CN104267354B - A kind of peak power Forecasting Methodology of electrokinetic cell - Google Patents

A kind of peak power Forecasting Methodology of electrokinetic cell Download PDF

Info

Publication number
CN104267354B
CN104267354B CN201410592570.4A CN201410592570A CN104267354B CN 104267354 B CN104267354 B CN 104267354B CN 201410592570 A CN201410592570 A CN 201410592570A CN 104267354 B CN104267354 B CN 104267354B
Authority
CN
China
Prior art keywords
battery
moment
fractional order
voltage
value
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201410592570.4A
Other languages
Chinese (zh)
Other versions
CN104267354A (en
Inventor
朱春波
李晓宇
魏国
裴磊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Harbin Institute of Technology
Original Assignee
Harbin Institute of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Harbin Institute of Technology filed Critical Harbin Institute of Technology
Priority to CN201410592570.4A priority Critical patent/CN104267354B/en
Publication of CN104267354A publication Critical patent/CN104267354A/en
Application granted granted Critical
Publication of CN104267354B publication Critical patent/CN104267354B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Abstract

A kind of peak power Forecasting Methodology of electrokinetic cell, is related to the peak power Predicting Technique of electrokinetic cell.The present invention is the accuracy in order to improve the prediction of the peak power of electrokinetic cell.The method of the present invention includes two parts:The battery peak power Forecasting Methodology that A is decomposed based on zero state response and zero input response based on the electrochemical impedance spectroscopy equivalent-circuit model simplifying and the parameter On-line Estimation of fractional order federated Kalman filtering, B.The present invention has selected the reference model that the simplification impedance spectrum model comprising fractional order element is predicted as battery peak power, therefore proposed by the present invention not only can accurately predict the maximum output in short-term of battery it is also possible to peak power fan-out capability in longer period for the Accurate Prediction battery based on the electrokinetic cell peak power Forecasting Methodology of this model.The present invention is applied to the prediction of electric automobile online peak power.

