Lithium ion battery parameter identification and SOC estimation method based on MSDBO-FFEKF

Abstract

The invention provides a parameter identification and SOC estimation method based on MSDBO-FFEKF, and belongs to the technical field of lithium ion batteries. The technical problem of large SOC estimation error caused by inaccurate parameter estimation of a classical swarm intelligence algorithm is solved. Comprising the following steps: 1) carrying out an intermittent constant-current discharge experiment on a lithium ion battery, and measuring terminal voltage and load current of the lithium ion battery; 2) establishing a second-order RC equivalent circuit model; step 3) proposing an MSDBO algorithm to carry out parameter identification on the second-order RC equivalent circuit model of the battery; 4) constructing an extended Kalman filter with a forgetting factor; and 5) inputting parameters obtained by MSDBO algorithm identification into the FFEKF, and carrying out accurate estimation of the SOC of the battery. According to the method, high-precision parameters can be obtained by using the improved algorithm, so that the estimation precision of the SOC is improved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of lithium ion batteries, and in particular to a lithium ion battery parameter identification and SOC estimation method based on MSDBO-FFEKF. BACKGROUND

[0002] Lithium ion battery state of charge (SOC) estimation is one of the core functions of the battery management system, and accurate SOC can ensure the safe and reliable operation of the battery system, optimize the battery system, and provide a basis for energy management and safety management of electric vehicles.

[0003] Establishing an accurate and appropriate battery model is crucial for SOC estimation. Common battery models include electrochemical models, equivalent circuit models, fractional order models, and black box models. Among them, the equivalent circuit model uses resistors, capacitors, and constant voltage sources to form a network to describe the characteristics of the battery. The model is relatively simple, the parameters are easy to identify, and a certain accuracy can be guaranteed, so it is widely used. Parameter identification methods are mainly divided into two types: offline identification and online identification. Although the traditional online identification method can correct the parameters and estimate the state in real time, it may have large errors or fail in some cases. Offline identification can use a large amount of historical data to identify parameters with relatively high accuracy, which has a significant advantage in accuracy. Various swarm intelligence algorithms are widely used in offline identification. How to design an algorithm with higher accuracy and faster convergence speed is an important research topic in this field.

[0004] Choosing the right SOC estimation method is also very important. The most common methods are direct measurement and model-based methods. The most typical method in direct measurement is ampere-hour integration, but this method is highly dependent on the initial value of SOC, and the estimation error is also large. Model-based methods include neural network algorithms and Kalman filter algorithms. Neural network algorithms require a large amount of data as support, while Kalman filter algorithms are relatively mature and widely used in battery SOC estimation, but there are still accuracy problems.

[0005] How to solve the above technical problems is the challenge faced by the present application. SUMMARY

[0006] The purpose of the present application is to provide a lithium ion battery parameter identification and SOC estimation method based on MSDBO-FFEKF, on the basis of the original DBO algorithm, the mutation, crossover and selection of the differential evolution algorithm are added, and on this basis, dynamic parameters are introduced, a multi-strategy scarab beetle optimization algorithm is proposed, which effectively solves the problem of poor precision caused by the heuristic algorithm easily falling into local optimum, and on the basis of the traditional extended Kalman filter algorithm, a forgetting factor is added, which reduces the influence of old data points on the current state estimation and improves the response speed of the algorithm to sudden changes in state or noise. The combination of MSDBO algorithm and FFEKF algorithm can make the SOC estimation accuracy of the battery higher and the robustness stronger.

[0007] The present application is realized by the following measures: a lithium ion battery parameter identification and SOC estimation method based on MSDBO-FFEKF, comprising the following steps:

[0008] Step 1) Through intermittent constant current discharge experiment, the real-time data of lithium ion battery current and voltage are obtained, the static voltage is open circuit voltage OCV, and on this basis, nine polynomials are used for curve fitting of SOC and open circuit voltage OCV, and the OCV value at each time is calculated;

[0009] Step 2) According to the characteristics of the battery, a second-order RC equivalent circuit model of the battery is established;

[0010] Step 3) The MSDBO algorithm is proposed to identify the parameters of the second-order RC equivalent circuit model of the battery;

[0011] Step 4) A forgetting factor is added to the traditional EKF to construct FFEKF;

[0012] Step 5) The parameters identified by the MSDBO algorithm are input into the FFEKF to realize the estimation of the battery SOC.

