阅读说明：本技术 基于混合滤波的锂离子电池状态计算方法 (Lithium ion battery state calculation method based on hybrid filtering ) 是由 罗磊鑫 欧名勇 陈剑 张兴伟 刘立洪 谭丽平 陈娟 齐增清 唐利松 李达伟 陈柯 于 2020-04-10 设计创作，主要内容包括：本发明公开了一种基于混合滤波的锂离子电池状态计算方法,包括建立锂电池二阶等效电路模型并得到空间状态方程；采用扩展卡尔曼滤波对二阶等效电路模型进行在线参数辨识得到SOH估计值；采用滑动可变结构滤波算法对锂电池的SOC值进行估计；采用粒子群优化算法修正混合滤波器的参数并得到锂电池的精确的的SOH估计值和SOC估计值。本发明能够在线实时估计电池的状态,而且本发明方法的可靠性高、稳定性好且实施简单方便。(The invention discloses a lithium ion battery state calculation method based on hybrid filtering, which comprises the steps of establishing a lithium battery second-order equivalent circuit model and obtaining a space state equation; performing online parameter identification on the second-order equivalent circuit model by adopting extended Kalman filtering to obtain an SOH estimated value; estimating the SOC value of the lithium battery by adopting a sliding variable structure filtering algorithm; and correcting the parameters of the hybrid filter by adopting a particle swarm optimization algorithm and obtaining an accurate SOH estimated value and an accurate SOC estimated value of the lithium battery. The method can estimate the state of the battery in real time on line, and has high reliability, good stability and simple and convenient implementation.)

1. A lithium ion battery state calculation method based on hybrid filtering comprises the following steps:

s1, establishing a second-order equivalent circuit model of the lithium battery so as to obtain a space state equation;

s2, performing online parameter identification on the second-order equivalent circuit model established in the step S1 by adopting extended Kalman filtering so as to obtain an SOH estimated value;

s3, estimating the SOC value of the lithium battery by adopting a sliding variable structure filtering algorithm;

and S4, correcting the parameters of the hybrid filter by adopting a particle swarm optimization algorithm, so as to obtain an accurate SOH estimation value and an accurate SOC estimation value of the lithium battery.

2. The lithium ion battery state calculation method based on hybrid filtering according to claim 1, wherein the step S1 of establishing a lithium battery second-order equivalent circuit model to obtain a spatial state equation, specifically, the following model is used as the lithium battery second-order equivalent circuit model, and the following formula is used as the spatial discrete state equation:

lithium battery second order equivalent circuit model:

the voltage of the voltage source of the model is V_OC(SOC); the output current of the model is i_b(k) (ii) a The impedance of the model comprises a first impedance, a second impedance and a third impedance which are sequentially connected in series; the first impedance is R_sThe resistance of (1); the second impedance is R_stHas a resistance and a capacitance value of C_stThe terminal voltage at both ends of the second impedance is V_st(k) (ii) a The third impedance is R_ltHas a resistance and a capacitance value of C_ltThe terminal voltage at both ends of the third impedance is V_lt(k) (ii) a The output voltage of the model is V_cell(k)；

Spatial discrete state equation:

y(k)＝V_cell(k)＝V_oc(SOC(k))-V_st(k)-V_lt(k)-R_si_B(k)

V_OC(SOC)＝-a₀exp(-a₁SOC)+a₂+a₃SOC-a₄SOC²+a₅SOC³-a₆SOC⁴

wherein x (k +1) ═ SOC (k +1) V_st(k+1) V_lt(k+1)]^TIs a state variable, k is a time index, η is coulombic efficiency, C_maxIs the maximum capacity of the battery, T_sIs a sampling period, i_B(k) Instantaneous current of the battery and positive in discharge mode, α₁＝exp(-T_s/τ s) and τ s ═ R_st·C_st；β₁＝R_st(1-α₁)；α₂＝exp(-T_sτ l) and τ l ═ R_lt·C_lt；β₂＝R_lt(1-α₂)，a₀～a₆The correlation coefficient of the OCV curve.