Description

A kind of peak power Forecasting Methodology of electrokinetic cell
Technical field
The present invention relates to the peak power Predicting Technique of electrokinetic cell.
Background technology
The content of the relevant battery peak power Forecasting Methodology disclosed in domestic patent is less, in paper periodical both domestic and external The method having the prediction of some electrokinetic cell peak powers.
These methods mostly adopt the single order RC equivalent-circuit model of battery, using SOC, terminal voltage, electric current as its peak work The restrictive condition of rate prediction, the power forecasting method of early stage adopts the complete battery pair of offline battery model supplemental characteristic to export/return Feedback peak power prediction, but battery with the change of the environment such as aging or temperature, operating mode, model parameter also can change therewith Become, therefore, the power estimation method reliability based on off-line model supplemental characteristic is poor;
In the up-to-date paper delivered, battery model supplemental characteristic is obtained online by method for parameter estimation, can effectively carry The accuracy of high estimation, but because the error of single order RC model itself is larger, power prediction can only be confined to the shorter time It is inside effective.
In order to reduce discretization error when for power prediction for the state equation, when adopting less discretization in paper Between be spaced (such as taking time interval to be 1s or 0.1s), by the method for recurrence calculation, obtain battery 10s after current time Predictive value at the peak power predictive value at place or 60s, the shortcoming of this method is that there is substantial amounts of recurrence calculation, real-time Difference.In addition this method also cannot improve the power inaccurate problem of estimation causing due to model oneself error.
Content of the invention
The present invention is the peak power forecasting reliability in order to improve electrokinetic cell, reduces amount of calculation, thus providing one Plant the peak power Forecasting Methodology of electrokinetic cell.
A kind of peak power Forecasting Methodology of electrokinetic cell, it is realized by following steps:
Step A, the step of battery model parameter On-line Estimation, specially:
Step A1, when modeling to secondary cell, due to electric automobile operating condition frequency characteristic, therefore battery electrochemical The impedance operator of the medium frequency in impedance spectrum model can be by conventional purely resistive element R and the letter of normal phase element Q parallel circuit Turn to purely resistive element R to describe, the battery electrochemical impedance spectrum equivalent-circuit model after being simplified;
Electrochemical impedance spectroscopy equivalent-circuit model after this simplification includes open-circuit voltage OCVe, ohmic internal resistance RoWith weber resistance Anti- ZW
Step A2, according to step A1 obtain simplification after electrochemical impedance spectroscopy equivalent-circuit model set up fractional order karr State equation needed for graceful wave filter and observational equation, specially:
Take total current I flowing through secondary cellLIt is on the occasion of data sampling period is 1s in electric discharge;
Wherein △rFor differential operator, r is differential order, when r is for decimal, △rRepresent fractional order differential operator, when r is During integer, △rFor integer differential operator;
Take fractional order element ZWBe both end voltage be UWQuantity of state, have:
For battery model parameter, diffusion parameter XW, open-circuit voltage OCVeWith ohmic internal resistance RoWith battery charge state (SOC) change is slow, therefore:
Aforementioned four equation is rewritten as matrix form, obtains the state equation of fractional order Kalman integrated filter:
Take ULFor the observed quantity of system, then have:
UL=OCVe-ILRo-UW
ILThe total current representing and flowing through battery;
Take:
Y=UL
Obtain the observational equation of fractional order Kalman integrated filter:
After equation discretization, have:
Wherein, w, v represent state-noise and the observation noise of system respectively;
Series definition (being also called the definition of Gr ü nwald-Letnikov fractional order differential) according to fractional order differential:
Wherein,
Separately take:Obtain the discretization recursion expression-form of Fractional Differential Equation by above formula:
Definition:
According to the series definition of fractional order differential, wherein:Amount of calculation will increase over time And constantly increase, this situation is not suitable for engineer applied, for this reason, above formula is rewritten as following form:
Step A3, the state equation needed for fractional order Kalman filter using step A2 structure and observational equation, right State, parameter and covariance matrix carry out time renewal and measurement updaue according to fractional order federated Kalman filtering algorithm:
It is specially:
Initialization:
Wherein, E [x] represents the mathematic expectaion of x, is experience preset value when method calculates,Represent x in initial time (k =0) estimated value,Represent the estimated value of the noise covariance in initial time (k=0) for the x;
The time of state, parameter and covariance matrix updates:
Wherein, QkIt is noise wkCovariance,For k moment state and model parameter xkPredictive value,During for k-1 Quarter state and model parameter xk-1Correction value,Noise covariance matrix P for k moment xkPredictive value,During for k-1 Carve the noise covariance matrix P of xk-1Correction value;
The measurement updaue of state, parameter and covariance matrix:
Wherein, RkIt is noise vkCovariance, LkIt is k moment Kalman filter gain size;
Step A4, terminal voltage U of collection batteryLWith total current I flowing through secondary cellL, using the simplification of step A1 acquisition Rear electrochemical impedance spectroscopy equivalent-circuit model and step A3 update after the state equation needed for fractional order Kalman filter With observational equation, obtain open-circuit voltage OCVe, ohmic internal resistance Ro, diffusion parameter XWEstimated value, will obtain open-circuit voltage OCVe, ohmic internal resistance Ro, diffusion parameter XWEstimated value as battery estimated result, complete based on fractional order joint karr The secondary cell of graceful filtering simplifies impedance spectrum model parameter On-line Estimation;
