Battery SOC estimation method based on rasterized wolf pack behavior algorithm-unscented Kalman filtering combined algorithm

Abstract

The invention provides a battery SOC estimation method based on a rasterized wolf pack behavior algorithm-unscented Kalman filtering combined algorithm. The method comprises the steps of performing a mixed pulse power characteristic test on a target battery to obtain current and voltage of the target battery; performing working condition test on the target battery to obtain current and voltage, and adding noise to the current and voltage to simulate a real sampling environment; a second-order RC equivalent battery model is initialized, and an SOC true value is calculated; the method comprises the following steps of: randomly generating an initial parameter group, optimizing the parameter group by using a rasterized wolf pack behavior algorithm, estimating by using an unscented Kalman filtering algorithm to verify an optimization effect, and combining the optimal parameter group with the unscented Kalman filtering algorithm to obtain an unscented Kalman filtering estimator after multiple rounds of iteration; and applying the unscented Kalman filter estimator to the SOC estimation of the subsequent battery. According to the method, the optimal unscented Kalman filter fitting parameter can be quickly and accurately found, and the SOC estimation precision of the battery is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of battery state estimation, and in particular to a method for estimating battery SOC based on a grid-based wolf pack behavior algorithm-unscented Kalman filter joint algorithm. Background Art

[0002] In order to achieve accurate estimation of battery SOC, researchers at home and abroad widely adopt algorithms based on Kalman filter. Since battery SOC estimation belongs to a non-linear system, unscented Kalman filter is usually used to handle its non-linear characteristics. In addition, batteries are often affected by external interferences in actual applications, such as current fluctuations and environmental noise, which pose higher requirements for the accuracy of SOC estimation.

[0003] The unscented Kalman filter algorithm has significant advantages in dealing with non-linear models, but its performance depends to a large extent on the accurate setting of parameters.

[0004] Traditional parameter optimization methods often have the defects of slow convergence speed and easy to fall into local optimal solutions, and are more sensitive to external noise, which limits the further improvement of SOC estimation accuracy.

[0005] Therefore, it is urgent to develop a more efficient and robust unscented Kalman filter estimator to optimize the estimation performance of SOC and enhance the reliability and efficiency of the battery management system in complex environments. Summary of the Invention

[0006] Aiming at the deficiencies of the prior art, the present invention provides a method for estimating battery SOC based on a grid-based wolf pack behavior algorithm-unscented Kalman filter joint algorithm. By introducing the grid-based wolf pack behavior algorithm and combining the idea of full cross of variables between individuals and local optimization, the present invention constructs a better unscented Kalman filter estimator, realizing accurate estimation of the battery state of charge and precise tracking of the working state.

[0007] The technical solution of the present invention is as follows: A method for estimating battery SOC based on a grid-based wolf pack behavior algorithm-unscented Kalman filter joint algorithm, comprising the following steps:

[0008] S1: Conduct a hybrid pulse power characteristic test on the target battery to obtain its current and voltage;

[0009] S2: Conduct a working condition test on the target battery to obtain its current and voltage, and add noise to the current and voltage to simulate a real sampling environment;

[0010] S3: Initialize the second-order RC equivalent battery model and calculate the true value of SOC;

[0011] S4: randomly generate an initial parameter group, use the gridded wolf pack behavior algorithm to optimize the parameter group, and use the unscented Kalman filter algorithm to estimate to verify the optimization effect. After multiple rounds of iterations, combine the optimal parameter group with the unscented Kalman filter algorithm to obtain the unscented Kalman filter estimator;

[0012] S5: Use the unscented Kalman filter estimator obtained in step S4 to estimate the SOC of the battery.

[0013] Preferably, in step S1, the specific steps of performing a mixed pulse power characteristic test on the target battery to obtain its current and voltage are as follows:

[0014] S1-1: Set 10 test points at 10% intervals between battery SOC 100% and 10%, and perform the following tests:

[0015] S1-1-1: Charge the target battery to the specified SOC point and let it stand for one hour to ensure that the test starts from a stable state;

[0016] S1-1-2: The target battery is discharged at a constant current of 75% of the maximum current of the battery for 1000s, then the charge and discharge is stopped for 400s, and then the constant current is charged at 75% of the maximum current of the battery for 1000s, and finally the battery is left to stand until the voltage is stable, and the test voltage and current data are recorded throughout the process;

[0017] S1-1-3: After each charge and discharge cycle, the battery is left to rest for 60 minutes and the battery voltage is recorded once. This voltage is considered to be the open circuit voltage of the battery at that SOC.