3. The hybrid filtering-based state calculation method for lithium ion batteries according to claim 2, wherein the step S2 is performed by performing online parameter identification on the second-order equivalent circuit model established in the step S1 by using extended kalman filtering, so as to obtain an SOH estimation value, and specifically, the step of calculating the SOH estimation value is performed by using the following steps:

A. the model is established by adopting the following formula:

θ_k+1＝θ_k+r_k

y_k＝h(x_k,i_B,k,θ_k)+e_k

where θ is a parameter vector and θ ═ α₁β₁α₂β₂1/C_maxR_s]The corresponding state error covariance matrix is P; r is_kIs white Gaussian noise with mean value of 0 and covariance of Q; e.g. of the type_kTo measure noise;

B. calculating a pre-parameter vector using the following equationSum error covariance P_k|k-1：

P_k|k-1＝P_k-1+Q

C. The estimated measurements are calculated:

in the formulaKF gain at time k;

D. the SOH estimated value is calculated by the following formula

4. The hybrid filtering-based lithium ion battery state calculation method according to claim 3, wherein the calculation parameters of step CSpecifically, the following differential is repeatedly calculated to obtain the final calculation parameter

WhereinMust be guaranteed in the calculation

5. The lithium ion battery state calculating method based on hybrid filtering of claim 4, wherein the step S3 is performed by using a sliding variable structure filtering algorithm to estimate the SOC value of the lithium battery, specifically, the step of estimating the SOC value is performed by using:

a. the dynamic process of the sliding variable structure filtering algorithm is expressed by the following formula:

in the formulaIn order to predict an estimate of the state,f is the vector field, which is the estimated value of the current state;

b. the previous state vector is calculated using the following equationAnd time update is carried out, thereby obtaining calculation

In the formulaIs a linearized measurement matrix and

c. to ensure the stability of the values, the gain is calculated by the following equation

In the formula e_y,k|k-1The measurement error in the previous step; psi is the smooth boundary layer width; gamma is convergence rate and is more than 0 and less than 1; omicron is the product of shuer; i is₃An identity matrix of 3 x 3; omega_dIs a damping coefficient;

d. correcting the state using the following equationEstimated value of (a):

in the formulaIs the corrected state estimate in step k at the current time.

6. The hybrid filtering-based lithium ion battery state calculation method according to claim 5, wherein in step S4, the particle swarm optimization algorithm is used to modify the parameters of the hybrid filter, so as to obtain the accurate SOH estimation value and SOC estimation value of the lithium battery, specifically, the following function is used as the fitness function j (Z), and the best Z value is searched to minimize the fitness function j (Z):

in the formula of omega₁Is a V_cell,error,k(Z) a weighting factor for the estimation error; omega₂Is SOC_error,k(Z) a weighting factor for the estimation error; omega₃Is C_max,error,k(Z) estimating a weighting factor for the error.

7. The hybrid filtering-based lithium ion battery state calculation method according to claim 6, wherein the particle swarm optimization algorithm specifically adopts the following steps:

(1) defining a problem space, and extracting boundaries from offline battery tests under different operating conditions;

(2) initializing a particle swarm with random positions and speeds in a problem space;

(3) evaluating an adaptive value function;

(4) the current position Z of each particle_iAnd Z based on health assessment thereof_i,pbestAnd (3) comparison:

if Z is_iIs superior to Z_i,pbestThen use Z_iIn place of Z_i,pbest；

(5) If Z is updated_i,pbestThen, according to the evaluation result of the fitness function, the Z of the particle is calculated_i,pbestAnd Z_gbestAnd (3) comparison:

if Z is_i,pbestIs superior to Z_gbestThen use Z_i,pbestIn place of Z_gbest；

(6) In iteration L, the new velocity V for each particle is calculated using the following equation_iHe-Xin position X_i

V_i(l+1)＝ωV_i(l)+c₁r₁(Z_i,pbest(l)-Z_i(l))+c₂r₂(Z_gbest(l)-Z_i(l))

X_i(l+1)＝X_i(i)+V_i(l+1)