Step B, the battery model parameter On-line Estimation result being obtained according to step A, the step carrying out power prediction:
Step B1, the weber impedance Z of derivation fractional order elementWWhether it is linear time invariant element:
If both end voltage U of fractional order elementWInitial value be 0, be I when applying amplitudeLStep current excitation When, both end voltage U of fractional order elementWVoltage response situation after Ns is calculated as follows, and N is positive number:
If the open-circuit voltage OCV of battery modele, ohmic internal resistance Ro, diffusion parameter XWNumerical values recited during power prediction It is constant;
Then UWVoltage response in k=1s is:
Wherein, symbol ^ represents predictive value;
UWVoltage response when k >=2 is:
Calculate:
After calculating:
By the voltage response that this recursion obtains 10s and 60s it is:
Thus, when battery model parameter XWWhen constant or slowly varying, output valve UWWith ILFor linear relationship, infer and divide Number rank element ZWFor linear time invariant element;
Then, the power forecasting method of fractional order element is:
In order to predict voltage response at k+ Δ Ts for the fractional order elementThis voltage response is divided into zero state ResponseAnd zero input response
Wherein zero state response is:
For k+10s moment, a=3.524;For k+60s moment, a=8.722;
Zero input responseDetermined by the data before the k moment, take time memory length L=of fractional order differential 60;
The zero input response in k+1 moment is:
The response of k+ Δ T moment fractional order element zero input voltage be can get by this recursion:
Wherein,It is battery model fractional order element ZWEstimation between (k+1-L)~k moment for the two ends terminal voltage Value, when predicting k+10 moment, b=α, α are one group of constant coefficient matrix;
When predicting k+60 moment, b=β, β are another set constant coefficient matrix;
Thus, can get battery constant-current discharge electric current is ImaxWhen battery the k+ Δ T moment terminal voltage predictive value:
Electric discharge peak value power prediction method be:
If battery charge state (SoC or SOC) is the restrictive condition of battery limit working condition, SoCminIt is battery discharge Terminate the minima of state-of-charge, then obtain maximum discharge current value now:
Capacity is battery capacity value, and unit is ampere * hour (Ah), SoCkFor the battery charge state in k moment, SoCminThe minimum state-of-charge limiting for battery discharge, the minimum state-of-charge that such as certain power battery for hybrid electric vehicle adopts For SoCmin=30%;
If battery terminal voltage ULFor the restrictive condition of battery limit working condition, if now with maximum discharge current ImaxRight Battery discharge, battery model parameter OCVePredictive value in the k+ △ T moment is:
In above formula,It is by calculated battery model parameter OCV of fractional order federated Kalman filtering algorithme In the estimated value in k moment,For OCVePredictive value in the k+ △ T moment;ImaxIt is as battery limit work in terminal voltage Make maximum discharge current value during state limit condition, be value to be solved,It is the maximum discharge current of battery;OCVkIt is in k The battery open circuit voltage values in moment, can be calculated by OCV (SoC) pre-defined function and obtain it is generally recognized that OCV and SoC has clearly Corresponding relation;The functional relationship of the pre-defined function of OCV and SoC can be obtained by handbook of batteries or test;For electricity Pond in the k+ △ T moment it is assumed that withCorresponding battery open circuit voltage values during constant-current discharge, equally can be predefined by OCV (SoC) Function calculates and obtains;Due in most cases, the change of OCV is all less and slow, it can thus be assumed that opening within the △ T period Road voltage variety is linear with discharge current;
And then calculate it is assumed that battery is I in discharge currentmaxWhen, estimated value U of battery terminal voltageL,k+△T
Wherein, the estimated value of terminal voltage be k moment open-circuit voltage estimated value, open circuit voltage variations value, ohm in the △ T period Internal resistance voltage difference, fractional order element ZWZero state response magnitude of voltage, zero input response magnitude of voltage sum;
Thus release, when with terminal voltage ULAs the extreme working position of battery constraints when, the maximum work of battery As electric current it is:
When the k+ Δ T moment reaches crest discharge power, consider above-mentioned restrictive condition, maximum discharge current value is:
The crest discharge power in k+ Δ T moment is:
The Forecasting Methodology of feedback current peak power of battery is with the Forecasting Methodology of above-mentioned discharge power in the same manner:
If ILBe on the occasion of;
If SoC is the restrictive condition of battery limit working condition, SoCmaxIt is the maximum SOC of battery, then obtain Minimum feedback current value now:
If ULFor the restrictive condition of battery limit working condition, then obtain minimum feedback current value now:
When the k+ Δ T moment reaches peak value feedback power, consider above-mentioned restrictive condition, minimum feedback current value is:
Thus, obtain the peak power of battery current feedback:
Complete the peak power prediction of electrokinetic cell.
In step A, due to the electrochemical impedance spectroscopy test data based on battery for the battery model, battery mould used in method Shape parameter has clear and definite physical significance, ohmic internal resistance RoPhysical significance be:
Ro≈RΩ+RSEI+Rct
Wherein, RΩFor high frequency ohmage, RSEIFor SEI membrane impedance, RctFor Charge-transfer resistance;
In addition, weber impedance to be defined by below equation:
Wherein, W is ionic diffusion coefficient, for the ease of impedance parameter On-line Estimation, takes:
Obtain:
Beneficial effects of the present invention:
1st, the present invention has selected the ginseng that the simplification impedance spectrum model comprising fractional order element is predicted as battery peak power Examine model, because fractional order element has the memory characteristic of long period, the power based on this model therefore proposed by the present invention Battery peak power Forecasting Methodology not only can accurately predict the maximum output in short-term of battery it is also possible to Accurate Prediction is electric Peak power fan-out capability in longer period for the pond.
