CN103413033A - Method for predicting storage battery faults - Google Patents

Method for predicting storage battery faults Download PDF

Info

Publication number
CN103413033A
CN103413033A CN2013103228674A CN201310322867A CN103413033A CN 103413033 A CN103413033 A CN 103413033A CN 2013103228674 A CN2013103228674 A CN 2013103228674A CN 201310322867 A CN201310322867 A CN 201310322867A CN 103413033 A CN103413033 A CN 103413033A
Authority
CN
China
Prior art keywords
battery
fault
symptom
membership
degree
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.)
Pending
Application number
CN2013103228674A
Other languages
Chinese (zh)
Inventor
张彦琴
卢明哲
刘汉雨
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.)
Beijing University of Technology
Original Assignee
Beijing University of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing University of Technology filed Critical Beijing University of Technology
Priority to CN2013103228674A priority Critical patent/CN103413033A/en
Publication of CN103413033A publication Critical patent/CN103413033A/en
Pending legal-status Critical Current

Links

Images

Landscapes

  • Secondary Cells (AREA)

Abstract

The invention discloses a method for predicting storage battery faults, and belongs to the field of storage battery fault prediction and diagnosis. The method is characterized in that whether a battery is broken down is predicted according to related data during the battery use process; according to the knowledge-based method for predicting the health status of the storage battery, namely, a fuzzy prediction method, degrees of membership of various faults are solved according to degrees of membership of certain symptoms and used for representing tendency of various faults, and scientific basis is provided for predicting the health status of the storage battery; and the method is proposed in order to effectively predict the faults of the battery and a battery pack, experimental data is obtained with a nickel-cadmium battery serving as an experimental subject, the DOH (degree of health) of the battery or the battery pack in the whole operation process can be obtained through analysis and processing for real-time data of the nickel-cadmium battery, and simultaneously maintenance information of the battery is given. The method is simple and effective, the disassembly and the discharge testing of the battery are avoided, and the accurate battery health level is obtained.

