CN104007390B - Battery state of charge tracking, equivalent circuit selection and reference test method and system - Google Patents

Battery state of charge tracking, equivalent circuit selection and reference test method and system Download PDF

Info

Publication number
CN104007390B
CN104007390B CN201410062166.6A CN201410062166A CN104007390B CN 104007390 B CN104007390 B CN 104007390B CN 201410062166 A CN201410062166 A CN 201410062166A CN 104007390 B CN104007390 B CN 104007390B
Authority
CN
China
Prior art keywords
soc
battery
estimation
equivalent
circuit
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
CN201410062166.6A
Other languages
Chinese (zh)
Other versions
CN104007390A (en
Inventor
B·巴乐兴艾姆
Y·巴-莎伦
B·弗伦奇
B·派迪派缇
K·R·派迪派缇
詹姆斯·米查姆
特莱维斯·威廉斯
G·V·艾沃瑞
黄泰植
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.)
Fairchild Semiconductor Suzhou Co Ltd
University of Connecticut
Fairchild Semiconductor Corp
Original Assignee
Fairchild Semiconductor Suzhou Co Ltd
University of Connecticut
Fairchild Semiconductor Corp
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
Priority to US201361768472P priority Critical
Priority to US61/768,472 priority
Priority to US14/185,790 priority
Priority to US14/185,835 priority
Priority to US14/185,790 priority patent/US20140244193A1/en
Priority to US14/185,835 priority patent/US10664562B2/en
Application filed by Fairchild Semiconductor Suzhou Co Ltd, University of Connecticut, Fairchild Semiconductor Corp filed Critical Fairchild Semiconductor Suzhou Co Ltd
Publication of CN104007390A publication Critical patent/CN104007390A/en
Application granted granted Critical
Publication of CN104007390B publication Critical patent/CN104007390B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Abstract

The application is related to battery state of charge tracking, equivalent circuit selection and reference test method and system.This method is included in the first estimation state of charge for the very first time calculating battery(SOC)The magnitude of voltage for the measurement voltage for representing the battery both ends is received in the second time, filter gain is calculated in second time, and calculates the second of the battery based on the described first estimation SOC, the magnitude of voltage and the filter gain in second time and estimates SOC.Another method includes the storehouse that storage in memory represents the equivalent-circuit model of battery, the operational mode of battery is determined based on the load related to the battery, one of described equivalent-circuit model is selected based on identified operational mode, and the state of charge of the battery is calculated using selected equivalent-circuit model(SOC).