2nd, the method for the present invention, under multiple electric automobile operating conditions, can keep preferable peak power prediction energy Power, the adaptability for working condition of the inventive method is better than the online power prediction algorithm based on single order RC model for the tradition.
3rd, the method adopts zero state response and the method for zero input response decomposition to predict in k+10 moment and k+60 moment Voltage response situation, eliminate k~k+10, the state recursion link between k~k+60, greatly reduce peak power prediction Amount of calculation it is adaptable to electric automobile online peak power prediction.
Brief description
Fig. 1 is the battery impedance spectroscopy equivalent-circuit model simplifying;
Fig. 2 is that the current value emulation schematic diagram collecting when work operating mode is run simulated by electrokinetic cell battery;
This simulated condition is made up of the United States Federal's city operating mode (FUDS operating mode) and discharge and recharge pulse operation, for authentication The reliability of method.
Fig. 3 is the comparison diagram that battery terminal voltage predicted the outcome and surveyed terminal voltage;
Wherein UL surveys terminal voltage for battery, and ULd10 assumes that battery reaches during electric discharge peak value working condition in 10s Terminal voltage predictive value, ULd60 assumes that battery reaches terminal voltage predictive value during electric discharge peak value working condition, ULr60 in 60s Assume that battery reaches terminal voltage predictive value during energy feedback peak value working condition in 10s.Contrast is it is found that ULd10 With ULd60 respectively at 10s discharge pulse operating mode and 60s discharge pulse operating mode with UL closely, thus it can be extrapolated that the present invention Method has preferable battery discharge characteristic predictive ability.
Fig. 4 is 10s electric discharge peak power and the 60s electric discharge peak value power prediction value of battery.
As shown in Figure 4, the inventive method has preferable electric discharge peak value power prediction ability.
Specific embodiment
Specific embodiment one, a kind of peak power Forecasting Methodology of electrokinetic cell,
The invention discloses a kind of peak estimation method of electrokinetic cell, the method includes two parts:A is based on the electricity simplifying Chemical impedance composes the parameter On-line Estimation of equivalent-circuit model and fractional order federated Kalman filtering, B is based on zero state response and The battery peak power Forecasting Methodology that zero input response is decomposed.
The method be specially:Based on the electrochemical impedance spectroscopy equivalent-circuit model simplifying, the filter of fractional order federated Kalman The power of battery Forecasting Methodology that ripple, zero state response and zero input response are decomposed.It is realized by following steps:
Step one, the electrochemical impedance spectroscopy test result according to battery, due in impedance spectrum, by electrochemical impedance spectroscopy etc. Effect circuit model has done and has simplified further, the electrochemical impedance spectroscopy equivalent-circuit model after being simplified, as shown in figure 1, UtAnd IL Represent the terminal voltage of battery and the total current flowing through battery respectively.
The impedance spectrum equivalent-circuit model of this simplification includes OCVe、RoAnd ZWThree elements.
Wherein, OCVeFor being combined open-circuit voltage, main reflection battery open circuit voltage characteristic, because equivalent-circuit model simplifies Many processes of cell dynamics processes, and have ignored each dynamic (dynamical) boundary condition of battery charge and discharge process, because This is due to this battery model error of itself, OCVeIt is the approximation of OCV, numerically mainly contain OCV and least a portion of Other chemical reaction potential value such as ion diffusion polarization potential.
OCVe≈OCV
RoFor being combined ohmic internal resistance, this parameter mainly reflects the medium-high frequency ohm impedance characteristic of battery electrochemical impedance spectrum (frequency is more than 0.5Hz), this parameter numerically approximates high frequency ohmage (RΩ), SEI membrane impedance (RSEI), electric charge transfer Impedance (Rct) impedance sum.
Ro≈RΩ+RSEI+Rct
ZWIt is used to describe the weber impedance (Warburg) of the ion diffusion polarization characteristic of battery, UWFor weber impedance two ends Voltage.Many phenomenons of nature meet fractional order characteristic, and the ion diffusion property process of battery charge and discharge process is especially such as This.It is that knowable to Qwest's figure, ion diffusion process meets fractional order differential characteristic, and this characteristic is normal from the electrochemical impedance spectroscopy of battery Represented with fractional order physical component weber impedance.Weber impedance to be defined by below equation:
Wherein, W is ionic diffusion coefficient, for the ease of impedance parameter On-line Estimation, takes:Obtain:
The feature of this impedance spectrum equivalent-circuit model is a simplified Conventional impedance and composes the high frequency (frequency in equivalent-circuit model More than 1kHz) and intermediate frequency impedance (frequency is more than 0.5Hz, and is less than 1kHz), from the associated description of test data and other paper From the point of view of, above-mentioned simplification impedance spectrum model can effectively reduce model parameter quantity, is suitable for the On-line Estimation of model parameter.
Step 2:State equation according to needed for above-mentioned equivalent-circuit model sets up fractional order Kalman filter and observation Equation:
Quantity of state based on fractional order Kalman integrated filter estimating circuit and parameter value, specific method is as follows.
First, take ILIt is on the occasion of data sampling period is 1s in electric discharge.
1st, row write state equation and the observational equation of fractional order Kalman integrated filter:
Wherein:△rFor differential operator, r is differential order, when r is for decimal, △rRepresent fractional order differential operator, when r is During integer, △rFor integer differential operator.
Take fractional order element ZWIt is both end voltage UWFor quantity of state, have:
For parameter XW, OCVe, RoChange with battery charge state (SoC) is slow, therefore:
Aforementioned four equation is rewritten as matrix form, has:
Take ULFor the observed quantity of system, then have:
UL=OCVe-ILRo-UW
Take:
Y=UL
Then have:
After above-mentioned equation discretization, have:
Wherein, w, v represent state-noise and the observation noise of system respectively, generally it will be assumed that both are independent noise.
Defined according to Gr ü nwald-Letnikov fractional order differential:
Wherein:
Separately take:
Arrive the discretization recursion expression-form of Fractional Differential Equation as available from the above equation:
Definition:
According to Gr ü nwald-Letnikov fractional order differential definition, whereinAmount of calculation will be with The increase of time and constantly increase, this situation is not suitable for engineer applied, for this reason, above formula is rewritten as following form:
Step 2, utilize fractional order Kalman integrated filter estimated state and parameter value
Initialization:
Wherein, E [x] represents the mathematic expectaion of x, is experience preset value when method calculates,Represent x in initial time (k =0) estimated value,Represent the estimated value of the noise covariance in initial time (k=0) for the x.
The time of state, parameter and covariance matrix updates:
Wherein, QkIt is noise wkCovariance,For k moment state and model parameter xkPredictive value,For the k-1 moment State and model parameter xk-1Correction value,Noise covariance matrix P for k moment xkPredictive value,For k-1 moment x Noise covariance matrix Pk-1Correction value.
The measurement updaue of state, parameter and covariance matrix:
Wherein, RkIt is noise vkCovariance, LkIt is k moment Kalman filter gain size.
Power prediction part:
The basic thought of power prediction:
In use, the size of output power value is determined electrokinetic cell by load characteristic.Different load classes Type, load operating mode all can lead to the power of battery to export the difference of situation.The purpose of maximum output prediction is in reliability While protection battery system, predict the maximum power output ability in certain moment for the battery, to have given play to the maximality of battery Energy.The premise of peak power prediction is reliably to protect battery system, prevents its super-charge super-discharge.
Key point has two:The reference operating mode of the constraints of cell operating status and power prediction;
The limiting constraint of cell operating status:The instruction manual of reference battery, the major constraints of cell operating status Condition has:Maximum charging current, maximum discharge current, charge cutoff voltage, discharge cut-off voltage, SoC operation interval etc..At this In invention, obtaining current, voltage, SoC are as constraints.
The reference operating mode of power prediction:If prediction battery, in the power output capacity in k+ Δ T moment, needs to know k~k+ Battery applying working condition between Δ T, but this operating mode is unknown.Conventional standard charged/discharged operating mode has:Constant current, constant voltage, Constant power load operating mode.Wherein, the amount of calculation minimum of battery peak power is solved based on constant current operating mode and battery can be reflected Power output capacity, therefore takes constant current operating mode can effectively prevent over-charging of battery mistake as the work operating mode between k~k+ Δ T Put.
1st, the Z of derivation fractional order elementWWhether it is linear time invariant element:
Assume UWInitial value be 0, be I when applying amplitudeLStep current excitation when, voltage after Ns for the Uw rings Situation is answered to be calculated as follows:
Assume battery model parameter OCVe、Ro、XWDuring power prediction, numerical values recited is constant.
UWVoltage response in k=1s is:
UWVoltage response when k >=2 is:
Calculate:
Can obtain after calculating:
By the voltage response that this recursion obtains 10s and 60s it is:
Thus it is easy to infer ZWElement is linear time invariant element.
Power prediction:
Due to ZWElement is linear time invariant element, and response under current excitation for the element can be according to the superposition theorem of circuit Decomposed.In order to predict ZWVoltage response at k+ Δ Ts, is classified as zero state response and zero input response:
Wherein:
Wherein zero state response is:
For k+10s moment, a=3.524;For k+60s moment, a=8.722.
Zero input responseDetermined by the data before the k moment, take time memory length L=of fractional order differential 60.
The zero input response in k+1 moment is:
K+ Δ T moment battery model element Z be can get by this recursionWLocate zero input voltage response:
Wherein, when the k+10 moment, b=α, α are constant coefficient matrix.
When the k+60 moment, b=β, β are constant coefficient matrix.
Thus, can get battery constant-current discharge electric current is ImaxWhen battery the k+ Δ T moment voltage prediction value.
For the peak power prediction of energy type electrokinetic cell, take SoC bound, bound blanking voltage, maximum charge and discharge Electric current is as the extreme working position of battery.
When wherein, any one parameter value preferentially reaches during its restrained boundary it is believed that battery is assigned for this numerical value in this parameter To extreme working position, the now peak power predictive value in calculated input/output performance number as k+ Δ T moment.
The step of peak power prediction is as follows:
Electric discharge peak value power prediction:
If SoC is the restrictive condition of battery limit working condition,:
If ULFor the restrictive condition of battery limit working condition, then:
Thus release, when terminal voltage UtAs the extreme working position of battery constraints when, operating current is:
When the k+ Δ T moment reaches crest discharge power, discharge current value is:
The crest discharge power in k+ Δ T moment is:
Similar with electric discharge peak value power method, the feedback current peak power of battery can be calculated:
Now ILBe on the occasion of.
If SoC is the restrictive condition of battery limit working condition,:
If ULFor the restrictive condition of battery limit working condition, then:
When the k+ Δ T moment reaches peak value feedback power, feedback current value is:
Thus, obtain the peak power of battery current feedback:
If it follows that battery is run with constant current mode in k to k+ Δ T time, the electric discharge its peak work in k+ Δ T moment Rate, current feedback peak power are obtained by above formula prediction respectively.
If battery operated under other mode of operations, such as constant voltage or constant power mode, made with above-mentioned peak power Extreme working position for battery can effectively prevent from over-charging of battery from crossing putting.