Description

A kind of method of predicting accumulator failure
One, technical field
The present invention relates to failure prediction and the diagnostic field of accumulator.
Two, background technology
Along with the fast development of China's economy, to the requirement of the numerous areas such as the energy, traffic, environment also in continuous raising.Battery pack does not allow the electric power system of cutting off the power supply all, is all an indispensable power-supply system.And the application of battery system in all trades and professions is also more and more extensive.Whether the operation of accumulator is normal, will directly affect normal, the reliable and safe operation of various device in this field.
The battery failures diagnostic techniques is the new technology just grown up in recent years, and the whole world not yet accomplishes tangible results to the research of battery failure diagnostic system at present.In the applicating maintenance of battery, with the most use is artificial testing and diagnosing method, the discharge test namely battery removed by certain multiplying power carries out cubic content measurement, to determine the active volume of accumulator, becomes behindhand battery thereby judge which batteries capacity loss.The discharge capacity test that this diagnostic mode reality is only a kind of off-line, do not carry out comprehensive diagnostic to battery.The defect of its method be the operation element amount large, take time manyly, thereby be not easy to repetitive operation, and repeat cubic content measurement, be disadvantageous to battery, accumulator can not Emergency use.In addition, capacity test needs special test equipment for the high voltage accumulator, may be dangerous.So should avoid as much as possible, directly by electric discharge, carry out measuring capacity and battery is made to diagnosis.
The Forecasting Methodology that we propose is change and consider that other factors diagnose battery by the parameter between the different monomers battery of Real-Time Monitoring, more same electric battery, its thought is: each cell of same electric battery is under identical charging and discharging currents, the performance great majority of each cell are close, but also have the performance of some batteries to exist inconsistent.By considering each battery, in the period, in the change in voltage of this section in the time, can estimate the quality of battery performance to the skew of average voltage and each battery at this section.Be a kind of online real-time predicting method by this method, can effectively avoid many drawbacks of manual testing's diagnosis.
Three, summary of the invention
The objective of the invention is to propose a kind of online, real-time nickel-cadmium battery failure prediction system.Utilize this method can obtain the probability of various faults appearance and health status DOH and the health level of this accumulator (or electric battery), can provide thus the maintenance information of battery.
Whether a kind of method of predicting accumulator failure, is characterized in that the method is for the method for cadmium-nickel storage cell based on expertise, utilize the related data prediction battery in the battery use procedure to break down.Based on the accumulator health status Forecasting Methodology of knowledge, be also the fuzzy prediction method, according to the degree of membership of some symptom, obtain the degree of membership of various faults, the tendentiousness existed in order to characterize various faults, for the prediction that judges cell health state provides scientific basis.Its reasoning process as shown in Figure 1
1) at first utilize battery historical archives data and health data and calculate battery symptom degree of membership in conjunction with the various symptom degree of membership computing formula of battery.Sympotomatic set U={ symptom 1, symptom 2, symptom 3 ... symptom n}, wherein the degree of membership of each symptom forms fuzzy vector α=[μ 1, μ 2, μ 3... μ n] T.
The μ in α wherein nBy following formula, calculate: μ n=f sta(I) f cha(X);
F sta(I) be function of state,
1. static:
f sta ( I ) = 1 I = 0 ; 0 I ≠ 0 ;
2. charging:
f sta ( I ) = 1 I > 0 ; 0 I ≤ 0 ;
3. electric discharge:
f sta ( I ) = 1 I < 0 ; 0 I &GreaterEqual; 0 ;
Annotate: I is current value.
F cha(X) be the external characteristic data function,
f cha ( X ) = X / ( X a * C r ) X &le; ( X a * C r ) 1 X > ( X a * C r )
F cha(X) be the degree of membership value of certain state of nickel-cadmium battery performance data;
X is the measured value of certain variation of the external characteristic data of battery;
X aMean value for certain variation of the external characteristic data of all batteries;
C rFor constant, C r=1.5, the relative coefficient of expression external characteristic data.
So just obtained the vectorial α that each symptom degree of membership obtains=[μ 1, μ 2, μ 3... μ n] T
2) utilize each symptom degree of membership of battery and in conjunction with weighting Judgement Matrix R, calculate the various fault degrees of membership of battery.
The fault collection V={ fault 1 of battery, fault 2, fault 3 ... fault n}, n herein and the n in the first step are same n.
A) weighting Judgement Matrix R's determines
The weighting Judgement Matrix define two kinds of modes, they are all to take this domain expert's experience as basis.A kind of is when the expert has larger assurance to definite fuzzy matrix, can provide the flexible strategy of each symptom, then carries out normalization or regular, forms the weighting Judgement Matrix; Another kind is when expert's oneself assurance is not very big, can provide the importance ranking that each symptom produces certain fault in conjunction with other a plurality of experts, then this significance sequence qualitatively is mapped as to a kind of flexible strategy and distributes, as the weighting Judgement Matrix.Considering complicacy, the ambiguity of Battery Diagnostic, be namely that veteran expert also is difficult to provide definite weights, so we adopts the second way.