Description

Battery state of charge tracking, equivalent circuit selection and reference test method and system
Technical field
Embodiment is related to the state of charge for calculating battery.
Background technology
Electrochemical energy storage device plays an important role in future source of energy strategy.In fact, battery be at present and soon Energy storage technology feasible in the future.Diversified equipment such as portable electric appts, mobile household electrical appliance, aerospace equipment Deng being increasingly battery powered.It is likely difficult to accurately estimate the state of charge of such as battery using known system and method. Therefore, it is necessary to solve the deficiency of current technology and provide other novel and character of innovation system, methods and apparatus.
The content of the invention
One embodiment includes a kind of method.This method is included in the first estimation state of charge for the very first time calculating battery (SOC) magnitude of voltage for the measurement voltage for representing battery both ends, is received in the second time, calculating wave filter in second time increases Benefit, and calculate the second of battery based on the first estimation SOC, magnitude of voltage and filter gain in second time and estimate SOC.
Another embodiment includes a kind of system.The system includes battery and battery meter module, and it is configured as making The estimation state of charge (SOC) of the battery is calculated with Reduced Order Filter, the Reduced Order Filter is single state filter, described Single state filter is configured as estimating SOC described in recursive calculation based on the SOC estimation calculated before.
Another embodiment includes a kind of computer-readable medium, and it includes code segment.Code segment is by computing device When the processor is calculated the estimation state of charge (SOC) of battery, estimation SOC is stored in buffer, filtered using depression of order Device calculates the estimation SOC after the renewal of battery, and the Reduced Order Filter is single state filter, single state filter by with The estimation SOC being set to after being updated based on estimation SOC come recursive calculation.
Further embodiment includes a kind of method.This method includes the equivalent circuit mould that storage in memory represents battery The storehouse of type, the operational mode of battery is determined based on the load related to the battery, selected based on identified operational mode One of equivalent-circuit model, and calculate the state of charge of battery (SOC) using the equivalent-circuit model of selection.
Another embodiment includes a kind of system.The system includes being configured as the equivalent-circuit model that storage represents battery Storehouse data storage, be configured as selecting the Model selection module of equivalent-circuit model based on the operational mode of battery, And it is configured as calculating the filter module of the estimation state of charge (SOC) of battery based on the equivalent-circuit model of selection.
Another embodiment includes a kind of computer-readable medium, and it includes code segment.Code segment is by computing device When the processor is selected equivalent circuit mould from the storehouse of equivalent-circuit model for representing battery based on the operational mode of battery Type;And the state of charge (SOC) of the battery is calculated using the equivalent-circuit model of selection.
Brief description of the drawings
According to this paper detailed description given below and accompanying drawing, exemplary embodiment will be more fully understood, in accompanying drawing Similar element is denoted by like reference characters, and these reference numbers are only provided in a manner of illustration, therefore is not to showing The limitation of example property embodiment, and wherein:
Fig. 1 and 2 shows the block diagram of the battery management system (BMS) according at least one exemplary embodiment.
Fig. 3 shows the frame for being used to select the signal stream of battery equivalent model according at least one exemplary embodiment Figure.
Fig. 4 shows the signal stream for being used to calculate battery state of charge (SOC) according at least one exemplary embodiment Block diagram.
Fig. 5 shows the block diagram of battery meter (BFG) system according at least one exemplary embodiment.
Fig. 6 shows the frame of the signal stream of the parameter module for BFG systems according at least one exemplary embodiment Figure.
Fig. 7 shows the frame of the signal stream of the SOC module for BFG systems according at least one exemplary embodiment Figure.
Fig. 8 shows the block diagram of the SOC module according at least one exemplary embodiment.
Fig. 9 shows the frame of total least square (TLS) module of the SOC module according at least one exemplary embodiment Figure.
Figure 10 shows recurrence least square (RLS) module of the SOC module according at least one exemplary embodiment Block diagram.
Figure 11 and 12 shows the flow chart of the method according at least one exemplary embodiment.
Figure 13 A-13D show the schematic diagram of the battery equivalent model according at least one exemplary embodiment.
Figure 14 is the characteristic schematic diagrames of OCV-SOC for showing Portable lithium ion cell.
Figure 15 A and 15B are the figure for showing load curve.
Figure 16 A and 16B are the figure for showing fictitious load curve.
Figure 17 is to show the schematic diagram that example system is realized.
Figure 18 is to show the schematic diagram that the user interface being used in combination can be realized with system.
Figure 19 A and 19B include the figure for showing exemplary discharge voltage/current curve.
Figure 20 A and 20B are the figure for showing exemplary coulomb count evaluation method.
Figure 21 A and 21B are the figure for showing shut-in time (TTS) appraisal procedure.
Figure 22 A, 22B and 22C are the table for representing voltameter reading.
It should be noted that these figures are intended to show that method used in some exemplary embodiments and/or structure General features, and supplement the printed instructions being provided below.However, these accompanying drawings are not necessarily drawn to scale and may be inaccurate Reflect the precise structural characteristics or performance characteristic of any given embodiment, and should not be construed as defining or limiting by exemplary reality Apply the scope of value that example covered or property.For example, for clarity, may reduce or exaggerate structural detail relative thickness and Positioning.
Embodiment
Although exemplary embodiment may include various modifications and alternative form, embodiment is in the accompanying drawings with citing Mode shows and will be described in detail herein.It will be appreciated, however, that it is not intended to exemplary embodiment being limited to the present invention Particular forms disclosed, but antithesis, exemplary embodiment will cover all modifications fallen within the scope of the appended claims Form, equivalents and alternative form.In the whole description to accompanying drawing, similar numeral refers to similar element.
The accurate estimation of battery status such as state of charge (SOC), health status (SOH) and remaining life (RUL) For reliable, safety and to widely use the device being battery powered be vital.Estimate that these quantity are referred to as battery electric quantity Measure (BFG).Different from the HC fuel in many current automobiles, the memory capacity of battery is not constant.Generally, battery capacity Change with battery service life, use pattern and temperature, so as to cause challenging ART network problem to BFG, Need in consideration temperature change, SOC to change and be modeled battery behavior on the basis of service life and on-line parameter identifies.
Fig. 1 and 2 shows the block diagram of the system 100 according at least one exemplary embodiment.As shown in figure 1, system 100 Including battery 105, battery management system (BMS) 110, display 120, unrestricted power supply 125 (such as wall outlet, automobile Charging station etc.) and switch 130.
BMS110 can be configured as managing utilization and/or the state of battery 105.For example, BMS110 can be configured with Unrestricted power supply 125 is connected or disconnected to be charged to battery 105 by switch 130 with battery 105.For example, BMS110 can quilt It is configured to that load (not shown) is connected or disconnected with battery 105.For example, BFG115 can be configured as calculating the electricity of battery 105 Lotus state (SOC) and/or health status (SOH).SOC and/or SOH can be shown (for example, in the form of percentage, during with residue Between form etc.) on display 120.
As shown in Fig. 2 BMS110 comprises at least analog-digital converter (ADC) 205,220, wave filter 210,225, numeral amplification Device 215 and battery meter (BFG) 115.BFG115 includes memory 230, processor 235 and controller 240.ADC205、 220, wave filter 210,225, at least one of digital amplifier 215 and BFG115 can be, such as application specific integrated circuit (ASIC), digital signal processor (DSP), field programmable gate array (FPGA) and/or processor, etc..Or BMS110 Can be to include ASIC, DSP, FPGA and/or processor of shown functional block, etc..Or system 100 can be realized to be stored in On memory and by the software of such as computing device.
BMS110 can be configured with digital quantizer (ADC) 205,220, wave filter 210,225 and digital amplifier 215 combination is by analog measurement (e.g., IbAnd Vb) be converted to digital value (for example, so that BFG115 calculates SOC and/or SOH). For example, digital amplifier 215 can be difference amplifier, it is according to the voltage drop V at the both ends of battery 105b(for example, positive and negative terminal Between magnitude of voltage difference) generation (e.g., produce) analog signal, then using ADC220 and wave filter 225 by the analog signal It is converted into the digital value of filtering.
System 100 can be the subsystem using battery powered any system or electronic equipment.In some specific implementations, Electronic equipment can be or may include (such as) have conventional laptop desktop, laptop laptop devices.In some specific implementations In, electronic equipment can be or may include (such as) wireline equipment and/or wireless device (for example, supporting Wi-Fi equipment), calculate Entity (for example, personal computing devices), server apparatus (for example, web server), toy, mobile phone, audio frequency apparatus, electricity Machine control device, power supply (for example, offline power supply), personal digital assistant (PDA), tablet device, electronic reader, TV and/ Or automobile, etc..In some specific implementations, electronic equipment can be or may include (such as) display device is (for example, liquid crystal display (LCD) monitor, for user's display information), keyboard, pointing device (for example, mouse, Trackpad, pass through the equipment, use Family can provide input to computer).
Fig. 3 shows the block diagram for being used to select the signal stream of battery equivalent model according at least one exemplary embodiment 300.As shown in figure 3, model selection block 310 receives input 320 (voltage and/or electric current letter e.g., from battery and/or load Number) and select using 320 (or its some modification) of input the equivalent mould of representative (or corresponding to) battery from equivalent model storehouse 305 Type.Then by state of charge calculator block 315 using the equivalent model come calculated charge state (SOC) 325.Equivalent model storehouse 305 may include at least one equivalent model for representing battery.Each equivalent model can be based on the operation of battery (or battery eliminator) Pattern.Operational mode can be based on the load related to battery.For example, operational mode can the voltage drop based on load both ends.For example, Operational mode can based on load both ends voltage drop be of a relatively high or relatively low, be relative constancy or dynamic and/or they Combination.
Equivalent model may include (referring to figure 13 below A-13D) resistor, voltage (e.g., voltage drop or voltage source), resistance- Any combinations of electric current (RC) circuit and/or impedance circuit etc..Therefore, the mathematics that can establish battery equivalent model is (e.g., public Formula) equivalents.Equivalent model storehouse 305 can store the mathematical equivalent form related to battery operation pattern.The mathematical equivalent shape Formula can be used for carrying out SOC calculating by state of charge calculator block 315.For example, mathematical equivalent form can be used for determining variable, make For the input of the formula for calculating SOC (or SOC estimate).Therefore, BFG systems can be based on operational mode and select equivalent mould Type with improve computational efficiency and reduce processing time.More details are provided below for Fig. 5-12.
Fig. 4 shows the signal stream for being used to calculate battery state of charge (SOC) according at least one exemplary embodiment Block diagram 400.As shown in figure 4, block diagram 400 include extended Kalman filter (EKF) block 405, state of charge (SOC) block 410, 420th, filter gain parameter block 41,425 and buffer 430.EKF blocks 505 can be configured as based on the electric charge shape calculated before State 410 and filter gain parameter 415, calculate SOC420 and determine filter gain parameter 425 (e.g., reading and/or calculating SOC variances, the voltage measured, the capacity calculated and/or variable related to equivalent circuit, etc.).Therefore, buffer 430 The SOC calculated before and filter gain parameter in such as processing cycle can be configured as storing.In other words, can be based on The SOC calculated before at least one calculates current (or next) SOC.In other words, the SOC calculated in the very first time can For calculating the SOC in the second (later) time.
In exemplary specific implementation, the set of at least two SOC410 and/or filter gain parameter 415 can be used. Therefore, at least two SOC vector, at least two SOC array, the average and at least two SOC of at least two SOC are put down Average and corresponding filter gain parameter can be used for calculating next SOC420 (or SOC of the second time) and determining/calculating Corresponding filter gain parameter 425.Therefore, buffer 430 can be configured as storing it is multiple before the SOC410 that calculate and Calculate/determine filter gain parameter 415.Therefore, the SOC that BFG systems calculate before, which is improved, calculates effect Rate and reduction processing time.More details are provided below for Fig. 5-12.
Fig. 5 shows the block diagram of the system of battery meter (BFG) 115 according at least one exemplary embodiment.Such as Fig. 5 Shown, BFG115 includes estimation module 510, tracking module 520, prediction module 530, open-circuit voltage-state of charge (OCV-SOC) Characterization module 540, offline parameter estimation module 545 and battery life characterization module 550.In addition, the system includes offline number According to collection module 555 and fuel cell modelling module 560.
Off-line data collection module 555 can be configured as measuring battery behavior in relatively controlled test environment.For example, Open-circuit voltage (OCV) measured value and SOC measured values of battery 105 (or battery eliminator) can be collected in test laboratory environment. For example, battery 105 (or battery eliminator) can be initialized to it is fully charged (e.g., close to it is fully charged, substantially completely charge), (rested) state of standing.OCV and SOC measurements can be carried out.Then battery 105 (or battery eliminator) can be made slowly to discharge, simultaneously At regular intervals (e.g., periodically, periodically, irregularly, the scheduled time) carry out OCV and SOC measurements, until battery 105 (or waits Imitate battery) completely (or substantially) electric discharge.OCV and SOC measured values can be used for determining, calculate or estimate battery parameter (e.g., hereafter Described OCV parameters Ki□{K0;K1;K2;K3;K4;K5;K6;K7})。
Data from off-line data collection module 555 can be used in fuel cell modelling module 560 to determine such as battery 105 The equivalent model of (or battery eliminator) and/or the mathematical equivalent form of equivalent model.Number from off-line data collection module 555 According to that can be used in offline parameter estimation module 545 to determine and/or calculate the parameter related to battery 105 (or battery eliminator) (e.g., the value of the element related to above-mentioned equivalent model).Data from off-line data collection module 555 can be used in OVC-SOC To determine and/or calculate OCV and SOC battery parameters (e.g., OCV parameters K described below in characterization module 540i□{K0;K1; K2;K3;K4;K5;K6;K7}).Data from offline parameter estimation module 545 can be used in battery life characterization module 550.Example Such as, the data from offline parameter estimation module 545 can be used for calculating initial health by battery life characterization module 550 (SOH) characteristic (e.g., maximum SOC).
Display 120 is shown as showing that 565, SOH shows 570, shut-in time (TTS) display 575 and residue with SOC Service life (RUL) display 580.Each display may be, for example, the instrument of show percent.It can be calculated or determined by BFG115 often The value of individual display.For example, TTS can be shown as the time value (e.g., hour and/or minute) such as calculated by TTS module 532.
Estimation module 510 includes parameter module 512 and capacity module 510.Estimation module 510 can be configured as calculate and/ Or determine that battery 105 (or battery eliminator) is specifically worth (e.g., parameter and capability value).In stable environment (e.g., test experiments Room) in, parameter and capability value are probably fixed (e.g., not changing).However, in real world environments, parameter and capability value Be probably it is dynamic or change.For example, complete SOC tracking solution is usually directed to (1) and characterized by offline OCV to shape Into the estimation of the OCV parameters of a part for state-space model.It is stable that OCV-SOC, which is characterized on temperature change and cell degradation, 's.Once estimating these parameters, these parameters just form a part for the state-space model with known parameters.(2) move The estimation of state equivalent circuit parameter.It has been observed that these parameters change with the temperature of battery, SOC and service life, therefore should When adaptively being estimated while BFG is run.(3) estimation of battery capacity:Although the nominal capacity of battery is by manufacturing Specified by business, but known available battery capacity can change because of manufacturing process error, temperature change, use pattern and aging. And the SOC that (4) are constrained by model parameter is tracked.Once known models parameter, SOC tracking has reformed into nonlinear filtering and asked Topic.But, it has been observed that gained state-space model includes correlated process and measurement noise process.Appropriately handle these correlations Effect will draw more preferable SOC tracking accuracies.Therefore, in exemplary specific implementation, for calculating parameter and capacity, mould is tracked Block 520 can be by data feedback to estimation module 510.
In addition, typically the method for estimation battery capacity ignores hysteresis effect, and assume the cell voltage generation stood The true open-circuit voltage (OCV) of table battery.However, according to exemplary embodiment, estimation module 510 will be late by being modeled as battery Error in 105 OCV, and the error in OCV is compensated using the combination of real time linear parameter Estimation and SOC tracking techniques.
Tracking module 520 includes SOC module 522 and SOH modules 524." electricity " in SOC instruction batteries 105.As above institute State, SOC is active volume, and it is expressed as the percentage of some benchmark (e.g., rated capacity or current capacity).According to exemplary reality Example is applied, SOC module 522 is by using the error (being combined with parameter Estimation) in the tracing compensation OCV being described in more below To calculate SOC.SOH indicates state of the battery compared with new or ideal battery.SOH can be based on charge acceptance, internal resistance, electricity Pressure and/or self discharge, etc..
Prediction module 530 includes TTS module 532 and RUL modules 534.TTS module 532 and RUL modules 534 can be configured To calculate TTS and RUL based on SOC.
Fig. 6 shows the signal stream of the parameter module 412 for BFG115 according at least one exemplary embodiment Block diagram.BFG systems can be based on operational mode selection equivalent model to improve computational efficiency and reduce processing time.As shown in fig. 6, Parameter module 412 includes operational mode module 605 and Model selection module 610.Operational mode module 605 can be configured as being based on At least one input and/or next self-supported 615 at least one input from battery 105 determine (battery) operational mode. At least one input can based on in battery 105 and load 615 at least one related electric current and voltage at least one It is individual.For example, the voltage drop at 615 both ends of load.For example, operational mode can be of a relatively high based on the voltage drop at 615 both ends of load Or it is relatively low, be relative constancy or dynamic and/or combinations thereof.Model selection module 610 can the operation mould based on determination Formula selection equivalent model (or its mathematical equivalent form).For example, Model selection module 610 can be generated for searching for equivalent model storehouse 305 query term.
In some specific implementations, definable or the multiple operational modes of sign.In exemplary specific implementation, it is described below To battery and use related four operational modes of the system of battery.
In the first operational mode, battery 105 may connect to weight and the load changed.In other words, load 615 can combine Consumed (or the high voltage of consumption variable current loads) using of a relatively high voltage and dynamic or variable current.For example, in movement In phone, the first operational mode may include such use environment, wherein mobile phone using include prolonged video playback, Multimedia and game application etc..Equivalent circuit shown in figure 13 below A can represent the load for being connected to weight and change Battery.
In the second operational mode, battery 105 loads connectable to dynamic load and/or variable voltage.In other words, bear Carry 615 voltages that dynamic or change can be used.For example, in the mobile phone, the second operational mode may include such to use ring Border, wherein mobile phone, which use, includes call, web-browsing and/or the conventional use for playing video clipping.Figure 13 below B Shown in equivalent circuit can represent to be connected to the battery of dynamic load.
In the 3rd operational mode, battery 105 may be connected to or consume constant current.In other words, load 615 can towing ahead With constant (or substantial constant) load.Or battery 105 just can be electrically charged using constant current.For example, in charging week Phase, battery 105 can disconnect with load 615 (e.g., can be used switch 130 that unrestricted power supply 125 is connected into battery 105 with to electricity Charged in pond 105).Equivalent circuit shown in figure 13 below C can represent to be connected to the battery of constant current.
In the 4th operational mode, battery 105 may be connected to the load of relatively low voltage.Or battery 105 can be at Periodicity static condition, wherein battery 105 are subjected to gently loading, then charging, then stand, be minimum or non-loaded.In other words, Minimum voltage can be rarely employed in load 615.For example, in the mobile phone, the 4th operational mode may include such use environment, After wherein mobile phone use is included in (or substantially completely) charging completely, base station is carried out using the call not taken place frequently Conventional contact.Equivalent circuit shown in figure 13 below D can represent to be connected to the battery of dynamic load.