Claims (2)

1. a kind of peak power Forecasting Methodology of electrokinetic cell, is characterized in that:It is realized by following steps:
Step A, the step of battery model parameter On-line Estimation, specially:
Step A1, when modeling to secondary cell, due to electric automobile operating condition frequency characteristic, therefore battery electrochemical impedance The impedance operator of the medium frequency in spectrum equivalent-circuit model is by conventional purely resistive element R and the letter of normal phase element Q parallel circuit Turn to purely resistive element R to describe, the battery electrochemical impedance spectrum equivalent-circuit model after being simplified;
Battery electrochemical impedance spectrum equivalent-circuit model after this simplification includes open-circuit voltage OCVe, ohmic internal resistance RoAnd fractional order Element ZW
Open-circuit voltage OCVePositive pole and ohmic internal resistance RoOne end connect;Described ohmic internal resistance RoThe other end and fractional order unit Part ZWOne end connect;Open-circuit voltage OCVeNegative pole and fractional order element ZWThe other end as cell voltage ULTwo ends;
Step A2, according to step A1 obtain simplification after battery electrochemical impedance spectrum equivalent-circuit model set up fractional order joint State equation needed for Kalman filter and observational equation, specially:
Take total current I flowing through secondary cellLIt is on the occasion of data sampling period is 1s in electric discharge;
Δ r = d r dt r , r > 0
Wherein ΔrFor differential operator, r is differential order, when r is for decimal, ΔrRepresent fractional order differential operator, when r is integer When, ΔrFor integer differential operator;
Take fractional order element ZWBe both end voltage be UWQuantity of state, have:
Δ 0.5 U W = 1 W I L = X W I L
For battery model parameter, diffusion parameter XW, open-circuit voltage OCVeWith ohmic internal resistance RoWith battery charge state SoC's Change is slow, therefore:
Δ 1 X W ≈ 0 Δ 1 O C V e ≈ 0 Δ 1 R o ≈ 0
Will:
Δ 0.5 U W = 1 W I L = X W I L
Δ 1 X W ≈ 0 Δ 1 O C V e ≈ 0 Δ 1 R o ≈ 0
It is rewritten as matrix form, obtain the state equation of fractional order Kalman integrated filter:
Δ 0.5 1 1 1 U W X W OCV e R o = 0 I L 0 0 0 0 0 0 0 0 0 0 0 0 0 0 U W X W OCV e R o ;
Take ULFor the observed quantity of system, then have:
UL=OCVe-ILRo-UW
ILThe total current representing and flowing through battery;
Take:
Y=UL
Obtain the observational equation of fractional order Kalman integrated filter:
Δ N x = 0 I L 0 0 0 0 0 0 0 0 0 0 0 0 0 0 x y = [ - 1 0 1 - I L ] x
After the observational equation discretization of this fractional order Kalman integrated filter, have:
Δ N x k = 0 I L , k - 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 x k - 1 + w y k = [ - 1 0 1 - I L , k ] x k + v
Wherein, w, v represent state-noise and the observation noise of system respectively;
Series definition according to fractional order differential:
Δ N x k + 1 = Σ j = 0 k ( - 1 ) j N j x k - j
Wherein,
N j = d i a g [ 0.5 j 1 j 1 j 1 j ] ,
r j = 1 f o r j = 0 r ( r - 1 ) ... ( r - j + 1 ) / j ! f o r j > 0 ,
Separately take:Obtain the discretization recursion expression-form of Fractional Differential Equation by above formula:
Definition:
A k - 1 = ∂ f ( x k - 1 , I L , k - 1 ) ∂ x k - 1 | x k - 1 = x ^ k - 1 + = 0 I L , k - 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ,
C k = ∂ g ( x k , I L , k ) ∂ x k | x k = x ^ k - = [ - 1 0 1 - I L , k ]
According to the series definition of fractional order differential, wherein:Amount of calculation will increase over time and not Disconnected increase, this situation is not suitable for engineer applied, for this reason, above formula is rewritten as following form:
Σ j = 1 k ( - 1 ) j γ j x k + 1 - j = Σ j = 1 L ( - 1 ) j γ j x k + 1 - j , k ≤ 64 , L = k k > 64 , L = 64
Step A3, the state equation needed for fractional order Kalman integrated filter using step A2 structure and observational equation, right State, parameter and covariance matrix carry out time renewal and measurement updaue according to fractional order federated Kalman filtering algorithm:
It is specially:
Initialization:
x ^ 0 = E [ x ] , P 0 + = E [ ( x - x ^ 0 ) ( x - x ^ 0 ) T ]
Wherein, E [x] represents the mathematic expectaion of x, is experience preset value when method calculates,Represent the estimation in initial time for the x Value,Represent the estimated value of the noise covariance in initial time for the x;
The time of state, parameter and covariance matrix updates:
x ^ k - = f ( x ^ k - 1 + , I L , k - 1 )
P k - = ( A k - 1 + γ 1 ) P k - 1 + ( A k - 1 + γ 1 ) T + Q k + Σ j = 2 L γ j P k - j + γ j T
Wherein, QkIt is noise wkCovariance,For k moment state and model parameter xkPredictive value,For k-1 moment state With model parameter xk-1Correction value,Noise covariance matrix P for k moment xkPredictive value,For making an uproar of k-1 moment x Sound covariance matrix Pk-1Correction value;
The measurement updaue of state, parameter and covariance matrix:
L k = P k - ( C k ) T [ C k P k - ( C k ) T + R k ] - 1
x ^ k + = x ^ k - + L k x [ y k - g ( x ^ k - , I L , k ) ]
P k + = ( I k - L k C k ) P k -
Wherein, RkIt is noise vkCovariance, LkIt is k moment Kalman filter gain size;
Step A4, terminal voltage U of collection batteryLWith total current I flowing through secondary cellL, using after the simplification that step A1 obtains The state needed for fractional order Kalman integrated filter after battery electrochemical impedance spectrum equivalent-circuit model and the renewal of step A3 Equation and observational equation, obtain open-circuit voltage OCVe, ohmic internal resistance Ro, diffusion parameter XWEstimated value, will obtain open-circuit voltage OCVe, ohmic internal resistance Ro, diffusion parameter XWEstimated value as battery estimated result, complete based on fractional order joint karr The secondary cell of graceful filtering simplifies impedance spectrum model parameter On-line Estimation;
Step B, the battery model parameter On-line Estimation result being obtained according to step A, the step carrying out power prediction:
Step B1, derivation fractional order element ZWWhether it is linear time invariant element:
If both end voltage U of fractional order elementWInitial value be 0, be I when applying amplitudeLStep current excitation when, point Both end voltage U of number rank elementWVoltage response situation after N s is calculated as follows, and N is positive number:
If the open-circuit voltage OCV of battery modele, ohmic internal resistance Ro, diffusion parameter XWDuring power prediction, numerical values recited is not Become;
Then UWVoltage response in k=1s is:
U ^ W , 1 = X ^ W , 0 I L - ( - 1 ) 1 0.5 1 U W , 0
Wherein, symbol ^ represents predictive value;
UWVoltage response when k >=2 is:
U ^ W , k + 1 = X ^ W , 0 I L - Σ j = 1 L ( - 1 ) j 0.5 j U ^ W , k + 1 - j + w
Calculate:
U ^ W , 1 = X ^ W I L
U ^ W , 2 = X ^ W , 0 I L - ( ( - 1 ) 1 0.5 1 U ^ W , 1 + ( - 1 ) 2 0.5 2 U W , 0 )
After calculating:
U ^ W , 2 = 1.5 X ^ W , 0 I L
By the voltage response that this recursion obtains 10s and 60s it is:
U ^ W , 10 = 3.524 X ^ W , 0 I L
U ^ W , 60 = 8.722 X ^ W , 0 I L
Thus, when battery model parameter XWWhen constant or slowly varying, output valve UWWith ILFor linear relationship, infer fractional order Element ZWFor linear time invariant element;
Then, the power forecasting method of fractional order element is:
In order to predict voltage response at k+ Δ Ts for the fractional order elementThis voltage response is divided into zero state responseAnd zero input response
U ^ W , k + Δ T = U ^ W , k + Δ T z s + U ^ W , k + Δ T z i
Wherein zero state response is:
U ^ W , k + Δ T z s = a X ^ W , k I max
For k+10s moment, a=3.524;For k+60s moment, a=8.722;
Zero input responseDetermined by the data before the k moment, take time memory length L=60 of fractional order differential;
The zero input response in k+1 moment is:
U ^ W , k + 1 z i = - Σ j = 1 L ( - 1 ) j 0.5 j U ^ W , k + 1 - j z i
The response of k+ Δ T moment fractional order element zero input voltage be can get by this recursion:
U ^ W , k + Δ T z i = Σ j = 1 L b k + 1 - j U ^ W , k + 1 - j
Wherein,It is battery model fractional order element ZWEstimated value between (k+1-L)~k moment for the two ends terminal voltage, When predicting k+10 moment, b=α, α are one group of constant coefficient matrix;
When predicting k+60 moment, b=β, β are another set constant coefficient matrix;
Thus, can get battery constant-current discharge electric current is ImaxWhen battery the k+ Δ T moment terminal voltage predictive value:
Electric discharge peak value power prediction method be:
If battery charge state is the restrictive condition of battery limit working condition, SoCminIt is that battery discharge terminates state-of-charge Minima, then obtain maximum discharge current value now:
I m a x , S o C = ( SoC k - SoC m i n ) × C a p a c i t y Δ T
Capacity is battery capacity value, and unit is ampere * hour (Ah), SoCkFor the battery charge state in k moment, SoCminFor The minimum state-of-charge that battery discharge limits;
If battery terminal voltage ULFor the restrictive condition of battery limit working condition, if now with maximum discharge current ImaxBattery is put Electricity, battery model parameter OCVePredictive value in the k+ Δ T moment is:
In above formula,It is by calculated battery model parameter OCV of fractional order federated Kalman filtering algorithmeIn k The estimated value carved,For OCVePredictive value in the k+ Δ T moment;ImaxIt is as battery limit working condition in terminal voltage Maximum discharge current value during restrictive condition, is value to be solved,It is the maximum discharge current of battery;OCVkIt is in the k moment Battery open circuit voltage values;For battery in k+ ΔTMoment it is assumed that withCorresponding battery open circuit electricity during constant-current discharge Pressure value;Due in most cases, the change of OCV is all less and slow, it can thus be assumed that the open-circuit voltage within the Δ T period becomes Change amount is linear with discharge current;
And then calculate it is assumed that battery is I in discharge currentmaxWhen, estimated value U of battery terminal voltageL,k+ΔT
Wherein, the estimated value of terminal voltage be k moment open-circuit voltage estimated value, open circuit voltage variations value, ohmic internal resistance in the Δ T period Voltage difference, fractional order element ZWZero state response magnitude of voltage, zero input response magnitude of voltage sum;
Thus release, when with terminal voltage ULAs the extreme working position of battery restrictive condition when, the maximum operating currenbt of battery For:
When the k+ Δ T moment reaches electric discharge peak power, consider the restrictive condition of the extreme working position of above-mentioned battery, maximum Discharge current value is:
I m a x = m i n ( I m a x , S o C , I m a x , U L , I max lim )
The electric discharge peak power in k+ Δ T moment is:
The Forecasting Methodology of the Forecasting Methodology of current peak feedback power and above-mentioned electric discharge peak power is in the same manner:
If ILBe on the occasion of;
If SoC is the restrictive condition of battery limit working condition, SoCmaxIt is the maximum SOC of battery, then obtain now Minimum feedback current value:
I m i n , S o C = ( SoC k - SoC m a x ) × C a p a c i t y Δ T
If ULFor the restrictive condition of battery limit working condition, then obtain minimum feedback current value now:
When the k+ Δ T moment reaches current peak feedback power, consider above-mentioned restrictive condition, minimum feedback current value is:
I m i n = m a x ( I m i n , S o C , I m i n , U L , I min lim )
Thus, obtain battery current peak value feedback power:
Complete the peak power prediction of electrokinetic cell.
2. a kind of peak power Forecasting Methodology of electrokinetic cell according to claim 1 is it is characterised in that in step A,
Due to the electrochemical impedance spectroscopy test data based on battery for the battery model used in method, battery model parameter has clearly Physical significance, ohmic internal resistance RoPhysical significance be:
Ro≈RΩ+RSEI+Rct
Wherein, RΩFor high frequency ohmage, RSEIFor SEI membrane impedance, RctFor Charge-transfer resistance;
In addition, weber impedance to be defined by below equation:
Z W = 1 W ( j w ) 0.5
Wherein, W is ionic diffusion coefficient, for the ease of impedance parameter On-line Estimation, takes:
X W = 1 W
Obtain:
Z W = X W ( j w ) 0.5 .
CN201410592570.4A 2014-10-29 2014-10-29 A kind of peak power Forecasting Methodology of electrokinetic cell Active CN104267354B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201410592570.4A CN104267354B (en) 2014-10-29 2014-10-29 A kind of peak power Forecasting Methodology of electrokinetic cell