At first two set U and V are carried out to importance by from big to small order sequence, then renumber, now the U={ symptom 1, symptom 2, symptom 3, symptom n} and V={ fault 1, fault 2, fault 3 ... element in these two set of fault n} is arranged by importance, the set of definition when these two set have not been initial, order wherein changes.In addition, the vectorial α of each symptom degree of membership composition=[μ 1, μ 2, μ 3... μ n] TAlso to resequence according to U.
Then utilize the binomial coefficient weighted method to carry out determining of matrix R.
Known
Figure BDA0000358338185
N is consistent with above-mentioned n herein.Binomial coefficient obtains one group and be 1 set W thus, is W = { C n - 1 0 / 2 n - 1 , C n - 1 1 / 2 n - 1 , C n - 1 2 / 2 n - 1 , . . . , C n - 1 n - 1 / 2 n - 1 } , According to the character of binomial coefficient, this set is the symmetry set that an intermediate value is maximum and progressively reduce to both sides.
Order W={W 1, W 2..., W i..., W n.
Therefore, weighting Judgement Matrix R is:
R = r 11 r 12 . . . . . . r 1 n r 21 r 22 . . . . . . r 2 n . . . . . . r n 1 r n 2 . . . . . . r nm = ( r ij ) n &times; n
R on diagonal line in each element in R 11=r 22=r 33=...=r NnBe equal to the maximal value of set in W, the remaining position of every delegation is by from big to small order assignment successively.
B) calculating of fault degree of membership
Failure collection V={ fault 1, fault 2, fault 3 ... fault n}, wherein the set β of the degree of membership of each element composition=[f 1, f 2, f 3..., f n] T, and β=R α
3) utilize the fault degree of membership, provide the health status of battery after comprehensive.
The calculating of DOF and DOH:
Failure degree DOF, to fault degree of membership set β=[f 1, f 2, f 3..., f n] TIf there is f=[f 1, f 2, f 3F n] ∈ F and f i>=0.5,1≤i≤n, DOF=f 1⊕ f 2⊕ f 3⊕ ... ⊕ f nOtherwise DOF=f 1∨ f 2∨ f 3∨ ... ∨ f n.
Wherein operator ∨ is defined as and gets greatly, and operator ⊕ is defined as and gets union, namely meets:
x⊕y=x+y-x×y
Health status DOH, DOH=C 1* DOF+C 2* DOH '
C wherein 1, C 2For constant, C 1=0.7, C 2=0.3; DOF is the failure degree of this calculating; DOH' for last operation calculate DOH, if there is no logout last time, DOH'=0.
According to the value of the DOH calculated, according to form 1, provide the health status of battery.
Form 1 battery health rank and corresponding measure
Health level DOH Span Counter-measure
One-level <0.4 Health, without processing
Secondary 0.4-0.7 Strengthen safeguarding
Three grades >0.7 Change
Four, accompanying drawing explanation
Fig. 1: based on the cell health state prediction flow process of expertise
Fig. 2: based on the cell health state algorithm flow of fuzzy logic
Five, embodiment
For nickel-cadmium battery, its fault collection and sympotomatic set are arranged according to importance:
Fault V={ volume lowering, internal resistance increases, aging };
Symptom U={ is B group voltage/14-monomer voltage in the same time, Δ U/ Δ t, Δ U is very big }; Due to n=3;
Weighting Judgement Matrix like this R = 1 / 2 1 / 4 1 / 4 1 / 4 1 / 2 1 / 4 1 / 4 1 / 4 1 / 2
Its specific algorithm is as follows, and process flow diagram is illustrated in fig. 2 shown below.
Primary data comprises: the data such as the voltage of monomer and electric current in A+B assembled battery total voltage, A battery voltage, B battery group.
The battery initial data segment that 1) at first will collect.Battery data to annual March is analyzed, and by the data sectional of every day, we only consider the charge and discharge data in period here, and each sampled point is exactly a time period.
2) then calculate the function of state of battery charge and discharge each symptom in period in every day.Because obtained data here are in large electric current, put, fill period, that is to say during this period, its function of state value is constantly equal to 1.
3), after obtaining the function of state value of each symptom of battery, continue to calculate the external characteristic function of this each symptom of battery.The X that each symptom is got aShown in following form 2.Such as, for symptom 1, the measured value X=0.0475 of electric discharge time period in period, so now the external characteristic functional value f of symptom 1 cha(X)=X/ (X a* C r)=0.0475/ (0.05028 * 1.5)=0.6298
Form 2 each symptom X aValue
Figure BDA00003583381810
4) calculate the membership function of each symptom.By step 2 and 3, calculated function of state f sta(I) and external characteristic data function f chaSo the membership function F (t) of each symptom (X), n=f sta(I) f cha(X).
5) complete if the membership function of each symptom all calculates, so just can carry out the calculating of next time period.The namely calculating of point of next sampling time.
6) after the calculating of the symptom degree of membership of all time periods is complete, carry out average value processing, try to achieve this symptom degree of membership on the same day, then calculate the fault degree of membership on the same day.Such as, obtained symptom degree of membership set α=[0.3082,0.4370,0.3339] of certain day T, known R, the set of fault degree of membership β=R α=[0.3469,0.3791,0.3533] so T.
7) according to the fault degree of membership obtained, judge the health status of battery, provide simultaneously maintenance information.If β=[0.3469,0.3791,0.3533] T, DOF=0.3791 so, DOH=0.7 * 0.3791+0.3 * 0=0.2653.Due to now DOH<0.4, therefore deducibility goes out, battery now is healthy, can continue to use.

