CN102749585A - Multi-layer SVM (support vector machine) based storage battery on-line monitoring method - Google Patents

Multi-layer SVM (support vector machine) based storage battery on-line monitoring method Download PDF

Info

Publication number
CN102749585A
CN102749585A CN2011100996466A CN201110099646A CN102749585A CN 102749585 A CN102749585 A CN 102749585A CN 2011100996466 A CN2011100996466 A CN 2011100996466A CN 201110099646 A CN201110099646 A CN 201110099646A CN 102749585 A CN102749585 A CN 102749585A
Authority
CN
China
Prior art keywords
data
normal
svm2
svm1
battery
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.)
Granted
Application number
CN2011100996466A
Other languages
Chinese (zh)
Other versions
CN102749585B (en
Inventor
李昌
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.)
Individual
Original Assignee
Individual
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 Individual filed Critical Individual
Priority to CN201110099646.6A priority Critical patent/CN102749585B/en
Publication of CN102749585A publication Critical patent/CN102749585A/en
Application granted granted Critical
Publication of CN102749585B publication Critical patent/CN102749585B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Secondary Cells (AREA)

Abstract

The invention relates to a multi-layer SVM (support vector machine) based storage battery on-line monitoring method. Considering that the storage battery performance degradation process is closely related to the voltage and current variation condition and temperature variation condition, in order to reflect the intrinsic relationship, through setting up an SVM data category training set, and utilizing a multi-layer SVM storage battery performance degradation diagnosis flow by adopting an SVM algorithm, the storage battery is subjected to real-time intelligent fault predication, the monitoring of the health state of the storage battery has higher accuracy compared with the traditional method, and also early warning can be provided in case of the invalidity of the storage battery.