[0018] Preferably, in step S2, a working condition test is performed on the target battery to obtain its current and voltage, and noise is added to the current and voltage to simulate a real sampling environment. The specific steps are as follows:

[0019] S2-1: Set the temperature of the test environment within a standardized temperature range, ensure that the target battery is in a fully charged state, and let the battery stand at the test temperature for a certain period of time after being fully charged to achieve thermal equilibrium and the internal electrochemical reaction is relatively static;

[0020] S2-2: applying a predetermined pulse current of a corresponding bus dynamic street condition test condition to the battery and recording current and voltage data;

[0021] S2-3: Add random normally distributed noise of 1% of the rated maximum battery current to the current obtained in step S2-2, and add normally distributed noise of 1% of the rated voltage to the voltage to simulate external interference during actual sampling of the model.

[0022] Preferably, in step S3, the formula of the second-order RC equivalent battery model is as follows:

[0023] U out = U ocv - I 0 R 0 - U 1 - U 2 (13)

[0024]

[0025] Wherein, U out , U ocv represent the battery output voltage and the battery open-circuit voltage; I 0 is the main circuit current flowing through R 0 ; R 0 is the ohmic internal resistance, representing the internal contact resistance of the battery; U 1 is the voltage of the parallel connection of R 1 and C 1 , representing the electrochemical polarization voltage, and U 2 is the voltage of the parallel connection of R 2 and C 2 , representing the concentration polarization voltage; R 1 , R 2 , C 1 , C 2 respectively represent the resistances and capacitances of the two RC circuits.

[0026] Preferably, in step S3, the calculation expression of the true value of SOC is as follows:

[0027]

[0028] In formula (3), SOC original is the SOC of the target battery at the initial moment, t is the time, C represents the battery capacity, and I(t) is the charge and discharge current at time t.

[0029] Preferably, in step S3, using the data obtained from the hybrid pulse power characteristic test in step S1, the R 0 , R 1 , R 2 , C 1 , C 2 and OCV data of the second-order RC equivalent battery model of the target battery at each 10% test point between SOC 100% and 10% are obtained by the least squares method and substituted into the second-order RC equivalent battery model for model initialization.

[0030] Preferably, step S4 specifically includes the following steps:

[0031] S4-1: Determine the parameters in the system that need to be tuned and optimized, initialize these parameter pairs within the specified range, and call them Wolf Pack A;

[0032] S4-2: Initialize the second-order RC equivalent battery model and calculate the true value of the battery SOC using the test data from Step S2;

[0033] S4-3: Initialize the unscented Kalman filter algorithm, perform the first estimation on Wolf Pack A, and obtain the initial fitness of the wolf pack; Use the mean absolute error MAE calculated by the unscented Kalman filter as the fitness of the wolf pack for this parameter group;

[0034] S4-4: Use the grid-based wolf pack behavior algorithm to perform vertical reproduction on the individuals of Wolf Pack A;

[0035] S4-5: Simulate the migration and predation processes of the wolf pack and update the fitness of the wolf pack;

[0036] S4-6: Use the grid-based wolf pack behavior algorithm to perform horizontal reproduction on the individuals of Wolf Pack A;

[0037] S4-7: Repeat Steps S4-4 to S4-6 until the SOC estimation error is less than the specified value or the specified number of iterations is completed;

[0038] S4-8: Output the individual parameter group with the optimal fitness of the tuned wolf pack and its estimation error, and use this individual parameter group to perform subsequent battery SOC estimation for this system to obtain the unscented Kalman filter estimator.

[0039] Preferably, in Step S4-3, the calculation formula for the mean absolute error MAE is as follows:

[0040]

[0041] In Equation (4), t 0 is the total time of the unscented Kalman filter estimation, is the estimated value of the SOC at time i, x i is the true value of the SOC at time i.

