CN105093128A - Storage battery state of charge (SOC) estimation method based on extended Kalman filtering (EKF) - Google Patents

Abstract

The invention relates to a storage battery state of charge (SOC) estimation method based on extended Kalman filtering (EKF). The method comprises the following steps: performing performance testing on a storage battery through stepped discharge to obtain discharge currents and monomer voltage data in a discharge test process; based on test data obtained from step a, obtaining an open voltage-SOC function relation; based on the test data obtained from the step a, carrying out parameter identification to obtain storage battery model parameters of each discharge phase; based on the function relation obtained from the step b and storage battery model parameters of a first discharge phase, obtained from step c, estimating the SOC of the storage battery through an EKF algorithm; based on the SOC obtained from step d, after temperature compensation is performed, a final SOC is obtained; regularly performing internal resistance testing on the storage battery, and selecting the storage battery model parameters in the corresponding phases in the step c from measured internal resistance values; and obtaining a latest SOC through the step d and step e. The method provided by the invention can accurately and rapidly carry out SOC estimation.

Description

Based on the storage battery charge state evaluation method of EKF

Technical field:

The present invention relates to a kind of storage battery charge state evaluation method based on EKF, relate to electric system battery technology field.

Background technology:

Last line of defense that valve-regulated lead-acid battery is reliably supplied as electric power, plays an important role in standby power system.Owing to being in floating charge for a long time, and lack maintenance, accumulator can be deteriorated gradually along with usability.Accumulator in electric system is that series connection uses, performance difference between battery will cause voltage distribution during discharge and recharge uneven, thus cannot ensure often to save the discharge and recharge that accumulator reaches specification, accelerate the aging of accumulator further, reduce the reliability of whole Battery pack.Therefore, grasp and often save accumulator, the SOC of especially poor accumulator is vital.

Existing accumulator SOC estimating techniques mainly comprise discharge test method, ampere-hour integral method, open-circuit voltage method, neural network, Kalman filtering method etc.Discharge test method is the most reliable SOC estimation method of generally acknowledging at present, but it is as a kind of measuring method of complete off-line, the time that general needs are longer, and the reliable supply that cannot ensure back-up source after accumulator off-line, the test of the method simultaneously needs special nurse.Ampere-hour integral method needs the electric current in continuous accumulation charge and discharge process, even if very little current measurement errors, after long-term accumulation, also can cause very large SOC estimation error.Open-circuit voltage method, needs the off-line of electric battery through the long period to shelve, cannot realize estimation on line SOC.Neural network, needs a large amount of training samples and test sample book, and higher to the requirement of training sample, and its estimation error is subject to the impact of training sample and training method to a great extent.Traditional Kalman filtering method is only applicable to linear system, non-linear inapplicable to what present in battery use procedure.

In implementation procedure of the present invention, prior art at least exists that manual labor amount is large, poor real, operation easier are large and the defect such as precision is low.

Summary of the invention:

For the deficiency in solving the problems of the technologies described above, the present invention proposes a kind of storage battery charge state evaluation method based on EKF (EKF).

The technical solution used in the present invention is:

1. data acquisition, carries out stage electric discharge to accumulator, obtains discharge process data simultaneously.

2. Function Fitting, based on the test data in 1., by parameter identification, obtains the functional relation OCV=F (SOC) of open-circuit voltage (OCV) and state-of-charge (SOC).

3. identification model parameter, based on the test data in each stage in 1., by parameter identification, obtains each discharge regime battery model parameter.

4. estimation on line SOC, based on the battery model parameter of the first discharge regime in 3., by EKF algorithm, estimation on line accumulator SOC.

5. on-line amending battery model parameter, regularly carries out inner walkway to accumulator, the battery model parameter being chosen certain discharge regime 3. by surveyed internal resistance value as new battery model parameter, by EKF algorithm, estimation on line accumulator SOC.

Wherein concrete: to comprise the steps:

A. by stage electric discharge, performance test is carried out to accumulator, obtain the discharge current in discharge test process and monomer voltage data;

B. based on the test data of step a gained, the functional relation of open-circuit voltage-state-of-charge is obtained;

C. based on the test data of step a gained, carry out parameter identification, obtain the battery model parameter of each discharge regime;

D. based on the battery model parameter of the functional relation of step b gained and the first discharge regime of step c gained, by EKF algorithm, estimation accumulator SOC;

E., based on the SOC of steps d gained, after carrying out temperature compensation, final SOC is obtained;

F. regular inner walkway is carried out to accumulator, by the internal resistance value recorded, select the battery model parameter under corresponding discharge regime in step c;

G. by steps d and step e, up-to-date SOC is obtained.

Step a, specifically comprises:

To discharge shelving after accumulator off-line after 1h 1h with 0.1C10; Be circulated to accumulated discharge duration according to this, reach after stopping electric discharge when 10h or sparking voltage reach cut-off voltage and shelve 1h; Gather discharge current and the single battery voltage of whole discharge process, the time interval of data acquisition is 1min simultaneously.