Description

A kind of battery on-line monitoring method based on multi-layer SVM
Technical field
Multi-layer SVM (SVM is based on the present invention relates to a kind of accumulator property deterioration diagnosis method, more particularly to one kind：Support Vector Machine) battery on-line monitoring method, belong to power failure diagnosing field.
Background technology
Safe and stable electric power system is the important foundation of the national economic development, is the necessary condition of modern industrial society's harmonious development.VRLA（Full-sealing maintaining-free lead acid storage battery）Group as transformer station, computer room, mobile base station, UPS electric power systems backup battery, in the power-supply system for being widely used in the industries such as communication, electric power, traffic, finance, monitoring in real time is carried out to it can effectively ensure that the reliability of electric power system.
VRLA is a complicated electrochemical system, and battery electrode material, process, the change of active material, running status can all influence life-span and the performance of battery, be mainly shown as：Battery dehydration, sulfation, grid corruption deformation, active material softening.These changes occur in inside battery, it is impossible to which directly observation is obtained.At present, the valve-regulated maximally effective diagnostic method of Full-sealing maintaining-free lead acid storage battery performance degradation is discharge test method, and discharge test method has two kinds of tupes：Off-line type with it is online.The former needs to separate tested battery from system, and real operation feasibility is little；The latter has potential danger to system, repeatedly enters the cycle life for being about to reduce battery limited.Valve-regulated Full-sealing maintaining-free lead acid storage battery performance monitoring object is battery health status（HOS）, and battery health status is mainly reflected in the residual capacity of battery（SOC）.
SVMs (Support Vector Machine, SVM) is a kind of Intelligence Classifier based on Structural risk minization principle.Relative to artificial neural network（Artificial Neural Network, ANN）Deng traditional intelligence Processing Algorithm, the algorithm can set up out two graders with excellent Generalization Capability in the case of small sample amount, while avoiding Local Minimum and dimension disaster problem.At present, the algorithm causes extensive concern, and Successful utilization in system fault diagnosis field.Although also there are many scholars to propose the research method for implementing accumulator property deterioration monitoring using SVMs at present, following problem is still suffered from practice：1）Lack rational data prediction：Reflect voltage, electric current and the temperature parameter measurement data of remaining battery capacity, effective description could be carried out to battery health status by having to pass through rational pretreatment, part research lacks the pretreatment to initial data, also there is part preprocess method unreasonable, it is difficult to applied in actual monitoring system.2）Diagnosis decision process is required further improvement：SVMs is essentially a kind of two grader, because accumulator property deterioration is the process of a gradual change, therefore can not be the diagnosis that accumulator property deterioration can be achieved by a single SVM fault grader, come with some shortcomings part in traditional decision process, such as divide and there is blind area, these blind area data interfering well cluster processes, along with training sample amount does not possess symmetry, have impact on the reliability of diagnosis.3）The real-time diagnosis system based on SVMs is not set up：Majority research gives the thinking of fault diagnosis, seldom builds system architecture that is specific, being capable of real-time diagnosis.Therefore further raising is waited to the accuracy that maintenance-free lead accumulator performance monitoring, analysis judge, to meet practical application.
The content of the invention