Fig. 7 shows the signal stream of the SOC module 422 for BFG115 systems according at least one exemplary embodiment Block diagram.As shown in fig. 7, SOC module 422 includes buffer block 705, model estimation block 710, SOC tracking blocks 715 and voltage drop Prediction module or block 720.
In the exemplary embodiment, it will be late by the error being modeled as in the OCV of battery 105.Voltage drop vDIn [k] can be represented Portion battery model element R0、R1、R2And xhThe voltage (referring to Figure 13 A) at [k] both ends.xh[k] item can be used in the SOC of explanation prediction Error.In other words, xh[k] can be " instantaneous lagging ", and it can be modified to zero by Adjustable calculation or the SOC estimated. If the SOC for calculating or estimating is equal to SOC, the x for calculating or estimatingh[k] should be equal to zero.In other words, calculate or estimate The x gone outh[k] is not equal to zero, it indicates that the SOC for calculating or estimating has error.Voltage drop model parameter vector (b) include pair Should be in the x for calculating or estimatinghThe element of [k].
Therefore, in Fig. 7 flow, the SOC for tracking the current calculating of block 715 from SOC or estimating is used in voltage block To calculate voltage drop v in prediction block 720D[k].At least one passing voltage drop vD[k] is stored in buffer 705 and is used for Parameter vector b estimation.The x that correspondence in parameter vector b is calculated or estimatedh[k] is that nonzero value instruction has instantaneous lagging. This means SOC evaluated errors.The SOC track algorithms of SOC tracking blocks 715 are configured as in the x for calculating or estimatingh[k] is Non-zero whenever amendment SOC.Hereafter (mathematically) describe relevant voltage drop vD[k], hysteresis, estimation xh[k], voltage The more details of model parameter vector (b) and SOC tracking drop.Therefore, the SOC and SOC that BFG systems calculate before are missed Poor accurate estimation SOC, improve computational efficiency and reduce processing time.
Fig. 8 shows the block diagram of the SOC module 422 according at least one exemplary embodiment.As shown in figure 8, SOC module 422 include extended Kalman filter (EKF) block 805.EKF blocks can be configured as calculating SOC845 and SOC errors 840.EKF blocks 805, which can be configured with equation 1, calculates SOC845 as estimation SOC, and uses equation 2 to calculate SOC errors 840 as estimation SOC errors (or variance).In each of following equations, k refers to instantaneous iteration, k+1 | k refer to one, it is previous or it Preceding iteration, and k+1 | k+1 refers to current, renewal, next or successive iterations.
Wherein:
It is estimation SOC that is current or updating iteration;
It is the estimation SOC of upper one or prediction iteration;
G [k+1] is the filter gain of upper one or prediction iteration;And
vk+1It is the load voltage of upper one or prediction iteration.
Wherein:
Ps[k+1 | k+1] it is SOC evaluated errors or variance that be current or updating iteration;
G [k+1] is the filter gain of upper one or prediction iteration;
H [k+1] is the observed differential of linearisation;
Ps[k+1 | k] is upper one or predicts the SOC evaluated errors or variance of iteration;And
It is the voltage noise reducing in initialization with zero mean and correlation.
SOC module 422 includes OCV parameter blocks 810.OCV parameter blocks 810 can be configured as from OVC-SOC characterization modules 540 Storage and/or reception OCV parameters { Ki}.OCV parameters { KiIt is constant, because they are off-line measurement and made in battery 105 It is negligible (or in the absence of) with the change in the life-span.OCV parameters are used to calculate OCV according to equation 3 with SOC.
Wherein:
S [k] is SOC;And
Vo(s [k]) is open-circuit voltage (OCV);
SOC module 422 includes voltage drop model block 825.Voltage drop block 825 can be configured with according to equation 4 or 5 The voltage drop at voltage drop model (as discussed above) computational load both ends.
Zv[k]=V0(xs[k])+a[k]Tb+nD[k] (4)
Wherein:
Zv[k] is the voltage of measurement;
V0(xs[k]) it is open-circuit voltage (OCV);
a[k]TIt is voltage drop model;
B is voltage drop model parameter vector;
It is estimated voltage drop model;
It is estimated voltage drop model parameter vector;And
nD[k] is voltage drop observation noise.
As described above, voltage drop model can be based on selected equivalent-circuit model and change.Selected equivalent-circuit model and/or Voltage drop model can be read from data storage 855.For example, data storage 855 may include equivalent model storehouse 305.
EKF (module or) block 805 can be configured with equation 1 and calculate SOC845 as estimation SOC, and by gained SOC845 is stored in buffer 850.EKF blocks 805 can be configured with equation 2 and calculate the calculating of SOC errors 840 as estimation SOC errors (or variance), and the SOC errors 840 of gained are stored in buffer 850.SOC the and SOC errors of storage are readable SOC815 and SOC errors 820 are taken as, that is, SOC845 the and SOC errors 840 stored.Therefore, EKF blocks recursively (can e.g., follow In ring) calculate SOC845 and SOC errors 840 so that follow-up (be renewal in time, next and/or afterwards) SOC845 With SOC errors 840 calculate can based on (in time for current, upper one or before) SOC815 before at least one and SOC errors 820 calculate.
As shown in figure 8, recurrence least square (RLS) block 830 and total least square (TLS) block 835 can be generated to EKF blocks 805 input.RLS blocks can generate initial estimation voltage drop model parameter vector, and (it may include at least one voltage drop model ginseng Number), and TLS blocks 835 can generate initial estimation capacity.Each circulation can be directed to and generate initial estimation voltage drop model parameter Vector sum initial estimation capacity.In exemplary specific implementation, with iterations (k) increase, initial estimation voltage drop model The change of parameter vector and initial estimation capacity can become to ignore.
Fig. 9 shows total least square (TLS) block 835 according to the SOC module 422 of at least one exemplary embodiment Block diagram.As shown in figure 9, TLS blocks 635 include buffer 910 and TLS computing modules 915.Buffer 910 be configured as receive, Storage and output SOC data, such as Δ SOC data 920 (or changing value of SOC data 920), so that TLS computing modules 915 make With.Buffer 910 is additionally configured to receive, store and export Δ coulomb data 925 (or changing value of coulomb data 925), with Used for TLS computing modules 915.Buffer 910 can receive the current data based on the electric current related to battery 105 measured 905 are used as such as coulomb enumeration data.
TLS computing modules 915 can be configured as calculating the capacity of battery 105 based on Δ SOC920 and Δ coulomb 925 930.For example, equation 6 can be used to calculate capacity 930 for TLS computing modules 915.The derivation of equation 6 is shown in further detail below.
Wherein:
It is estimated capacity;
It is the covariance of augmentation observing matrix;And
ΔK(2,2) be non-negative feature value diagonal 2 × 2 matrix;
Figure 10 shows recurrence least square (RLS) block according to the SOC module 422 of at least one exemplary embodiment 830 block diagram.As shown in Figure 10, TLS blocks 635 include buffer 1005 and RLC computing modules 1010.Buffer 1005 is configured For receive and store SOC815, SOC error 820 and as the output from voltage drop model block 825 voltage-drop data (e.g., ZV[k] or OCV).Buffer 1005 is configured as output voltage drop 1015 and electric current and electric capacity (I&C) matrix 1020.
RLC computing modules 1010 can be configured as calculating initialization ginseng based on voltage drop 1015 and (I&C) matrix 1020 Number 1025.For example, equation 7 can be used to calculate initiation parameter 1025 for RLC computing modules 1010.The derivation of equation 7 is hereinafter It is shown in further detail.
Wherein:
αiIt is R1C1Current attenuation coefficient in circuit;
βiIt is R2C2Current attenuation coefficient in circuit;
It is R1Estimation resistance value;
It is R2Estimation resistance value;
It is the estimation lagging voltage of battery;And
xh[k] is instantaneous lagging;
It is worth noting that, as described above for described in Fig. 7, the estimation lagging voltage of battery should be zero.Therefore, in exemplary tool During body is implemented, (or substantially removing) is eliminated because being missed caused by hysteresis by SOC tracking due to tracking block 715 using SOC Difference, therefore the b (6) in equation 7 should be zero.Therefore, SOC estimation is more accurate, because can will be late by taking into account.
In figs. 8-10, the length of buffer 1005 can be the L for parameter Estimationb, and the length of buffer 905 can be to use In the L of capacity estimationc.EKF blocks 805 are iterated for each k, and RLS830 is for being each used as Lb, the k progress of integral multiple Iteration, and TLS835 is for being each used as LcThe k of integral multiple is iterated, and wherein k is time index.BFG estimation SOC tracking Required all required model parameters and battery capacity, but except OCV parameters (it estimates offline) and from measuring instrumentss The voltage and current measurement error standard deviation sigma of circuit calibrationviOutside.RLS blocks do not need any outside primary condition, need to only set λ=1 is put, it is initial value to provide for firm LS estimates, i.e.,WithWhereinWith For Mission Number.The mathematical proof of EKF blocks 805 is described below.
Figure 11 and 12 shows the flow chart of the method according at least one exemplary embodiment.For described in Figure 11 and 12 The step of can be carried out due to the execution of software code, the software code is stored in and equipment (e.g., shown in Fig. 1 and 2 BMS110) in related memory (e.g., memory 230) and by device-dependent at least one processor (e.g., processor 235) perform.It is contemplated, however, that alternate embodiment, such as the system for being embodied as application specific processor.Although following step quilt It is described as by such as computing device, but these steps need not be by same computing device.In other words, at least one processing The executable step described in below for Figure 11 and 12 of device.
Figure 11 describes selection and represents the equivalent model of battery for the flow chart of calculating estimation SOC method.Such as figure Shown in 11, in step S1105, the library storage of equivalent-circuit model of battery is represented in memory.For example, use offline number According to collection module 555, the data related to battery 105 (or battery eliminator) are collected.Use the data and universal circuit work Tool, can generate at least one equivalent circuit for representing battery.The equivalent circuit may include at least one equivalent voltage, resistance, electricity Any combinations of appearance and/or equiva lent impedance.See, for example, following Figure 13 A-13D.The mathematics of each equivalent circuit can also be generated Equivalents.Equivalent circuit and/or mathematical equivalent form are storable in such as equivalent model storehouse 305.
In step S1110, the operational mode of battery is determined based on the load related to battery.It is for example, each equivalent Model can be based on the operational mode of battery (or battery eliminator).Operational mode can be based on the load related to battery.For example, operation Pattern can the voltage drop based on load both ends.For example, operational mode can based on load both ends voltage drop be it is of a relatively high or It is relatively low, be relative constancy or dynamic and/or combinations thereof.Therefore, operational mode can based on the electric current related to battery and/ Or voltage and/or the load related to battery determine.
In step S1115, based on identified pattern come select for identified pattern equivalent-circuit model it One.For example, equivalent model storehouse 305 can be searched for based on identified operational mode.For example, represent battery equivalent circuit and/ Or mathematical equivalent form can use the mode corresponding with operational mode identification (e.g., unique name or unique identification number) to store In equivalent model storehouse 305.Accordingly, it is determined that operational mode may include to determine operational mode identification, operational mode identification is then For searching for equivalent model storehouse 305.Selection equivalent circuit may include selection by search for equivalent model storehouse 305 returned it is equivalent Circuit or mathematical equivalent form.
In step S1120, the state of charge of battery (SOC) or estimation are calculated using selected equivalent-circuit model SOC.For example, as described above, voltage drop model parameter vector (b) can be based on by calculating SOC.Voltage drop model parameter vector can have The parameter of equivalent circuit based on battery (referring to above-mentioned equation 7).Accordingly, it is determined that voltage drop model parameter vector can be based on etc. Effect circuit has high or low complexity.For example, as described below, equivalent circuit may not include RC circuit elements, because electric capacity Charge and bypass resistance.Therefore, b (3) can be unique remaining voltage drop model parameter vector element.So as to simplify SOC or Estimate SOC calculating.In addition, between the terminal of battery 105 voltage v [k] (can be used for calculate SOC or estimation SOC) can be based on it is equivalent Circuit model.Equation, SOC and equivalent-circuit model related v [k] is more particularly described hereinafter.
Figure 12 shows the flow chart for the method that estimation SOC is calculated using recursion filter.As shown in figure 12, in step In S1205, the estimation state of charge (SOC) of stored battery is read from buffer.For example, buffer 850 can be deposited wherein Store up at least one SOC errors calculated in the preceding an iteration for the step described in the flow chart and SOC.It can postpone Rush at least one in the reading storage SOC value of device 850.
In step S1210, the measurement voltage at battery both ends is read.For example, such as digital amplifier 215 can be used to read Or determine voltage (e.g., the v shown in figure below 13A-13D [k]).In an exemplary specific implementation, voltage is stored in buffering In device.Therefore, different voltage measuring values can be used in different iteration.In other words, previous (on the time) voltage measuring value can use It can be used for iteration k+2 in current iteration or v [k+1].
In step S1215, filter gain is calculated.For example, it is as briefly described above and be described in more below, calculate The filter gain (e.g., G [k+1]) of EKF blocks 805.Filter gain can be based on being calculated using weighted least square algorithm At least one capability value.For example, filter gain can be based on using weighting recurrence least square (RLS) algorithm and an overall most young waiter in a wineshop or an inn Multiply at least one capability value that at least one of (TLS) algorithm calculates.Filter gain can be based on using weighting RLS algorithm The capability value calculated, the weighting RLS algorithm are based on SOC tracking errors covariance and current measurement errors standard deviation.Filtering Device gain can be based on estimation SOC variances.Filter gain can be based on the capability value calculated using TLS algorithms, the TLS algorithms Recurrence renewal based on covariance matrix.Filter gain can be based on searching the capability value calculated using open-circuit voltage (OCV). In SOC tracking errors covariance, current measurement errors standard deviation, SOC variances, covariance matrix and OCV each below The description (as mathematically) in more detail.
In step S1220, storage SOC, the voltage at battery both ends and filter gain based on battery calculate battery Estimation SOC.For example, the SOC of estimation, which can be equal to filter gain, is multiplied by estimation SOC of the digital voltage value plus storage.In step In rapid S1225, the estimation SOC calculated is stored in buffer (e.g., buffer 850).If necessary and/or need into One step calculates estimation SOC (S1230), then processing is back to step S1205.For example, if battery 105 is continuing in use, such as Fruit SOC errors, which exceed desirable value and further iteration, can reduce error, and/or if battery testing is carried out, etc., then It may be necessary to and/or need further to calculate.
Figure 13 A-13D show the schematic diagram of the battery equivalent model according at least one exemplary embodiment.Below will Optionally reference picture 13A-13D is to describe one or more exemplary specific implementations.As shown in Figure 13 A-13D, battery is represented Equivalent model 1300-1,1300-2,1300-3,1300-4 may include resistor 1315,1325,1340, capacitor 1330, 1345 and any combinations of equivalent voltage source 1305,1310.Voltage 1355 represents the voltage drop at battery both ends during loading.Electric current 1320th, 1335 and 1350 represent to flow through the electric current of the element of (or flowing to) equivalent model.For example, electric current 1350 represents to flow to load Electric current.
Resistor and the RC circuits of capacitor definable one.For example, resistor 1325 and 1330 defines a RC circuits.At some In exemplary specific implementation, capacitor can be completely charged and short circuit, causes RC circuits actually to be disappeared from equivalent model.Example Such as, in the equivalent model 1300-2 of battery is represented, the RC circuits defined by resistor 1340 and capacitor 1345 are not in model In, because capacitor 1345 is completely charged, form short circuit.In some exemplary embodiments, there is no (or few) and electricity The related hysteresis in pond (e.g., battery is in standing or traction seldom loads).Therefore, in the equivalent model 1300-4 for such as representing battery Shown, due in the absence of hysteresis, equivalent voltage source 1310 is not in a model.
Next the application describes the details of exemplary specific implementation.The details may include at least one above-mentioned equation Foundation (e.g., mathematical proof or simplification).For clarity, these equations are repeated, however, these equations will retain square brackets Equation numbering shown in ([]).Since identifying real-time model, it refers to following annotation.
a[k]TObservation model
AκA [k] in batch κTContinuous Observation, stack in a matrix
B [k] observation model parameter
The LS estimations of model parameter
The RLS estimations of model parameter
I [k] flows through the electric current of battery
KiOCV parameters:K0,K1,K2,K3,K4,K5,K6 K7
LbBatch length for parameter Estimation
ni[k] current measurement errors
nv[k] voltage measurement error
PbThe covariance matrix of [κ] LS estimators
R0Inside battery series resistance
R1 R1C1The internal resistance of cell in circuit
R2 R2C2The internal resistance of cell in circuit
The voltage at v [k] batteries both ends
vD[k] voltage drop
The Continuous Observation voltage drop of the vector form of kth batch
VDThe z-transform form of [z] voltage drop
Flow through R1Electric current
Flow through R2Electric current
xs[k] state of charge (SOC) s [k]
xsThe estimation of [k]
zi[k] flows through the measurement electric current of battery
zvThe measurement voltage at [k] battery both ends
Δ sample time
LS errors of fitting
CiPercentage evaluated error
RiPercentage evaluated error
Voltage drop observation noise covariance
The element of SOC track algorithms may include:
A. the estimation of OCV parameters:It is normalized when by using the time limit and dependent on the battery capacity of service life When, it is stable that OCV-SOC, which is characterized and varied with temperature with cell degradation,.
B. the estimation of dynamic equivalent circuit parameter:It has been observed that these parameters are with the temperature of battery, SOC and service life And change, therefore should adaptively estimate while BFG is run.
C. the estimation of battery capacity:Although the nominal capacity of battery is as specified by manufacturer, known available battery capacity It can change because of manufacturing process error, temperature change, loading mode and aging.
D. the SOC by model parameter constraint is tracked:Once known models parameter, SOC tracking has reformed into nonlinear filtering Problem.
Exemplary embodiment allows to carry out real time linear estimation to the dynamic equivalent circuit parameter of battery.It is following by solving Problem, the method for improving existing battery equivalent circuit modeling and parameter Estimation is realized in this exemplary embodiment:
A. some models only consider resistance, are not suitable for dynamic load.
B. they carry out system identification using nonlinear method.
C. the initial parameter estimation for method of model identification is needed.
D. assume that single dynamic equivalent model represents all battery operation patterns.
In the exemplary specific implementation, solve the problems, such as aforementioned four and be summarized as follows:
A. it is used for the online linear method of model parameter estimation, without estimating that the exact physical of battery equivalent circuit represents shape The parameter of formula.SOC tracking mode spatial models make use of it is amended can Linear Estimation parameter estimation.
B. it is applied to various batteries, without any initial value or calibration:Due to exemplary status spatial model Adaptivity, the SOC trackings proposed do not need any offline initialization of model parameter.Least square (LS) method carries Supplied needs whenever the initialization (or reinitializing) to parameter, block recurrence least square (RLS) is used to hold Continuous trace model parameter.Additionally, it has been suggested that amended open-circuit voltage (OCV) model different battery models, different temperatures and It is effective under the conditions of different loads.This just makes the battery that exemplary BFG can be in a manner of plug and play applied to broad range, nothing Need any other relative other information.
C. different battery modes are carried out with the possibility of seamless SOC tracking.Can recognize that four different battery equivalent models with Reflection very light load or static condition, constant current or low-frequency load, dynamic load and the heavy duty of change.Also identify Four (somewhat) different dynamic equivalent models are with these patterns of best match.These models track available for seamless SOC, without By the patterns of change of battery operation.
D. hysteresis modeling, it eliminates the needs of hysteresis modeling:Exemplary specific implementation is recognized hardly possible (complete Hysteresis modeling beauteously) is carried out offline, because hysteresis is relevant with SOC ∈ [0 1] and load current I ∈ R.Therefore, according to exemplary Embodiment, in voltage drop model, it will be late by being modeled as the error in OCV, and online filtering method constantly tries to pass through tune Whole SOC carrys out fill in a gap (to correction value).
The real-time model parameter Estimation that real-time model identification is carried out including the use of equivalent circuit.Figure 13 A are exemplary battery The equivalent circuit of (e.g., battery 105).When battery, which is in, to be stood, V0(s [k]) is the OCV of battery.OCV uniquely depends on battery SOC, s [k] ∈ [0,1].When battery is in active state, such as when current active be present, the behavior of battery, which passes through, to be moved State equivalent circuit represents that the dynamic equivalent circuit is by lag element h [k], series resistance R0And two be connected in series are simultaneously Join RC circuits (R1,C1) and (R2,C2) composition.Discrete time uses [k] instruction.
In figure 13a, the measurement electric current for flowing through battery is write as:
zi[k]=i [k]+ni[k] (8)
Wherein i [k] is the real current for flowing through battery, and ni[k] is current measurement noise, it is assumed that current measurement noise For zero mean and there is standard deviation (s.d.) σiThe measurement voltage at battery both ends is:
zv[k]=v [k]+nv[k] (9)
Wherein v [k] is the real voltage at battery both ends, and nv[k] is voltage measurement noise, it is assumed that voltage measurement noise For the zero mean with s.d..σv
According to following form writing inner member R0,R1,R2Dropped with the cell voltage at h [k] both ends:
Wherein flow through resistor R1And R2Electric current can be write according to following form