[0042] Preferably, in Step S4-4, when using the grid-based wolf pack behavior algorithm to perform vertical reproduction on the individuals of Wolf Pack A, the specific steps are as follows:

[0043] S4-4-1: Read the existing wolf pack and its fitness;

[0044] S4-4-2: Perform wolf pack grid vertical reproduction, and its expression is:

[0045] M op (m, D 1 ) = a × X(m, D 1)+(1 - a)×X(m,D 2 )+b(X(m,D 1 ) - X(m,D 2 )) (17)

[0046] where a and b are random numbers between 0 and 1, X(m,D 1 ) and X(m,D 2 ) are the parent optimization parameters, M op (m,D 1 ) is the offspring optimization parameter generated by the vertical reproduction of the parent, the parameter m = 1, 2, …, N, N is the wolf pack size, D 1 ,D 2 = 1, 2, …, B, B is the number of dimensions of the wolf pack;

[0047] S4 - 4 - 3: Check whether the new individual meets the conditions and calculate the fitness of the new individual; use equations (6) and (7) to correct each dimension that exceeds the upper and lower bounds of the constraint range respectively, so that the wolf pack can approach the boundary more smoothly;

[0048]

[0049] where E D is the parameter to be adjusted in the D dimension, is the adjusted parameter in the D dimension, boundary up and boundary down are the upper and lower bounds of the parameter respectively;

[0050] S4 - 4 - 4: For the extreme case where the wolf pack crosses zero in a certain dimension, use equations (8) and (9) to approximate the upper and lower boundaries of the wolf pack in this dimension respectively, so that the wolf pack will not be exactly equal to the boundary;

[0051] E' D = boundary down ×a down (20)

[0052] E' D = boundary up ×a up (21)

[0053] where a up is a random number between 1 and 10, a down is a random number between 0.1 and 1;

[0054] S4 - 4 - 5: Judge whether it is better than the original wolf individual. If it is better, replace it; otherwise, keep it, so that the whole wolf pack develops in a better direction;

[0055] S4-4-6: Complete the vertical reproduction of the current grid.

[0056] Preferably, in step S4-5, the migration and predation processes of the wolf pack are simulated to update the fitness of the wolf pack. The specific steps are as follows:

[0057] S4-5-1: Read the existing wolf pack and its fitness.

[0058] S4-5-2: Select the best three wolf individuals as Alpha (α), Beta (β), and Delta (δ) according to the fitness values. These three wolf individuals will guide the movement of the wolf pack.

[0059] When the wolf pack migrates and preys, it will try to surround the prey, which is simulated by the following formula:

[0060]

[0061] where r linearly decreases from 2 to 0, rand() is a random number in the range [0, 1], A and C are coefficient vectors used to calculate the next position of the wolf pack. Then all wolf individuals update their positions according to α, β, and δ. The position update can be calculated by the following formula, where A 1 、A 2 and A 3 as well as C 1 、C 2 and C 3 are the vectors calculated by the coefficient formula (10):

[0062]

[0063] where X is the current position of the wolf individual, X α 、X β and X δ represent the positions of the current α, β, and δ wolves respectively, D α 、D β and D δ is the distance vector between the target and the current wolf individual;

[0064] S4-5-4: Complete the migration and predation of the wolf pack.

[0065] The beneficial effects of the present invention are as follows:

[0066] 1. The present invention uses a grid-based wolf pack behavior algorithm to optimize the unscented Kalman filter parameters, which can quickly find a better parameter set and significantly improve the accuracy and reliability of SOC estimation;

[0067] 2. The present invention enhances the anti-noise ability and improves the system robustness. By simulating the real environmental noise and adopting a robust optimization algorithm, the present invention effectively reduces the influence of noise on the SOC estimation, and improves the stability and reliability of the battery management system in complex environments.

