CN110095723A - A kind of Li-ion battery model parameter and SOC online joint estimation method - Google Patents

Abstract

The invention belongs to the field of electric vehicle power battery management, and relates to an online joint estimation method of battery model parameters and SOC. The method mainly includes the following steps: firstly collect experimental data and build a battery model; secondly, use two SOC calculation modules to estimate SOC in parallel, one uses the extended Kalman filter (EKF) algorithm, the other uses the EKF algorithm first, and then adds mutation after a period of time Disturbance, and then use the unscented Kalman filter (UKF) algorithm to estimate the SOC. Then the estimation results of the two modules are weighted and averaged to obtain the current SOC estimation result. Finally, the forgetting factor least squares (FFRLS) method is used to identify the battery model parameters online, and the EKF, UKF and FFRLS algorithms are integrated to realize real-time update of model parameters and online estimation of SOC, effectively eliminating the influence of model errors and improving SOC estimation. Accuracy and Algorithmic Stability.

Description

An online joint estimation method of lithium-ion battery model parameters and SOC

technical field

The invention belongs to the field of electric vehicle battery management, and relates to an online joint estimation method for battery model parameters and SOC.

Background technique

In recent years, with the continuous development of new energy vehicles, the power battery as its core component has become a research hotspot in various countries. In the management of power batteries, the state of charge (SOC) of the battery is an important indicator to reflect the remaining capacity and working capability of the battery, and SOC estimation is the core technology of battery management system development. It has important theoretical significance and application value to improve the safety and reliability of the battery, improve the energy utilization rate of the battery, and prolong the service life.

As the internal state of the power battery, SOC cannot be measured directly, but can only be estimated by testing parameters such as battery voltage, current, and internal resistance. At present, the typical power battery SOC estimation methods mainly include: ampere-hour integration method, open circuit voltage method, neural network method, Kalman filter method, etc. Among them, the ampere-hour integration method is simple to implement, but the accumulated error in the integration process cannot be eliminated, which has a great impact on the estimation accuracy; the open-circuit voltage method requires the battery to stand for a period of time to measure and estimate, and is not suitable for online real-time estimation; the neural network method requires A large amount of data is used for training, which is complex and difficult to achieve; the core idea of the Kalman filter method is to make the optimal estimation of the state of the dynamic system in the sense of the least mean square, and the error correction ability is strong, but the estimation accuracy is very important to the accuracy of the battery model. The battery is a complex nonlinear system, and the parameters of the battery model change in real time during use, and the model uncertainty leads to low Kalman filter accuracy.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a method for on-line joint estimation of lithium-ion battery model parameters and SOC in view of the above-mentioned deficiencies of the prior art, so as to eliminate the battery model error and solve the problem of difficulty in SOC on-line estimation.

The object of the present invention can be achieved by the following technical solutions: a method for online joint estimation of lithium-ion battery model parameters and SOC, which is used for online identification of battery model parameters and real-time estimation of SOC, including steps:

S1: Carry out a discharge static experiment on the battery to obtain the relationship between the open circuit voltage (OCV) and SOC; according to the basic experimental data such as voltage, current, temperature, etc., identify the initial values of the model parameters offline and establish the battery 1-RC equivalent circuit model , set the initial values of the matching coefficients of the state space equation A ₀ , B ₀ , C ₀ , D ₀ ;

S2: SOC calculation module 1, using the extended Kalman filter (EKF) algorithm to estimate the current state of charge SOC ₁ of the battery;

S3: Determine whether the time exceeds the set value T (sudden change interval), if it does not exceed the time T, the first SOC calculation module continues to calculate, if it exceeds the time T, while the SOC calculation module 1 continues to calculate, go to the next step;

S4: When time t=T, according to the current SOC value SOC _T calculated by the SOC calculation module 1, a new SOC value SOC _m is obtained by adding sudden disturbance, and SOC _m is calculated as the second SOC The initial value of the module. Among them, the mutation method used is:

Among them, R is a random number that obeys the normal distribution N(0, 1), and "||" means to take the absolute value.

