes the following steps:

[0018] S4-3-1: Calculate using the following formula:

[0019]

[0020] Among them, J p (·) is the objective function, To solve J p (·) Minimum parameters ω, b, ξ i , ξ i and is the introduced slack variable, c is the penalty coefficient, ε is the approximate accuracy, and the constraints are as follows:

[0021]

[0022] Among them, y i is the battery SOC value of the i-th sample in the training set D;

[0023] S4-3-2: Simplify the formula of step S4-3-1;

[0024] The equality constraint and the two-norm loss function are simplified as follows:

[0025]

[0026] Among them, x iis the set of measured values ​​of the battery current, voltage and temperature of the i-th sample in the training set D; To transform the sample x i Mapping The mapping function is e i is the error variable, γ is the regularization parameter;

[0027] S4-3-3: Construct the Lagrangian function as follows:

[0028]

[0029] Among them, α i is the Lagrange multiplier, For sample x i The mapping function, J p (·) is the objective function, y i is the battery SOC value of the i-th sample in the training set D;

[0030] S4-3-4: Optimize the Lagrangian function according to the optimization conditions;

[0031] S4-3-5: Assume the kernel function is as follows:

[0032]

[0033] K is the kernel function, For x j The mapping function, x j is the set of measured values ​​of the current, voltage and temperature of the battery of the jth sample in the training set D;

[0034] S4-3-6: Substitute the function in step S4-3-5 into the Lagrangian function optimized in step S4-3-4 for calculation, and obtain the following equation:

[0035]

[0036] Where E = {1,1,···,1}∈R n ,I∈R n ×R n is the unit matrix, α={α1,α2,···,α n} T ∈R n ;

[0037] y={y1,y2,···,y n} T ∈R n , K∈R n ×R n is the kernel matrix formed by the kernel function, whose constituent elements Kij =K(x i ,x j ), where i,j∈[1,2,···,n];

[0038] S4-3-7: Solve the equation in S4-3-6 to obtain the parameters α and b, and construct the prediction model in the dual space as follows:

[0039]

[0040] In the above scheme: step S4-2 also includes the following steps:

[0041] S4-2-1: According to x i The number of samples to construct the kernel matrix; since x i The number of samples is n, then the kernel matrix K NN The scale is N×N;

[0042] S4-2-2: From the kernel matrix K NN If M samples are drawn from the kernel matrix K of M samples, MM The scale is M×M;

[0043] Then the kernel matrix K MM The eigenvector matrix is ​​as follows:

[0044] in is the eigenvector matrix, is the diagonal matrix of eigenvalues,

[0045]

[0046] Then the kernel matrix K NN The eigenvector matrix is ​​as follows:

[0047] in is the eigenvector matrix, is the diagonal matrix of eigenvalues,

[0048]

[0049] S4-2-3: Decompose the vector;

[0050] S4-2-4: Calculate using the following formula:

[0051]

[0052] Where K(x k ,x l ) is the kernel matrix K NN The element in the kth row and lth column, k,l=1,L,N ,λ i and φ i are the eigenvalues ​​and the corresponding eigenfunctions, respectively;

[0053] S4-2-5: The formula in step S4-2-4 is transformed by integral transformation. The transformed formula is as follows:

[0054] ∫K(x k ,x)φ i (x)P(x)dx=λ i φ i (x k )

[0055] Where P(x) is the input variable x k The probability density function, λ i and φ i are the eigenvalues ​​and the corresponding eigenfunctions, respectively;

[0056] S4-2-6: Extract M samples X from N samples s ,s=1,2,···,M, and set the sample X s The characteristic function of is as follows:

[0057]

[0058] S4-2-7: Combine the vector decomposed in step S4-2-3 with the function in step S4-2-6 to obtain a high-dimensional vector The relationship between the i-th element and the i-th characteristic function is as follows:

[0059]

[0060] Where M is the number of samples drawn, is the eigenvalue λ i The corresponding estimated value, K(x s ,x k ) is the matrix K MM The element at row s and column k in for The eigenvector of the sth row and ith column;

[0061] And obtain The estimated values ​​are as follows:

[0062] S4-2-8: The value obtained in step S4-2-7 is The estimated value of is substituted into the prediction model constructed in S4-3-7, and the prediction model is rewritten as follows:

[0063]

[0064] The original space model parameters ω and b are obtained by the following formula:

[0065]

[0066] In the above scheme: step S4-3-4 includes the following steps:

[0067] The Karush-Kuhn-Tucker conditions are as follows:

[0068]

[0069] Based on the Karush-Kuhn-Tucker condition, we can get:

[0070]

[0071] In the above scheme: step S4-2-3 also includes the following steps:

[0072] K NN The expression is as follows:

[0073]

[0074] K MM The expression is as follows:

[0075]

[0076] The expression after decomposing the vector is as follows:

[0077]

[0078] Among them, λ i and φ i are the eigenvalues ​​and the corresponding eigenfunctions, respectively. is the characteristic function φ i The corresponding estimated value is is the eigenvalue λ i The corresponding estimated value.