[0068] 3. The present invention improves the optimization efficiency and shortens the iteration time. The proposed grid-based wolf pack behavior algorithm has an efficient global search ability, avoids falling into local optimal solutions, can quickly converge to the optimal parameter group, significantly shortens the optimization time, enhances the real-time response ability of the SOC estimation system, and meets the requirements of fast and accurate SOC estimation in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 is a flowchart of the battery SOC estimation method of the present invention;

[0070] Figure 2 is a circuit diagram of the second-order RC equivalent circuit model of the battery model selected by the present invention;

[0071] Figure 3 is a schematic diagram of the mean absolute error (MAE) of 50 estimations implemented by the estimator of the present invention;

[0072] Figure 4 is a schematic diagram of the mean absolute percentage error (MAPE) of 50 estimations implemented by the estimator of the present invention;

[0073] Figure 5 is a schematic diagram of the root mean square error (RMSE) of 50 estimations implemented by the estimator of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0074] The following further describes the specific embodiments of the present invention with reference to the drawings:

[0075] As Figure 1 shown, a battery SOC estimation method based on a grid-based wolf pack behavior algorithm - unscented Kalman filter joint algorithm according to this embodiment is characterized by including the following steps:

[0076] S1: Conduct a hybrid pulse power characteristic test on the target battery to obtain its current and voltage; the specific steps are as follows:

[0077] S1-1: Average set 10 test points at intervals of 10% between 100% and 10% of the battery SOC, and respectively conduct the following tests:

[0078] S1-1-1: Charge the target battery to the specified SOC point and then let it stand for one hour to ensure the test starts from a stable state;

[0079] S1-1-2: Constant current discharge the target battery at 75% of the maximum battery current for 1000 s, then stop charge and discharge for 400 s, then constant current charge at 75% of the maximum battery current for 1000 s, and finally let it stand until its voltage is stable. Record the test voltage and current data throughout the process;

[0080] S1-1-3: After each charge and discharge cycle, let the battery stand for 60 minutes, record the battery voltage once, and consider this voltage as the open circuit voltage of the battery at this SOC.

[0081] S2: Conduct the Beijing bus dynamic street condition test on the target battery to obtain its current and voltage, and add noise to the current and voltage to simulate the real sampling environment; the specific steps are as follows:

[0082] S2-1: Set the temperature of the test environment within the standardized temperature range, usually 20 °C to 25 °C, ensure that the target battery is in a fully charged state, and after full charge, let the battery stand for a certain time at the test temperature to reach thermal equilibrium and relatively static internal electrochemical reactions;

[0083] S2-2: Apply the pulsed current specified in the Beijing bus dynamic street condition test condition to the battery and record the current and voltage data;

[0084] S2-3: Add noise with a random normal distribution of 1% of the rated maximum battery current to the current obtained in step S2-2, and add noise with a normal distribution of 1% of the rated voltage to the voltage to simulate the external interference during the actual sampling operation of the model.

[0085] S3: Initialize the second-order RC equivalent battery model and calculate the true value of SOC; the specific steps are as follows:

[0086] S3-1: The calculation expression of SOC is as follows:

[0087]

[0088] In formula (3), SOC original is the SOC of the target battery at the initial moment, t is the time, C represents the battery capacity, and I(t) is the charge and discharge current at time t;

[0089] S3-2: As Figure 2 shown, the battery model in this embodiment selects a second-order RC equivalent battery model, and its formula is:

[0090] U out = U ocv -I 0 R 0 -U 1 -U 2 (26)

[0091]

[0092] Wherein, U out and U ocv represent the battery output voltage and the battery open-circuit voltage; I 0 is the main circuit current flowing through R 0 ; R 0 is the ohmic internal resistance, representing the internal contact resistance of the battery; U 1 is the voltage across the parallel connection of R 1 and C 1 , representing the electrochemical polarization voltage, U 2 is the voltage across the parallel connection of R 2 and C 2 , representing the concentration polarization voltage; R 1 , R 2 , C 1 , C 2 respectively represent the resistances and capacitances of two RC circuits.

[0093] S3-3: Use the data obtained from the hybrid pulse power characteristic test in step S1, and use the least squares method to find the R 0 , R 1 , R 2 , C 1 , C 2 and OCV data of the second-order RC equivalent battery model of the target battery at test points every 10% between SOC 100% and 10%, and substitute them into the above battery model for model initialization.