Wherein,
And (14)
Δ is sampling interval.
By using measurement electric current zi[k] replaces i [k], the electric current in (11) and (12) again book can be written as into form:
(8) are used now, (10) can book be written as form again in z domains:
Next, write (15) again in z domains:
Draw
And for (16) similarly,
By the way that (19) and (20) are substituted into (17):
Rearrange (21) and convert it back to time domain:
Wherein,
α=α12 (23)
β=α1α2 (24)
-h[k]=Xh[k]-αXh[l of k mono-]+β Xh[k-2], (27)
And (28)
(29)
(22) are write into following form again now:
vD[k]=[k]T+nD[k] (30)
Wherein observation model a [k]TProvided with model parameter vector b as follows:
Wherein subscript 4 indicates above-mentioned model corresponding with the model 4 of four models shown in Figure 13 A-13B.
(30) in voltage drop observation noise writing into:
It has the autocorrelation being given below:
And assume in length LbTime interval batch during lagging component be constant, for example,
Now, four different battery operations " pattern " can describing and match these patterns appropriate battery it is equivalent Model.
A. pattern 1-light load or static condition:When battery is only gently loaded, then it is electrically charged, it is stagnant when then standing Component will be so small as to negligible afterwards.The example of the pattern will be cell phone, and it is after fully charged, when expending nearly all Between contact base station until charge event next time, in addition to possible a small number of calls.Single resistor is non-(referring to Figure 13 D) Often it is adapted to the pattern.
B. 2-constant current of pattern is run:When the electric current for flowing through battery is constant, the capacitor in RC circuits is changed into complete Charging.Therefore, in terms of parameter Estimation viewpoint, the circuit of gained can be considered as single resistor and hysteresis/biasing element (referring to figure 13C).The constant current charge of battery is the good example of the pattern.
C. 3-dynamic load of pattern:When battery is in the pattern, different size of a large amount of loads be present.Example:Periodically Ground is used for the smart phone of call, web-browsing, video clipping etc..Battery eliminator shown in Figure 13 B is adapted to this very much Scene.
D. pattern 4-heavy and change use:For mobile phone, video when heavy and change use includes long Broadcasting, multimedia and game application etc..Figure 13 A match the scene very much.
Pay attention to, the different model complexities of dynamic equivalent circuit can be by only changing [k]TTo represent.It shown below pin To [k] of each modelTDefinition.For each in above-mentioned model complexity, noise item nD[k] according to following form withWithRepresent:
Wherein,
α=α12 (38)
β=α1α2 (39)
The least-squares estimation of constant dynamic model parameters when following discussion is related to.True SOC at time k is represented For:
The SOC track algorithms of foundation can be used for obtaining, i.e. xsThe more new estimation of [k].Now, the voltage in (10) V dropsD[k] can write into:
Wherein v0(xs[k | k]) the estimation open-circuit voltage (OCV) of battery is represented, it can be described as the function for estimating SOC. Following OCV-SOC relations can be used:
Can be by step as described below to OCV parameters Ki∈{K0,K1,K2,K3,K4,K5,K6,K7Estimated offline Meter.By considering LbBatch is observed, can book be written as form again by (30):
Wherein κ is Mission Number,
And noiseWith following covariance
WhereinFor five diagonal toeplitz matrixs, wherein diagonal entry, the first and second off diagonal elements point Do not pass throughWithProvide (see (32)).Now, dynamic model parameters vector can pass through a most young waiter in a wineshop or an inn Multiply (LS) optimization to be estimated as follows by (42):
The covariance matrix of LS estimators provides as follows:
When obtaining the measured value of new lot, LS estimates can be updated by repeating (50)-(51) recurrence
Wherein λ is to forget (Attenuation Memory Recursive) factor, ()TRepresent transposition, ()-1Expression is inverted, and Pb -1[κ] is claimed For information matrix, the information matrix can be multiplied by suitable constant to initialize by appropriately sized unit matrix.It can pay attention to Arrive, as λ=0,It is changed into.May be present makesApproximate a variety of methods.May be selected following two approximation methods be used for than Compared with:
A. following approximation method can be carried out:
B. the estimate before use to be to build electric current covariance matrix, such as using:
In
Following discussion is related to least mean-square error (MMSE) estimation of the dynamic model parameters of time-varying.Assuming that dynamic model is joined Number is the stochastic variable that following slowly varying Wiener-Hopf equation occurs:
xb[k+1]=xb[k]+wb[k] (54)
Wherein wb[k] is with covariance ∑bZero mean white Gaussian noise.Now, (30) are used to be used as measurement model And (53) are used as process model, Kalman filter provides b MMSE estimations.SOC can be used for determining vD[k] (see (42)), SOC tracking/smooth iterative algorithm and pass through the parameter Estimation based on Kalman filtering that the observation window of sufficient length is carried out can For improving the precision of SOC tracking and parameter Estimation.
Following discussion is related to open-circuit voltage (OCV) parameter Estimation.SOC estimation can utilize battery open-circuit voltage (OCV) with Uniqueness and stable relations between SOC simultaneously allow to calculate SOC for the OCV measured.However, only when battery, which is in, to be stood, Just can direct measurement OCV.When battery is in use, the dynamic relationship between cell voltage and electric current must pass through parameter and shape State method of estimation illustrates.State of charge method of estimation based on OCV-SOC includes and following related error:(1) battery The modeling of dynamic equivalent electric model and the uncertainty of parameter Estimation;And the error of the voltage and current of (2) measurement.(43) The parameter that middle OCV-SOC is characterized can be estimated by gathering OCV characterize datas on Sample Cell as follows:
A. since battery that is fully charged, standing completely
B. its open-circuit voltage V is recordedbatt=Vfull
C. k=1 is set
D. v [k]=V is recordedbatt;Record SOC [k]=1
E. k=k+1 is set
F. using minimal amount of (usual C/30 or C/40, wherein C are the battery capacity represented with Ah) constant current i [k] Make battery continuous discharge, until battery discharges completely.Once electric discharge completely, battery is just set to keep standing, and hereafter charging is straight It is fully charged to battery.Then
1. battery terminal voltage is measured, the V per the Δ secondbatt
2. record v [k]=Vbatt
G. SOC [k]=SOC [k-1]+c is recordedhi[k]Δ
Now, OCV models (43) vector format can represent as follows for all measured value v [k]:
V=Aocvk (55)
Wherein
V=[v [1], v [2] ..., v [Nv]]T (56)
Aocv=[aocv(1),aocv(2),…,aocv(Nv)]T (57)
K=[K0 K1 K2 K3 K4 K5 K6 K7 R0]T (58)
Then assignment s [k]=SOC [k] is passed through
Now, the least-squares estimation value and internal resistance of cell R of OCV parameters0Provide as follows:
Following discussion is related to four exemplary equivalent circuit models.
Table 1
Pattern number Equivalent circuit
Model 1 Figure 13 D
Model 2 Figure 13 C
Model 3 Figure 13 B
Model 4 Figure 13 A
The voltage drop at the circuit element both ends of each can following form in four equivalent-circuit models shown in table 1 Writing:
vD[k]=[k]T[k]+nD[k] (61)
Wherein,
For model 3:
And for model 4:
α=α12 (66)
β=α1α2 (67)
It is related to the derivation of Noise Correlation below.In this part, for | l |=0,1,2 and be directed to | l |>2 auto-correlation Property (27) can by (33) derive it is as follows:
L=0
L=1
L=2
Now, it is above-mentioned to represent as follows for each model:
The discussion continues real time capacity estimation, and it refers to following annotation.
CbattBattery capacity
chCoulomb meter number system number
The LS estimates of battery capacity
The RLS estimates of battery capacity
The TLS estimates of battery capacity
The vector of SOC differences
The vector of SOC differences between two standings
Coulomb vector
Coulomb vector between two standings
HκAugmentation observing matrix
I [k] flows through the electric current of battery
KiOCV parameters:K0,K1,K2,K3,K4,K5,K6 K7
niNoise in [k] measurement electric current
Ps[k | k] SOC estimation error variances
Qc[κ] capacitance variability variance
Rc[κ] uses the capacity estimation error of TLS, RLS or OCV method
RLS capacity estimation error variances
RTLS[κ] TLS capacity estimation error variances
ws[k] process noise
Differential SOC errors
The process noise accumulation between standing and follow-up standing at k
The differential SOC errors searched based on OCV
The vector of differential SOC errors
The vector of differential SOC errors based on OCV
xcThe battery capacity of [κ] fusion
SOC estimates
Error in SOC estimations
xs[k+1] SOC
The SOC estimations searched based on OCV
Error in the SOC estimations searched based on OCV
zcThe capacity estimation that [κ] is carried out by TLS, RLS or OCV method
zi[k] measures electric current
Capacity estimation error
ηcCharge efficiency (η=ηc)
ηdDischarging efficiency (η=ηd)
The covariance of augmentation observing matrix
The covariance of differential SOC errors
The state of charge (SOC) of battery, is defined as:
Formula 78 provides the information about battery status.When the understanding of SOC and battery capacity is used to estimate the closing of battery Between (TTS) or fully charged time (TTF).Battery capacity generally varies with temperature and its root Ju use pattern and service life Weaken over time.Accurate battery capacity tracking is the key element of battery electric quantity metering.
In the exemplary specific implementation, online capacity estimation can be based on:
A. there is weighting recurrence least square (RLS) estimation of the capacity of accurate weight derivation.For online capacity estimation Weighting RLS methods include based in the updated SOC tracking the whole time in variance and covariance and current measurement Error to standard deviation derives the expression formula of weight.
B. it is used for the TLS methods of real-time tracking battery capacity.TLS methods provide to be reached for the closed meter of capacity estimation Formula.This method can be used for the change of continuous tracking battery capacity.
C. the adaptive capacity estimation that the OCV based on resting batteries is searched.For tracking the TLS side of battery capacity online The SOC that method by using battery standing point search based on OCV estimates.
D. the fusion of the capacity estimation obtained by distinct methods.
Following discussion is related to battery capacity estimation and fusion.The instantaneous charge state (SOC) of battery can be written as following mistake Journey model, it is also referred to as coulomb Counting Formula, is represented according to following form with measuring electric current:
Wherein xs[k] ∈ [0,1] represents the SOC, C of batterybattFor the battery capacity represented with ampere-hour (Ah), and zi[k] is measurement electric current
zi[k]=i [k]+ni[k] (80)
It is by with standard deviation (s.d.) σiZero mean white noise ni[k] is damaged.(79) process noise in (80) measurement noise in is related, such as
ws[k]=- chΔni[k] (81)
And for zero mean, and s.d. is:
σs=chΔσi (82)
Wherein Coulomb meter number system number is
Here, η be depending on battery charging or discharge constant, for example,
And Δ is (constant) sampling interval.
Following discussion is directed to use with the online battery capacity estimation of recurrence least square (RLS).Estimation SOC can be based on voltage And current measurement value.Two continuous SOC value xs[k] and xs[k+1] is written as with its estimate:
Wherein evaluated errorWithThere is zero mean and variance P respectivelys[k | k] and Ps[k+1| k+1].Covariance between two continuous evaluated errors is:
Wherein G [k+1] is scalar Kalman gain, and H [k+1] is the scalar linearisation observation model at time k+1. Now, write again in the form of following (80):
(84) and (85) substitution (87) is obtained:
Wherein,
And differential error provide it is as follows:
For with zero mean, and variance is as follows:
Wherein S [k+1] is the new breath of Kalman filterCovariance.By considering LcSample batch, (88) following form can be write into the form of vectors:
Wherein,
κ is Mission Number,
LcFor batch length,
AndFor the white Gaussian noise vector with following covariance:
It is Lc×LcDiagonal matrix, its n-th diagonal element provide as follows:
Now, battery capacity LS estimations reciprocal provide as follows:
And the variance of LS capacity estimation reciprocal is:
It is new a collection of when obtainingPair when, LS estimation can recurrence renewal it is as follows:
WhereinFor the L for capacity estimationc×LcInformation matrix, and λ are Attenuation Memory Recursive constant.It should be noted that , in (92)By known noisy measurement current value structure, and above-mentioned LS and RLS methods of estimation are assumedIt has been It is complete known.For actual solution, it is considered asIn uncertainty.Next, describe based on overall minimum Two multiply the method for (TLS) optimization, which solveIn error.
Following discussion is directed to use with the online battery capacity estimation of adaptive total least square (TLS).In this part, Online capacity estimation method is established based on TLS, it is assumed in (92)WithIn exist uncertainty.Structure is following Augmentation observing matrix:
The information matrix related to augmentation observing matrix be:
Write in the form of followingFeature decomposition:
Wherein,
ΛκFor by diagonal 2 × 2 matrix from the non-negative characteristic value for being up to minimum arrangement, i.e. Λκ(1,1) represent maximum special Value indicative, and Λκ(2,2) minimal eigenvalue is represented.
2 × 2 matrixesEach column have that character pair is vectorial, i.e. first rowFor corresponding to maximum feature The characteristic vector of value, and secondary seriesFor the characteristic vector corresponding to minimal eigenvalue.
Then battery capacity TLS estimations reciprocal pass throughThe ratio of component provides, i.e.,
WhereinForThe i-th element, andFor(i, j) element.(105) derivation show as Under.
For smooth estimation, the information matrix in (103) can use Attenuation Memory Recursive to update, as follows:
Now, it is based on [85], TLS evaluated errors covariance (approximation):
WhereinFor HκThe i-th row, M HκIn line number, and
Following discussion is related to the battery capacity estimation based on open-circuit voltage (OCV).For given standing voltage zv[k], Corresponding SOC estimationsCan be by (33) inverted acquisition.Due to the hysteresis in battery, SOC estimationsWill Different from actual SOC xs[k], obtain OCV and search errorOCV search error will be always during electric discharge it is negative and It is always just during charging.The open-circuit voltage (OCV) of battery can be written as SOC nonlinear function, such as
Wherein can be by gathering the voltage and current measured value for battery slowly charge and then discharging and obtaining, come to being Number K0,K1,K2,K3,K4,K5,K6And K7Estimated offline.No matter battery whether sufficient standing, the OCV-SOC characteristics of gained are bent Line can be used in obtaining SOC measured value.For given standing terminal voltage (it is also open-circuit voltage) zvThe battery of [k] SOC is written as:
It OCV-SOC can be used to characterize to be calculated by calculating the inverse of (109).Have multiple for calculating nonlinear function Method reciprocal, such as Newton method and binary chop.This can be described as the SOC estimations searched based on OCV.(110) SOC in estimates Meter is damaged as follows by lagging voltage:
Wherein OCV searches errorCaused by the hysteresis effect in OCV.It should be noted that when battery is discharging Process is changed into standing at time k, and OCV searches errorIt should be negative.Similarly, when battery charging process after when Between be changed into standing at k, OCV searches errorTo be always just.However, the size of error will change with hysteresis size Become, it is size of current, SOC and the function of time before standing.Now, write again in the form of following (79):
xs[k+1]=xs[k]+chΔzi[k]+ws[k] (112)
xs[k+2]=xs[k+1]+chΔk+1zi[k+1]+ws[k+1] (113)
·
·
·
xs[k+N]=xs[k+N-1]+chΔk+N-1zi[k+N-1]+ws[k+N-1] (114)
By being separately summed (112) to (114) on both sides, following result is obtained:
Wherein,
For with standard deviationZero mean.
Stood assuming that battery is in time k and k+N, then (115) can be written as:
Wherein,
It should be noted that no matter OCV searches errorSign deviation battery mode ∈ (charging, electric discharge } this One is true, " differential error "(defined in (119)) can be positive or negative.By considering a large amount of differential errors, it is assumed that It is approximately white.Assuming that the differential of kth batchRespectively in first group of standing pointIt is quiet with second group Put a littleBetween gather, and provide as follows:
Wherein,
Now, it can be seen that (120) have with (92) identical form, whereinWithReplace respectively WithSo as to which the capacity estimation based on RLS and TLS can be derived for the observation based on OCV, following institute Show.RLS and the TLS estimation of capacity based on OCV are expressed asWithIt should be noted that given be interspersed with The given N of electric dischargerThe static condition of quantity, N can be carried outr(Nr- 1)/2 differential is observed.For example, for Nr=4, it is assumed that battery exists Time point t1,t2,t3And t4It is in static condition.
It is related to below by merging the capacity estimation carried out.In this part, describe the TLS for merging capacity and estimate The exemplary specific implementation of value.In this part, for the capacity estimation value based on TLS, establish and derive.These derivations can also answer For merging the capacity estimation value based on RLS.
Online capacity estimation valueIt is measured caused by the error in electric currentThe uncertainty of (referring to (92)) Caused by the error in SOC track algorithmsUncertain damage.Similarly, the capacity estimation value based on OCVIt is measured caused by the error in electric currentThe uncertainty and OCV of (referring to (120)) search differential error institute Caused byUncertain damage.It is assumed that online capacity estimation value etThe error of [κ] and the capacity estimation value searched based on OCV etoThe error of [κ] is uncorrelated.Based on these it is assumed that capacity fusion becomes the fusion of two independent flight paths.
First, it should be noted that capacity estimate reciprocal, such as capacity estimation valueWithRespectively 1/CbattEstimate.Correspondingly, respective evaluated error covarianceWithAlso correspond to capacity estimation reciprocal Value.Based on Taylor series expansion, the desired value of the dynamic capacity estimate based on TLS and corresponding estimation error variance are approximately:
WhereinFor the C based on dynamic databattEstimate, and RTLS[κ] is estimation error variance.By following Identical program, the capacity estimation value C based on OCV can be obtainedTO[κ] and corresponding evaluated error covariance RTO[κ].Now, it is false It is stochastic variable to determine battery capacity, and it undergoes following slowly varying Wiener-Hopf equation:
xc[κ+1]=xc[κ]+wc[κ] (126)
Wherein wc[κ] is assumed to have variance QcThe zero mean white Gaussian noise of [κ].Capacity estimation valueWithIt is adapted to following observation model:
zc[κ]=xc[κ]+nc[κ] (127)
Wherein,
κ ' and κ " is according to the time index of the latest estimated value of corresponding algorithm (being respectively TLS and TO), and nc[κ] is false It is set to the zero mean white noise with following variance:
Now, new measured value no matter when is receivedWherein κ '=κ or κ ' '=κ, melts The capacity estimation value of conjunction obtains as follows:
WhereinFor the previous renewal of capacity estimation value, and Pc[κ -1 | κ -1] it is previous estimation Error variance, it is updated to:
Above-mentioned fusion method can be similarly used for merging the capacity estimation value based on RLS.
It is related to the derivation of capacity estimation error covariance below.In this part, differential error in (90) is derived Covariance.For convenience's sake, differential error (90) is rewritten as following form
Purpose is to calculate variance:
According to following form writing process equation (79):
xs[k+1]=xs[k]+chΔzi[k]+ws[k](134)
WhereinFor xsThe Kalman filter estimate of [k+1], ν [k+1] is that wave filter newly ceases, and G [k+1] is kalman gain.(134) difference between (135) is:
It can be rearranged as following form:
So as to,
Wherein,
S [k+1] is new breath covariance.
The following describe the replacement or second method for confirming above-mentioned derivation.Deploy (132):
Wherein,
E1=Ps[k|k] (139)
E2=Ps[k+1|k+1] (140)
E5=0 (143)
The enclosed for being related to total least square (TLS) capacity estimation value below derives.By 2 × 2 matrixesWrite into:
A characteristic value meets
| A- λ I |=0 (147)
It is reduced to
Wherein λ1For eigenvalue of maximum and λ2For minimal eigenvalue.Corresponding to λ2Characteristic value meet
Wherein
For example,
And
It is related to the conversion of capacity estimation reciprocal below.Exemplary specific implementation includes estimating based on estimate reciprocal and inverse The method that error variance draws capacity estimation value and estimation error variance.It is simple for capacity estimation value reciprocal and error variance assignment Variable, for example,
Definition:
Our purpose be find E { y } andApproximation.
It is related to the desired value for determining y below.Second order Taylor series approximation value is by given below:
E { y } Two-order approximation value is by given below:
It is related to the variance for the desired value for determining y below.In actual value x0F (x) is expanded into first order Taylor series around
Y=f (x)=f (x0)+f'(x0)(x-x0) (162)
Y variance is by given below:
Now, the desired value of capacity estimation value and its estimation error variance are by given below:
Present disclosure tracks followed by state of charge (SOC), and it refers to following symbol.
a[k]TVoltage drop model
Amended voltage drop model for SOC tracking
BFG battery meters
B voltage drop model parameters
Amended voltage drop model parameter for SOC tracking
CbattBattery capacity
chCoulomb meter number system number
G [k+1] SOC tracking filter gains
The lagging voltage of h [k] battery
I [k] flows through the electric current of battery
i1[k] flows through R1Electric current
i2[k] flows through R2Electric current
KiOCV parameters:K0,K1,K2,K3,K4,K5,K6 K7
nD[k] voltage drop observation noise
niNoise in [k] measurement electric current
nvNoise in [k] measurement voltage
Noise in measurement model
OCV open-circuit voltages
Ps[k+1 | k] SOC estimation error variances
Ps[k+1 | k+1] SOC estimation error variances
R0The internal resistance of cell of series connection
R1 R1C1The internal resistance of cell in circuit
R2 R2C2The internal resistance of cell in circuit
The autocorrelation of voltage drop observation noise
SOC state of charge
S [k] state of charge (SOC)
Cross correlations of the U [k] between measurement noise and process noise
The voltage at v [k] battery terminals both ends
V0(s [k]) open-circuit voltage (OCV)
xh[k] corresponds to h [k] state component
Corresponding to i1The state component of [k]
Corresponding to i2The state component of [k]
xs[k] corresponds to s [k] state component
The SOC estimation of prediction
Update SOC estimation
The state of x [k] vector form
zi[k] measures electric current
zv[k] measures voltage
αi(k) the current attenuation coefficient in RC circuits,
Time difference between Δ neighboring samples
SOC tracking errors
ηcCharge efficiency
ηdDischarging efficiency