[0079] In the above scheme: step S5 includes the following steps:

[0080] S5-1: Grouping the data information collected in step S1 into a training set D and a test set X, where the test set X is the measured values ​​of battery voltage and current;

[0081] S5-2: Test the model trained in step S4 according to the test set X.

[0082] In the above scheme: step S5-2 also includes the following steps:

[0083] S5-2-1: Calculate the root mean square error value using the following formula:

[0084]

[0085] Among them, RMSE is the root mean square error, N is the data length, y i is the battery SOC value of the i-th sample in the training set D, is the estimated value of SOC;

[0086] S5-2-2: Calculate the maximum absolute error value using the following formula:

[0087]

[0088] Among them, MAE is the maximum absolute error, N is the data length, y i is the battery SOC value of the i-th sample in the training set D, y$ i is the estimated SOC value.

[0089] In summary, the beneficial effects of the present invention are: the original SVMR model can be directly solved by the Nystrom algorithm, the computational efficiency can be improved, the training time can be reduced, the computational resources can be saved, the battery SOC estimation performance can be effectively improved, and the battery SOC estimation accuracy can be improved. The training set and the test set are set to test the learned model to verify the modeling ability of the learning machine to ensure the prediction performance and estimation accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] Figure 1 This is the current change diagram of INR18650 lithium battery under DST conditions.

[0091] Figure 2 This is the voltage change diagram of INR18650 lithium battery under DST conditions.

[0092] Figure 3 This is the SOC change diagram of INR18650 lithium battery under DST conditions.

[0093] Figure 4 It is the estimated SOC value of INR18650 lithium battery based on LS-SVMR method.

[0094] Figure 5 is the estimated variance of INR18650 lithium battery based on LS-SVMR method.

[0095] Figure 6 It is the estimated SOC value of the INR18650 lithium battery under the estimation method of this technical solution.

[0096] Figure 7 is the estimated variance of INR18650 lithium battery in the estimation method of this technical solution. DETAILED DESCRIPTION

[0097] The present invention will be further described below by way of embodiments and in conjunction with the accompanying drawings:

[0098] The present invention provides a battery SOC estimation method based on the NYSR-SVMR model, comprising the following steps:

[0099] S1: Collect the battery voltage, current and battery SOC estimation information under various working conditions;

[0100] S2: Select the training set D from the collected information = (x i ,y i ),i=1,2,···,n, where x i is the set of measured values ​​of current, voltage and temperature of the i-th sample battery, y i is the battery SOC value of the i-th sample, and n represents the number of samples;

[0101] S3: Normalize the selected data.

[0102] S4: Initialize the model and train the model;

[0103] S4-1: Set the model as follows:

[0104]

[0105] Where, y(t) is the estimated value of the battery SOC, x is the set of measured values ​​of the battery current, voltage and temperature; ω T is the coefficient, b is the deviation value;

[0106] S4-2: Calculate the original space of LS-SVMR model through NYSR-SVMR

[0107] S4-2-1: According to x i The number of samples to construct the kernel matrix; since x i The number of samples is n, then the kernel matrix K NN The scale is N×N, where n=N;

[0108] S4-2-2: From the kernel matrix K NN If M samples are drawn from the kernel matrix K of M samples, MM The scale is M×M;

[0109] Then the kernel matrix K MM The eigenvector matrix is ​​as follows:

[0110] in is the eigenvector matrix, is the diagonal matrix of eigenvalues,

[0111]

[0112] Then the kernel matrix K NN The eigenvector matrix is ​​as follows:

[0113] in is the eigenvector matrix, is the diagonal matrix of eigenvalues,

[0114]

[0115] S4-2-3: Decompose the vector;

[0116] K NN The expression is as follows:

[0117]

[0118] K MM The expression is as follows:

[0119]

[0120] right The decomposition vector is as follows:

[0121]

[0122] Among them, λ i and φ i are the eigenvalues ​​and the corresponding eigenfunctions, respectively. is the characteristic function φ i The corresponding estimated value is is the eigenvalue λ i The corresponding estimated value;

[0123] S4-2-4: Calculate using the following formula:

[0124]

[0125] Where K(x k ,x l ) is the kernel matrix K NN The k-th row and l-th column element, k, l = 1, L, N, λ i and φ i are the eigenvalues ​​and the corresponding eigenfunctions, respectively;

[0126] S4-2-5: The formula in step S4-3-4 is transformed by integral transformation. The transformed formula is as follows:

[0127] ∫K(x k ,x)φ i (x)P(x)dx=λ i φ i (x k )

[0128] Where P(x) is the input variable x k The probability density function, λ i and φ i are the eigenvalues ​​and the corresponding eigenfunctions, respectively;

[0129] S4-2-6: Extract M samples X from N samples s ,s=1,2,···,M, and set the sample X s The characteristic function of is as follows:

[0130]

[0131] S4-2-7: Combine the vector decomposed in step S4-2-3 with the function in step S4-2-6 to obtain a high-dimensional vector The relationship between the i-th element and the i-th characteristic function is as follows:

[0132]

[0133] Where M is the number of samples drawn, is the eigenvalue λ i The corresponding estimated value, K(x s ,x k ) is the matrix K MM The element in the sth row and kth column of for The eigenvector of the sth row and ith column;

[0134] And obtain The estimated values ​​are as follows:

[0135] S4-2-8: The value obtained in step S4-2-7 is The estimated value of is substituted into the prediction model constructed in S4-2-7, and the prediction model is rewritten as follows:

[0136]

[0137] The original space model parameters ω and b are obtained by the following formula:

[0138]

[0139] S4-3: Substitute the selected training set D into the set model and train the model through the support vector machine;

[0140] S4-3-1: Calculate using the following formula:

[0141]

[0142] Among them, J p (·) is the objective function, To solve J p (·) Minimum parameters ω, b, ξ i , ξ i and is the introduced slack variable, c is the penalty coefficient, ε is the approximate accuracy, and the constraints are as follows:

[0143]

[0144] Among them, y i is the battery SOC value of the i-th sample in the training set D;

[0145] Slack variables are a technique used to solve the linear inseparable problem. In the case of linear inseparable, SVMR allows some data points to be on the wrong side of the hyperplane and tolerates these errors by introducing slack variables. These slack variables allow data points to have a certain distance between the wrong side and the hyperplane, thereby improving the robustness of the model; c is the penalty coefficient, also known as the regularization parameter, which controls the degree of fit of the model to the training data and the degree of penalty for complexity. A larger penalty coefficient will cause the model to fit the training data more strictly, which may produce a simpler model and avoid overfitting, but may cause a larger training error. A smaller penalty coefficient will allow the model to fit the training data more flexibly, which may lead to a more complex model, which may overfit, but the training error will be smaller. Therefore, the choice of penalty coefficient needs to be adjusted according to the specific problem and data set to strike a balance between bias and variance to obtain the best generalization performance; ε is the approximate accuracy, which is one of the indicators used to evaluate the performance of the support vector machine model. It usually refers to the performance of the model on data outside the training set. A higher approximate accuracy means that the model performs better on unseen data and has better generalization ability. S4-3-2: The formula of step S4-2-1 is simplified as follows by using equality constraints and the second norm as the loss function:

[0146]

[0147] Among them, x iis the set of measured values ​​of the battery current, voltage and temperature of the i-th sample in the training set D, To transform the sample x i Mapping The mapping function is e i is the error variable, γ is the regularization parameter;

[0148] S4-3-3: Construct the Lagrangian function as follows:

[0149]

[0150] Among them, α i is the Lagrange multiplier, For sample x i The mapping function, J p (·) is the objective function, y i is the battery SOC value of the i-th sample in the training set D;

[0151] S4-3-4: Optimize according to the optimization conditions;

[0152] The optimization conditions are as follows:

[0153]

[0154] The optimized Lagrangian function is as follows:

[0155]

[0156] S4-3-5: Assume the kernel function is as follows:

[0157]

[0158] K is the kernel function, For x j The mapping function, x j is the set of measured values ​​of the current, voltage and temperature of the battery of the jth sample in the training set D;

[0159] S4-3-6: Substitute the function in step S4-3-5 into the Lagrangian function optimized in step S4-3-4 for calculation, and obtain the following equation:

[0160]

[0161] Where E = {1,1,···,1}∈R n ,I∈R n ×R n is the unit matrix, α={α1,α2,···,α n} T ∈Rn ;

[0162] y={y1,y2,···,y n} T ∈R n ,K∈R n ×R n is the kernel matrix formed by the kernel function, whose constituent elements K ij =K(x i ,x j ), where i,j∈[1,2,···,n];

[0163] S4-3-7: Solve the equation in S4-3-6 to obtain the parameters α and b, and construct the prediction model in the dual space as follows:

[0164]

[0165] S5: Validate the model;

[0166] S5-1: Grouping the data information collected in step S1 into a training set D and a test set X, where the test set X is the measured values ​​of battery voltage and current;

[0167] S5-2: Test the model trained in step S4 according to the test set X.