S5: SOC calculation module 2, using the unscented Kalman filter (UKF) algorithm to start to estimate the current state of charge SOC2 of the battery, while the SOC calculation module 1 continues to use the extended Kalman filter (EKF) algorithm to estimate the current SOC value;

S6: Determine whether the difference ΔU between the predicted voltage and the measured voltage of the battery exceeds the set threshold U _th when the SOC calculation module 2 uses the unscented Kalman filter algorithm to estimate the SOC. If it does not exceed the threshold U _th , the SOC calculation module 2 continues to calculate , if it exceeds U _th , go to the next step;

S7: Perform a weighted average of the estimation results of the SOC calculation module 1 and the calculation module 2 at this time to obtain a revised SOC value: SOC _w =W ₁ *SOC ₁ +W ₂ *SOC ₂ , where W ₁ and W ₂ are corresponding weight;

S8: Determine whether the revised SOC value SOC _w is valid, if valid, set the current SOC of the battery as the revised SOC value SOC _w , if invalid, set the current SOC of the battery as the estimated result SOC of the SOC calculation module 1 ₁ ;

S9: According to the current SOC value of the battery and the relationship between the open circuit voltage and SOC calculated in the previous step, use the forgetting factor least squares (FFRLS) to identify the model parameters R ₀ , R ₁ , C ₁ online and update the A _k , R 1 and C 1 in the system state equation. B _k , C _k , D _k .

In the step S7, the method for determining the weight when correcting the SOC value is: W ₁ =D ₂ /(D ₁ +D ₂ ), W ₂ =D ₁ /(D ₁ +D ₂ ), D ₁ =(Z _EKF -U) ² , D ₂ =(Z _UKF -U) ² . Among them, Z _EKF represents the current moment SOC calculation module 1 uses the extended Kalman filter (EKF) algorithm to predict the voltage value, Z _UKF represents the current moment SOC calculation module 2 uses the unscented Kalman filter (UKF) algorithm to predict the voltage value, U represents the current moment The measured voltage of the battery at time, D ₁ and D ₂ represent the degree of offset between Z _EKF and Z _UKF and U, respectively.

In the step S8, the method for judging the validity of the corrected SOC value is: comparing the corrected SOC value with the SOC value calculated by the open circuit voltage method, if the difference between the corrected SOC value and the SOC value calculated by the open circuit voltage method is within a specified range Within the range, the corrected SOC value is valid, otherwise it is invalid.

The beneficial effects of the present invention are:

1. By adding abrupt disturbance, the out-of-range SOC value can be quickly processed, avoiding the filter divergence phenomenon caused by abnormal data; at the same time, by judging whether the difference between the predicted voltage and the measured voltage exceeds the threshold value, further reducing the The possibility of filtering divergence improves the stability of the algorithm.

2. The battery model is identified online by the forgetting factor least square method, which effectively ensures the accuracy of the model, avoids the influence of model parameter changes on the estimation accuracy of the algorithm, and improves the robustness of the algorithm.

3. The battery model parameters and SOC online joint estimation algorithm adopted in the present invention can avoid the high computational complexity of the unscented Kalman filter algorithm and the low convergence speed of the extended Kalman filter algorithm, and can simultaneously have high precision, low computational complexity and Fast convergence speed.

Description of drawings

Figure 1: The flow chart of the joint estimation algorithm of lithium-ion battery model parameters and SOC

Figure 2: Lithium-ion battery 1-RC equivalent circuit model diagram

Detailed ways

The specific embodiments of the present invention will be further described below with reference to the accompanying drawings.

The present invention provides a method for joint online estimation of lithium-ion battery model parameters and SOC. The flow chart of the method is shown in FIG. 1 and includes steps S1-S9. Each implementation step is described in detail below:

S1: Carry out a discharge static experiment on the battery to obtain the relationship between the open circuit voltage (OCV) and SOC; according to the basic experimental data such as voltage, current, temperature, etc., identify the initial values of the model parameters offline and establish the battery 1-RC equivalent circuit model , set the initial values of the matching coefficients of the state space equation A ₀ , B ₀ , C ₀ , D ₀ ;

Based on the electrochemical principle, the established 1-RC battery equivalent circuit model is shown in Figure 2, which is composed of an ideal voltage source U _OC , an ohmic internal resistance R ₀ and an RC network that reflects the polarization characteristics of the battery. The current I is the input quantity (charging is positive), the terminal voltage U is the output quantity, and the system discretized state equation is:

The output equation is:

U(k)=U _oc (k)+U ₁ (k)+R ₀ I(k);

where Δt is the sampling interval, Q _v is the actual capacity of the battery, τ=R ₁ C ₁ , w and v represent the process noise and measurement noise, respectively; the state variable x=[SOC, U ₁ ] ^T is selected, corresponding to the matching coefficient of the state equation for:

S2: SOC calculation module 1, using the extended Kalman filter (EKF) algorithm to estimate the current state of charge SOC ₁ of the battery; the main steps of the EKF algorithm are as follows:

S201: Algorithm initialization

Set the initial state x ₀ , the initial state error covariance P ₀ , the process noise covariance Q and the measurement noise covariance R;

S202: Time update at time k by the state of k-1 and its error covariance

S203: Kalman filter gain calculation

S204: measure and update the state and its error covariance with the measured value at time k

in the formula Indicates the error between the measured value of the battery terminal voltage and the predicted value;

S205: Output the optimal estimated value of the state at time k

S3: determine whether the time exceeds the set value T, if it does not exceed the time T, the first SOC calculation module continues to calculate, if it exceeds the time T, the next step is entered while the SOC calculation module 1 continues to calculate; where T= 6τ is a parameter related to the system structure.

S4: When time t=T, according to the current SOC value SOC _T calculated by the SOC calculation module 1, a new SOC value SOC _m is obtained by adding sudden disturbance, and SOC _m is used as the SOC calculation module 2 initial value. Among them, the mutation method used is:

Among them, R is a random number that obeys the normal distribution N(0, 1), and "||" means to take the absolute value.

Since SOC is a variable between [0, 1], if the estimated value of SOC calculation module 1 at time T is 0 < SOC _T < 1, it means that it is within the normal range, and this value is directly used as the initial state value of the UKF algorithm; if SOC If _T ≥ 1 or SOC _T ≤ 0, it belongs to the abnormal value out of range. Using the formula SOC _m =|SOC _T +R-max(SOC _T , R)|, the abnormal value is quickly restored to [0, 1], The correct operation of the UKF algorithm is guaranteed, and the possibility of filter divergence of the two SOC calculation modules is reduced.

S5: SOC calculation module 2, using the unscented Kalman filter (UKF) algorithm to start estimating the current state of charge SOC ₂ of the battery, while the SOC calculation module 1 continues to use the extended Kalman filter (EKF) algorithm to estimate the current SOC value; wherein, The main steps of the UKF algorithm are:

S501: Initialize state variable mean x ₀ and mean square error P ₀

The state variable here expands the state variable in step S1, that is, x=[SOC, U ₁ , w, v] ^T , and the dimension of the expanded state is n.

S502: Obtain sampling points x _i and corresponding weights w.

In the formula, λ=α ² (n+κ)-n, w ^m , w ^c are the weights corresponding to the mean and variance of the particle points, respectively, and α and β control the distribution distance of the particle points in the sampling points and the error size of the higher-order terms respectively. .

S503: Time update of state estimation and mean square error

Status estimate time update:

In the formula, f( ^* ) represents the nonlinear function of the state equation;

Mean squared error time update:

System output time update:

In the formula, g(*) represents the nonlinear function of the measurement equation;

S504: Calculate the filter gain matrix

L _k =P _xy,k P _y,k ^-1

In the formula,

S505: State estimation and measurement update of mean square error

State estimation measurement update:

Mean squared error measurement update:

P _k =P _k/k-1 -L _k P _y,k L _k ^T

S506: Output the optimal estimated value of the state at time k

S6: Determine whether the difference ΔU between the predicted voltage and the measured voltage of the battery exceeds the set threshold U _th when the SOC calculation module 2 uses the unscented Kalman filter algorithm to estimate the SOC. If it does not exceed the threshold U _th , the SOC calculation module 2 continues to calculate , if it exceeds U _th , go to the next step; the threshold U _th is set to 0.05, that is, the predicted voltage error cannot exceed 0.05V.