Step b, specifically comprises:

Adopt least squares identification to go out open-circuit voltage-state-of-charge functional relation F (SOC), process is as follows:

F(SOC)＝a ₀|a ₁*SOC|a ₂*SOC ²|a ₃*SOC ³|…|a _N*SOC ^N(1)

$Y_{N+1}=\left[\begin{array}{c}{\mathrm{OCV}}_0\\ {\mathrm{OCV}}_1\\ .\\ .\\ .\\ {\mathrm{OCV}}_N\end{array}\right]---\left(3\right)$

$X_{N+1}=\left[\begin{array}{c}{a}_0\\ a_1\\ .\\ .\\ .\\ a_N\end{array}\right]={\[{A_{N+1}}^T{A}_{N+1}\]}^{-1}{A}_{N+1}{Y}_{N+1}---\left(4\right)$

In formula (1) (2) (3) (4): SOC ₀for initial state-of-charge; SOC _ifor the state-of-charge after each stage discharge; OCV ₀for initial open circuit voltage; OCV _ifor the open-circuit voltage after each stage discharge; a _ifor the coefficient in formula (1); N is discharge regime number; I=1,2,3 ..., N

Step c, specifically comprises:

Select accumulator Thevenin single order equivalent model, go out battery model parameter by the least squares identification improved, its process is as follows:

In formula (5): y _k: the open-circuit voltage OCV in k moment _kwith single battery voltage V _kdifference, by discharge current I, electric discharge duration estimate SOC, then correspond to F (SOC) and obtain OCV _k; I _k: the discharge current in k moment; α ₁, β ₀, β ₁: coefficient, scalar; X: matrix of coefficients; a _k: the calculation matrix in k moment; P _k: the intermediate variable matrix in k moment; K _k: the intermediate variable matrix in k moment; X _k: the matrix of coefficients in k moment.

C1. initialization

In formula (6): V ₀: the single battery voltage of electric discharge moment first stage; E _{3 × 3}: 3 × 3 unit matrixs; T _s: sampling period, unit s.

C2. parameter identification process

Recursion is carried out until k=30, the α obtained with formula (7) ₁, β ₀, β ₁as the net result of identification.

C3. R, R is calculated _f, C _f

R, R is calculated by formula (8) _f, C _f.

R, R _f, C _falong with the residual capacity of accumulator changes, therefore R, R of each discharge regime of this process identification _f, C _f, for the on-line amending of model.

Steps d, specifically comprises:

With the state-of-charge SOC of accumulator, electric capacity C _fon voltage U _cfor state variable, electric current I is input quantity, and monomer voltage V is output quantity, draws formula (9):

V _k＝F(SOC _k)-U _C，k-RI _k(10)

In formula (9) (10): T: sampling interval, unit s; Q _n: battery rating; I _k: the current value that the k moment gathers is negative during electric discharge, is just during charging; τ _f: time constant, value is τ _f=R _fc _f; η: charge efficiency, when one section, to determine electric current η be 0.65, when two sections of constant-current charge η are 0.8, when two sections determine electric current, constant-voltage charge η is 0.85, η is 1 upon discharging.

After formula (10) linearization, obtain following matrix in conjunction with Kalman filtering algorithm:

D＝-R， $X_k=\left[\begin{array}{c}{\mathrm{SOC}}_k\\ U_{C,k}\end{array}\right]$

Recursive process is as follows:

D1. initialization

$X_0=\left[\begin{array}{c}{\mathrm{SOC}}_0\\ 0\end{array}\right],$ P ₀＝10 ^-5E _2×2，G＝10

Wherein: G: observation noise covariance;

D2. status predication matrix

X _k′＝AX _k-1+BU _C，k-1(11)

D3. predicted value error covariance

P _k′＝AP _k-1A ^T(12)

D4. the prediction of output

V _k′＝F(SOC _k′)-U _C，k-RI _k(13)

D5. optimal filtering gain matrix

L _k＝P _k′C ^T(CP _k′C ^T+DGD ^T) ^-1(14)

D6. state optimization estimated matrix

X _k＝X _k′+L _k(V _k-V _k′)(15)

D7. optimal estimation error covariance

P _k＝(E _2×2-L _kC)P _k′(16)。

Step e, specifically comprises:

Through type (17), draws final SOC;

SOC＝SOC _k(1|K _T(TT ₀)(17)

In formula (17): K _t: temperature coefficient, is taken as 0.08; T: the environment temperature corresponding to current internal resistance; T ₀: the environment temperature 1. corresponding to discharge process.