[0013] As the parameter identification and SOC estimation method based on MSDBO-FFEKF of the present application, the step 2) comprises the following contents:

[0014] U oc and U L respectively represent the open circuit voltage and the terminal voltage of the lithium ion battery, R omc is the equivalent ohmic resistance of the lithium ion battery, R D1 and C D1 in parallel represent the electrochemical polarization reaction of the lithium ion battery, R D2 and C D2 in parallel represent the concentration polarization reaction of the lithium ion battery, I L represents the charge and discharge current, C D1 and CD2 The voltage at both ends is represented by U D1 and U D2 , the equation of the second-order RC equivalent circuit model is represented as:

[0015]

[0016] The definition of SOC is the ratio of the remaining capacity to the nominal capacity, and the calculation formula is as follows:

[0017]

[0018] In the formula: SOC(t) is the SOC value at time t, SOC(t0) is the SOC value at time t0, Q n is the rated capacity of the battery;

[0019] According to the function equation of the dual polarization model and the ampere-hour integration method, the current is taken as the input, and the terminal voltage is taken as the output, and the continuous state space expression of the system is obtained:

[0020]

[0021] U L (t)=U OC (SOC(t))-U D1 (t)-U D2 (t)-I(t)R omc (4)。

[0022] As the parameter identification and SOC estimation method based on MSDBO-FFEKF of the application, the step 3) comprises the following contents:

[0023] Step 3-1) dung beetle optimization algorithm:

[0024] The dung beetle optimization algorithm simulates the behaviors of rolling balls, reproduction, small dung beetle foraging and stealing in nature, and quickly performs optimization in a global range;

[0025] The steps of the dung beetle optimization algorithm are as follows:

[0026] a. Initialization

[0027] The number of population individuals P, the dimension N of individuals, the maximum number of iterations M, the lower limit LI and the upper limit UI of the global range are set,

[0028] The initial population P is represented as:

[0029] P i,j =LI i,j +(UI i,j -LI i,j )·rand i,j (0,1) (5)

[0030] where rand(0, 1) denotes a random number in (0, 1), i and j denote the jth parameter of the ith individual;

[0031] b. Rolling dung beetle

[0032] The strategy simulates the rolling process of the dung beetle, and the position update formula is as follows:

[0033] x k (z+1) = x k (z) + aFx k (z-1) + cAx (6)

[0034] Ax = |x k (z) - X ω | (7)

[0035] x k (z+1) = x k (z) + tan 0 |x k (z) - x k (z-1) | (8)

[0036] where k denotes the current iteration number, x k (z) denotes the position information of the kth dung beetle at the zth iteration, F e (0, 0.2] denotes the deflection coefficient, which is a constant value; c denotes a constant value belonging to (0, 1), a is a natural coefficient assigned as -1 or 1, X ω denotes the global worst position, Ax is used to simulate the change of light intensity, 0 is the deflection angle, 0 e [0, n], |x k (z) - x k (z-1) | denotes the difference between the position of the kth dung beetle at the zth iteration and its position at the (z-1)th iteration, if 0 = 0, n / 2, n, the position of the dung beetle is not updated;

[0037] c. Reproductive dung beetle

[0038] The strategy simulates the female dung beetle selecting a suitable breeding site to lay eggs, and the position update formula is as follows:

[0039]

[0040] B k (z+1) = X * + b1 x (B k (z) - LI * ) + b2 x (B k (z) - UI * ) (10)

[0041] where X *LI is the current local best position. * and UI * These are the lower and upper boundaries of the spawning area, respectively, where G = 1 - t / T. max T max B represents the maximum number of iterations, LI and UI represent the lower and upper bounds of the optimization problem, respectively. k (z) represents the position information of the k-th egg in the z-th iteration, b1 and b2 represent two independent random vectors of size 1×M, and M represents the dimension of the optimization problem;

[0042] d. Foraging dung beetles

[0043] This strategy simulates the behavior of dung beetles that have grown into adults and emerged from the ground in search of food. The location update formula is as follows:

[0044]

[0045] D k (z+1)=D k (z)+C1(D k (z)-LI * )+C2×(D k (z)-UI * (12)

[0046] In the formula: D k (z) represents the position information of the kth dung beetle in the z-th iteration, C1 represents a random number that follows a normal distribution, and C2 represents a random vector belonging to (0,1);