What the mesh of the present invention overcame is the not high enough defect of judgment accuracy existed for existing accumulator property degradation failure diagnostic techniques, and proposes a kind of accuracy rate of diagnosis mutually higher battery on-line monitoring method based on multi-layer SVM.
The battery on-line monitoring method based on multilayer SVM algorithm of the present invention comprises the following steps：
A kind of battery on-line monitoring method based on multilayer SVM, it is characterised in that the monitoring method comprises the following steps：
1）Set up SVM data classification based training collection
Data acquisition is carried out with the single battery different to several known, health status of continuous sampling mode at times by data collecting system, the parameter of collection is single battery discharge voltage variable u, current variable i and battery surface temperature temp during whole discharge cycle, piezoelectric voltage variable u, discharge current variable i and the battery surface temperature temp of same single battery are a sample set, and according to " normal " or the mark of "abnormal" on the single battery health status table, the sample set of all single batteries constitutes SVM data classification based training collection；
    2）The acquisition of SVM data classification input variable
The real-time discharge voltage variable u of single battery to be diagnosed, discharge current variable i and real-time battery surface temperature temp are gathered in continuous sampling mode at times in real time by data collecting system, above-mentioned variable simultaneously delivers to the SVM data classification input variable of storage composition in SQL database system, is designated as respectively
                                                 、
Figure 600693DEST_PATH_IMAGE002
、
Figure 2011100996466100002DEST_PATH_IMAGE003
；
3）Data prediction：
Single battery discharge cycle data to be diagnosed in SQL database are read, discharge time interval is determined
Figure 752057DEST_PATH_IMAGE004
, the voltage in the discharge cycle period is obtained in continuous sampling mode at times
Figure 2011100996466100002DEST_PATH_IMAGE005
, electric current
Figure 2011100996466100002DEST_PATH_IMAGE007
And temperature
Figure 484914DEST_PATH_IMAGE008
Numerical value,（
Figure 2011100996466100002DEST_PATH_IMAGE009
）If, data sampling at intervals of, then continuous sampling size is data at times
Figure 2011100996466100002DEST_PATH_IMAGE011
, it is designated as：
Figure 821404DEST_PATH_IMAGE001
、
Figure 702641DEST_PATH_IMAGE002
、
Figure 481765DEST_PATH_IMAGE003
The electric discharge real-time power on each sampled point is calculated accordingly
Figure 651716DEST_PATH_IMAGE012
；
Calculate voltage
Figure 984608DEST_PATH_IMAGE005
, electric current, temperature
Figure 746076DEST_PATH_IMAGE008
And electric discharge real-time power
Figure 2011100996466100002DEST_PATH_IMAGE013
Average value：
,
Figure 2011100996466100002DEST_PATH_IMAGE015
,And
Figure 2011100996466100002DEST_PATH_IMAGE017
The deviation of each measurement data is obtained by following equation：
Figure 292006DEST_PATH_IMAGE018
,
Figure 2011100996466100002DEST_PATH_IMAGE019
,
Figure 803146DEST_PATH_IMAGE020
And
Figure 2011100996466100002DEST_PATH_IMAGE021
Thus pretreated data are obtained
Figure 541164DEST_PATH_IMAGE022
4）Data diagnosis
Using 3 SVM classifiers：SVM1 is used to judge battery tension normal condition and deterioration state, and SVM2 is used to judge single battery temperature normal condition and deterioration state, and SVM3 is used to judge battery comprehensive state normal condition and deterioration state；Following 6 paths of diagnostic process point： 
Path 1：
By step 2）The voltage gradient data sequence of the continuous multiple discharge cycles of gainedSVM1 is delivered to, when SVM1 diagnostic results are " normal ", then by the temperature gradients data of continuous multiple discharge cyclesFurther deliver to SVM2；When SVM2 diagnostic results are " normal ", then last diagnostic result is " normal "；
Path 2：
By step 2）The voltage gradient data sequence of the continuous multiple discharge cycles of gained
Figure 483603DEST_PATH_IMAGE001
SVM1 is delivered to, when SVM1 diagnostic results are "abnormal", then by the temperature gradients data of continuous multiple discharge cycles
Figure 849862DEST_PATH_IMAGE003
Further deliver to SVM2；When SVM2 diagnostic results are "abnormal", then last diagnostic result is "abnormal"；