In the formula c₁Is the cognitive learning rate of the particle; c. C₂Social learning rate as a particle; ω is the inertial weight and decreases with increasing number of iterations; r is₁And r₂All are random numbers uniformly distributed between 0 and 1; n is the number of particles in the population;

(7) repeating the steps (3) to (6) until the iteration is finished, thereby obtaining the final optimal solution Z_gbest。

8. The lithium ion battery state calculation method based on hybrid filtering according to claim 7, wherein the particle swarm optimization algorithm has the corresponding algorithm parameters: the population size was 20; the number of iterations is 20; ω at the beginning is 0.9; ω at end is 0.4; c. C₁Is 2.05; c. C₂Is 2.05; omega₁Is 10; omega₂Is 1; omega₃Is 1.

Technical Field

The invention belongs to the field of lithium batteries, and particularly relates to a lithium ion battery state calculation method based on hybrid filtering.

Background

The lithium ion battery has been widely used in the fields of power grid energy storage, electric vehicle power battery and the like due to the characteristics of high energy ratio, long cycle life, low self-discharge rate, high conversion efficiency and the like. The state of the battery mainly refers to the state of charge (SOC of charge) and the state of health (SOH of health) of the battery. Generally, for a state estimation method of a lithium ion battery, a model-based method is adopted, and the model includes an electrochemical model, an equivalent circuit model, a data driving model and the like. Currently, the commonly used SOC state estimation methods include: open circuit voltage method, ampere-hour integral method, neural network method, Kalman filtering method, etc. The open-circuit voltage method is simple and easy to implement, but the battery needs to be kept still for a period of time before estimation, and the process needs a long time and is not suitable for on-line estimation; the initial error of the ampere-hour integration method cannot be corrected, and the estimation precision of the SOC is not high along with the time; estimating the SOC of the battery by a Kalman Filtering (KF) method, wherein closed-loop estimation is adopted and is the optimal estimation of least mean square; the Extended KF (EKF) needs to carry out linearization processing on a nonlinear model, so that a linearization error is introduced into a system, Unscented Kalman Filtering (UKF) has the defect of unstable estimation, the semipositive nature of state covariance cannot be determined, the error caused by noise covariance cannot be reduced, and the final estimation precision is influenced; the neural network method has wide application range and is suitable for various lithium batteries, but a large amount of experimental data are required for accumulation. The SOH estimation method mainly includes an open-loop method based on an endurance model and a closed-loop method based on a battery model. Complex algorithms such as Kalman filtering and the like are not mature at present, and the system is difficult to set. One key challenge of algorithms such as neural network method and kalman filter is how to adjust the algorithm parameters.

Therefore, a lithium ion battery state calculation method which is high in reliability, good in stability and simple and convenient to implement does not exist at present.

Disclosure of Invention

The invention aims to provide a lithium ion battery state calculation method based on hybrid filtering, which has high reliability and good stability and is simple and convenient to implement.

The invention provides a lithium ion battery state calculation method based on hybrid filtering, which comprises the following steps:

s1, establishing a second-order equivalent circuit model of the lithium battery so as to obtain a space state equation;

s2, performing online parameter identification on the second-order equivalent circuit model established in the step S1 by adopting extended Kalman filtering so as to obtain an SOH estimated value;

s3, estimating the SOC value of the lithium battery by adopting a sliding variable structure filtering algorithm;

and S4, correcting the parameters of the hybrid filter by adopting a particle swarm optimization algorithm, so as to obtain an accurate SOH estimation value and an accurate SOC estimation value of the lithium battery.