Step f, specifically comprises:

F1. heavy-current discharge technology is adopted to measure accumulator internal resistance value R ' to accumulator;

F2. contrasted the R value of each discharge regime in step R ' by surveyed internal resistance value, get R ' the differs minimum discharge regime correction of model parameter as current battery model parameter implementation model with R.

The invention has the beneficial effects as follows:

Can estimation on line accumulator SOC fast and accurately, exploitativeness is high, compensate for the deficiency of existing SOC estimation method simultaneously.

Accompanying drawing explanation

Fig. 1 is the voltage-time curve in battery discharging process;

Fig. 2 is the current versus time curve in battery discharging process;

Fig. 3 is accumulator single order equivalent model circuit theory diagrams.

Embodiment

The present invention includes following steps:

A. by stage electric discharge, performance test is carried out to accumulator, obtain the discharge current in discharge test process and monomer voltage data;

B. based on the test data of step a gained, the functional relation of open-circuit voltage-state-of-charge is obtained;

C. based on the test data of step a gained, carry out parameter identification, obtain the battery model parameter of each discharge regime;

D. based on the battery model parameter of the functional relation of step b gained and the first discharge regime of step c gained, by EKF algorithm, estimation accumulator SOC;

E., based on the SOC of steps d gained, after carrying out temperature compensation, final SOC is obtained;

F. regular inner walkway is carried out to accumulator, by the internal resistance value recorded, select the battery model parameter under corresponding discharge regime in step c;

G. by steps d and step e, up-to-date SOC is obtained.

2, the storage battery charge state evaluation method based on EKF according to claim 1, it is characterized in that, described step a, specifically comprises:

To discharge shelving after accumulator off-line after 1h 1h with 0.1C10; Be circulated to accumulated discharge duration according to this, reach after stopping electric discharge when 10h or sparking voltage reach cut-off voltage and shelve 1h; Gather discharge current and the single battery voltage of whole discharge process, the time interval of data acquisition is 1min simultaneously.

3, the storage battery charge state evaluation method based on EKF according to claim 1, it is characterized in that, described step b, specifically comprises:

Adopt least squares identification to go out open-circuit voltage-state-of-charge functional relation F (SOC), process is as follows:

F(SOC)＝a ₀|a ₁*SOC|a ₂*SOC ²|a ₃*SOC ³|…|a _N*SOC ^N(1)

$Y_{N+1}=\left[\begin{array}{c}{\mathrm{OCV}}_0\\ {\mathrm{OCV}}_1\\ .\\ .\\ .\\ {\mathrm{OCV}}_N\end{array}\right]---\left(3\right)$

$X_{N+1}=\left[\begin{array}{c}{a}_0\\ a_1\\ .\\ .\\ .\\ a_N\end{array}\right]={\[{A_{N+1}}^T{A}_{N+1}\]}^{-1}{A}_{N+1}{Y}_{N+1}---\left(4\right)$

In formula (1) (2) (3) (4): SOC ₀for initial state-of-charge; SOC _ifor the state-of-charge after each stage discharge; OCV ₀for initial open circuit voltage; OCV _ifor the open-circuit voltage after each stage discharge; a _ifor the coefficient in formula (1); N is discharge regime number; I=1,2,3 ..., N

4, the storage battery charge state evaluation method based on EKF according to claim 1, it is characterized in that, described step c, specifically comprises:

Select accumulator Thevenin single order equivalent model, go out battery model parameter by the least squares identification improved, its process is as follows:

In formula (5): y _k: the open-circuit voltage OCV in k moment _kwith single battery voltage V _kdifference, by discharge current I, electric discharge duration estimate SOC, then correspond to F (SOC) and obtain OCV _k; I _k: the discharge current in k moment; α ₁, β ₀, β ₁: coefficient, scalar; X: matrix of coefficients; a _k: the calculation matrix in k moment; P _k: the intermediate variable matrix in k moment; K _k: the intermediate variable matrix in k moment; X _k: the matrix of coefficients in k moment.

C1. initialization

In formula (6): V ₀: the single battery voltage of electric discharge moment first stage; E _{3 × 3}: 3 × 3 unit matrixs; T _s: sampling period, unit s.