S7: Perform a weighted average of the estimation results of the SOC calculation module 1 and the calculation module 2 at this time to obtain a revised SOC value: SOC _w =W ₁ *SOC ₁ +W ₂ *SOC ₂ , where W ₁ and W ₂ are corresponding The weights are determined as follows: W ₁ =D ₂ /(D ₁ +D ₂ ), W ₂ =D ₁ /(D ₁ +D ₂ ), D ₁ =(Z _EKF -U) ² , D ₂ =(Z _UKF -U) ² , X _EKF indicates that the SOC calculation module 1 at the current moment uses the extended Kalman filter (EKF) algorithm to predict the voltage value, and Z _UKF indicates that the SOC calculation module 2 adopts the unscented Kalman filter (UKF) at the current moment ) algorithm predicts the voltage value, U represents the measured voltage of the battery at the current moment, and D ₁ and D ₂ represent the degree of deviation between Z _EKF and Z _UKF and U, respectively.

S8: Determine whether the corrected SOC value SOC _w is valid, if valid, set the current SOC of the battery as SOC _w , if invalid, set the current SOC of the battery as the estimation result SOC ₁ of the SOC calculation module 1; The method for the validity of the SOC value is: compare the corrected SOC value with the SOC value calculated by the open circuit voltage method. If the difference between the corrected SOC value and the SOC value calculated by the open circuit voltage method is within the specified range, the corrected SOC value is valid, otherwise invalid.

S9: According to the current SOC value of the battery and the relationship between the open circuit voltage and SOC calculated in the previous step, use the forgetting factor least squares (FFRLS) to identify the model parameters R ₀ , R ₁ , C ₁ online and update the A _k , R 1 and C 1 in the system state equation. B _k , C _k , D _k .

The battery equivalent circuit shown in Figure 2 can be expressed as: U=[R ₁ /(R ₁ C ₁ s+1)+R ₀ ]I+U _oc , which can be transformed into: y(k) =a ₁ y(k-1)+b ₀ I(k)+b ₁ I(k-1), where the output quantity y=UU _oc represents the difference between the battery terminal voltage and the open circuit voltage, the current I is the input quantity, let Φ(k)=[y(k-1), I(k), I(k-1)], then θ=[a ₁ , b ₀ , b ₁ ] ^T is the parameter to be identified. Then use the FFRLS algorithm to identify θ and update the model parameters and equation of state matching coefficients A _k , B _k , C _k , D _k . Among them, the main steps of the FFRLS algorithm are:

S901: Determine the least squares covariance P ₀ and the initial value θ ₀ of the parameter matrix

S902: Calculate the least square gain matrix K(k)

K(k)=P(k-1)Φ(k)[λ+Φ ^T (k)P(k-1)Φ(k)] ^-1

where λ is the least squares weighting factor, take λ=0.98;

S903: Calculate the parameter estimation matrix

S904: Covariance matrix update

P(k)=[IK(k)Φ ^T (k)]P(k-1)

S905: Output the estimated state value at time k

The specific embodiments to which the present invention relates are given above, but the present invention is not limited to the described embodiments. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention and in a manner easily conceived by those skilled in the art should be included within the protection scope of the present invention.

Claims (6)

1. a kind of Li-ion battery model parameter and SOC online joint estimation method, the on-line identification for battery model parameter With the real-time estimation of SOC, comprising steps of

S1: battery carries out electric discharge and stands experiment, obtains the relational expression of open-circuit voltage (OCV) and SOC；According to voltage, electric current, temperature Etc. basic experiments off-line data pick out the initial value of model parameter and establish battery 1-RC equivalent-circuit model, set state space The matching factor initial value A of equation₀、B₀、C₀、D₀；

S2:SOC computing module 1 starts the state-of-charge for estimating current time battery with Extended Kalman filter (EKF) algorithm SOC₁；

S3: judging whether the time is more than (the mutation interval) setting value T, if being not above time T, first SOC computing module Continue to calculate, if it exceeds time T, then while SOC computing module 1 continues to calculate, into next step；

S4: as time t=T, according to the 1 calculated current SOC value SOC of institute of SOC computing module_T, by the way that mutation disturbance is added Mode obtains a new SOC value SOC_m, and by SOC_mInitial value as SOC computing module 2；Wherein, mutation side used Method are as follows:

R is the random number of (0,1) Normal Distribution N, " | | " indicate to take absolute value；

S5:SOC computing module 2 starts the state-of-charge SOC for estimating present battery with Unscented kalman filtering (UKF) algorithm₂, SOC computing module 1 continues with Extended Kalman filter (EKF) algorithm and estimates current SOC value simultaneously；