[0168] S5-2-1: Calculate the root mean square error value using the following formula:

[0169]

[0170] Among them, RMSE is the root mean square error, N is the data length, y i is the battery SOC value of the i-th sample in the training set D, is the estimated value of SOC;

[0171] S5-2-2: Calculate the maximum absolute error value using the following formula:

[0172]

[0173] Among them, MAE is the maximum absolute error, N is the data length, y i is the battery SOC value of the i-th sample in the training set D, is the estimated SOC value.

[0174] In order to verify the effectiveness of the method proposed in this technical solution, the following is a verification using the charge and discharge data of a INR18650 lithium battery under dynamic stress test (DST) conditions. The rated capacity of the selected INR18650 lithium battery is 2000mAh and the rated voltage is 3.7V. The selected battery charge and discharge data comes from the open source data of UNIVERSITY OF MARYLAND. The open source data is normalized and 1062 groups are selected for testing with a sampling interval of 10s. The current change curve, voltage change curve and battery SOC change curve in the selected data are as follows Figures 1 to 3 shown.

[0175] In addition, the SOC estimation based on the LS-SVMR method is added as a comparative experiment to better verify the reliability and superiority of the proposed algorithm. The battery SOC estimation value and estimation error obtained through the experiment are shown in the figure below: Figures 4 to 7 shown.

[0176] Figure 4 is the SOC curve estimated by LS-SVMR method, Figure 5 The error is estimated by Figure 4 , 5 It can be seen that when using the LS-SVMR method for SOC estimation, the estimated value can track the true value well, but there is an obvious error.

[0177] By comparing the performance indicators of the battery SOC value estimated using each model, the performance comparison of the battery SOC estimated by each model is shown in Table 1.

[0178] Table 1 Comparison of battery SOC performance estimated by each model

[0179]

[0180] pass Figures 4 to 7 As can be seen from Table 1, compared with the LS-SVMR method for estimating battery SOC, this technical solution has smaller RMSE, MAE and maximum error when estimating battery SOC, so this technical solution has better performance and can effectively improve the estimation performance of battery SOC and improve the estimation accuracy of battery SOC.

Claims

1. A battery SOC estimation method based on the NYSR-SVMR model, characterized in that: The following steps are involved: S1: Collect the battery voltage, current and battery SOC information under various working conditions; S2: Select the training set D from the collected information = (x i ,y i ),i=1,2,···,n, where x i is the set of measured values ​​of current, voltage and temperature of the i-th sample battery, y i is the battery SOC value of the i-th sample, and n represents the number of samples; S3: normalize the selected data; S4: Initialize the model and train the model; S5: Validate the model.

2. The battery SOC estimation method based on the NYSR-SVMR model according to claim 1, characterized in that: Step S4 also includes the following steps: S4-1: Set the model as follows: Where y(t) is the estimated value of the battery SOC, x is the set of measured values ​​of the battery current, voltage and temperature; ω T is the coefficient, b is the deviation value; S4-2: Accelerate the training speed of the support vector machine model through the Nystrom approximation algorithm to obtain NYSR-SVMR; S4-3: Substitute the selected training set D into the set model and train the support vector machine model.

3. The battery SOC estimation method based on the NYSR-SVMR model according to claim 2 is characterized in that: Step S4-3 also includes the following steps: S4-3-1: Calculate using the following formula: Among them, J p (·) is the objective function, To solve J p (·) Minimum parameters ξ i and is the introduced slack variable, c is the penalty coefficient, ε is the approximate accuracy, and the constraints are as follows: Among them, y i is the battery SOC value of the i-th sample in the training set D; S4-3-2: Simplify the formula of step S4-3-1; The equality constraint and the two-norm loss function are simplified as follows: Among them, x i is the set of measured values ​​of the current, voltage and temperature of the battery of the i-th sample in the training set D; To transform the sample x i Mapping The mapping function is e i is the error variable, γ is the regularization parameter; S4-3-3: Construct the Lagrangian function as follows: Among them, α i is the Lagrange multiplier, For sample x i The mapping function, J p (·) is the objective function, y i is the battery SOC value of the i-th sample in the training set D; S4-3-4: Optimize the Lagrangian function according to the optimization conditions; S4-3-5: Assume the kernel function is as follows: K is the kernel function, For x j The mapping function, x j is the set of measured values ​​of the current, voltage and temperature of the battery of the jth sample in the training set D; S4-3-6: Substitute the function in step S4-3-5 into the Lagrangian function optimized in step S4-3-4 for calculation, and obtain the following equation: Where E = {1,1,···,1}∈R n ,I∈R n ×R n is the unit matrix, α={α1,α2,···,α n } T ∈R n ; y={y1,y2,···,y n } T ∈R n , K∈R n ×R n is the kernel matrix formed by the kernel function, whose constituent elements K ij =K(x i ,x j ), where i,j∈[1,2,···,n]; S4-3-7: Solve the equation in S4-3-6 to obtain the parameters α and b, and construct the prediction model in the dual space as follows:

4. The battery SOC estimation method based on the NYSR-SVMR model according to claim 1, characterized in that: Step S4-2 also includes the following steps: S4-2-1: According to x i The number of samples to construct the kernel matrix; since x i The number of samples is n, then the kernel matrix K NN The scale is N×N; S4-2-2: From the kernel matrix K NN If M samples are drawn from the kernel matrix K of M samples, MM The scale is M×M; Then the kernel matrix K MM The eigenvector matrix is ​​as follows: in is the eigenvector matrix, is the diagonal matrix of eigenvalues, Then the kernel matrix K NN The eigenvector matrix is ​​as follows: in is the eigenvector matrix, is the diagonal matrix of eigenvalues, S4-2-3: Decompose the vector; S4-2-4: Calculate using the following formula: Where K(x k ,x l ) is the kernel matrix K NN The element in the kth row and lth column, k,l=1,L,N,λ i and φ i are the eigenvalues ​​and the corresponding eigenfunctions, respectively; S4-2-5: The formula in step S4-3-4 is transformed by integral transformation. The transformed formula is as follows: ∫K(x k ,x)φ i (x)P(x)dx=λ i φ i (x k ) Where P(x) is the input variable x k The probability density function, λ i and φ i are the eigenvalues ​​and the corresponding eigenfunctions, respectively; S4-2-6: Extract M samples X from N samples s ,s=1,2,···,M, and set the sample X s The characteristic function of is as follows: S4-2-7: Combine the vector decomposed in step S4-2-3 with the function in step S4-2-6 to obtain a high-dimensional vector The relationship between the i-th element and the i-th characteristic function is as follows: Among them, M is the number of samples drawn, λ i is the eigenvalue λ i The corresponding estimated value, K(x s ,x k ) is the matrix K MM The element at row s and column k in for The eigenvector of the sth row and ith column; And obtain The estimated values ​​are as follows: S4-2-8: The value obtained in step S4-2-7 is The estimated value of is substituted into the prediction model constructed in S4-2-7, and the prediction model is rewritten as follows: The original space model parameters ω and b are obtained by the following formula:

5. The battery SOC estimation method based on the NYSR-SVMR model according to claim 3 is characterized in that: Step S4-2-4 includes the following steps: The Karush-Kuhn-Tucker conditions are as follows: Based on the Karush-Kuhn-Tucker condition, we can get:

6. The battery SOC estimation method based on the NYSR-SVMR model according to claim 4 is characterized in that: Step S4-2-3 also The following steps are involved: K NN The expression is as follows: K MM The expression is as follows: The expression after decomposing the vector is as follows: Among them, λ i and φ i are the eigenvalues ​​and the corresponding eigenfunctions, respectively. is the characteristic function φ i The corresponding estimated value, λ is the eigenvalue λ i The corresponding estimated value.

7. The battery SOC estimation method based on the NYSR-SVMR model according to claim 1, characterized in that: Step S5 includes the following steps: S5-1: Grouping the data information collected in step S1 into a training set D and a test set X, where the test set X is the measured values ​​of battery voltage and current; S5-2: Test the model trained in step S4 according to the test set X.

8. The battery SOC estimation method based on the NYSR-SVMR model according to claim 7, characterized in that: Step S5-2 also includes the following steps: S5-2-1: Calculate the root mean square error value using the following formula: Among them, RMSE is the root mean square error, N is the data length, y i is the battery SOC value of the i-th sample in the training set D, is the estimated value of SOC; S5-2-2: Calculate the maximum absolute error value using the following formula: Among them, MAE is the maximum absolute error, N is the data length, y i is the battery SOC value of the i-th sample in the training set D, is the estimated SOC value.

Images