In the exemplary specific implementation, electrochemistry is tracked based on instantaneous terminal voltage, load current and measured temperature The state of charge (SOC) of energy storage device (battery).SOC track algorithms use above-mentioned model parameter estimation and battery capacity estimation Understand.Exemplary SOC tracking will be late by the error being modeled as in open-circuit voltage (OCV) and using parameter Estimation and SOC tracking skills The combination of art compensates to it.This, which is eliminated, will be late by off-line modeling as SOC and the needs of the function of load current.The example Property model cause for SOC tracking depression of order (e.g., single status) filtering, no matter wherein the complexity level of battery equivalent model How, it need not track supplementary variable.Identify the presence of correlated noise and use it for improving SOC tracking.With routine " one Model adaptation owns " unlike strategy, four different equivalent models of battery are identified, these models represent representative cells fortune Capable four unique patterns simultaneously establish the framework for seamless SOC tracking based on appropriate model.
It is related to including SOC with the typical reduced order state filtering method for combining (recurrence) estimation of (unnecessary) amount of other redundancies Expensive matrix operation and the precision of SOC estimations is reduced in calculating.In the exemplary specific implementation, having used does not increase The depression of order filtering of state space dimension, so as to the computational complexity for obtaining more preferable SOC precision and reducing.By will be late by modeling For the error in OCV, the needs that hysteresis models are eliminated, and online filtering method is constantly tried to by adjusting SOC (to amendment Value) carry out fill in a gap.So as to which hysteresis is modeled as time-varying biasing.Using noise whitening program, and derive amended state Spatial model, to ensure that SOC track algorithms draw result as well as possible in least mean-square error meaning.Using battery not With " pattern " tracking SOC.At least four different battery equivalent models are used to reflect very light load or static condition, constant electricity Stream operation or low frequency loading (e.g., charging), dynamic load and heavy duty.Also identify four (somewhat) different dynamic equivalent moulds Type is with these patterns of best match.The depression of order filtering method provided ensure that seamless SOC tracking, regardless of whether the mould of battery operation Formula changes.
Following discussion is related to exemplary system models.Battery equivalent circuit model considered here is shown in figure 13a. When battery, which is in, to be stood, V0(s [k]) is the OCV of battery.OCV uniquely depends on the SOC, s [k] of battery.Lived when battery is in During jump state, such as when current active be present, the behavior of battery is represented by dynamic equivalent circuit, the dynamic equivalent circuit By lag element h [k], series resistance R0And two parallel RC circuits, (R being connected in series1,C1) and (R2,C2) composition.It is discrete Time uses [k] instruction.
The battery equivalent circuit model that this part is considered is shown in figure 13a.With regard to the element in battery equivalent circuit The terminal voltage v [k] of speech is by given below:
V [k]=V0(s[k])+i[k]R0+i1[k]R1+i2[k]R2+h[k] (166)
Wherein V0(s [k]) represents the open-circuit voltage (with voltmeter) of battery at time k, and it is write as at time k here SOC function, s [k] ∈ [0,1];H [k] illustrates the hysteresis in cell voltage;i1[k] and i2[k] is respectively to flow through R1And R2's Electric current.
Some non-linear representations be present, it makes the function that OCV is similar to SOC.In the exemplary specific implementation, use In the multinomial log-linear model reciprocal that OCV is represented with SOC:
Wherein K0,K1,K2,K3,K4,K5,K6And K7Offline estimation can be characterized by OCV-SOC.SOC transient change can book Write as following form (introduce under x and be marked with instruction state component):
Wherein i [k] is in terms of amperage;
ch=η/3600Cbatt (169)
For with ampere-1Second-1The Coulomb meter number system number of meter, CbattFor the battery capacity in terms of ampere-hour (Ah), Δ be with The sampling interval of second meter, and η be constant, and it is charging or electric discharge depending on battery, for example,
It should be noted that (168) draw the instantaneous SOC of battery.Calculate SOC the technology be referred to as coulomb count and/ Or " the SOC " of prediction.Coulomb count assume solution initial charge state and completely understand battery capacity so as to consider transfer certainly/ The coulomb amount for being transferred into battery calculates residual charge state afterwards.Coulomb counting error includes the understanding of (1) to initial SOC not Certainty;(2) uncertainty of the understanding to battery capacity;And (3) because measurement electric current error and because of timing oscillator The error of coulomb is measured caused by the error of the time difference caused by inaccuracy/drift.
Electric current i [k] is measured and error easily occurs in current measurement value.Will measurement electric current zi[k] is write as:
zi[k]=i [k]+ni[k] (171)
Wherein ni[k] is current measurement noise, and it is considered to have white zero mean and with known standard deviation (s.d.)σi..Can be by using z as followsi[k] replaces i [k] and carrys out rewrites status equation (168):
xs[k+1]=xs[k]+chΔzi[k]-chΔni[k] (172)
It can as follows write and flow through resistor R1And R2Electric current:
Wherein
By using measurement electric current zi[k] replaces i [k], can rewrite the electric current in (172) and (173) as follows:
Lagging voltage h [k] is the load current of battery and SOC nonlinear function.Delay Process can be write as:
Wherein nh[k] is the process noise of lag model, it is assumed that it is zero mean white Gaussian noise and has s.d. σh。 (166) voltage in for measurement amount and measures voltage zvEasily there is error in [k].Measurement voltage is write as:
zv[k]=v [k]+nv[k]=V0(s[k])+i[k]R0+i1[k]R1+i2[k]R2+h[k]+nv[k] (179)
Wherein nv[k] is assumed to have zero mean and s.d. σvWhite Gaussian noise.Now, by by (171), (172), (173) and (178) are substituted into (179), derive following measurement model:
Wherein,
Now, instantaneous voltage and current measurement value z are providedv[k] and zi[k], BFG purpose is the instantaneous of tracking battery SOCxs[k].." unnecessary " variable in observation model (180)And xhThere are problems that of [k] causes Combined estimator, That is SOC and these variables must Combined estimators.This can be by forming the multidimensional mistake of vector form as shown in (183)-(189) Journey and measurement model and/or by application Bayes's nonlinear filtering technique recursively to estimate to realize:
All measured values until time k are provided, { [0], [1], [2] ..., [k] }, wherein by (171) and (180) group Into.This can pass through nonlinear filtering technique known to application (such as extended Kalman filter (EKF), Unscented kalman filtering Device (UKF) or particle filter) effectively carry out.Can be in the form of vectors by process equation (172), (176), (177) and (178) Write into:
Or, it is abbreviated as:
X [k+1]=Fkx[k]+u[k]+ΓLw[k] (184)
Wherein,
For with the white noise of zero mean and following covariance vector:
Correspondingly, measuring equation (180) can write into:
Wherein,
And the n with zero mean and following s.d.z[k] noise vector.
It is additionally noted that state-space representation form (183)-(189) involve a need to pass through system identification technique Come the following model parameter estimated, including battery capacity:Cbatt, open-circuit voltage model parameter:K0,K1,K2,K3,K4,K5,K6,K7, Dynamic equivalent circuit model parameter:R0,R1,C1,R2,C2, it is charged and discharged efficiency:ηcd, process-noise variance:And Measurement noise variance
Requirement to all model parameter estimations makes SOC tracking problems have more challenge.In addition, the chemical property meeting of battery Change because of temperature change, aging and use pattern, therefore these model parameters can time to time change.So as to be pushed away with the time Move, it is necessary to reevaluate model parameter.
In exemplary specific implementation, it is assumed that the OCV parameters K of battery0,K1,…,K7Estimate offline.Described above is Estimate the program of these parameters.It is assumed that voltage and current error to standard deviation, i.e., respectively σvAnd σi, can be designed to from measuring circuit Arrive.It is assumed that it is charged and discharged efficiency, i.e., respectively ηcAnd ηd, known by calibration.So as to, it is therefore an objective to by assuming electrolytic cell Capacity CbattWith the electrical equivalent model parameter R of battery0,R1,R2,C1And C2To establish online SOC track algorithms.
Discussion below is related to SOC tracking.The purpose of depression of order filtering is tracking xs[k], while redundant variables need not be trackedAnd xh[k].First, rewritten (172) according to following form:
xs[k+1]=xs[k]+chΔzi[k]+ws[k] (190)
Wherein,
Ws[k]=- chΔni[k] (191)
For process noise, it is the white noise with following s.d.:
σs=chΔσi (192)
Now, voltage measuring value (180) is rewritten as:
zv[k]=V0(xs[k])+a[k]Tb+nD[k] (193) [4]
Wherein,
a[k]TB=[vD[k-l]vD[k-2]zi[k]zi[k-1]-zi[k-2]1] (194)
And voltage drop is by being given below:
vD[k]=zv[k]-V0(xs[k]) (195)
Wherein b is parameter vector to be estimated, and nD[k] is measurement noise.It should be noted that [k] in (194) With voltage drop vD[k-1] and vD[k-2] is defined.Described above is the estimation parameter b based on voltage drop observation model.By estimation Parameter is derived as according to the parameter of the battery equivalent model in Figure 13 A:
B (3)=Ro (198)
Measurement noise nD[k] is zero mean and has the autocorrelation by being given below
Next, voltage drop is described in the meaning for estimating the parameter b in (193).Use (193), voltage drop (195) can Write as:
vD[k]=a [k]Tb+nD[k] (203)
Given voltage drop observation, above-mentioned model (203) can be used for Linear Estimation b.However, in order to obtain the electricity as observation Pressure drop, the predicted value that SOC can be used areOr the SOC estimation after renewal isSOCxsThe understanding of [k], For example,
Describe below and how to be predictedAfter renewal(respectively referring to (209) and (215)).BFG existing method uses voltage and current observation zv[k] and zi[k] carries out Model Identification and SOC Tracking.Consider conventional voltage observation model (180).xhItem in [k] represents lagging voltage, and it is electricity as shown in (178) Flow i [k], SOCxs[k] and time k. function.For example, when battery experience 1A load (this is heavy duty in Mobile solution) During the several seconds, compared with when load continuous 30 minutes for 1A situation, caused by the amplitude that lags it is smaller.In addition, the amplitude of hysteresis Also it is SOC function in the time.
Due to observing (180) using for the voltage between the battery terminal of Model Identification, it is also necessary to lagging xh[k] is modeled And model parameter must be estimated.The model of the hysteresis represented using SOC, electric current and time is complete to be nonlinear, and not yet Understand.Attempt to model hysteresis and another shortcoming of estimation is that it make it that on-time model identification is hardly possible.Due to stagnant It is SOC function afterwards, Model Identification needs the data across whole SOC scopes.This there may come a time when it is impossible, because some Using may be never by battery from completely using sky.Because hysteresis or electric current function, Model Identification need across should Use data for the possibility load current of various duration.So as to which complete hysteresis modeling and Model Identification become not Reality.
It is also important that, it is noted that model parameter is estimated offline using Sample Cell, then these parameters are used for electricity Metering may be unsatisfactory;Some battery parameters have been notified based on use pattern and changed.Exemplary embodiment passes through introducing Above-mentioned voltage drop model avoids hysteresis modeling.Voltage drop vD[k] represents internal cell model element R0,R1,R2And xh[k] two The voltage at end.Item xh[k] is deliberately introduced to consider to be used for derive voltage drop " measured value "PredictionIn error.xh[k] is referred to alternatively as " instantaneous lagging ", and it should pass through adjustment according to exemplary embodiment SOC estimationTo correct to zero.As described above for described in Fig. 7, the use of voltage drop model can be used for eliminating hysteresis effect Should.Existing SOC is understood for calculating voltage drop.Estimating in a buffer and for parameter b is collected in a collection of passing voltage drop Meter.(determined by such as in model estimation module or block 710)Nonzero value instruction instantaneous lagging presence, this Mean SOC evaluated errors be present.SOC track algorithms are designed in estimationFor non-zero whenever, (SOC with In track block 715) amendment SOC.
OCV-SOC models (167) represent OCV-SOC relations.It is assumed that in voltage drop (195)(201) InEstimate will beHowever,Mean to be used to calculate voltage drop observation vDThe SOC of [k] estimates Error be present in evaluation.So as to which SOC track algorithms need corresponding adjustmentThis is by using following amended observation mould Type substitutes (193) to realize.
Wherein,
By removing a [k] respectivelyTObtained with last element in b.In other words, hysteresis item is removed.Retouch below The meaning of amended observation model is stated.In addition, the process noise w in (190)sMeasurement noise n in [k] and (205)D[k] Between following covariance be present.
The estimate of given state of chargeWith related variance Ps[k | k], following EKF recurrence (referring to Fig. 8) uses Voltage and current measured value zv[k+1],zi[k],zi[k+1] is drawnRenewal after SOC estimation and its Related variance Ps[k+1|k+1]..These steps also ensure SOC estimation by optimum apjustment to consider the covariance of (208). Filtering recurrence is made up of following:
Wherein ch[k] andIt is the latest estimated value of Coulomb meter number system number and model parameter vector respectively.It should describe to make With in state-space model (190)-(205)WithCarry out the importance of SOC tracking.Hysteresis can be considered as OCV- Error in SOC indicatrixes.It is likely difficult to modeling and accurate estimation lags, because it can becomes with previous electric current and SOC Change (referring to (178)).However, true OCV-SOC relations can be estimated.In fact, the V in (205)0(xs[k]) it is based on true OCV- SOC models.For example it is assumed that the hysteresis of estimation isThis means wave filter, " OCV " of perception is true with battery OCV differs 10mV.For BFG algorithms, the OCV of perception, V0(xs[k]), with SOC estimationDirectly (and monotonously) phase Close.In other words, if the OCV of wave filter is different from actual OCV, then wave filter estimateAlso different from battery True SOC.So as to when wave filter finds that it predicts terminal voltage in (212)Decline 10mV, it can be adjusted in (215) Its whole SOC estimationSo that " the OCV errors of perception " (or hysteresis H of estimation) (little by little) is adjusted to zero.From And the good instruction of the normal operation of the method proposed is estimationAll the time close to zero.
In the absence of the reliable apparatus and method for verifying electric quantity metering algorithm in desired manner.Voltameter is assessed using simulation It is infeasible, in default of can for example allow simulated battery dynamically reliable mathematical modeling.For example, the self-correcting positive model of enhancing The influence of cell degradation can not be considered.Because the uncertainty in state of charge, battery capacity and the true value of internal resistance (all may be used For the amount constantly drifted about), may be difficult to verify voltameter using single measurement or verification method.Need to calculate multiple checking degree Measure to understand the overall picture of voltameter precision.
Present disclosure is followed by benchmark test, and it is since measurement.In the exemplary specific implementation, use is described In multiple reference test methods of the electric quantity metering algorithm of checking electrochemical energy storage device (battery).Relatively accurate electric quantity metering (FG) cycle life of battery can be extended.The embodiment also describes accurate and objective voltameter evaluation scheme.This Measurement described in text can be used for measuring FG precision in many aspects and returning may indicate that at least one of voltameter overall performance Numeral.Benchmark test as described herein can be applied to a variety of electric quantity metering algorithms.E.g., including in this specific embodiment Details can be with entitled " Methods and Apparatus Related to Tracking Battery State of Charge:The A Reduced Order Filtering Approach (method and apparatus related to tracking battery state of charge: Depression of order filtering method) " description described in any concept combine.
Benchmark test described in the embodiment can for example, by be calculated as follows in three kinds of measurements of definition one Person or more persons are carried out:
First Exemplary metrology is coulomb counting error.With reference to the battery capacity to experiment and the state of charge of starting (SOC) understanding of point, coulomb counting methods and/or equipment can provide the accurate estimation to battery state of charge.Based on Coulomb meter Several SOC estimations and voltameter can be used as benchmark survey through the error (such as root mean square (RMS)) between SOC after a while estimates First measurement of examination.Error between SOC based on FG and the SOC based on coulomb counting might mean that the FG phases with being verified One or more problems of pass:
- the model characterized for open-circuit voltage state of charge (OCV-SOC) may be not accurate enough (assuming that FG uses OCV- SOC is characterized)
- the battery capacity estimation carried out by FG may be inaccurate
- dynamic equivalent circuit model may have problem in terms of for the selection of FG model and parameter Estimation scheme.
Second Exemplary metrology is OCV-SOC errors.(can be used what one or more methods and/or equipment were carried out) battery OCV-SOC characterize can provide for search SOC search program.So as to by making battery be in (or at least part) completely Static condition and the voltage by measuring battery, can by voltameter preset time SOC estimation can be characterized with OCV-SOC into Row compares to obtain error.OCV-SOC errors may indicate that one or more problems with of voltameter:
- the dynamic model for battery equivalent model may be invalid
- minimum OCV-SOC errors may indicate that the actual dynamic property of dynamic model and battery used is well matched with.
3rd Exemplary metrology reaches voltage time (TTV) error for prediction.Provide constant load/charging current, voltameter (can using one or more methods and/or equipment), prediction reaches a certain voltage spent time (TTV).Shut-in time (TTS) and Fully charged time (TTF) can be the special circumstances of TTV estimations.Error in TTV estimations can reach the actual electricity that is considered Calculated after pressure.The TTV errors may indicate that it is related to the voltameter being evaluated it is following one or more:
The impedance estimation precision of-battery,
The battery capacity estimation precision of-voltameter,
Understanding (such as information) of-voltameter to battery SOC,
The precision that-OCV-SOC is characterized.
Battery can show different qualities in response to temperature change.For example, the impedance of battery can it is higher at low temperature (and So as to which available horsepower may be relatively low).In response to the load under relatively low SOC, compared with the identical load of higher SOC level, OCV rates of change may be larger and nonlinearity.High-quality voltameter can have in the range of wide in range temperature and SOC level The ability of operational excellence.Benchmark test as described herein can be configured to ensure that during Performance Evaluation by these key elements extremely Some are included in test less.
Following discussion is related to the measurement characterized based on OCV-SOC.The state of charge of battery can uniquely with its open-circuit voltage (OCV) it is relevant.Figure 14 illustrates for curve map for one example of such relation.
A variety of methods can be used for obtaining OCV characterize datas.A kind of illustrative methods are summarized as follows:
1) since battery that is fully charged, standing completely
2) its open-circuit voltage V is recordedBAT=Vfull
3) i=1 is set
4) OCV (i)=V is recordedBAT
Record SOC (i)=1
5) i=i+1 is set
6) battery discharge is made up to Δ T duration using constant current I
7) make battery sufficient standing (such as at least 2 hours)
8) battery terminal voltage, V are measuredBAT
9) OCV (i)=V is recordedBAT
Calculate SOC (i)=SOC (i-1)+chI (i) Δ T, wherein ch=η/(3600Cbatt), η is instruction charge/discharge effect The constant of rate, CbattBattery capacity and I (i) to be represented with ampere-hour (Ah) are to flow into the electric current of battery.
10) repeat step 5 to 9 closes voltage VSD until OCV (i) reaches battery.
Now, { OCV (i) is used;SOC (i) } it is right, SOC ∈ [0 can be directed to;1] OCV is obtained to characterize.
Note:
1) OCV-SOC, which is characterized, to change because temperature is different.(if drying method can be used to calculate in certain use range OCV at any temperature is characterized).
2) in some specific implementations, battery capacity CbattBeing available from manufacturer's tables of data or its can estimate.
3) OCV-SOC signs can be with unchanged, regardless of whether the service life of battery.
4) characterized using above-mentioned OCV-SOC, the OCV of given SOC s battery can be calculated, write as υ=OCV (s).
5) characterized, can be calculated for given standing terminal voltage υ using above-mentioned OCV-SOCrBattery SOC, write as s= OCV-1r).If the SOC estimation of certain electric quantity metering algorithm at time k is reported asThen corresponding error can It is calculated as
Wherein υr[k] is the terminal voltage (after battery standing) at time k.
The checking characterized based on OCV-SOC can be used to calculate the OCV errors (217) of whole temperature and/or SOC region.
Checking in of a relatively high SOC region can be by being carried out as follows:First since full rechargeable battery and apply time-varying Load up to one time for being enough to consume 1/2 capacity that (about) is less than tested battery.Similarly, testing in low SOC region Card can be by being carried out as follows:First since the battery after full rechargeable battery or high SOC checkings, and apply and be enough battery band To the dynamic load of relatively low SOC region.
Average OCV-SOC errors (in terms of %) can be calculated as
Wherein ∈OCV(sL, Ti) represent at low SOC region and in temperature TiUnder the error that calculates.It is specific real at some Shi Zhong,Lower, FG algorithms are better.
Following discussion is related to the measurement based on relative coulomb counting error.Battery SOC can pass through coulomb as follows (CC) is counted to calculate:
It is assumed that battery capacity CbattWith sufficiently exact startingIt is known.So as to Coulomb meter The related FG errors of number may be defined as (in terms of %):
Wherein T is the duration (in seconds) for performing validation test.
1) in some specific implementations, some in battery measuring equipment and/or method may include that coulomb counting is used as it Part.However, above-mentioned measurement still can be considered verification tool, this is due to it has been assumed that understanding of battery capacity and checking BeginAnd FG methods may not assume that the understanding.
2) can be used for by the way that battery is discharged into empty (or basic overhead) completely from full (or substantially full) to pre-estimate CheckingBattery capacity.Or validation test can be carried out from full (or substantially full) to sky (or basic overhead), and And it can be updated with the battery capacity Cbatt of latest estimated, can be in this mode in some specific implementations Hysteresis and relaxation factor are considered in battery capacity estimation.In some specific implementations, FG algorithms can be forbidden so to do.
3) in some specific implementations, the temperature for the battery just assessed is in whole (e.g., substantially entirely) verification process It can keep constant.