S6: when judging SOC computing module 2 with Unscented kalman filtering algorithm estimation SOC, the predicted voltage and actual measurement electricity of battery Whether the difference DELTA U of pressure is more than given threshold U_thIf being not above threshold value U_th, SOC computing module 2 continues to calculate, if super U is crossed_th, then enter in next step；

S7: SOC computing module 1 at this time and the estimation result of computing module 2 are weighted and averaged, revised SOC value is obtained: SOC_W=W₁*SOC₁+W₂*SOC₂, wherein W₁、W₂For corresponding weight；

S8: judge revised SOC value SOC_wWhether effectively, if effectively, the current SOC of battery is set as revised SOC value SOC_wIf in vain, the current SOC of battery to be set as to the estimated result SOC of SOC computing module 1₁；

S9: according to the current SOC value of the calculated battery of previous step and open-circuit voltage and SOC relationship, forgetting factor minimum is utilized Square law (FFRLS) on-line identification model parameter R₀、R₁、C₁And update the A in system state equation_k、B_k、C_k、D_k。

2. a kind of Li-ion battery model parameter according to claim 1 and SOC online joint estimation method, feature exist In in the step S3, when with EKF algorithm estimation SOC, mutation interval T=6 τ is and system structure SOC computing module 1 Related parameter, system model parameter is real-time update, therefore T is also real-time update.

3. a kind of Li-ion battery model parameter according to claim 1 and SOC online joint estimation method, feature exist In in the step S4, the mode of mutation disturbance is added in the t=T moment are as follows:

R is the random number of (0,1) Normal Distribution N, " | | " indicate to take absolute value；Since SOC is the variable between [0,1], If the 0 < SOC of estimated value of T moment SOC computing module 1_T< 1, indicates in the normal range, this value is calculated directly as UKF The state initial value of method, is effectively reduced the initial error of algorithm；If SOC_T>=1 or SOC_T≤ 0, belong to off-limits exception Value, utilizes formula S OC_m=| SOC_T+R-max(SOC_T, R) |, exceptional value is restored between [0,1] rapidly, guarantees UKF algorithm Correct operation, while improving the stability of algorithm.

4. a kind of Li-ion battery model parameter according to claim 1 and SOC online joint estimation method, feature exist In in the step S5, two SOC computing modules are run parallel, and first module is estimated with EKF algorithm always It calculates, second module first uses EKF algorithm, after T after a period of time, mutation disturbance is added, then estimated with UKF algorithm It calculates, then the calculated result of two modules is weighted and averaged, is effectively guaranteed the stabilization of SOC estimation precision and algorithm Property；Weighting algorithm used in it are as follows: SOC_W=W₁*SOC₁+W₂*SOC₂, wherein weight W₁=D₂/(D₁+D₂), W₂=D₁/(D₁+ D₂), D₁=(Z_EKF-U)², D₂=(Z_UKF-U)², Z_EKFIndicate that current time SOC computing module 1 uses EKF algorithm predicted voltage value, Z_UKFIndicate that current time SOC computing module 2 uses UKF algorithm predicted voltage value, U indicates current time battery measurement voltage, D₁ With D₂Respectively indicate Z_EKFAnd Z_UKFWith the degrees of offset of U.

5. a kind of Li-ion battery model parameter according to claim 1 and SOC online joint estimation method, feature exist In judging the method for revised SOC value validity are as follows: more revised SOC value and open circuit voltage method in the step S8 The SOC value of calculating, it is revised if revised SOC value differs within the specified scope with the SOC value that open circuit voltage method calculates Effectively, otherwise in vain SOC value is；The method for obtaining current time SOC optimal estimation value are as follows: if repaired obtained in step S7 Positive value SOC_wEffectively, current SOC is set as SOC_wIf in vain, current SOC to be set as to the estimation of SOC computing module 1 As a result SOC₁。

6. a kind of Li-ion battery model parameter according to claim 1 and SOC online joint estimation method, feature exist In utilizing forgetting factor least squares algorithm (FFRLS) on-line identification model parameter R in the step S9₀、R₁、C₁And it updates and is The matching factor A of system state equation_k、B_k、C_k、D_k, then repeatedly step S1-S9, so that EKF, UKF and FFRLS algorithm are merged, Realize the real-time update of battery model parameter and the On-line Estimation of SOC.