Step S1, establishing a second-order equivalent circuit model of the lithium battery, so as to obtain a spatial state equation, specifically, using the following model as the second-order equivalent circuit model of the lithium battery, and using the following formula as the spatial discrete state equation:

lithium battery second order equivalent circuit model:

the voltage of the voltage source of the model is V_OC(SOC); the output current of the model is i_b(k) (ii) a The impedance of the model comprises a first impedance, a second impedance and a third impedance which are sequentially connected in series; the first impedance is R_sThe resistance of (1); the second impedance is R_stHas a resistance and a capacitance value of C_stThe terminal voltage at both ends of the second impedance is V_st(k) (ii) a The third impedance is R_ltHas a resistance and a capacitance value of C_ltThe terminal voltage at both ends of the third impedance is V_lt(k) (ii) a The output voltage of the model is V_cell(k)；

Spatial discrete state equation:

y(k)＝V_cell(k)＝V_oc(SOC(k))-V_st(k)-V_lt(k)-R_si_B(k)

V_OC(SOC)＝-a₀exp(-a₁SOC)+a₂+a₃SOC-a₄SOC²+a₅SOC³-a₆SOC⁴

wherein x (k +1) ═ SOC (k +1) V_st(k+1) V_lt(k+1)]^TIs a state variable, k is a time index, η is coulombic efficiency, C_maxIs the maximum capacity of the battery, T_sIs a sampling period, i_B(k) Instantaneous current of the battery and positive in discharge mode, α₁＝exp(-T_s/τ s) and τ s ═ R_st·C_st；β₁＝R_st(1-α₁)；α₂＝exp(-T_sτ l) and τ l ═ R_lt·C_lt；β₂＝R_lt(1-α₂)，a₀～a₆Phase of OCV curveAnd (4) a correlation coefficient.

In step S2, performing online parameter identification on the second-order equivalent circuit model established in step S1 by using extended kalman filtering, so as to obtain an SOH estimate value, specifically, calculating the SOH estimate value by using the following steps:

A. the model is established by adopting the following formula:

θ_k+1＝θ_k+r_k

y_k＝h(x_k,i_B,k,θ_k)+e_k

where θ is a parameter vector and θ ═ α₁β₁α₂β₂1/C_maxR_s]The corresponding state error covariance matrix is P; r is_kIs white Gaussian noise with mean value of 0 and covariance of Q; e.g. of the type_kTo measure noise;

B. calculating a pre-parameter vector using the following equationSum error covariance P_k|k-1：

P_k|k-1＝P_k-1+Q

C. The estimated measurements are calculated:

in the formulaKF gain at time k;

D. the SOH estimated value is calculated by the following formula

Calculating parameters of step CSpecifically, the following differential is repeatedly calculated to obtain the final calculation parameter

Wherein, it must guarantee in the calculation

The step S3, which is to estimate the SOC value of the lithium battery by using a sliding variable structure filter algorithm, specifically includes the following steps:

a. the dynamic process of the sliding variable structure filtering algorithm is expressed by the following formula:

in the formulaIn order to predict an estimate of the state,f is the vector field, which is the estimated value of the current state;

b. the previous state vector is calculated using the following equationAnd time update is carried out, thereby obtaining calculation

In the formulaIs a linearized measurement matrix and

c. to ensure the stability of the values, the gain is calculated by the following equation

In the formula e_y,k|k-1The measurement error in the previous step; psi is the smooth boundary layer width; gamma is convergence rate and is more than 0 and less than 1;is the product of Shu Er; i is₃An identity matrix of 3 x 3; omega_dIs a damping coefficient;

d. correcting the state using the following equationEstimated value of (a):

in the formulaIs the corrected state estimate in step k at the current time.

Step S4, modifying the parameters of the hybrid filter by using a particle swarm optimization algorithm, so as to obtain an accurate SOH estimation value and SOC estimation value of the lithium battery, specifically, using the following function as a fitness function j (Z), and searching for an optimal Z value so that the fitness function j (Z) is minimum:

in the formula of omega₁Is a V_cell,error,k(Z) a weighting factor for the estimation error; omega₂Is SOC_error,k(Z) a weighting factor for the estimation error; omega₃Is C_max,error,k(Z) estimating a weighting factor for the error.