[0047] e. Dung beetle

[0048] This strategy simulates the thieving behavior of dung beetles, and the position update formula is as follows:

[0049] E k (z+1)=X v +M×g×(|E k (z)-X v |+|E k (z)-X * |) (13)

[0050] In the formula: E k (z) represents the position information of the k-th dung beetle in the z-th iteration, g is a 1×M dimensional random vector following a normal distribution, M represents a constant value, and X v The globally optimal position;

[0051] Step 3-2) Multi-strategy dung beetle optimization algorithm:

[0052] A multi-strategy algorithm is introduced into the algorithm, which adds mutation, crossover and selection strategies in the differential evolution algorithm, and introduces a mechanism of adaptive dynamic parameter adjustment to jump out of local optimum.

[0053] a. Mutation strategy

[0054] Three individuals are randomly selected from the population to generate a new solution And i≠z1,z2,z3, the mutation method is:

[0055]

[0056] In the formula: h ij (z+1) is the jth dimensional parameter value of the mutation vector generated by the mutation operation for the ith individual in the z+1th iteration, is the mutation term, and H is the mutation factor;

[0057] b. Crossover strategy

[0058] In order to increase the diversity of the population, the following operations are performed:

[0059]

[0060] In the formula: C r represents the crossover factor, and the variable C r The value of x ij (z) is the jth dimensional original parameter of the ith individual in the zth iteration;

[0061] c. Selection strategy

[0062] In order to determine whether x ij (z) can enter the next generation, the evaluation functions of v ij (z+1) and x ij (z) are compared:

[0063]

[0064] Repeat the above steps until the maximum number of iterations is reached;

[0065] d. Adaptive dynamic parameter adjustment strategy

[0066] This strategy can realize the dynamic balance between global exploration and local search by real-time sensing of the convergence state of the algorithm, increasing adaptive dynamic parameters w, adaptive mutation factor F and adaptive crossover probability C r , The formula is as follows:

[0067]

[0068] In the formula: winitial is the initial weight, is the attenuation coefficient, F max , F min is the maximum and minimum value of the mutation factor, and β is the attenuation coefficient, is the initial mutation probability, χ is the growth coefficient, iter is the current iteration number, and iter_max is the maximum iteration number;

[0069] Adaptive mutation factor F and adaptive crossover probability C r is a further improvement on the increased mutation and crossover strategy, and the adaptive dynamic parameter w mainly improves the rolling ball dung beetle strategy of the dung beetle optimization algorithm. The improved position update is as formula (20):

[0070] x k (z+1)=x k (z)+w·tanθ|x k (z)-x k (z-1)| (20).

[0071] As the parameter identification and SOC estimation method based on MSDBO-FFEKF of the application, the step 4) comprises the following contents:

[0072] A forgetting factor is introduced, and an extended Kalman filter algorithm with a forgetting factor is constructed:

[0073] The state equation and the observation equation of the nonlinear system are:

[0074]

[0075] In the formula, x(t) is a state variable, u(t) is an input variable, y(t) is an observation variable, and ω(t) and v(t) are Gaussian white noises, and the variances thereof are Q and R respectively;

[0076] The algorithm steps are as follows:

[0077] Step 4-1) parameter initialization;

[0078] The state variable x(t), the error covariance matrix P(t), Q and R are given initial values;

[0079] Step 4-2) state update;

[0080] a. State prediction

[0081]

[0082] In the formula: I(t)=u(t)

[0083] b. Error covariance matrix prediction

[0084]

[0085] where λ is a forgetting factor;

[0086] c. Kalman gain

[0087]

[0088] where:

[0089] d. Output equation

[0090]

[0091] e. State estimate update

[0092]

[0093] f. Error covariance matrix update

[0094]

[0095] As the MSDBO-FFEKF-based parameter identification and SOC estimation method of the application, the step 5) comprises the following contents:

[0096] Step 5-1) Parameter identification

[0097] The data obtained by the intermittent constant-current discharge test are divided into ten segments, and each segment has a process of standing, discharging and standing again, so that the result of parameter identification is more accurate, and the MSDBO algorithm is used to identify the parameters of each segment respectively, and ten groups of different parameters are obtained.