Following discussion is related to based on the measurement up to voltage time (TTV).If the SOC of the electricity calculating method at time k estimates Evaluation is sFG[k], then its reach the time of voltage υ needs can book be written as form:
Wherein electric current I (I in charging process>0 and discharge process in I<0) keep constant (or substantial constant) until up to To voltage υ.
Once reach terminal voltage υ in the process of running, you can record reaches the real time during voltage.When specific Between when reaching voltage υ, following TTV checkings measurement can be calculated
Wherein Tv[i]=T-i are the real times for reaching voltage υ needs from time i.∈TTVValue can be in minutes.One In a little specific implementations, metric can be calculated (in terms of %):
The measurement of combination may be defined as:
WhereinIn terms of %.In some specific implementations, the value is lower, and voltameter is better.
In some specific implementations, benchmark may include to use (e.g., allusion quotation for evaluated battery loading reflection battery Type use) one or more different current loadings, and record voltameter report SOC and TTV readings.This method may include These steps are repeated under different temperatures, until for example Table I, II and III are filled.
In some specific implementations, simulation and actual loading curve can be used in verification process.Fictitious load curve The advantages of be the amount (e.g., correct amount) that can calculate the coulomb obtained from battery, so as to avoid by sampling and current sense institute One or more errors of cause.This can be based on such hypothesis:Load simulating device may not introduce one or more notable Error.Various actual loading curves and fictitious load curve can be created.
Actual loading curve:Can be used for example smart phone as load, come create actual loading curve (such as Figure 15 A and Shown in 15B).When load is connected to the battery being verified, following activity can perform:Call (15 minutes), web punchings Wave, e-mail is read, game etc. (20 minutes) is played, sends short messages (10 message), music is listened using loudspeaker or sees (30 points of video Clock video), it is standby-to enable cellular radio electrical communication base station (1 hour).
Load curve shown in Figure 15 A and 15B shows such scene, wherein three kinds can be calculated in single test (the input of each during e.g., 1515) OCV measurements 1505, TTV measure 1510 and CC measurements of type measurement.The experiment is from complete Rechargeable battery starts, and dynamic applies about 3 hours 15 minutes again using load.Battery is in leisure state 2 hours afterwards.5 Hour-symbols, which provide, calculates OCV-SOC error metrics ∈OCV(sH, Ti) chance.Constant current when experiment closes to an end Load allows to calculate ∈TTV(sL, Ti) coulomb count metric ∈ can be calculated according to whole dataCC(Ti)。
The load curve shown in Figure 15 A and 15B can be used to calculate temperature TiUnder metric:
OCV-OSC error metrics ∈ at high SOC regionOCV(sH, Ti)
TTC measurements ∈ at low SOC regionTTV(sL, Ti)
Coulomb count metric ∈CC(Ti)
Fictitious load curve:Short duration Δ can be usedsThe piecewise constant current loading I of interior different amplitudesmTo create Fictitious load curve.These segment loads can be mixed and be connected together, so as to obtain the fictitious load shown in Figure 16 A and 16B Curve.It note that fictitious load can occur between about 3.5 hours to 6.5 hours of test.Example can be used in the load curve Such as Δs=2 seconds and Im={ 40,120,130,160,300,400,440,520,600,640,800,880 } (unit:MA) Kikusui programmable loads devices is simulated.
Fictitious load curve shown in Figure 16 A and 16B shows such scene, wherein can be calculated in single test (the input of each during e.g., 1615) OCV measurements 1605, TTV measure 1610 and CC measurements of three types measurement.The experiment Since full rechargeable battery, and apply constant 500mA and load about 1.5 hours.Battery is in leisure state 2 hours afterwards, so After-applied dynamic load curve 3 hours.Mark within 15 minutes to provide again within 3 hours and calculate OCV-SOC error metrics ∈OCV(sH, Ti) Chance.Constant current load when experiment closes to an end allows to calculate ∈TTV(sL,Ti).Storehouse can be calculated according to whole data Logical sequence count metric ∈CC(Ti)。
In some specific implementations, the fictitious load curve shown in Figure 16 A and 16B can be used to calculate temperature TjUnder under Row measurement:
OCV-OSC error metrics ∈ at high SOC regionocv(sH, Ti)
TTC measurements ∈ at low SOC regionTTV(sL, Ti)
Coulomb count metric ∈CC(Ti)
Above-mentioned example specific implementation describe SOC suitable for battery powdered device (e.g., portable mobile device) with Track.The exemplary embodiment realizes linear method, and computationally inexpensive and effectiveness of performance is better than now the linear method Some is used for on-time model and knows method for distinguishing.Describe the weighted least-squares method for parameter Estimation.Parameter is described to estimate It is attested in weight (being based on variance) and parameter Estimation in the LS methods of meter to significantly improve.For the different operation moulds of battery The applicability of formula includes four different equivalent models that identification represents the battery of representative cells operational mode, and establishes seamless SOC The framework of tracking.Methods described will be late by the error being modeled as in open-circuit voltage (OCV), will be late by being modeled as so as to eliminate The needs of the function of SOC and load current.This method additionally aids from the SOC of mistake and initializes fast quick-recovery.
Above-mentioned example specific implementation is described for battery capacity estimation to promote the feature of battery electric quantity metering progress. Weighting recurrence least square (RLS) estimation of the capacity derived with accurate weight.The formula of weight can be based on SOC tracking errors Covariance and current measurement errors standard deviation calculate.Describe the TLS methods for battery capacity real-time tracking.TLS estimates Value with it is closed derive and available for by using Attenuation Memory Recursive renewal covariance matrix carry out ART network.It is electric based on standing The technology for the adaptive capacity estimation that the OCV in pond is searched.The OCV in deriving is considered to search the source of error (hysteresis) and describe The method that the adaptive capacity searched by OCV is estimated.Described based on capacity estimation value and evaluated error covariance by not The method of the optimum fusion of the capacity estimation value obtained with method, the method proposed is carried out adaptive using Kalman filter Answer optimum fusion.
Above-mentioned example specific implementation describe SOC suitable for battery powdered device (e.g., portable mobile device) with Track.For by the way that other nuisance parameter is stacked in state vector to estimate other nuisance parameter together with SOC Routine techniques is computationally high and effectiveness of performance is low.In order to avoid these problems, exemplary embodiment describes depression of order filter Wave pattern, for carrying out SOC tracking by new state-space model.Describe the state space with decorrelation noise model Model.SOC tracking problems are related to two measurement amounts, i.e. voltage and current, and this allows for the state of SOC tracking problems and measurement is made an uproar There is correlation between acoustic model.Describe and showed with the amended state space of uncorrelated state and measurement noise process Form.Exemplary embodiment describes the different operational modes of battery, and identifies the battery for representing representative cells operational mode At least four different equivalent models, and establish the framework of seamless SOC tracking.Exemplary embodiment describes a kind of method, the party Method will be late by the error being modeled as in open-circuit voltage (OCV), so as to eliminate the function that will be late by being modeled as SOC and load current Needs.This method additionally aids from the SOC of mistake and initializes fast quick-recovery.
Above-mentioned example specific implementation describes the SOC realized by some strategies and tracked.First, built by minimum battery Mould.The method proposed uniquely needs to carry out off-line modeling to open-circuit voltage (OCV) characteristic of battery.Parameter needed for every other Estimated by firm means.Due to the single set equipped with OCV parameters, the method proposed can perform at any temperature SOC is tracked, without any other parameter.Second, voltage drop observation model.The voltage drop model of observation allow online SOC with Track, without worrying that the lag element to battery models.This allows for proposed method and can obtain more preferable precision and sane Property.3rd, estimated by stability parameter.Identify the effect of the correlated noise structure in the least square model for parameter Estimation Should.For parameter estimation algorithm, this has just obtained obvious preferably precision and has enhanced robustness.4th, pass through battery capacity Estimation.Total least square (TLS) method for capacity estimation proposed ensure that the excellent precision of capacity estimation.And Finally, by using filtering, depression of order EKF methods consider noise process in state-space model correlation (be derived with Tracked for SOC) and correlation filter is gone so that the error in SOC tracking minimizes using appropriate.
In a specific embodiment, the benchmark based at least three battery meter quantity algorithms for assessing measurement is described to survey Method for testing.First assesses open-circuit voltage (OCV) sign that measurement can be based on battery.Second assesses measurement can the phase based on voltameter To coulomb counting error, and the 3rd benchmark can reach the calculating of the time needed for specific voltage based on battery.Each checking degree Amount may include to calculate some measurements at various SOC levels, different temperatures and/or voltage regime, such place.
Some SOC trackings include at least following weak point:(1) some models only consider resistance, are not suitable for dynamic Load;(2) they carry out system identification using nonlinear method;(3) need to estimate for the initial parameter of method of model identification Meter;(4) assume that single dynamic equivalent model represents all battery operation patterns;(5) importance of online capacity estimation is not solved; (6) existing online battery capacity estimation technology is influenceed by SOC and parameter estimating error, i.e., they are built on the sand;(7) remove Outside SOC, also using the online tracking of many amount of redundancys, (this causes to increase computational complexity and reduces SOC tracking essences for they Degree);(8) they need individually to model battery hysteresis, and it is that the function of SOC and load current (therefore is nothing that battery, which lags, Limit model), it is only possible to carry out approximate modeling to hysteresis;(9) it is any in existing method not recognize the process and measurement Correlation be present in noise process;And any one is not recognized due to temperature, aging, SOC and born in (10) existing method The fact that change and single equivalent model in battery behavior caused by carrying change may not be adapted to all these conditions.
Therefore, specific implementation as described herein can have short design time (in a couple of days), can have comparatively faster algorithm Convergence, and may have about in " real world " use condition 1% SOC and the precision of battery capacity report precision. In some specific implementations, can not (or seldom) need to customize battery model or data, and may include to have relatively fast SOC with The convergent adaptive learning algorithm of track.Some specific implementations may include automatic temperature-adjusting, service life and load compensation.
Some specific implementations can be based on such as decoupling of reduced-order EKF device, correlated measurement noise, online electric mould Shape parameter estimates and real time capacity estimation.
Depression of order Kalman filter may include the accurate SOC estimations (tracking) for being related to common four different parameters of estimation: SOC, the electric current for flowing through two different resistors in dynamic equivalent model and hysteresis, it is all these battery be in load/ It can change during charging.This is related to the complex matrix operation of commonly referred to as recursive filtering.The depression of order filtering method enters in this way Simplification is gone:SOC is only estimated by recursive filtering program.Three other specifications are by mathematical operation by marginalisation.The one of gained Individual or multiple SOC track algorithms are computationally feasible in voltameter SOC now.
Correlated measurement noise decoupling may include Kalman filter, its Kernel-based methods noise and the incoherent vacation of measurement noise And if run.In voltameter application, intrinsic measurement noise can be coupled to SOC and voltage in depression of order EKF in current measurement Both the measurement noise of variable and process noise.A kind of unique method is employed, this method is by electric current induced noise and karr The process noise decoupling of graceful wave filter.
Online electrical model parameters estimation may include the dynamic estimation of model parameter (coefficient).The dynamic of model parameter (coefficient) Estimation may include when it changes with SOC, changing currents with time load curve, temperature, charge-discharge cycles and/or the like Estimation.EKF wave filters are applicable, and precondition is that the model parameter of the dynamic equivalent circuit of battery is known, however, waiting Effect circuit represents the inner member of battery;These model parameters, which can also be used, is available from the measured value of battery to estimate:Voltage and Electric current.Solution as described herein estimates model parameter in real time fashion, so as to use kalman filter method.
Real-time or online capacity estimation may include to update (e.g., based on actual service conditions, load, temperature, service life Continuous updating) active volume one or more algorithms.Some specific implementations may include coulomb counting methods, and this method reports FG The SOC of announcement based on coulomb compared with counting the SOC that (bookkeeping method) calculates.Some specific implementations may include that TTE (use by electric power Time to the greatest extent) method, FG can be used to predict TTE and compared with actual conditions in this method.Some specific implementations may include SOC/OCV profile lookup methods, the SOC that this method can report FG is compared with SOC/OCV curves.
In some specific implementations, various batteries can be supported, it has a variety of particular battery models and every kind of electricity Chemical composition, battery manufacturers and/or the aging data in pond, it is such.In some specific implementations, voltameter appraisal procedure Alterable.For example, state of charge precision assessment method and test program, dynamic load and/or detailed test request, all such Class alterable.In some specific implementations, to requiring specification (such as key system perameter and required precision and/or the system integration It is required that such) feedback alterable.In some specific implementations, operating system driving requires alterable.
Figure 17 is to show the schematic diagram that example system is realized.System 1700 includes voltameter evaluation module 1705, electricity Pond 1710, battery meter 1715 and computing device 1715.Voltameter evaluation module 1705 can be realized as the software in BMS110 Module or ASIC.In other words, voltameter evaluation module 1705 can be code, the code be stored in memory 230 and by Processor 235 and/or another module related to BMS110 perform.Computing device 1715 can be from voltameter evaluation module 1705 Receive information and by the presentation of information on a graphical user interface (as shown in figure 18).
Figure 18 is to show the schematic diagram that the user interface being used in combination can be realized with system.Figure 19 A and 19B include showing Exemplary discharge voltage/current curve is to illustrate the figure that SOC lookup verifies 1905 and TTS tests 1910.Figure 20 A and 20B is to show exemplary CC appraisal procedures to illustrate electricity calculating method 2005 and the figure of the proximity of coulomb counting 2010. Figure 21 A and 21B are the schematic diagram for showing TTS appraisal procedures, and it illustrates voltameter 2105 and actual SOC2110 obvious weight Folded and TTS errors 2115.
Some in above-mentioned example embodiment are described as the process or method illustrated with flow.Although flow chart will Operation is described as sequential process, but many operations parallel, can be performed concurrently or simultaneously.In addition, order of operation can be rearranged.This A little processes can terminate when completing its operation, but can also have the other step not included in figure.These process sides of may correspond to Method, function, program, subroutine, subprogram etc..
The above method (some of them are illustrated by flow) can be retouched by hardware, software, firmware, middleware, microcode, hardware Predicate is sayed or any combination of them is realized.When being realized with software, firmware, middleware or microcode, necessary task is performed Program code or code segment are storable in machine or computer-readable medium such as storage medium.It can be handled by one or more Device performs necessary task.
Concrete structure and function detail disclosed herein are only representational, to describe exemplary embodiment.However, Exemplary embodiment is embodied as many alternative forms and should not be construed as limited to embodiments set forth herein.
It should be appreciated that although term first, second etc. can be used to describe various elements herein, these elements are not Should be limited by these terms.These terms are only used for making an element distinguish with another element.For example, the first element can To be referred to as the second element, and similarly, the second element is properly termed as the first element, without departing from the model of exemplary embodiment Enclose.As used herein, term "and/or" includes any and all combination of one or more of the continuous item listed.
It should be appreciated that when element be referred to as " connecting " or " coupling " to another element when, it can be directly connected to or be coupled to Intermediary element may be present in another element.On the contrary, when element is referred to as " being directly connected to " or " direct-coupling " to another element When, in the absence of intermediary element.Other words for describing the relation between element should explain in a similar manner (e.g., " between ... Between " and " directly between ... between ", " adjacent " and " direct neighbor " etc.).
It should also be noted that in some replacement specific implementations, it is indicated that function/action can not press it is pointed in figure Order occurs.For example, two figures shown in a continuous manner actually can concurrently be performed or can performed sometimes with inverted order, specifically take Certainly in involved function/action.
The specific implementation of various techniques described herein can be in Fundamental Digital Circuit or in computer hardware, firmware, soft Realized in part or in combinations thereof.Specific implementation can be achieved as computer program product, i.e., tangible embodiment is in information carrier In, such as machine-readable storage device, (computer-readable medium, non-transitorycomputer readable storage medium, tangible computer can Read storage medium) in or the signal propagated in computer program, for passing through data processing equipment (such as programmable processing Device, computer or multiple stage computers) processing, or the operation of the control data processing equipment.Computer program for example above-mentioned one Individual or multiple computer programs (including compiling or interpretative code) can be write in the form of any programming language, and can be with any shape Formula is disposed, including stand-alone program or the module suitable for computing environment, component, subroutine or other units.
Each several part in above-mentioned example embodiment and corresponding embodiment is according to in computer storage The software or algorithm and symbol of operation in data bit are represented to propose.Algorithm is (when the term is here using and when it with one As mode when using) be envisioned for result needed for drawing be in harmony sequence of steps certainly.These steps are that requirement carries out thing to physical quantity The step of reason manipulates.Generally, although it is not necessary, this tittle uses can store, shift, combining, comparing and with its other party The form of optical signal, electric signal or magnetic signal that formula manipulates.Have proven to be more conveniently sometimes, particularly for general original Cause, these signals are referred to as bit, value, element, symbol, character, item, numeral etc..
In exemplary embodiment above, what is referred to can be achieved the operation for program module or function course (e.g., with flow The form of figure) action and symbol represent to include execution specific tasks or realize the routine of specific abstract data type, be program, right As, component, data structure etc., and the existing hardware at existing structure element can be used to describe and/or realize.Some are existing There is hardware to may include one or more CPU (CPU), digital signal processor (DSP), application specific integrated circuit, scene Programmable gate array (FPGA) computer etc..
However, it should be remembered that all these and similar terms are related to appropriate physical quantity, and be only to be applied to these The facilitate label of amount.Unless otherwise specified, otherwise such as from discussion it will be evident that term for example " handles " or " calculating " (computing) or " calculating " (calculating) or " it is determined that " or " display " etc. refer to computer system or similar electronics The action of computing device and process, the data represented with the physics in computer system register and memory, amount of electrons are grasped Make and be transformed into similarly with computer system memory or register or the storage of other this type of informations, transmission or display device Physical quantity represent other data.
It shall yet further be noted that the aspect realized by software of exemplary embodiment is generally encoded in some form of non-transitory journey Realize in sequence storage medium or over some type of transmission medium.Program recorded medium can be (e.g., floppy disk or the hard disk of magnetic Driver) or it is optical (e.g., compact disc read-only memory, or " CDROM "), and be read-only or random-access.Similarly, Transmission medium can be twisted-pair feeder, coaxial cable, optical fiber or some other suitable transmission mediums.Exemplary embodiment is not by any Limitation in terms of these of specified specific implementation.
Although some features of the specific implementation have been shown as described herein, those skilled in the art is existing will It is contemplated that many modifications, replacement, change and equivalents.It will thus be appreciated that appended claims are intended to fall into All such modifications and modification in the range of specific implementation.It should be appreciated that the embodiment is only with the side of citing Formula rather than present in a restricted way, and various changes can be carried out in terms of form and details.Device as described herein and/ Or any portion of method can be combined with any combinations, but except mutually exclusive combination.It is as described herein specific real Apply the function of may include the different specific implementations, part and/or the various combinations of feature and/or sub-portfolio.