[0094] S4: Randomly generate an initial parameter group, optimize the parameter group using the grid-based wolf pack behavior algorithm, and use the unscented Kalman filter algorithm for estimation to verify the optimization effect. After multiple rounds of iteration, combine the optimal parameter group with the unscented Kalman filter algorithm to obtain a better unscented Kalman filter estimator; the specific steps are as follows:

[0095] S4-1: Determine the parameters in the system that need to be tuned and optimized, and initialize these parameter pairs within the specified range and call them wolf pack A;

[0096] S4-2: Initialize the second-order RC equivalent battery model, and use the test data in step S2 to calculate the true value of the battery SOC;

[0097] S4-3: Initialize the unscented Kalman filter algorithm, perform the first estimation on wolf pack A, and obtain the initial wolf pack fitness; use the mean absolute error MAE (Mean Absolute Error) calculated by the unscented Kalman filter as the wolf pack fitness of this parameter group. Among them, the calculation formula of the mean absolute error MAE is as follows:

[0098]

[0099] In Equation (4), t 0 is the total time of the unscented Kalman filter estimation, is the estimated value of SOC at time i, and x i is the true value of SOC at time i.

[0100] S4-4: Use the grid-based wolf pack behavior algorithm to perform vertical reproduction on the individuals of wolf pack A. The specific steps are as follows:

[0101] S4-4-1: Read the existing wolf pack and its fitness;

[0102] S4-4-2: Perform vertical reproduction of the wolf pack grid. Its expression is:

[0103] M op (m, D 1 ) = a × X(m, D 1 ) + (1 - a) × X(m, D 2 ) + b(X(m, D 1 ) - X(m, D 2 )) (29)

[0104] where a and b are random numbers between 0 and 1, X(m, D 1 ) and X(m, D 2 ) are the parent optimization parameters, and M op (m, D 1 ) is the offspring optimization parameter generated by the vertical reproduction of the parent. The parameter m = 1, 2,..., N, where N is the size of the wolf pack, and D 1 , D 2 = 1, 2,..., B, where B is the number of dimensions of the wolf pack;

[0105] S4-4-3: Check whether the new individual meets the conditions and calculate the fitness of the new individual; In the present invention, Equations (6) and (7) are used to correct each dimension that exceeds the upper and lower bounds of the constraint range respectively, so that the wolf pack can approach the boundary more smoothly;

[0106]

[0107] where E D is the parameter to be adjusted in dimension D, is the adjusted parameter in dimension D, and boundary up and boundary down are the upper and lower bounds of the parameter respectively;

[0108] S4-4-4: For the extreme case where the wolf pack crosses zero in a certain dimension, the wolf pack uses equations (8) and (9) to approximate the upper and lower boundaries respectively in this dimension, so that the wolf pack will not be exactly equal to the boundary, retaining a certain possibility of random evolution. Among them, a up is a random number between 1 and 10, and a down is a random number between 0.1 and 1;

[0109] E' D = boundary down ×a down (32)

[0110] E' D = boundary up ×a up (33)

[0111] S4-4-5: Judge whether it is better than the original wolf individual. If it is better, replace it; otherwise, retain it, so that the entire wolf pack develops in a better direction;

[0112] S4-4-6: Complete the vertical reproduction of this grid;

[0113] S4-5: Simulate the migration and predation process of the wolf pack and update the fitness of the wolf pack. The specific steps are as follows:

[0114] S4-5-1: Read the existing wolf pack and its fitness;

[0115] S4-5-2: According to the fitness value, select the best three wolf individuals as Alpha (α), Beta (β), and Delta (δ). These three wolf individuals will guide the movement of the wolf pack;

[0116] S4-5-3: When migrating and preying, the wolf pack will try to surround the prey, which is simulated by the following formula:

[0117]

[0118] r linearly decreases from 2 to 0, and rand() is a random number in the range [0,1]. Here, A and C are coefficient vectors used to calculate the next position of the wolf pack. Then all wolf individuals update their positions according to α, β, and δ. The position update can be calculated by the following formula, where A 1 、A 2 and A 3 as well as C 1 、C 2 and C 3 are vectors calculated by coefficient formula (10):