[0098] Step 5-2) SOC estimation

[0099] The identified parameters are input into the FFEKF, and the current and voltage data of the intermittent constant-current discharge test are combined to obtain the estimation result of the SOC.

[0100] Compared with the prior art, the application has the following beneficial effects:

[0101] 1. On the basis of the original DBO algorithm, the mutation, crossover and selection strategies of the differential evolution algorithm are added, and the mechanism of adaptive dynamic parameter adjustment is added, so that the problem that the original algorithm is easily trapped in local optimum in the process of parameter identification is effectively solved, and compared with the traditional particle swarm optimization (PSO) algorithm, the accuracy and convergence speed of the method are significantly improved.

[0102] 2、The application introduces the strategy of forgetting factor on the basis of the traditional extended Kalman filter algorithm, can effectively reduce the influence of old data points on the current state estimation, improve the response speed of the algorithm to the sudden change of state or noise, make the algorithm have higher precision and stronger robustness. BRIEF DESCRIPTION OF DRAWINGS

[0103] The accompanying drawings are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and are used to explain the application, but do not constitute a limitation on the application.

[0104] Figure 1 The voltage and current curve diagram of the intermittent constant current discharge experiment of the application.

[0105] Figure 2 The schematic diagram of the second-order RC circuit equivalent model used in the application.

[0106] Figure 3 The flowchart of the MSDBO algorithm of the application.

[0107] Figure 4 The voltage fitting curve diagram of the MSDBO algorithm of the application.

[0108] Figure 5 The voltage prediction error comparison curve diagram of the MSDBO algorithm of the application, DBO algorithm and PSO algorithm.

[0109] Figure 6 The SOC estimation curve diagram of the application.

[0110] Figure 7 The SOC estimation error comparison curve diagram of the FFEKF and EKF of the application. DETAILED DESCRIPTION

[0111] In order to make the purpose, technical scheme and advantages of the application more clear and obvious, the application will be further described in detail below in combination with the drawings and examples. Of course, the specific examples described here are only used to explain the application, and not used to limit the application.

[0112] Example 1

[0113] The Panasonic lithium ion battery NCR-18650B is taken as the research object battery in the application, the rated voltage of which is 3.7V, and the capacity of which is 3400mAh. The battery is charged in a constant current charging mode (0.5C) until the cut-off voltage is reached, and the battery reaches the full charge state after standing for a period of time.

[0114] Referring to Figures 1 to 7 , the application provides a lithium ion battery parameter identification and SOC estimation method based on MSDBO-FFEKF. It includes the following steps:

[0115] Step 1) Real-time data of lithium ion battery current and voltage are obtained through intermittent constant current discharge experiment, the end voltage after standing is open circuit voltage OCV, on this basis, nine polynomial is used to curve fit SOC and open circuit voltage OCV, and the OCV value at each moment is calculated;

[0116] Step 2) According to the characteristics of the battery, a second-order RC equivalent circuit model of the battery is established;

[0117] Step 3) The MSDBO algorithm is proposed to identify the parameters of the second-order RC equivalent circuit model of the battery;

[0118] Step 4) The forgetting factor is added to the traditional EKF to build FFEKF;

[0119] Step 5) The parameters identified by the MSDBO algorithm are input into the FFEKF to realize the estimation of the battery SOC.

[0120] As the parameter identification and SOC estimation method based on MSDBO-FFEKF of the application, the step 2) includes the following contents:

[0121] U oc and U L respectively represent the open circuit voltage and the end voltage of the lithium ion battery, R omc is the equivalent ohmic resistance of the lithium ion battery, R D1 and C D1 in parallel represent the electrochemical polarization reaction of the lithium ion battery, R D2 and C D2 in parallel represent the concentration polarization reaction of the lithium ion battery, I L represents the charge and discharge current, C D1 and C D2 voltage between the two ends is represented by U D1 and U D2 , and the equation of the second-order RC equivalent circuit model is represented as:

[0122]

[0123] The definition of SOC is the ratio of the remaining capacity to the nominal capacity, and the calculation formula is as follows:

[0124]

[0125] In the formula: SOC(t) is the SOC value at time t, SOC(t0) is the SOC value at time t0, Q n is the rated capacity of the battery;

[0126] According to the function equation of the dual polarization model and the ampere-hour integration method, taking the current as the input and the terminal voltage as the output, the continuous state space expression of the system is obtained as follows:

[0127]

[0128] U L (t)=U OC (SOC(t))-U D1 (t)-U D2 (t)-I(t)R omc (4)。

[0129] As the parameter identification and SOC estimation method based on MSDBO-FFEKF of the application, the step 3) comprises the following contents:

[0130] Step 3-1) Melworm optimization algorithm:

[0131] The melworm optimization algorithm quickly performs optimization in a global range by imitating the behaviors of rolling, breeding, small melworm foraging and stealing of melworms in nature;

[0132] The steps of the melworm optimization algorithm are as follows:

[0133] a. Initialization

[0134] The number of population individuals P, the dimension N of individuals, the maximum number of iterations M, the lower limit LI and the upper limit UI of the global range are set,

[0135] The initial population P is represented as:

[0136] P i,j =LI i,j +(UI i,j -LI i,j )·rand i,j (0,1) (5)

[0137] In the formula, rand(0,1) represents a random number in (0,1), and i and j represent the j-dimensional parameters of the i-th individual;

[0138] b. Rolling melworm

[0139] This strategy simulates the rolling process of melworms, and the position update formula is as follows:

[0140] x k (z+1)=x k (z)+αFx k (z-1)+cΔx (6)

[0141] Δx=|x k (z)-X ω | (7)

[0142] x k (z+1) = x k (z) + tan 0 | x k (z) - x k (z - 1) | (8)

[0143] where k denotes the current iteration number, x k (z) denotes the position information of the kth beetle at the zth iteration, F e (0, 0.2] denotes the deflection coefficient, which is a constant value; c denotes a constant value belonging to (0, 1), a is a natural coefficient assigned as -1 or 1, x ω denotes the global worst position, Ax is used to simulate the change of light intensity, 0 is the deflection angle, 0 e [0, π], | x k (z) - x k (z - 1) | denotes the difference between the position of the kth beetle at the zth iteration and its position at the (z - 1)th iteration, if 0 = 0, π / 2, π, the position of the beetle is not updated;

[0144] c. Reproducing beetles

[0145] This strategy simulates the behavior of female beetles choosing a suitable breeding site to lay eggs, and the position update formula is as follows:

[0146]

[0147] B k (z + 1) = X * + b1 x (B k (z) - LI * ) + b2 x (B k (z) - UI * ) (10)

[0148] where X * is the current local best position, LI * and UI * are the lower and upper bounds of the egg-laying area, respectively, G = 1 - t / T max , T max denotes the maximum number of iterations, LI and UI represent the lower and upper bounds of the optimization problem, respectively, where B k (z) is the position information of the kth egg at the zth iteration, b1 and b2 are two independent random vectors with a size of 1 x M, and M denotes the dimension of the optimization problem;

[0149] d. Foraging beetles

[0150] This strategy simulates the behavior of adult beetles that have grown up from the ground to search for food, and the position update formula is as follows:

[0151]

[0152] D k (z+1) = D k (z) + C1(D k (z) - LI * ) + C2 x (D k (z) - UI * ) (12)

[0153] where D k (z) represents the position information of the kth mule in the zth iteration, C1 represents a random number subject to normal distribution, and C2 represents a random vector belonging to (0, 1);

[0154] e. Stealing mule

[0155] The strategy simulates the stealing behavior of mules, and the position update formula is as follows:

[0156] E k (z+1) = X v + M x g x (|E k (z) - X v | + |E k (z) - X * |) (13)

[0157] where E k (z) represents the position information of the kth stealing mule in the zth iteration, g is a random vector with a size of 1 x M, subject to normal distribution, M represents a constant value, and X v is the global optimal position;

[0158] Step 3-2) Multi-strategy mule optimization algorithm:

[0159] A multi-strategy mule optimization algorithm is introduced, which adds the mutation, crossover and selection strategies in the differential evolution algorithm, and introduces a mechanism for self-adaptive dynamic parameter adjustment to jump out of the local optimum;

[0160] a. Mutation strategy

[0161] A new solution is generated by randomly selecting three individuals from the population and i≠z1,z2,z3, and the mutation method is:

[0162]

[0163] where h ij (z+1) is the jth dimensional parameter value of the mutation vector generated by the mutation operation for the ith target individual in the z+1th iteration, is the mutation term, and H is the mutation factor;

[0164] b. Crossover strategy

[0165] To increase the diversity of the population, the following operations are performed:

[0166]

[0167] where: C r represents the crossover factor, the variable C r has a value range between 0 and 1, x ij (z) is the jth dimension of the original parameters of the ith individual in the zth iteration;

[0168] c. Selection strategy

[0169] To determine whether x ij (z) can enter the next generation, the evaluation function of v ij (z+1) and x ij (z) is compared:

[0170]

[0171] The above steps are repeatedly performed until the maximum number of iterations is reached.