Claims (32)

1. a kind of method for the state of charge SOC for calculating battery, this method include:
The first estimation SOC of the battery is calculated in the very first time;
Voltage drop is calculated based on the described first estimation SOC;
The magnitude of voltage for the measurement voltage for representing the battery both ends is received in the second time;
Filter gain is calculated in second time;And
SOC, the magnitude of voltage and the filter gain are estimated to calculate the electricity based on described first in second time The second estimation SOC in pond,
Wherein, this method further comprises:
At least one voltage drop model parameter is calculated based on the equivalent model for representing the battery, the equivalent model represents institute The operation of battery is stated, at least one voltage drop model parameter includes indicating the instantaneous of the error in the second estimation SOC Hysteresis;And
By making at least one voltage drop model parameter change reduce the second estimation SOC to remove instantaneous lagging In error, according to described second estimation SOC calculate SOC.
2. the method according to claim 11, in addition to:
By the described second estimation SOC storages in a buffer;And
Before the second estimation SOC of the battery is calculated, the first estimation SOC is read from the buffer.
3. the method according to claim 11, in addition to:
By the described first estimation SOC storages in a buffer, before the second estimation SOC of the battery is calculated, from institute State buffer and read the first estimation SOC as one of following:
At least two SOC vector,
At least two SOC array,
At least two SOC average and
At least two SOC average.
4. the method according to claim 11, in addition to:Estimation SOC variances are read from buffer, wherein the wave filter increases Benefit is based on the estimation SOC variances.
5. according to the method for claim 1, wherein, calculating filter gain includes:Use weighting recurrence least square RLS At least one in algorithm and total least square TLS algorithms calculates at least one capability value.
6. according to the method for claim 1, wherein, calculating filter gain includes:Using based on SOC tracking errors association side The weighting RLS algorithm of difference and current measurement errors standard deviation calculates capability value.
7. according to the method for claim 1, wherein, calculating filter gain includes:Use passing based on covariance matrix Return the TLS algorithms of renewal to calculate capability value.
8. according to the method for claim 1, wherein, calculating filter gain includes:Searched using open-circuit voltage OCV to count Calculate capability value.
9. the method according to claim 11, wherein
The second estimation SOC is multiplied by the magnitude of voltage plus the described first estimation SOC equal to the filter gain.
10. the method according to claim 11, in addition to:
Reaching in voltage time TTV error metrics at least for error metrics, OCV-SOC error metrics and prediction is counted based on coulomb One, benchmark test is carried out to the described second estimation SOC.
11. a kind of system for being configured as calculating the state of charge SOC of battery, the system include:
Battery;And
Battery meter module, it is configured with the estimation SOC that Reduced Order Filter calculates the battery, the depression of order filtering Device is single state filter, and single state filter is configured as based on the SOC estimation calculated before come recursive calculation The estimation SOC,
Wherein, the battery meter module includes:
Voltage drop prediction module, it is configured as calculating voltage drop based on the estimation SOC;
Model estimation module, it is configured as calculating at least one voltage drop model based on the equivalent model for representing the battery Parameter, the equivalent model represent the operation of the battery, and at least one voltage drop model parameter includes estimating described in instruction Count the instantaneous lagging of the error in SOC;And
Tracking module, its be configured as by make at least one voltage drop model parameter change with remove instantaneous lagging so as to The error in the estimation SOC is reduced, SOC is calculated according to the estimation SOC.
12. system according to claim 11, wherein, the battery meter includes capacity module, the capacity module The total least square TLS algorithms for being configured with the renewal of the recurrence based on covariance matrix calculate the capacity of the battery Value.
13. system according to claim 11, wherein, the battery meter includes capacity module, the capacity module Open-circuit voltage OCV is configured with to search to calculate the capability value of the battery.
14. system according to claim 11, wherein
The battery meter includes buffer, and the buffer is configured to store at few estimation SOC value, and
The SOC estimation calculated before described is read from the buffer.
15. system according to claim 11, in addition to data storage, the data storage is configured as storing OCV parameters, voltage measurement error standard deviation and current measurement errors standard deviation, wherein it is related to the Reduced Order Filter at least One voltage drop model parameter is by the OCV parameters, the voltage measurement error standard deviation and the current measurement errors standard One of difference defines.
16. system according to claim 11, in addition to:Display, the display are configured as showing the estimation SOC。
17. a kind of method for the state of charge SOC for calculating battery, this method include:
Storage represents the storehouse of the equivalent-circuit model of battery in memory;
The operational mode of battery is determined based on the load related to the battery;
One of described equivalent-circuit model is selected based on identified operational mode;And
The SOC of the battery is calculated using selected equivalent-circuit model,
Wherein, this method further comprises:
At least one voltage drop model parameter is calculated based on the equivalent-circuit model selected in the storehouse from equivalent-circuit model, The equivalent-circuit model represents the operation of the battery, and at least one voltage drop model parameter includes instruction estimation SOC In error instantaneous lagging;And
By making at least one voltage drop model parameter change be reduced to remove instantaneous lagging in calculated SOC Error, SOC is calculated according to the estimation SOC.
18. the method according to claim 11, wherein
High voltage load of the operational mode based on consumption variable current,
The operational mode includes corresponding equivalent-circuit model, and the equivalent-circuit model includes first resistor-electric current RC electricity Road and the 2nd RC circuits.
19. the method according to claim 11, wherein
The operational mode is loaded based on variable voltage,
The operational mode includes corresponding equivalent-circuit model, and the equivalent-circuit model includes resistance-electric current RC circuits.
20. the method according to claim 11, wherein
The operational mode is loaded based on variable voltage,
The operational mode includes corresponding equivalent-circuit model, and the equivalent-circuit model includes resistor and biasing element.
21. the method according to claim 11, wherein
The operational mode is loaded based on low-voltage,
The operational mode includes corresponding equivalent-circuit model, and the equivalent-circuit model includes resistor.
22. the method according to claim 11, wherein
It is at least one including lag element in the equivalent-circuit model, and
The lag element is modeled as open-circuit voltage OCV calculation errors.
23. the method according to claim 11, in addition to:
During battery eliminator discharges, the terminal voltage of the battery eliminator is measured;
Linear equation is determined based on measured terminal voltage;And
At least one parameter is calculated using weighted least square algorithm based on the linear equation, to described in the battery SOC calculating is based at least one parameter.
24. the method according to claim 11, in addition to:
Reaching in voltage time TTV error metrics at least for error metrics, OCV-SOC error metrics and prediction is counted based on coulomb One, benchmark test is carried out to the SOC.
25. a kind of system for being configured as calculating the state of charge SOC of battery, the system include:
Data storage, it is configured as the storehouse that storage represents the equivalent-circuit model of battery;
Model selection module, it is configured as selecting equivalent-circuit model based on the operational mode of the battery;And
Filter module, it is configured as the estimation SOC that the battery is calculated based on selected equivalent-circuit model;
Wherein, the system further comprises:
Model estimation module, it is configured as selecting from the storehouse of equivalent-circuit model based on the Model selection module equivalent Circuit model calculates at least one voltage drop model parameter, and the equivalent-circuit model represents the operation of the battery, described At least one voltage drop model parameter includes indicating the instantaneous lagging of the error in the estimation SOC;And
Tracking module, its be configured as by make at least one voltage drop model parameter change with remove instantaneous lagging so as to The error of the estimation SOC is reduced, SOC is calculated according to the estimation SOC.
26. system according to claim 25, in addition to:
Operational mode module, it is configured as in the voltage related based on the electric current related to the battery and with the battery It is at least one to determine the operational mode of the battery.
27. system according to claim 25, wherein
High voltage load of the operational mode based on consumption variable current,
The operational mode includes corresponding equivalent-circuit model, and the equivalent-circuit model includes first resistor-electric current RC electricity Road and the 2nd RC circuits.
28. system according to claim 25, wherein
The operational mode is loaded based on variable voltage,
The operational mode includes corresponding equivalent-circuit model, and the equivalent-circuit model includes resistance-electric current RC circuits.
29. system according to claim 25, wherein
The operational mode is loaded based on variable voltage,
The operational mode includes corresponding equivalent-circuit model, and the equivalent-circuit model includes resistor and biasing element.
30. system according to claim 25, wherein
The operational mode is loaded based on low-voltage,
The operational mode includes corresponding equivalent-circuit model, and the equivalent-circuit model includes resistor.
31. system according to claim 25, wherein
It is at least one including lag element in the equivalent-circuit model, and
The lag element is modeled as open-circuit voltage OCV calculation errors.
32. system according to claim 25, in addition to display, the display is configured as showing the estimation SOC.
CN201410062166.6A 2013-02-24 2014-02-24 Battery state of charge tracking, equivalent circuit selection and reference test method and system Active CN104007390B (en)