C2. parameter identification process

Recursion is carried out until k=30, the α obtained with formula (7) ₁, β ₀, β ₁as the net result of identification;

C3. R, R is calculated _f, C _f

R, R is calculated by formula (8) _f, C _f;

R, R _f, C _falong with the residual capacity of accumulator changes, therefore R, R of each discharge regime of this process identification _f, C _f, for the on-line amending of model.

5, the storage battery charge state evaluation method based on EKF according to claim 1, it is characterized in that, described steps d, specifically comprises:

With the state-of-charge SOC of accumulator, electric capacity C _fon voltage U _cfor state variable, electric current I is input quantity, and monomer voltage V is output quantity, draws formula (9):

V _k＝F(SOC _k)-U _C，k-RI _k(10)

In formula (9) (10): T: sampling interval, unit s; Q _n: battery rating; I _k: the current value that the k moment gathers is negative during electric discharge, is just during charging; τ _f: time constant, value is τ _f=R _fc _f; η: charge efficiency, when one section, to determine electric current η be 0.65, when two sections of constant-current charge η are 0.8, when two sections determine electric current, constant-voltage charge η is 0.85, η is 1 upon discharging;

After formula (10) linearization, obtain following matrix in conjunction with Kalman filtering algorithm:

D＝-R， $X_k=\left[\begin{array}{c}{\mathrm{SOC}}_k\\ U_{C,k}\end{array}\right]$

Recursive process is as follows:

D1. initialization

$X_0=\left[\begin{array}{c}{\mathrm{SOC}}_0\\ 0\end{array}\right],$ F ₀＝10 ^-5E _2×2，G＝10

Wherein: G: observation noise covariance;

D2. status predication matrix

X _k′＝AX _k-1+BU _C，k-1(11)

D3. predicted value error covariance

P _k′＝AP _k-1A ^T(12)

D4. the prediction of output

V _k′＝F(SOC _k′)-U _C，k-RI _k(13)

D5. optimal filtering gain matrix

L _k＝P _k′C ^T(CP _k′C ^T+DGD ^T) ^-1(14)

D6. state optimization estimated matrix

X _k＝X _k′+L _k(V _k-V _k′)(15)

D7. optimal estimation error covariance

P _k＝(E _2×2-L _kC)P _k′(16)。

6, according to claim 1 or 5 based on the storage battery charge state evaluation method of EKF, it is characterized in that, described step e, specifically comprises:

Through type (17), draws final SOC;

SOC＝SOC _k(1|K _T(TT ₀))(17)

In formula (17): K _t: temperature coefficient, is taken as 0.08; T: the environment temperature corresponding to current internal resistance; T ₀: the environment temperature 1. corresponding to discharge process.

7, the storage battery charge state evaluation method based on EKF according to claim 1, is characterized in that step f specifically comprises:

F1. heavy-current discharge technology is adopted to measure accumulator internal resistance value R ' to accumulator;

F2. contrasted the R value of each discharge regime in step R ' by surveyed internal resistance value, get R ' the differs minimum discharge regime correction of model parameter as current battery model parameter implementation model with R.

Claims (7)

1., based on a storage battery charge state evaluation method for EKF, it is characterized in that, comprise the steps:

A. by stage electric discharge, performance test is carried out to accumulator, obtain the discharge current in discharge test process and monomer voltage data;

B. based on the test data of step a gained, the functional relation of open-circuit voltage-state-of-charge is obtained;

C. based on the test data of step a gained, carry out parameter identification, obtain the battery model parameter of each discharge regime;

D. based on the battery model parameter of the functional relation of step b gained and the first discharge regime of step c gained, by EKF algorithm, estimation accumulator SOC;

E., based on the SOC of steps d gained, after carrying out temperature compensation, final SOC is obtained;

F. regular inner walkway is carried out to accumulator, by the internal resistance value recorded, select the battery model parameter under corresponding discharge regime in step c;

G. by steps d and step e, up-to-date SOC is obtained.

2. the storage battery charge state evaluation method based on EKF according to claim 1, it is characterized in that, described step a, specifically comprises:

To discharge shelving after accumulator off-line after 1h 1h with 0.1C10; Be circulated to accumulated discharge duration according to this, reach after stopping electric discharge when 10h or sparking voltage reach cut-off voltage and shelve 1h; Gather discharge current and the single battery voltage of whole discharge process, the time interval of data acquisition is 1min simultaneously.