[0172] d. Adaptive dynamic parameter adjustment strategy

[0173] This strategy increases the adaptive dynamic parameters w, adaptive mutation factor F and adaptive crossover probability C r by real-time sensing of the convergence state of the algorithm, to achieve a dynamic balance between global exploration and local search, as follows:

[0174]

[0175] where: w initial is the initial weight, is the decay coefficient, F max , F min is the maximum and minimum value of the mutation factor, β is the decay coefficient, is the initial mutation probability, χ is the growth coefficient, iter is the current iteration number, and iter_max is the maximum iteration number.

[0176] The adaptive mutation factor F and the adaptive crossover probability C r are further improvements to the increased mutation and crossover strategies. The adaptive dynamic parameter w mainly improves the rolling ball strategy of the Scarab Optimization Algorithm. The improved position update is as follows:

[0177] x i (t+1) = x i (t) + w tan θ |xi (t)-x i (t-1)| (20).

[0178] As the MSDBO-FFEKF-based parameter identification and SOC estimation method of the application, the step 4) comprises the following contents:

[0179] The forgetting factor is introduced, and an extended Kalman filtering algorithm with the forgetting factor is constructed:

[0180] The state equation and the observation equation of the nonlinear system are as follows:

[0181]

[0182] In the formula, x(t) is a state variable, u(t) is an input variable, y(t) is an observation variable, ω(t) and v(t) are Gaussian white noises, and the variances of ω(t) and v(t) are Q and R respectively;

[0183] The algorithm steps are as follows:

[0184] Step 4-1) parameter initialization;

[0185] The initial values of the state variable x(t), the error covariance matrix P(t), Q and R are given;

[0186] Step 4-2) state update;

[0187] a. State prediction

[0188]

[0189] In the formula: I(t)=u(t)

[0190] b. Error covariance matrix prediction

[0191]

[0192] In the formula, λ is a forgetting factor;

[0193] c. Kalman gain

[0194]

[0195] In the formula:

[0196] d. Output formula

[0197]

[0198] e. State estimate value update

[0199]

[0200] f.error covariance matrix update

[0201]

[0202] As the MSDBO-FFEKF-based parameter identification and SOC estimation method of the application, the step 5) comprises the following contents:

[0203] Step 5-1) parameter identification

[0204] The data obtained by the intermittent constant current discharge test are divided into ten segments, each of which has a process of standing, discharging and standing again, which makes the result of parameter identification more accurate. The MSDBO algorithm is used to perform parameter identification on each segment of data respectively, and ten groups of different parameters are obtained. According to the obtained parameter identification results, the predicted value of the battery terminal voltage can be calculated, and it is compared with the real terminal voltage, as shown in Figure 4 . Then the unimproved DBO algorithm and PSO algorithm are used to perform parameter identification respectively, and the terminal voltage fitting errors of them and the terminal voltage fitting error of MSDBO are compared, as shown in Figure 5 . It is found by comparison that the MSDBO algorithm has certain advantages in convergence speed and identification accuracy, and can effectively avoid the algorithm from falling into local optimum.

[0205] Step 5-2) SOC estimation

[0206] The identified parameters are input into the FFEKF, and the current and voltage data of the intermittent constant current discharge test are combined to obtain the estimation result of SOC, as shown in Figure 6 . The SOC estimated by the EKF algorithm is compared, as shown in Figure 7 . The results show that the FFEKF algorithm has certain accuracy advantage compared with the EKF algorithm.

[0207] In summary, it can be seen that the MSDBO-FFEKF-based parameter identification and SOC estimation method has good effect, the model parameter identification result is accurate, the SOC estimation accuracy is high, and it has engineering value.

[0208] The above only describes the preferred embodiments of the application, and is not intended to limit the application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application shall be included in the protection scope of the application.