Claims (1)

1. method of predicting accumulator failure is characterized in that:
1) at first utilize battery historical archives data and health data and calculate battery symptom degree of membership in conjunction with the various symptom degree of membership computing formula of battery; Sympotomatic set U={ symptom 1, symptom 2, symptom 3 ... symptom n}, wherein the degree of membership of each symptom forms fuzzy vector α=[μ 1, μ 2, μ 3... μ n] T
The μ in α wherein nBy following formula, calculate: μ n=f sta(I) f cha(X);
F sta(I) be function of state,
1. static:
f sta ( I ) = 1 I = 0 ; 0 I &NotEqual; 0 ;
2. charging:
f sta ( I ) = 1 I > 0 ; 0 I &le; 0 ;
3. electric discharge:
f sta ( I ) = 1 I < 0 ; 0 I &GreaterEqual; 0 ;
Annotate: I is current value;
F cha(X) be the external characteristic data function,
f cha ( X ) = X / ( X a * C r ) X &le; ( X a * C r ) 1 X > ( X a * C r )
F cha(X) be the degree of membership value of certain state of nickel-cadmium battery performance data;
X is the measured value of certain variation of the external characteristic data of battery;
X aMean value for certain variation of the external characteristic data of all batteries;
C rFor constant, C r=1.5, the relative coefficient of expression external characteristic data;
So just obtained the vectorial α that each symptom degree of membership obtains=[μ 1, μ 2, μ 3... μ n] T
2) utilize each symptom degree of membership of battery and in conjunction with weighting Judgement Matrix R, calculate the various fault degrees of membership of battery;
The fault collection V={ fault 1 of battery, fault 2, fault 3 ... fault n}, n herein and the n in the first step are same n;
A) weighting Judgement Matrix R's determines
Definite a plurality of experts of weighting Judgement Matrix provide the importance ranking that each symptom produces certain fault, then this significance sequence qualitatively are mapped as to a kind of flexible strategy and distribute, as the weighting Judgement Matrix;
At first two set U and V are carried out to importance by from big to small order sequence, then renumber, now the U={ symptom 1, symptom 2, symptom 3, symptom n} and V={ fault 1, fault 2, fault 3 ... element in these two set of fault n} is arranged by importance, the set of definition when these two set have not been initial, order wherein changes; In addition, the vectorial α of each symptom degree of membership composition=[μ 1, μ 2, μ 3... μ n] TAlso to resequence according to U;
Then utilize the binomial coefficient weighted method to carry out determining of matrix R;
Known C n - 1 0 + C n - 1 1 + C n - 1 2 + . . . + C n - 1 n = 2 n - 1 , N is consistent with above-mentioned n herein; Binomial coefficient obtains one group and be 1 set W thus, is W = { C n - 1 0 / 2 n - 1 , C n - 1 1 / 2 n - 1 , C n - 1 2 / 2 n - 1 , . . . , C n - 1 n - 1 / 2 n - 1 } , According to the character of binomial coefficient, this set is the symmetry set that an intermediate value is maximum and progressively reduce to both sides;
Order W={W 1, W 2..., W i..., W n;
Therefore, weighting Judgement Matrix R is:
R = r 11 r 12 . . . . . . r 1 n r 21 r 22 . . . . . . r 2 n . . . . . . r n 1 r n 2 . . . . . . r nn ( r ij ) n &times; n
R on diagonal line in each element in R 11=r 22=r 33=...=r NnBe equal to the maximal value of set in W, the remaining position of every delegation is by from big to small order assignment successively;
B) calculating of fault degree of membership
Failure collection V={ fault 1, fault 2, fault 3 ... fault n}, wherein the set β of the degree of membership of each element composition=[f 1, f 2, f 3..., f n] T, and β=R α
3) utilize the fault degree of membership, provide the health status of battery after comprehensive;
The calculating of DOF and DOH:
Failure degree DOF, to fault degree of membership set β=[f 1, f 2, f 3..., f n] TIf there is f=[f 1, f 2, f 3..., f n] ∈ F and f i>=0.5,1≤i≤n, DOD=f 1⊕ f 2⊕ f 3⊕ ... ⊕ f nOtherwise DOF=f 1∨ f 2∨ f 3∨ ... ∨ f n
Wherein operator ∨ is defined as and gets greatly, and operator ⊕ is defined as and gets union, namely meets:
x⊕y＝x+y-x×y
Health status DOH, DOH=C 1* DOF+C 2* COH '
C wherein 1, C 2For constant, C 1=0.7, C 2=0.3; DOF is the failure degree of this calculating; DOH' for last operation calculate DOH, if there is no logout last time, DOH'=0;
According to the value of the DOH calculated, according to following mode, provide the health status of battery:
Health status DOH span counter-measure
One-level<0.4 health, without outer reason
Secondary 0.4-0.7 strengthens safeguarding
Three grades > 0.7 replacing.
CN2013103228674A 2013-07-29 2013-07-29 Method for predicting storage battery faults Pending CN103413033A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2013103228674A CN103413033A (en) 2013-07-29 2013-07-29 Method for predicting storage battery faults