Priority Applications (6)

Application Number Priority Date Filing Date Title
US201361768472P true 2013-02-24 2013-02-24
US61/768,472 2013-02-24
US14/185,790 US20140244193A1 (en) 2013-02-24 2014-02-20 Battery state of charge tracking, equivalent circuit selection and benchmarking
US14/185,835 US10664562B2 (en) 2013-02-24 2014-02-20 Battery state of charge tracking, equivalent circuit selection and benchmarking
US14/185,790 2014-02-20
US14/185,835 2014-02-20

Publications (2)

Publication Number Publication Date
CN104007390A CN104007390A (en) 2014-08-27
CN104007390B true CN104007390B (en) 2018-03-20

Family

ID=51368136

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201410062166.6A Active CN104007390B (en) 2013-02-24 2014-02-24 Battery state of charge tracking, equivalent circuit selection and reference test method and system

Country Status (1)

Country Link
CN (1) CN104007390B (en)

Families Citing this family (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140244193A1 (en) 2013-02-24 2014-08-28 Fairchild Semiconductor Corporation Battery state of charge tracking, equivalent circuit selection and benchmarking
US10371755B2 (en) 2014-10-21 2019-08-06 Fairchild Semiconductor Corporation Reported state-of-charge scaling
CN105807226B (en) * 2014-12-31 2018-07-10 北京航天测控技术有限公司 Lithium ion battery SOC Forecasting Methodologies and device based on equivalent-circuit model
JP2016176780A (en) * 2015-03-19 2016-10-06 エスアイアイ・セミコンダクタ株式会社 Battery residual amount prediction device and battery pack
CN105629173B (en) * 2015-12-22 2019-04-12 东软集团股份有限公司 Estimate the method and apparatus of the state-of-charge SOC of battery
FR3050278B1 (en) * 2016-04-15 2018-03-30 Saft METHOD FOR DETERMINING THE VALUE OF PARAMETERS RELATING TO THE STATUS OF AN BATTERY BATTERY BATTERY BATTERY AND ELECTRONIC BATTERY MANAGEMENT SYSTEM
CN106908730B (en) * 2017-01-19 2019-08-23 浙江零跑科技有限公司 A method of estimation lithium battery OCV_SOC calibration curve SOC error
US10670662B2 (en) * 2017-03-02 2020-06-02 Mediatek Inc. Method and apparatus for calibrating coulomb counting based state-of-charge estimation
CN107064816A (en) * 2017-04-13 2017-08-18 绵阳世睿科技有限公司 It is a kind of to strengthen the method that battery status estimates robustness
CN107425826B (en) * 2017-07-19 2020-12-15 广东美芝制冷设备有限公司 Adaptive filtering method, adaptive filter, and storage medium
CN107607878A (en) * 2017-08-30 2018-01-19 苏州精控能源科技有限公司 Kalman's iterative algorithm, device and the battery management system of battery SOC
CN108490361B (en) * 2018-03-22 2020-07-24 深圳库博能源科技有限公司 Cloud feedback-based SOC (state of charge) calculation method
CN108957337B (en) * 2018-06-20 2020-12-01 东软睿驰汽车技术(上海)有限公司 Method and device for determining state of health of battery, storage medium and electronic equipment
CN108828461B (en) * 2018-09-25 2020-11-17 安徽江淮汽车集团股份有限公司 Power battery SOH value estimation method and system
CN109839599A (en) * 2018-11-29 2019-06-04 西安科技大学 Lithium ion battery SOC estimation method based on second order EKF algorithm
CN112068003B (en) * 2020-11-16 2021-02-19 中南大学 Method and device for predicting service life of cadmium-nickel storage battery based on linear wiener process

Family Cites Families (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE10107583A1 (en) * 2001-02-17 2002-08-29 Vb Autobatterie Gmbh Method for determining the performance of a storage battery
CN1601295A (en) * 2004-10-25 2005-03-30 清华大学 Estimation for accumulator loading state of electric vehicle and carrying out method thereof
US8103485B2 (en) * 2004-11-11 2012-01-24 Lg Chem, Ltd. State and parameter estimation for an electrochemical cell
JP4967362B2 (en) * 2006-02-09 2012-07-04 トヨタ自動車株式会社 Secondary battery remaining capacity estimation device
KR100804698B1 (en) * 2006-06-26 2008-02-18 삼성에스디아이 주식회사 The method of assuming the state of charge of the battery, battery management system using the method and the driving method of the battery management system using the method
KR101189150B1 (en) * 2008-01-11 2012-10-10 에스케이이노베이션 주식회사 The method for measuring SOC of a battery in Battery Management System and the apparatus thereof
US9726732B2 (en) * 2010-06-22 2017-08-08 GM Global Technology Operations LLC Adaptive battery parameter extraction and SOC estimation for lithium-ion battery
CN102088118B (en) * 2010-12-28 2013-09-18 深圳市航盛电子股份有限公司 Battery management system, electric vehicle and state-of-charge estimation method
CN102645637B (en) * 2012-04-25 2015-02-25 中航锂电(洛阳)有限公司 Method for estimating SOC (state of charge) of equalized batteries
US20140244193A1 (en) * 2013-02-24 2014-08-28 Fairchild Semiconductor Corporation Battery state of charge tracking, equivalent circuit selection and benchmarking

Also Published As

Publication number Publication date
CN104007390A (en) 2014-08-27

Similar Documents

Publication Publication Date Title
Lai et al. A comparative study of different equivalent circuit models for estimating state-of-charge of lithium-ion batteries
Richardson et al. Gaussian process regression for forecasting battery state of health
Hannan et al. Neural network approach for estimating state of charge of lithium-ion battery using backtracking search algorithm
Hu et al. Co-estimation of state of charge and state of health for lithium-ion batteries based on fractional-order calculus
Nejad et al. A systematic review of lumped-parameter equivalent circuit models for real-time estimation of lithium-ion battery states
He et al. State of charge estimation for Li-ion batteries using neural network modeling and unscented Kalman filter-based error cancellation
Chaoui et al. Lyapunov-based adaptive state of charge and state of health estimation for lithium-ion batteries
Tang et al. A novel framework for Lithium-ion battery modeling considering uncertainties of temperature and aging
Wei et al. Enhanced online model identification and state of charge estimation for lithium-ion battery with a FBCRLS based observer
He et al. A new model for State-of-Charge (SOC) estimation for high-power Li-ion batteries
Zou et al. Electrothermal dynamics-conscious lithium-ion battery cell-level charging management via state-monitored predictive control
Wang et al. An adaptive remaining energy prediction approach for lithium-ion batteries in electric vehicles
CN103907030B (en) The estimation unit of battery charge rate and the method for estimation of battery charge rate
Zhong et al. An SOC estimation approach based on adaptive sliding mode observer and fractional order equivalent circuit model for lithium-ion batteries
Meng et al. Lithium polymer battery state-of-charge estimation based on adaptive unscented Kalman filter and support vector machine
Schwunk et al. Particle filter for state of charge and state of health estimation for lithium–iron phosphate batteries
JP2018513378A (en) Method and apparatus for determining the health and state of charge of a lithium sulfur battery
Lee et al. Soft computing for battery state-of-charge (BSOC) estimation in battery string systems
Chang et al. A new hybrid method for the prediction of the remaining useful life of a lithium-ion battery
Chan et al. The available capacity computation model based on artificial neural network for lead–acid batteries in electric vehicles
Pérez et al. Enhanced closed loop State of Charge estimator for lithium-ion batteries based on Extended Kalman Filter
Yang et al. Power system state estimation using PMUs with imperfect synchronization
Wang et al. State-of-charge estimation for lithium-ion batteries based on a nonlinear fractional model
Xia et al. Accurate Lithium-ion battery parameter estimation with continuous-time system identification methods
Lin et al. An ordinal optimization theory-based algorithm for solving the optimal power flow problem with discrete control variables

Legal Events

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