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201410592570.4A CN104267354B (en) 2014-10-29 2014-10-29 A kind of peak power Forecasting Methodology of electrokinetic cell

Publications (2)

Publication Number Publication Date
CN104267354A CN104267354A (en) 2015-01-07
CN104267354B true CN104267354B (en) 2017-03-01

Family

ID=52158894

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201410592570.4A Active CN104267354B (en) 2014-10-29 2014-10-29 A kind of peak power Forecasting Methodology of electrokinetic cell

Country Status (1)

Country Link
CN (1) CN104267354B (en)

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104950263B (en) * 2015-06-12 2017-09-19 安徽江淮汽车集团股份有限公司 Automobile power cell SOC evaluation method
CN105277895B (en) * 2015-09-30 2018-01-26 上海凌翼动力科技有限公司 A kind of series battery power rating SOP On-line Estimation method and its application
CN106249170B (en) * 2016-08-31 2018-11-23 简式国际汽车设计(北京)有限公司 A kind of electrokinetic cell system power rating estimation method and device
CN108287310B (en) * 2017-01-09 2020-03-31 宝山钢铁股份有限公司 Lithium battery load detection device and load calculation method
CN107102271A (en) * 2017-05-25 2017-08-29 宁德时代新能源科技股份有限公司 The evaluation method of battery pack peak power, device and system
CN108983108B (en) * 2018-08-10 2021-04-02 山东大学 Power battery pack peak power estimation method
CN110927592B (en) * 2018-08-31 2021-05-14 华为技术有限公司 Method and device for estimating peak power of battery
CN109738809A (en) * 2019-01-10 2019-05-10 上海动力储能电池系统工程技术有限公司 The estimation method of power and energy-storage battery power characteristic
CN110308395A (en) * 2019-06-28 2019-10-08 安徽贵博新能科技有限公司 A kind of power rating evaluation method based on multi-constraint condition battery pack
CN110470994A (en) * 2019-09-26 2019-11-19 重庆长安新能源汽车科技有限公司 A kind of prediction technique of power battery peak power
CN111443290A (en) * 2019-12-30 2020-07-24 南京航空航天大学 SOP estimation method for power battery of electric vehicle with closed-loop control

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US3969667A (en) * 1972-08-23 1976-07-13 The United States Of America As Represented By The Secretary Of The Navy Device for determining the state of charge in batteries
CN102158134A (en) * 2011-05-27 2011-08-17 深圳珈伟光伏照明股份有限公司 Peak power tracking method for solar generating system
CN102323553A (en) * 2011-05-31 2012-01-18 惠州市亿能电子有限公司 Method for testing battery peak power
CN103675707A (en) * 2013-12-13 2014-03-26 国家电网公司 Method for evaluating lithium ion battery peak power online
CN103995232A (en) * 2014-04-21 2014-08-20 中通客车控股股份有限公司 Lithium iron phosphate power battery pack peak value charge and discharge performance detection method

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20130134507A (en) * 2012-05-31 2013-12-10 현대중공업 주식회사 Method for measuring output property parameter of solar cell module

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US3969667A (en) * 1972-08-23 1976-07-13 The United States Of America As Represented By The Secretary Of The Navy Device for determining the state of charge in batteries
CN102158134A (en) * 2011-05-27 2011-08-17 深圳珈伟光伏照明股份有限公司 Peak power tracking method for solar generating system
CN102323553A (en) * 2011-05-31 2012-01-18 惠州市亿能电子有限公司 Method for testing battery peak power
CN103675707A (en) * 2013-12-13 2014-03-26 国家电网公司 Method for evaluating lithium ion battery peak power online
CN103995232A (en) * 2014-04-21 2014-08-20 中通客车控股股份有限公司 Lithium iron phosphate power battery pack peak value charge and discharge performance detection method

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Online peak power prediction based on a parameter and state estimator for lithium-ion batteries in electric vehicles;Lei Pei et al;《Energy》;20140222;第66卷;766-778 *
动力电池组峰值功率估计算法研究;张彩萍等;《系统仿真学报》;20100630;第22卷(第6期);1524-1527 *

Also Published As

Publication number Publication date
CN104267354A (en) 2015-01-07

Similar Documents

Publication Publication Date Title
Cacciato et al. Real-time model-based estimation of SOC and SOH for energy storage systems
Ye et al. A model-based adaptive state of charge estimator for a lithium-ion battery using an improved adaptive particle filter
CN105301509B (en) The combined estimation method of charge states of lithium ion battery, health status and power rating
TWI708068B (en) Method and apparatus for determining the state of health and state of charge of lithium sulfur batteries
CN106443474B (en) A kind of electrokinetic cell system service life Decline traits quickly know method for distinguishing
CN106026260B (en) A kind of series winding battery pack SOC estimation method with equalizing circuit
Hu et al. Robustness analysis of State-of-Charge estimation methods for two types of Li-ion batteries
Hentunen et al. Time-domain parameter extraction method for Thévenin-equivalent circuit battery models
CN104704380B (en) For estimating device and the method for estimation of battery parameter
Dong et al. An online model-based method for state of energy estimation of lithium-ion batteries using dual filters
CN103852727B (en) Method and device for estimating power battery charge state on line
Huria et al. High fidelity electrical model with thermal dependence for characterization and simulation of high power lithium battery cells
CN106909716B (en) Lithium iron phosphate battery modeling and SOC estimation method considering capacity loss
Charkhgard et al. State-of-charge estimation for lithium-ion batteries using neural networks and EKF
CN103926538B (en) Change exponent number RC equivalent-circuit model based on AIC criterion and implementation method
JP5946436B2 (en) Battery parameter estimation apparatus and parameter estimation method
CN103399277B (en) A kind of power battery actual capacity estimation method
Chen et al. A new state-of-health estimation method for lithium-ion batteries through the intrinsic relationship between ohmic internal resistance and capacity
Zou et al. State-space model with non-integer order derivatives for lithium-ion battery
Zhang et al. An on-line estimation of battery pack parameters and state-of-charge using dual filters based on pack model
Wang et al. Modeling and state-of-charge prediction of lithium-ion battery and ultracapacitor hybrids with a co-estimator
CN104773086B (en) The method and system of battery impedance parameter is estimated using rolling time horizon regression analysis
CN106451600B (en) For the device and method of quickly charging battery
Sun et al. Adaptive unscented Kalman filtering for state of charge estimation of a lithium-ion battery for electric vehicles
Gholizadeh et al. Estimation of state of charge, unknown nonlinearities, and state of health of a lithium-ion battery based on a comprehensive unobservable model

Legal Events

Date Code Title Description
PB01 Publication
C06 Publication
SE01 Entry into force of request for substantive examination
C10 Entry into substantive examination
GR01 Patent grant
GR01 Patent grant