Path 3：
By step 2）The voltage gradient data sequence of the continuous multiple discharge cycles of gained
Figure 328248DEST_PATH_IMAGE001
SVM1 is delivered to, when SVM1 diagnostic results are " normal ", then by the temperature gradients data of continuous multiple discharge cycles
Figure 830774DEST_PATH_IMAGE003
Further deliver to SVM2；When SVM2 diagnostic results are "abnormal", then by the preprocessed data of gained
Figure 621400DEST_PATH_IMAGE022
Further deliver to SVM3；When SVM3 diagnostic results are "abnormal", then last diagnostic result is "abnormal"；
Path 4：
By step 2）The voltage gradient data sequence of the continuous multiple discharge cycles of gainedSVM1 is delivered to, when SVM1 diagnostic results are " normal ", then by the temperature gradients data of continuous multiple discharge cycles
Figure 491452DEST_PATH_IMAGE003
Further deliver to SVM2；When SVM2 diagnostic results are "abnormal", then by the preprocessed data of gained
Figure 176381DEST_PATH_IMAGE022
Further deliver to SVM3；When SVM3 diagnostic results are " normal ", then last diagnostic result is " normal "；
Path 5：
By step 2）The voltage gradient data sequence of the continuous multiple discharge cycles of gained
Figure 620131DEST_PATH_IMAGE001
SVM1 is delivered to, when SVM1 diagnostic results are "abnormal", then by the temperature gradients data of continuous multiple discharge cycles
Figure 960983DEST_PATH_IMAGE003
Further deliver to SVM2；When SVM2 diagnostic results are " normal ", then by the preprocessed data of gained
Figure 45087DEST_PATH_IMAGE022
Further deliver to SVM3；When SVM3 diagnostic results are " normal ", then last diagnostic result is " normal "；
Path 6：
By step 2）The voltage gradient data sequence of the continuous multiple discharge cycles of gained
Figure 335254DEST_PATH_IMAGE001
SVM1 is delivered to, when SVM1 diagnostic results are "abnormal", then by the temperature gradients data of continuous multiple discharge cycles
Figure 136857DEST_PATH_IMAGE003
Further deliver to SVM2；When SVM2 diagnostic results are " normal ", then by the preprocessed data of gained
Figure 761742DEST_PATH_IMAGE022
Further deliver to SVM3；When SVM3 diagnostic results are "abnormal", then last diagnostic result is "abnormal".
The present invention has the advantages that：
The present invention is due to consideration that accumulator property deterioration process and voltage, curent change situation and temperature variations are closely related, in order to more embody the contact of inherence, the present invention is to voltage, curent change situation and temperature variations data have carried out timing continuous acquisition, and pre-processed, using SVM algorithm, diagnostic process is deteriorated using multilayer SVM accumulator properties, real time intelligent failure judgement is carried out to battery, therefore the sorting of Full-sealing maintaining-free lead acid storage battery has higher accuracy compared with conventional method, and early warning can also be provided to the failure of Full-sealing maintaining-free lead acid storage battery.
Brief description of the drawings：
Fig. 1 is the on-line monitoring method flowage structure figure of the present invention.
Fig. 2 is that SVMs decision data separates principle of classification figure.
Fig. 3 is the failure modes diagnosis schematic diagram based on decision-directed figure method.
Fig. 4 is carried out using multilayer SVM classifier in failure modes diagnosis path schematic diagram, figure in the present invention, 1- normal conditions；0- is deteriorated（It is abnormal）State.
Embodiment
In the present embodiment, categorised decision one by one is carried out to warranty 480 section south all power supply model " GFM-1000E " type battery data records of more than 4 years.Monitor decision process structure chart as shown in Figure 1, first, SVM data classification based training collection is set up, method is to carry out off-line type verification property discharge test to being selected as each batteries of training set sample, according to the battery discharge capacity of water surveyed by residual capacity after floating charge（Percentage）Size is marked, and mark is carried out as follows：
Figure 2011100996466100002DEST_PATH_IMAGE023
   