[0119]

[0120]

[0121] X is the position of the current wolf individual, X α , X β and X δ represent the positions of the current alpha, beta, and delta wolves respectively, D α , D β and D δ are the distance vectors between the target and the current wolf individuals;

[0122] S4-5-4: Complete the migration and predation of the wolf pack;

[0123] S4-6: Use the grid-based wolf pack behavior algorithm to perform horizontal reproduction on the individuals of wolf pack A. The specific steps are as follows:

[0124] S4-6-1: Read the existing wolf pack and its fitness;

[0125] S4-6-2: Perform horizontal reproduction of the wolf pack grid. Its calculation expression is:

[0126]

[0127] where the parameter n = 1, 2,..., N;

[0128] S4-6-3: The same as step S4-4-3;

[0129] S4-6-4: The same as step S4-4-4;

[0130] S4-6-5: The same as step S4-4-5;

[0131] S4-6-6: Complete this horizontal reproduction of the grid;

[0132] S4-7: Repeat steps S4-4 to S4-6 until the SOC estimation error is less than the specified value or the specified number of iterations is completed;

[0133] S4-8: Output the individual parameter group with the optimal fitness of the tuned wolf pack and its estimation error, and use this individual parameter group for the subsequent battery SOC estimation of this system to obtain a better unscented Kalman filter estimator.

[0134] S5: Use the better unscented Kalman filter estimator obtained in S4 for the subsequent SOC estimation of this battery.

[0135] To demonstrate the advantages of the grid-based wolf pack behavior algorithm - unscented Kalman filter algorithm in battery SOC estimation, the pure unscented Kalman filter algorithm and the grid-based wolf pack behavior algorithm - unscented Kalman filter joint algorithm proposed in this paper are respectively used for SOC estimation in the MATLAB environment. The two algorithms perform 50 SOC estimations each. The noise data for the 50 estimations are all different, but the same noise is used for estimations with different models to eliminate the uncertainty caused by noise during model comparison. Under the same input noise conditions, the average errors of the 50 estimations are shown in Table 1. The MAE, MAPE (Mean Absolute Percentage Error, MAPE), and RMSE (Root Mean Square Error, RMSE) curves of the errors of the 50 estimations are respectively as Figure 3 、 Figure 4 and Figure 5 shown.

[0136] Table 1 Errors of 50 estimations

[0137]

[0138] As can be seen from Figures 3 - 5 , the estimation errors of the grid-based wolf pack behavior algorithm - unscented Kalman filter algorithm model proposed in the present invention in terms of MAE, MAPE, and RMSE under 50 different noise interferences are much smaller than those of the pure unscented Kalman filter algorithm model. From the average data, the errors of the grid-based wolf pack behavior algorithm - unscented Kalman filter algorithm model proposed in the present invention are more than one order of magnitude smaller than those of the pure unscented Kalman filter algorithm model, demonstrating the superiority of the battery SOC estimation method based on the grid-based wolf pack behavior algorithm - unscented Kalman filter joint algorithm proposed in the present invention.

[0139] The above embodiments and descriptions in the specification only illustrate the principles and the best embodiments of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed.

Claims

1. A battery SOC estimation method based on a gridded wolf pack behavior algorithm-unscented Kalman filter joint algorithm, characterized in that: The following steps are involved: S1: Perform a mixed pulse power characteristic test on the target battery to obtain its current and voltage; S2: Perform a working condition test on the target battery to obtain its current and voltage, and add noise to the current and voltage to simulate the real sampling environment; S3: Initialize the second-order RC equivalent battery model and calculate the true value of SOC; S4: randomly generate an initial parameter group, use the gridded wolf pack behavior algorithm to optimize the parameter group, and use the unscented Kalman filter algorithm to estimate to verify the optimization effect. After multiple rounds of iterations, combine the optimal parameter group with the unscented Kalman filter algorithm to obtain the unscented Kalman filter estimator; S5: Use the unscented Kalman filter estimator obtained in step S4 to estimate the SOC of the battery.