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2013103228674A CN103413033A (en) 2013-07-29 2013-07-29 Method for predicting storage battery faults

Publications (1)

Publication Number Publication Date
CN103413033A true CN103413033A (en) 2013-11-27

Family

ID=49606044

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2013103228674A Pending CN103413033A (en) 2013-07-29 2013-07-29 Method for predicting storage battery faults

Country Status (1)

Country Link
CN (1) CN103413033A (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106227942A (en) * 2016-07-25 2016-12-14 温州大学 Electrokinetic cell system method for designing based on Brittleness Theory of Complex System
CN107144722A (en) * 2017-04-16 2017-09-08 知豆电动汽车有限公司 Battery pressure drop rate alarm method
CN107340475A (en) * 2016-04-29 2017-11-10 株式会社日立制作所 Battery fault detection method and battery fault detection device
CN110069810A (en) * 2019-03-11 2019-07-30 北京百度网讯科技有限公司 Battery failures prediction technique, device, equipment and readable storage medium storing program for executing
CN110116625A (en) * 2019-05-16 2019-08-13 辽宁工业大学 A kind of automobile storage battery fault monitoring method for electric-controlled vehicle
CN111967191A (en) * 2020-08-24 2020-11-20 哈尔滨理工大学 Fuzzy Bayesian network-based lithium ion power battery safety degree evaluation method and device
CN114895196A (en) * 2022-07-13 2022-08-12 深圳市威特利电源有限公司 New energy battery fault diagnosis method based on artificial intelligence

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050119534A1 (en) * 2003-10-23 2005-06-02 Pfizer, Inc. Method for predicting the onset or change of a medical condition
CN101174715A (en) * 2007-09-28 2008-05-07 深圳先进技术研究院 Power battery management system with control and protection function and method thereof

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050119534A1 (en) * 2003-10-23 2005-06-02 Pfizer, Inc. Method for predicting the onset or change of a medical condition
CN101174715A (en) * 2007-09-28 2008-05-07 深圳先进技术研究院 Power battery management system with control and protection function and method thereof

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
刘文杰: "基于模糊理论的电池故障诊断专家系统", 《吉林大学学报》, vol. 21, no. 6, 30 November 2005 (2005-11-30), pages 1 *
刘文杰: "电池组故障诊断专家系统的研究与实现", 《中国优秀博硕士学位论文全文数据库（硕士）信息科技辑》, no. 7, 15 July 2006 (2006-07-15) *
梅胜敏: "故障诊断专家系统中的模糊推理方法", 《南京航空航天大学学报》, vol. 27, no. 4, 30 August 1995 (1995-08-30) *