Figure 510256DEST_PATH_IMAGE024
Depending on actual measurement, empirical value is 3 ~ 5；
In this specific embodiment, healthy battery in sample set（Ident value is 1）There are 100 sections, inferior health battery（Ident value is 0.5）100 sections, deteriorate battery（（Ident value is 0）50 sections.Using 2min as sampling time interval, discharge voltage u, discharge current i and cell body real-time temperature temp of the record single battery in normal course of operation in each discharge cycle T=[t1, t2]；Record period number is at least more than 500 times.Piezoelectric voltage variable u, discharge current variable i and the battery surface temperature temp of same single battery are a sample set, it is that each record adds a property value on the battery identification in record data table, the size of property value etc. is identified by the ident value in foregoing.The sample set of all single batteries constitutes SVM data classification based training collection.
Then, 480 section south all power supply model " GFM-1000E " type battery datas are acquired as SVM data classification input variable.
On-line data acquisition is carried out to 480 section south all power supply model " GFM-1000E " type batteries in continuous sampling mode at times by data collecting system, gather the real-time discharge voltage variable u of single battery to be diagnosed, discharge current variable i and real-time battery surface temperature temp in real time, above-mentioned variable simultaneously delivers to the SVM data classification input variable of storage composition in SQL database system, is designated as respectively
 