2. The battery SOC estimation method based on the gridded wolf pack behavior algorithm-unscented Kalman filter joint algorithm according to claim 1 is characterized by: In step S1, the specific steps of performing a mixed pulse power characteristic test on the target battery to obtain its current and voltage are as follows: S1-1: Set 10 test points at 10% intervals between battery SOC 100% and 10%, and perform the following tests: S1-1-1: Charge the target battery to the specified SOC point and let it stand for one hour to ensure that the test starts from a stable state; S1-1-2: The target battery is discharged at a constant current of 75% of the maximum current of the battery for 1000s, then the charge and discharge is stopped for 400s, and then the constant current is charged at 75% of the maximum current of the battery for 1000s, and finally the battery is left to stand until the voltage is stable, and the test voltage and current data are recorded throughout the process; S1-1-3: After each charge and discharge cycle, the battery is left to rest for 60 minutes and the battery voltage is recorded once. This voltage is considered to be the open circuit voltage of the battery at that SOC.

3. The battery SOC estimation method based on the gridded wolf pack behavior algorithm-unscented Kalman filter joint algorithm according to claim 1 is characterized by: In step S2, a working condition test is performed on the target battery to obtain its current and voltage, and noise is added to the current and voltage to simulate a real sampling environment. The specific steps are as follows: S2-1: Set the temperature of the test environment within a standardized temperature range, ensure that the target battery is in a fully charged state, and let the battery stand at the test temperature for a certain period of time after being fully charged to achieve thermal equilibrium and the internal electrochemical reaction is relatively static; S2-2: applying a predetermined pulse current of a corresponding bus dynamic street condition test condition to the battery and recording current and voltage data; S2-3: Add random normally distributed noise of 1% of the rated maximum battery current to the current obtained in step S2-2, and add normally distributed noise of 1% of the rated voltage to the voltage to simulate external interference during actual sampling of the model.

4. The battery SOC estimation method based on the gridded wolf pack behavior algorithm-unscented Kalman filter joint algorithm according to claim 1 is characterized by: In step S3, the formula of the second-order RC equivalent battery model is: U out =U ocv -I0R0-U1-U2 (1) Among them, U out , U ocv represents the battery output voltage and battery open circuit voltage; I0 is the main current flowing through R0; R0 is the ohmic internal resistance, representing the internal contact resistance of the battery; U1 is the voltage of R1 and C1 in parallel, representing the electrochemical polarization voltage, U2 is the voltage of R2 and C2 in parallel, representing the concentration polarization voltage; R1, R2, C1, C2 represent the resistance and capacitance of the two RC circuits respectively.

5. The battery SOC estimation method based on the gridded wolf pack behavior algorithm-unscented Kalman filter joint algorithm according to claim 4 is characterized by: In step S3, the calculation expression of the actual value of SOC is as follows: In formula (3), SOC original is the SOC of the target battery at the initial moment, t is the time, C represents the battery capacity, and I(t) is the charge and discharge current at time t.

6. The battery SOC estimation method based on the gridded wolf pack behavior algorithm-unscented Kalman filter joint algorithm according to claim 5 is characterized by: In step S3, using the data obtained from the mixed pulse power characteristic test in step S1, the least squares method is used to calculate the R0, R1, R2, C1, C2 and OCV data of the second-order RC equivalent battery model of the target battery between SOC 100% and 10% with every 10% as a test point, and the data are substituted into the second-order RC equivalent battery model for model initialization.

7. The battery SOC estimation method based on the gridded wolf pack behavior algorithm-unscented Kalman filter joint algorithm according to claim 1 is characterized by: Step S4 specifically includes the following steps: S4-1: Determine the parameters that need to be tuned and optimized in the system, initialize these parameter pairs within the specified range and call them wolf pack A; S4-2: Initialize the second-order RC equivalent battery model and use the test data of step S2 to calculate the actual value of the battery SOC; S4-3: Initialize the unscented Kalman filter algorithm, make the first estimate of wolf pack A, and obtain the initial wolf pack fitness; The mean absolute error MAE calculated by the unscented Kalman filter is used as the wolf pack fitness of this parameter group; S4-4: Use the gridded wolf behavior algorithm to perform vertical reproduction on individuals in wolf pack A; S4-5: Simulate the migration and predation process of wolves and update the fitness of wolves; S4-6: Use the gridded wolf behavior algorithm to perform horizontal reproduction on individuals in wolf pack A; S4-7: Repeat steps S4-4 to S4-6 until the SOC estimation error is less than a specified value or a specified number of iterations are completed; S4-8: Output the individual parameter group with the best fitness of the wolf pack and its estimation error, and use the individual parameter group to perform subsequent battery SOC estimation of the system to obtain an unscented Kalman filter estimator.