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107340475A (en) * 2016-04-29 2017-11-10 株式会社日立制作所 Battery fault detection method and battery fault detection device
CN106227942A (en) * 2016-07-25 2016-12-14 温州大学 Electrokinetic cell system method for designing based on Brittleness Theory of Complex System
CN106227942B (en) * 2016-07-25 2020-05-19 温州大学 Power battery system design method based on complex system brittleness theory
CN107144722A (en) * 2017-04-16 2017-09-08 知豆电动汽车有限公司 Battery pressure drop rate alarm method
CN110069810A (en) * 2019-03-11 2019-07-30 北京百度网讯科技有限公司 Battery failures prediction technique, device, equipment and readable storage medium storing program for executing
CN110069810B (en) * 2019-03-11 2023-04-07 北京百度网讯科技有限公司 Battery failure prediction method, device, equipment and readable storage medium
CN110116625A (en) * 2019-05-16 2019-08-13 辽宁工业大学 A kind of automobile storage battery fault monitoring method for electric-controlled vehicle
CN110116625B (en) * 2019-05-16 2020-07-28 辽宁工业大学 Automobile storage battery fault monitoring method for electric control vehicle
CN111967191A (en) * 2020-08-24 2020-11-20 哈尔滨理工大学 Fuzzy Bayesian network-based lithium ion power battery safety degree evaluation method and device
CN111967191B (en) * 2020-08-24 2024-03-19 哈尔滨理工大学 Lithium ion power battery safety evaluation method and device based on fuzzy Bayesian network
CN114895196A (en) * 2022-07-13 2022-08-12 深圳市威特利电源有限公司 New energy battery fault diagnosis method based on artificial intelligence

Similar Documents

Publication Publication Date Title
CN103413033A (en) Method for predicting storage battery faults
CN101067644B (en) Storage battery performance analytical expert diagnosing method
CN106033113A (en) Health state evaluation method for energy-storage battery pack
CN102520361B (en) State of health (SOH) value assessment method of battery pack
CN105277896B (en) Lithium battery method for predicting residual useful life based on ELM MUKF
CN108254696A (en) The health state evaluation method and system of battery
CN103412264B (en) The conforming evaluation method of cell in battery pack
CN106443461A (en) Battery energy storage system state assessment method
CN105652106A (en) Comprehensive performance testing method and apparatus of energy storage system
CN103033762B (en) Difference value differential based lithium iron phosphate power battery power loading capacity dynamic prediction method
JPH08233918A (en) Method and apparatus for estimation of residual unused capacity and residual unused time of battery during discharge
CN107367698A (en) The health status Forecasting Methodology of electric automobile lithium battery group
CN103941191A (en) Energy storing device integrated management method and energy storing device
CN104316879A (en) Method for forecasting service life of lead-acid storage battery
CN103886518A (en) Early warning method for voltage sag based on electric energy quality data mining at monitoring point
CN104680024A (en) Method for predicting remaining useful life of lithium ion battery based on GA (Genetic Algorithms) and ARMA (Auto Regressive and Moving Average) models
CN103176137B (en) Battery pack health condition evaluation method based on battery SOC (state of charge) inhomogeneity
CN114444370A (en) Method and device for predicting accumulated loss life of rechargeable battery by considering operation conditions, electronic equipment and readable storage medium
CN113687251B (en) Double-model-based lithium ion battery pack voltage abnormality fault diagnosis method
CN105354768A (en) Assessment method and system for electric transmission and transformation equipment states
CN106033114A (en) Health state evaluation method of energy storage battery
CN113112114A (en) Energy storage power station online evaluation method and device
CN115980612B (en) Satellite battery pack health state evaluation method, system and equipment
CN111509319A (en) PHM management system for energy storage power supply
CN112883634A (en) DC measurement system state prediction method and system based on multi-dimensional analysis

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20131127