Figure 844809DEST_PATH_IMAGE001
、
Figure 145210DEST_PATH_IMAGE002
、
Figure 8123DEST_PATH_IMAGE003
；
Continuous sampling mode refers to that by 4 hours before each discharge process of battery be standard time section at times, and continuous several times sampling is carried out to each variable of battery simultaneously in each period.It is specially in this specific embodiment：Sampling time is 1 minute.
Because data collecting system is highly developed technology, here is omitted.
Next by data handling system（Such as computer）Carry out data prediction：
Single battery discharge cycle data to be diagnosed in SQL database are read, discharge time interval is determined
Figure 419382DEST_PATH_IMAGE004
, the voltage in the discharge cycle period is obtained in continuous sampling mode at times, electric current
Figure 93126DEST_PATH_IMAGE007
And temperatureNumerical value,（）If, data sampling at intervals of
Figure 826486DEST_PATH_IMAGE010
, then continuous sampling size is data at times
Figure 406372DEST_PATH_IMAGE011
, it is designated as：
、
Figure 999869DEST_PATH_IMAGE002
、
Figure 160592DEST_PATH_IMAGE003
The electric discharge real-time power on each sampled point is calculated accordingly
Figure 911379DEST_PATH_IMAGE012
；
Calculate voltage
Figure 157553DEST_PATH_IMAGE005
, electric current
Figure 643199DEST_PATH_IMAGE007
, temperature
Figure 986324DEST_PATH_IMAGE008
And electric discharge real-time power
Figure 845696DEST_PATH_IMAGE013
Average value：
Figure 641482DEST_PATH_IMAGE014
,
Figure 676959DEST_PATH_IMAGE015
,And
Figure 780230DEST_PATH_IMAGE017
The deviation of each measurement data is obtained by following equation：
Figure 814045DEST_PATH_IMAGE018
,
Figure 712600DEST_PATH_IMAGE019
,
Figure 968001DEST_PATH_IMAGE020
And
Figure 44541DEST_PATH_IMAGE021
Thus pretreated data are obtained
Figure 552270DEST_PATH_IMAGE022
Finally carry out data diagnosis：As shown in figure 4, using 3 SVM in the present invention（SVMs）Grader, SVM is two graders, i.e., each SVMs can only be judged specific two kinds of running statuses A, B, and test sample is divided into the higher class of possibility.Because battery, particularly Full-sealing maintaining-free lead acid storage battery performance degradation can only survey parameter according to external：Voltage, electric current and battery cell temperature judge that, in order to improve judgement accuracy, system establishes 3 layers of SVM Decision Classfication devices：Wherein SVM1 is used to judge battery tension normal condition and deterioration state, and SVM2 is used to judge single battery temperature normal condition and deterioration state, and SVM3 is used to judge battery comprehensive state normal condition and deterioration state.
SVM decision datas principle of classification is as shown in Fig. 2 SVM uses kernel function by the inseparable data projection of low dimensional to higher dimensional space, the data set of one linear separability of formation, and data are classified by building largest interval Optimal Separating Hyperplane.If assuming that training set is
Figure DEST_PATH_IMAGE025
,Wherein
Figure DEST_PATH_IMAGE027
For input data,
Figure 239790DEST_PATH_IMAGE028
Output data is represented, SVM passes through Nonlinear Mapping first：
Figure DEST_PATH_IMAGE029
Data projection to higher-dimension separable space, in this spatial configuration largest interval Optimal Separating Hyperplane
Figure 864062DEST_PATH_IMAGE030
, and then classification problem is converted into following quadratic programming problem：
Minimize：
S.t.：
Figure 793841DEST_PATH_IMAGE032
,
Figure DEST_PATH_IMAGE033
Figure 909565DEST_PATH_IMAGE034
 