The particle swarm optimization algorithm specifically comprises the following steps of:

(1) defining a problem space, and extracting boundaries from offline battery tests under different operating conditions;

(2) initializing a particle swarm with random positions and speeds in a problem space;

(3) evaluating an adaptive value function;

(4) the current position Z of each particle_iAnd Z based on health assessment thereof_i,pbestAnd (3) comparison:

if Z is_iIs superior to Z_i,pbestThen use Z_iIn place of Z_i,pbest；

(5) If Z is updated_i,pbestThen, according to the evaluation result of the fitness function, the Z of the particle is calculated_i,pbestAnd Z_gbestAnd (3) comparison:

if Z is_i,pbestIs superior to Z_gbestThen use Z_i,pbestIn place of Z_gbest；

(6) In iteration L, the new velocity V for each particle is calculated using the following equation_iHe-Xin position X_i

V_i(l+1)＝ωV_i(l)+c₁r₁(Z_i,pbest(l)-Z_i(l))+c₂r₂(Z_gbest(l)-Z_i(l))

X_i(l+1)＝X_i(i)+V_i(l+1)

In the formula c₁Is the cognitive learning rate of the particle; c. C₂Social learning rate as a particle; ω is the inertial weight and decreases with increasing number of iterations; r is₁And r₂All are random numbers uniformly distributed between 0 and 1; n is the number of particles in the population;

(7) repeating the steps (3) to (6) until the iteration is finished, thereby obtaining the final optimal solution Z_gbest。

The particle swarm optimization algorithm comprises the following corresponding algorithm parameters: the population size was 20; the number of iterations is 20; ω at the beginning is 0.9; ω at end is 0.4; c. C₁Is 2.05; c. C₂Is 2.05; omega₁Is 10; omega₂Is 1; omega₃Is 1.

The invention provides a lithium ion battery state calculation method based on hybrid filtering, and provides a lithium ion battery state online estimation method based on hybrid filtering.

Drawings

FIG. 1 is a schematic process flow diagram of the process of the present invention.

Fig. 2 is a schematic circuit diagram of a second-order equivalent circuit model of a lithium battery according to the present invention.

Detailed Description

The invention provides a lithium ion battery state calculation method based on hybrid filtering, which comprises the following steps:

s1, establishing a second-order equivalent circuit model of the lithium battery so as to obtain a space state equation; specifically, the following model is used as a lithium battery second-order equivalent circuit model, and the following formula is used as a space discrete state equation:

second-order equivalent circuit model of lithium battery (as shown in fig. 2):

the voltage of the voltage source of the model is V_OC(SOC); the output current of the model is i_b(k) (ii) a The impedance of the model comprises a first impedance, a second impedance and a third impedance which are sequentially connected in series; the first impedance is R_sThe resistance of (1); the second impedance is R_stHas a resistance and a capacitance value of C_stThe terminal voltage at both ends of the second impedance is V_st(k) (ii) a The third impedance is R_ltHas a resistance and a capacitance value of C_ltThe terminal voltage at both ends of the third impedance is V_lt(k) (ii) a The output voltage of the model is V_cell(k)；

Spatial discrete state equation:

y(k)＝V_cell(k)＝V_oc(SOC(k))-V_st(k)-V_lt(k)-R_si_B(k)

V_OC(SOC)＝-a₀exp(-a₁SOC)+a₂+a₃SOC-a₄SOC²+a₅SOC³-a₆SOC⁴

wherein x (k +1) ═ SOC (k +1) V_st(k+1) V_lt(k+1)]^TIs a state variable, k is a time index, η is coulombic efficiency, C_maxIs the maximum capacity of the battery, T_sIs a sampling period, i_B(k) Instantaneous current of the battery and positive in discharge mode, α₁＝exp(-T_s/τ s) and τ s ═ R_st·C_st；β₁＝R_st(1-α₁)；α₂＝exp(-T_sτ l) and τ l ═ R_lt·C_lt；β₂＝R_lt(1-α₂)，a₀～a₆Correlation coefficient of OCV curve;

s2, performing online parameter identification on the second-order equivalent circuit model established in the step S1 by adopting extended Kalman filtering so as to obtain an SOH estimated value; specifically, the calculation of the SOH estimated value is carried out by adopting the following steps:

A. the model is established by adopting the following formula:

θ_k+1＝θ_k+r_k

y_k＝h(x_k,i_B,k,θ_k)+e_k

where θ is a parameter vector and θ ═ α₁β₁α₂β₂1/C_maxR_s]The corresponding state error covariance matrix is P; r is_kIs white Gaussian noise with mean value of 0 and covariance of Q; e.g. of the type_kTo measure noise;

B. calculating a pre-parameter vector using the following equationSum error covariance P_k|k-1：

P_k|k-1＝P_k-1+Q

C. The estimated measurements are calculated:

in the formulaKF gain at time k;

in practice, parametersFor repeatedly calculating the following differential to obtain the final calculation parameters

Wherein, it must guarantee in the calculation

D. The SOH estimated value is calculated by the following formula

S3, estimating the SOC value of the lithium battery by adopting a sliding variable structure filtering algorithm; specifically, the SOC value is estimated by adopting the following steps:

a. the dynamic process of the sliding variable structure filtering algorithm is expressed by the following formula:

in the formulaIn order to predict an estimate of the state,f is the vector field, which is the estimated value of the current state;

b. the previous state vector is calculated using the following equationAnd time update is carried out, thereby obtaining calculation

In the formulaIs a linearized measurement matrix and

c. to ensure the stability of the values, the gain is calculated by the following equation

In the formula e_y,k|k-1The measurement error in the previous step; psi is the smooth boundary layer width; gamma is convergence rate and is more than 0 and less than 1;is the product of Shu Er; i is₃An identity matrix of 3 x 3; omega_dIs a damping coefficient;

d. correcting the state using the following equationEstimated value of (a):

in the formulaThe state estimation value after the correction in the current time step k is obtained;

s4, correcting parameters of the hybrid filter by adopting a particle swarm optimization algorithm, so as to obtain an accurate SOH estimation value and an SOC estimation value of the lithium battery, specifically, adopting the following function as a fitness function J (Z), and searching an optimal Z value to enable the fitness function J (Z) to be minimum:

in the formula of omega₁Is a V_cell,error,k(Z) a weighting factor for the estimation error; omega₂Is SOC_error,k(Z) a weighting factor for the estimation error; omega₃Is C_max,error,k(Z) estimating a weighting factor for the error.

In specific implementation, the particle swarm optimization algorithm adopts the following steps to calculate:

(1) defining a problem space, and extracting boundaries from offline battery tests under different operating conditions;

(2) initializing a particle swarm with random positions and speeds in a problem space;

(3) evaluating an adaptive value function;

(4) the current position Z of each particle_iAnd Z based on health assessment thereof_i,pbestAnd (3) comparison:

if Z is_iIs superior to Z_i,pbestThen use Z_iIn place of Z_i,pbest；

(5) If Z is updated_i,pbestThen, according to the evaluation result of the fitness function, the Z of the particle is calculated_i,pbestAnd Z_gbestAnd (3) comparison:

if Z is_i,pbestIs superior to Z_gbestThen use Z_i,pbestIn place of Z_gbest；

(6) In iteration L, the new velocity V for each particle is calculated using the following equation_iHe-Xin position X_i

V_i(l+1)＝ωV_i(l)+c₁r₁(Z_i,pbest(l)-Z_i(l))+c₂r₂(Z_gbest(l)-Z_i(l))

X_i(l+1)＝X_i(i)+V_i(l+1)

In the formula c₁Is the cognitive learning rate of the particle; c. C₂Social learning rate as a particle; ω is the inertial weight and decreases with increasing number of iterations; r is₁And r₂All are random numbers uniformly distributed between 0 and 1; n is the number of particles in the population;

(7) repeating the steps (3) to (6) until the iteration is finished, thereby obtaining the final optimal solution Z_gbest；

Meanwhile, the particle swarm optimization algorithm has the corresponding algorithm parameters as follows: the population size was 20; the number of iterations is 20; ω at the beginning is 0.9; ω at end is 0.4; c. C₁Is 2.05; c. C₂Is 2.05; omega₁Is 10; omega₂Is 1; omega₃Is 1.