8. The battery SOC estimation method based on the gridded wolf pack behavior algorithm-unscented Kalman filter joint algorithm according to claim 7 is characterized by: In step S4, the calculation formula of the mean absolute error MAE is as follows: In formula (4), t0 is the total time of unscented Kalman filter estimation, is the estimated value of SOC at time i, x i is the true value of SOC at time i.

9. The battery SOC estimation method based on the gridded wolf pack behavior algorithm-unscented Kalman filter joint algorithm according to claim 7 is characterized by: In step S4-4, the individual members of wolf pack A are vertically propagated using the gridded wolf pack behavior algorithm, and the specific steps are as follows: S4-4-1: Read the existing wolf packs and their fitness; S4-4-2: Execute wolf pack grid vertical reproduction, the expression is: M op (m,D1)=a×X(m,D1)+(1-a)×X(m,D2)+b(X(m,D1)-X(m,D2)) (5) Among them, a and b are random numbers between 0 and 1, X(m,D1) and X(m,D2) are the parent generation optimization parameters, and M op (m, D1) is the optimal parameter of the offspring produced by the vertical reproduction of the parent generation, parameter m = 1, 2, ..., N, N is the size of the wolf pack, D1, D2 = 1, 2, ..., B, B is the number of dimensions of the wolf pack; S4-4-3: Check whether the new individual meets the conditions and calculate the fitness of the new individual; use formula (6) and formula (7) to correct each dimension that exceeds the upper and lower bounds of the constraint range respectively, so that the wolf pack can approach the boundary more smoothly; Among them, E D is the parameter to be adjusted in D dimension, E' D is the adjusted parameter of D dimension, boundary up and boundary down are the upper and lower bounds of the parameter respectively; S4-4-4: For the extreme case where the wolf pack passes through zero in a certain dimension, the wolf pack is approximated to the upper and lower boundaries in this dimension using equations (8) and (9) respectively, so that the wolf pack will not be completely equal to the boundary; E' D =boundary down ×a down (8) E' D =boundary up ×a up (9) Among them, a up is a random number between 1 and 10, a down A random number between 0.1 and 1; S4-4-5: Determine whether it is better than the original individual wolf. If so, replace it; otherwise, keep it, so that the entire wolf pack can develop in a better direction; S4-4-6: Complete the vertical reproduction of the grid.

10. The battery SOC estimation method based on the gridded wolf pack behavior algorithm-unscented Kalman filter joint algorithm according to claim 7, characterized in that: In step S4-5, the migration and predation process of the wolf pack is simulated to update the fitness of the wolf pack. The specific steps are as follows: S4-5-1: Read the existing wolf packs and their fitness; S4-5-2: According to the fitness value, select the best three wolf individuals as Alpha (α), Beta (β) and Delta (δ), these three wolf individuals will guide the movement of the wolf pack; S4-5-3: Wolves will try to surround their prey when migrating and hunting, which can be simulated by the following formula: Among them, r decreases linearly from 2 to 0, rand() is a random number in the range of [0,1], A and C are coefficient vectors used to calculate the position of the wolf pack in the next step, and then all wolf individuals update their positions according to α, β and δ. The position update can be calculated by the following formula, where A1, A2 and A3 and C1, C2 and C3 are the vectors calculated by coefficient formula (10): Among them, X is the current position of the wolf individual, X α , X β and X δ Represent the current positions of α, β and δ wolves respectively, D α , D β and D δ is the distance vector between the target and the current wolf individual; S4-5-4: Complete the migration and hunting of the wolf pack.

Images