Figure 936295DEST_PATH_IMAGE033
Wherein：
Figure DEST_PATH_IMAGE035
For hyperplane weight weight vector；
Figure 556238DEST_PATH_IMAGE036
For biasing；
Figure DEST_PATH_IMAGE037
It is slack variable；
Figure DEST_PATH_IMAGE039
For number of training；
Figure 301209DEST_PATH_IMAGE040
For penalty factor；
Figure 17361DEST_PATH_IMAGE042
Numbered for training sample；Similarly hereinafter；
Solving this optimization problem needs construction Lagrangian to solve dual problem：
Figure DEST_PATH_IMAGE043
 
Figure 104790DEST_PATH_IMAGE044
S.t.：
Figure DEST_PATH_IMAGE045
Wherein,,
Figure 785356DEST_PATH_IMAGE033
For the Lagrange multiplier of sample；
Figure DEST_PATH_IMAGE047
Accumulated for Kronecker；
Figure 495079DEST_PATH_IMAGE048
For kernel function；
When solving optimal solution
Figure DEST_PATH_IMAGE049
, hyperplane equation is substituted into, decision function is obtained
Figure 371768DEST_PATH_IMAGE050
, last classification results are by decision ruleProvide；When training SVM classifier, penalty factor value is [1.3,1200]；Selection of kernel function is RBF kernel functions, i.e.,
Figure 505815DEST_PATH_IMAGE052
,
Figure DEST_PATH_IMAGE053
Value is [0.3,2]；In C=100 in this method implementation, kernel function=1, SVM classifier is realized by the Matlab svmtrain functions carried.
Following 6 paths of diagnostic process point in the present invention： 
Path 1：
Data collecting system is gathered to the voltage gradient data sequence of continuous multiple discharge cycles obtained by single battery to be diagnosed
Figure 223946DEST_PATH_IMAGE001
SVM1 is delivered to, when SVM1 diagnostic results are " normal ", then by the temperature gradients data of continuous multiple discharge cycles
Figure 17459DEST_PATH_IMAGE003
Further deliver to SVM2；When SVM2 diagnostic results are " normal ", then last diagnostic result is " normal "；
Path 2：
Data collecting system is gathered to the voltage gradient data sequence of continuous multiple discharge cycles obtained by single battery to be diagnosed
Figure 948506DEST_PATH_IMAGE001
SVM1 is delivered to, when SVM1 diagnostic results are "abnormal", then by the temperature gradients data of continuous multiple discharge cycles
Figure 889786DEST_PATH_IMAGE003
Further deliver to SVM2；When SVM2 diagnostic results are "abnormal", then last diagnostic result is "abnormal"；
Path 3：
Data collecting system is gathered to the voltage gradient data sequence of continuous multiple discharge cycles obtained by single battery to be diagnosed
Figure 830060DEST_PATH_IMAGE001
SVM1 is delivered to, when SVM1 diagnostic results are " normal ", then by the temperature gradients data of continuous multiple discharge cycles
Figure 469290DEST_PATH_IMAGE003
Further deliver to SVM2；When SVM2 diagnostic results are "abnormal", then by the preprocessed data of gained
Figure 758189DEST_PATH_IMAGE022
Further deliver to SVM3；When SVM3 diagnostic results are "abnormal", then last diagnostic result is "abnormal"；
Path 4：
Data collecting system is gathered to the voltage gradient data sequence of continuous multiple discharge cycles obtained by single battery to be diagnosed
Figure 186765DEST_PATH_IMAGE001
SVM1 is delivered to, when SVM1 diagnostic results are " normal ", then by the temperature gradients data of continuous multiple discharge cycles
Figure 852101DEST_PATH_IMAGE003
Further deliver to SVM2；When SVM2 diagnostic results are "abnormal", then by the preprocessed data of gained
Figure 167676DEST_PATH_IMAGE022
Further deliver to SVM3；When SVM3 diagnostic results are " normal ", then last diagnostic result is " normal "；
Path 5：
Data collecting system is gathered to the voltage gradient data sequence of continuous multiple discharge cycles obtained by single battery to be diagnosed
Figure 958302DEST_PATH_IMAGE001
SVM1 is delivered to, when SVM1 diagnostic results are "abnormal", then by the temperature gradients data of continuous multiple discharge cycles
Figure 359328DEST_PATH_IMAGE003
Further deliver to SVM2；When SVM2 diagnostic results are " normal ", then by the preprocessed data of gained
Figure 828355DEST_PATH_IMAGE022
Further deliver to SVM3；When SVM3 diagnostic results are " normal ", then last diagnostic result is " normal "；
Path 6：
Data collecting system is gathered to the voltage gradient data sequence of continuous multiple discharge cycles obtained by single battery to be diagnosed
Figure 247704DEST_PATH_IMAGE001
SVM1 is delivered to, when SVM1 diagnostic results are "abnormal", then by the temperature gradients data of continuous multiple discharge cycles
Figure 957034DEST_PATH_IMAGE003
Further deliver to SVM2；When SVM2 diagnostic results are " normal ", then by the preprocessed data of gained
Figure 94623DEST_PATH_IMAGE022
Further deliver to SVM3；When SVM3 diagnostic results are "abnormal", then last diagnostic result is "abnormal".
By said process, this southern all power supply model " GFM-1000E " type battery of 480 section has obtained following result：
The present invention take into account accumulator property deterioration process and voltage, curent change situation and temperature variations are closely related, by setting up SVM data classification based training collection, using SVM algorithm, diagnostic process is deteriorated using multilayer SVM accumulator properties, real time intelligent failure judgement is carried out to battery, therefore there is higher accuracy compared with conventional method to the monitoring of the health status of battery, and early warning can also be provided to the failure of battery.