Claims

1. A parameter identification and SOC estimation method based on MSDBO-FFEKF, comprising the following steps: Step 1) Obtain real-time data of lithium-ion battery current and voltage through intermittent constant current discharge experiment. The terminal voltage after resting is the open circuit voltage OCV. Based on this, use a ninth-order polynomial to perform curve fitting on SOC and open circuit voltage OCV to calculate the OCV value at each moment. Step 2) Based on the characteristics of the battery, establish a second-order RC equivalent circuit model of the battery; Step 3) The MSDBO algorithm is proposed to identify the parameters of the second-order RC equivalent circuit model of the battery; Step 4) A forgetting factor was added to the traditional EKF to construct FFEKF; Step 5) Input the parameters identified by the MSDBO algorithm into FFEKF to estimate the battery SOC.

2. The parameter identification and SOC estimation method based on MSDBO-FFEKF according to claim 1, characterized in that, Step 2) includes the following: U oc and U L R represents the open-circuit voltage and terminal voltage of a lithium-ion battery, respectively. omc R is the equivalent ohmic internal resistance of a lithium-ion battery. D1 and C D1 The parallel connection represents the electrochemical polarization reaction of a lithium-ion battery, R D2 and C D2 The parallel connection represents the concentration polarization reaction in lithium-ion batteries, I L C represents the charging and discharging current. D1 and C D2 The voltage across the two ends is represented by U. D1 and U D2 The equations of the second-order RC equivalent circuit model are expressed as follows: SOC is defined as the ratio of remaining capacity to nominal capacity, and the calculation formula is as follows: In the formula: SOC(t) is the SOC value at time t, SOC(t0) is the SOC value at time t0, and Q... n This is the battery's rated capacity; Based on the functional equations of the dual-polarization model and the ampere-hour integral method, taking the current as the input and the terminal voltage as the output, the continuous state-space expression of the system is obtained as follows: U L (t)=U OC (SOC(t))-U D1 (t)-U D2 (t)-I(t)R omc (4)。 3. The parameter identification and SOC estimation method based on MSDBO-FFEKF according to claim 1, characterized in that, Step 3) includes the following: Step 3-1) Dung beetle optimization algorithm: The dung beetle optimization algorithm quickly finds the best option globally by mimicking the dung beetle's behavior in nature, such as rolling balls, reproducing, foraging for food, and stealing. The steps of the dung beetle optimization algorithm are as follows: a. Initialization Set the population size P, the dimension of each individual N, the maximum number of iterations M, and the lower bound LI and upper bound UI of the global scope. The initial population P is represented as: P i,j =LI i,j +(UI i,j -LI i,j )·rand i,j (0.1) (5); In the formula: rand(0,1) represents a random decimal number within (0,1), and i and j represent the j-th dimension parameter of the i-th individual; b. Dung beetle This strategy simulates the rolling process of a dung beetle, and the position update formula is as follows: x k (z+1)=x k (z)+αFx k (z-1)+cΔx (6); Δx=|x k (z)-X ω | (7); x k (z+1)=x k (z)+tanθ|x k (z)-x k (z-1)| (8); In the formula: k represents the current iteration number, x k (z) represents the position information of the k-th dung beetle at the z-th iteration, F∈(0,0.2] represents the deflection coefficient, which is a constant; c represents a constant belonging to (0,1), α is a natural coefficient assigned a value of -1 or 1, and X ω This represents the worst-case position globally, Δx is used to simulate changes in light intensity, θ is the deflection angle, θ∈[0,π], |x k (z)-x k (z-1)| represents the difference between the position of the k-th dung beetle in the z-th iteration and its position in the z-1-th iteration. If θ = 0, π / 2, π, the dung beetle's position is not updated. c. Breeding dung beetles This strategy simulates how female dung beetles choose suitable breeding sites to lay their eggs, and the location update formula is as follows: B k (z+1)=X * +b1×(B k (z)-LI * )+b2×(B k (z)-UI * ) (10); In the formula: X * LI is the current local best position. * and UI * These are the lower and upper boundaries of the spawning area, respectively, where G = 1 - t / T. max T max B represents the maximum number of iterations, LI and UI represent the lower and upper bounds of the optimization problem, respectively. k (z) represents the position information of the k-th egg in the z-th iteration, b1 and b2 represent two independent random vectors of size 1×M, and M represents the dimension of the optimization problem; d. Dung beetle foraging This strategy simulates the behavior of dung beetles