3. the storage battery charge state evaluation method based on EKF according to claim 1, it is characterized in that, described step b, specifically comprises:

Adopt least squares identification to go out open-circuit voltage-state-of-charge functional relation F (SOC), process is as follows:

F(SOC)＝a ₀|a ₁*SOC|a ₂*SOC ²|a ₃*SOC ³|…|a _N*SOC ^N(1)

$Y_{N+1}=\left[\begin{array}{c}{\mathrm{OCV}}_0\\ {\mathrm{OCV}}_1\\ .\\ .\\ .\\ {\mathrm{OCV}}_N\end{array}\right]---\left(3\right)$

$X_{N+1}=\left[\begin{array}{c}{a}_0\\ a_1\\ .\\ .\\ .\\ a_N\end{array}\right]={\[{A_{N+1}}^T{A}_{N+1}\]}^{-1}{A}_{N+1}{Y}_{N+1}---\left(4\right)$

In formula (1) (2) (3) (4): SOC ₀for initial state-of-charge; SOC _ifor the state-of-charge after each stage discharge; OCV ₀for initial open circuit voltage; OCV _tfor the open-circuit voltage after each stage discharge; a _ifor the coefficient in formula (1); N is discharge regime number; I=1,2,3 ..., N.

4. the storage battery charge state evaluation method based on EKF according to claim 1, it is characterized in that, described step c, specifically comprises:

Select accumulator Thevenin single order equivalent model, go out battery model parameter by the least squares identification improved, its process is as follows:

In formula (5): y _k: the open-circuit voltage OCV in k moment _kwith single battery voltage V _kdifference, by discharge current I, electric discharge duration estimate SOC, then correspond to F (SOC) and obtain OCV _k; I _k: the discharge current in k moment; α ₁, β ₀, β ₁: coefficient, scalar; X: matrix of coefficients; α _k: the calculation matrix in k moment; P _k: the intermediate variable matrix in k moment; K _k: the intermediate variable matrix in k moment; X _k: the matrix of coefficients in k moment;

C1. initialization

In formula (6): V ₀: the single battery voltage of electric discharge moment first stage; E _{3 × 3}: 3 × 3 unit matrixs; T _s: sampling period, unit s;

C2. parameter identification process

Recursion is carried out until k=30, the α obtained with formula (7) ₁, β ₀, β ₁as the net result of identification;

C3. R, R is calculated _f, C _f

R, R is calculated by formula (8) _f, C _f;

R, R _f, C _falong with the residual capacity of accumulator changes, therefore R, R of each discharge regime of this process identification _f, C _f, for the on-line amending of model.

5. the storage battery charge state evaluation method based on EKF according to claim 1, it is characterized in that, described steps d, specifically comprises:

With the state-of-charge SOC of accumulator, electric capacity C _fon voltage U _cfor state variable, electric current I is input quantity, and monomer voltage V is output quantity, draws formula (9):

V _k＝F(SOC _k)-U _z，k-RI _k(10)

In formula (9) (10): T: sampling interval, unit s; Q _n: battery rating; I _k: the current value that the k moment gathers is negative during electric discharge, is just during charging; τ _f: time constant, value is τ _f=R _fc _f; η: charge efficiency, when one section, to determine electric current η be 0.65, when two sections of constant-current charge η are 0.8, when two sections determine electric current, constant-voltage charge η is 0.85, η is 1 upon discharging;

After formula (10) linearization, obtain following matrix in conjunction with Kalman filtering algorithm:

Recursive process is as follows:

D1. initialization

Wherein: G: observation noise covariance;

D2. status predication matrix

X _k′＝AX _k-1+BU _c，k-1(11)

D3. predicted value error covariance

P _k′＝AP _k-1A ^T(12)

D4. the prediction of output

V _k′＝F(SOC _k′)-U _C，k-RI _k(13)

D5. optimal filtering gain matrix

L _k＝P _k′C ^T(CP _k′C ^T+DGD ^T) ^-1(14)

D6. state optimization estimated matrix

X _k＝X _k′+L _k(V _k-V _k′)(15)

D7. optimal estimation error covariance

P _k＝(E _2×2-L _kC)P _k′(16)。

6., according to claim 1 or 5 based on the storage battery charge state evaluation method of EKF, it is characterized in that, described step e, specifically comprises:

Through type (17), draws final SOC;

SOC＝SOC _k(1|K _T(TT ₀))(17)

In formula (17): K _t: temperature coefficient, is taken as 0.08; T: the environment temperature corresponding to current internal resistance; T ₀: the environment temperature 1. corresponding to discharge process.

7. the storage battery charge state evaluation method based on EKF according to claim 1, is characterized in that step f specifically comprises:

F1. heavy-current discharge technology is adopted to measure accumulator internal resistance value R ' to accumulator;

F2. contrasted the R value of each discharge regime in step R ' by surveyed internal resistance value, get R ' the differs minimum discharge regime correction of model parameter as current battery model parameter implementation model with R.