that have grown into adults and emerged from the ground in search of food. The location update formula is as follows: D k (z+1)=D k (z)+C1(D k (z)-LI * )+C2×(D k (z)-UI * ) (12); In the formula: D k (z) represents the position information of the kth dung beetle in the z-th iteration, C1 represents a random number that follows a normal distribution, and C2 represents a random vector belonging to (0,1); e. Dung beetle This strategy simulates the thieving behavior of dung beetles, and the position update formula is as follows: E k (z+1)=X v +M×g×(|E k (z)-X v |+|E k (z)-X * |) (13); In the formula: E k (z) represents the position information of the k-th dung beetle in the z-th iteration, g is a 1×M dimensional random vector following a normal distribution, M represents a constant value, and X v The globally optimal position; Step 3-2) Multi-strategy dung beetle optimization algorithm: A multi-strategy dung beetle optimization algorithm is introduced, which incorporates mutation, crossover, and selection strategies from differential evolution algorithms, and introduces an adaptive dynamic parameter adjustment mechanism to escape local optima. a. Mutation strategy A new solution is generated by randomly selecting three individuals from the population. And i≠z1,z2,z3, the mutation method is: Where: h ij (z+1) In the (z+1)th iteration, for the i-th target individual, the j-th dimension parameter value of the mutation vector generated by the mutation operation. H is the variable, and H is the variation factor; b. Cross strategy To increase population diversity, the following steps are taken: In the formula: C r Represents the cross factor, variable C r The value of x ranges from 0 to 1. ij (z) represents the j-th dimension of the original parameter of the i-th individual in the z-th iteration; c. Choosing a strategy To determine x ij (z) Whether it can enter the next generation, for v ij (z+1) and x ij Compare the evaluation functions of (z): Repeat the above steps until the maximum number of iterations is reached; d. Adaptive dynamic parameter adjustment strategy This strategy increases the adaptive dynamic parameter w, adaptive mutation factor F, and adaptive crossover probability C by sensing the convergence state of the algorithm in real time. r This achieves a dynamic balance between global exploration and local search, as shown in the following formula: In the formula: w initial As the initial weights, F is the attenuation coefficient. max F min Let be the maximum and minimum values ​​of the variation factor, and β be the decay coefficient. Here, χ is the initial mutation probability, χ is the growth coefficient, and iter is the current iteration number. _ `max` represents the maximum number of iterations. Adaptive mutation factor F and adaptive crossover probability C r This is a further improvement on the added mutation and crossover strategies. The adaptive dynamic parameter w mainly improves the rolling ball dung beetle strategy of the dung beetle optimization algorithm. The improved position update is shown in formula (20): x k (z+1)=x k (z)+w·tanθ|x k (z)-x k (z-1)| (20)。 4. The parameter identification and SOC estimation method based on MSDBO-FFEKF according to claim 1, characterized in that, Step 4) includes the following: An extended Kalman filter algorithm with a forgetting factor was constructed by introducing a forgetting factor: The state equations and observation equations of the nonlinear system are as follows: In the formula: x(t) is the state variable, u(t) is the input variable, y(t) is the observation variable, ω(t) and v(t) are Gaussian white noise with variances Q and R, respectively; The algorithm steps are as follows: Step 4-1) Parameter initialization; Assign initial values ​​to the state variable x(t), the error covariance matrix P(t), Q, and R; Step 4-2) Status update; a. State prediction In the formula: b. Error covariance matrix prediction In the formula: λ is the forgetting factor; c. Kalman gain In the formula: d. Output formula e. State estimate update f. Error covariance matrix update 5. The parameter identification and SOC estimation method based on MSDBO-FFEKF according to claim 1, characterized in that, Step 5) includes the following: Step 5-1) Parameter Identification The data obtained from the intermittent constant current discharge test were divided into ten segments. Each segment involved a process of resting, discharging, and then resting again. This approach made the parameter identification results more accurate. The MSDBO algorithm was used to identify the parameters of each segment separately, resulting in ten different sets of parameters. Step 5-2) SOC estimation The identified parameters are input into the FFEKF, and combined with the current and voltage data from the intermittent constant current discharge test, the estimated SOC is obtained.

Images