Claims (2)

1. a kind of battery on-line monitoring method based on multi-layer SVM, it is characterised in that the monitoring method comprises the following steps：
1）Set up SVM data classification based training collection by data collecting system in continuous sampling mode at times to known to several, the different single battery of health status carries out data acquisition, the parameter of collection is single battery discharge voltage variable u during whole discharge cycle, current variable i and battery surface temperature temp, the piezoelectric voltage variable u of same single battery, discharge current variable i and battery surface temperature temp are a sample set, and according to " normal " or the mark of "abnormal" on the single battery health status table, the sample set of all single batteries constitutes SVM data classification based training collection；
    2）The acquisition of SVM data classification input variable gathers the real-time discharge voltage variable u of single battery to be diagnosed, discharge current variable i and real-time battery surface temperature temp in real time by data collecting system in continuous sampling mode at times, above-mentioned variable simultaneously delivers to the SVM data classification input variable of storage composition in SQL database system, is designated as respectively
                                                 
Figure 2011100996466100001DEST_PATH_IMAGE001
、
Figure 619954DEST_PATH_IMAGE002
、
Figure 2011100996466100001DEST_PATH_IMAGE003
；
3）Data prediction：
Single battery discharge cycle data to be diagnosed in SQL database are read, discharge time interval is determined
Figure 102888DEST_PATH_IMAGE004
, the voltage in the discharge cycle period is obtained in continuous sampling mode at times
Figure DEST_PATH_IMAGE005
, electric current
Figure DEST_PATH_IMAGE007
And temperature
Figure 824725DEST_PATH_IMAGE008
Numerical value,（
Figure DEST_PATH_IMAGE009
）If, data sampling at intervals of, then continuous sampling size is data at times
Figure DEST_PATH_IMAGE011
, it is designated as：
Figure 36974DEST_PATH_IMAGE001
、
Figure 198965DEST_PATH_IMAGE002
、
The electric discharge real-time power on each sampled point is calculated accordingly
Figure 62589DEST_PATH_IMAGE012
；
Calculate voltage
Figure 662067DEST_PATH_IMAGE005
, electric current
Figure 362170DEST_PATH_IMAGE007
, temperature
Figure 730703DEST_PATH_IMAGE008
And electric discharge real-time power
Figure DEST_PATH_IMAGE013
Average value：
Figure 310589DEST_PATH_IMAGE014
,
Figure DEST_PATH_IMAGE015
,
Figure 275658DEST_PATH_IMAGE016
And
Figure DEST_PATH_IMAGE017
The deviation of each measurement data is obtained by following equation：
Figure 91036DEST_PATH_IMAGE018
,
Figure DEST_PATH_IMAGE019
,
Figure 328175DEST_PATH_IMAGE020
And
Figure DEST_PATH_IMAGE021
Thus pretreated data are obtained
4）Data diagnosis is using 3 SVM classifiers：SVM1 is used to judge battery tension normal condition and deterioration state, and SVM2 is used to judge single battery temperature normal condition and deterioration state, and SVM3 is used to judge battery comprehensive state normal condition and deterioration state；Following 6 paths of diagnostic process point： 
Path 1：
By step 2）The voltage gradient data sequence of the continuous multiple discharge cycles of gainedSVM1 is delivered to, when SVM1 diagnostic results are " normal ", then by the temperature gradients data of continuous multiple discharge cycles
Figure 9452DEST_PATH_IMAGE003
Further deliver to SVM2；When SVM2 diagnostic results are " normal ", then last diagnostic result is " normal "；
Path 2：
By step 2）The voltage gradient data sequence of the continuous multiple discharge cycles of gainedSVM1 is delivered to, when SVM1 diagnostic results are "abnormal", then by the temperature gradients data of continuous multiple discharge cyclesFurther deliver to SVM2；When SVM2 diagnostic results are "abnormal", then last diagnostic result is "abnormal"；
Path 3：
By step 2）The voltage gradient data sequence of the continuous multiple discharge cycles of gained
Figure 886031DEST_PATH_IMAGE001
SVM1 is delivered to, when SVM1 diagnostic results are " normal ", then by the temperature gradients data of continuous multiple discharge cyclesFurther deliver to SVM2；When SVM2 diagnostic results are "abnormal", then by the preprocessed data of gained
Figure 506423DEST_PATH_IMAGE022
Further deliver to SVM3；When SVM3 diagnostic results are "abnormal", then last diagnostic result is "abnormal"；
Path 4：
By step 2）The voltage gradient data sequence of the continuous multiple discharge cycles of gained
Figure 601943DEST_PATH_IMAGE001
SVM1 is delivered to, when SVM1 diagnostic results are " normal ", then by the temperature gradients data of continuous multiple discharge cycles
Figure 635758DEST_PATH_IMAGE003
Further deliver to SVM2；When SVM2 diagnostic results are "abnormal", then by the preprocessed data of gainedFurther deliver to SVM3；When SVM3 diagnostic results are " normal ", then last diagnostic result is " normal "；
Path 5：
By step 2）The voltage gradient data sequence of the continuous multiple discharge cycles of gained
Figure 337183DEST_PATH_IMAGE001
SVM1 is delivered to, when SVM1 diagnostic results are "abnormal", then by the temperature gradients data of continuous multiple discharge cycles
Figure 662991DEST_PATH_IMAGE003
Further deliver to SVM2；When SVM2 diagnostic results are " normal ", then by the preprocessed data of gained
Figure 184103DEST_PATH_IMAGE022
Further deliver to SVM3；When SVM3 diagnostic results are " normal ", then last diagnostic result is " normal "；
Path 6：
By step 2）The voltage gradient data sequence of the continuous multiple discharge cycles of gainedSVM1 is delivered to, when SVM1 diagnostic results are "abnormal", then by the temperature gradients data of continuous multiple discharge cyclesFurther deliver to SVM2；When SVM2 diagnostic results are " normal ", then by the preprocessed data of gained
Figure 31646DEST_PATH_IMAGE022
Further deliver to SVM3；When SVM3 diagnostic results are "abnormal", then last diagnostic result is "abnormal".
2. the battery on-line monitoring method according to claim 1 based on multi-layer SVM, it is characterised in that：Described SVM classifier is that the long-term charge and discharge data record of battery is trained using SVM algorithm to obtain, and training parameter selection is as follows：
Penalty factor value is [1.3,1200]；Selection of kernel function is RBF kernel functions, i.e.,
Figure DEST_PATH_IMAGE023
,
Figure 164687DEST_PATH_IMAGE024
Value is [0.3,2].
CN201110099646.6A 2011-04-21 2011-04-21 Multi-layer SVM (support vector machine) based storage battery on-line monitoring method Expired - Fee Related CN102749585B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201110099646.6A CN102749585B (en) 2011-04-21 2011-04-21 Multi-layer SVM (support vector machine) based storage battery on-line monitoring method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201110099646.6A CN102749585B (en) 2011-04-21 2011-04-21 Multi-layer SVM (support vector machine) based storage battery on-line monitoring method

Publications (2)

Publication Number Publication Date
CN102749585A true CN102749585A (en) 2012-10-24
CN102749585B CN102749585B (en) 2014-11-05

Family

ID=47029934

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201110099646.6A Expired - Fee Related CN102749585B (en) 2011-04-21 2011-04-21 Multi-layer SVM (support vector machine) based storage battery on-line monitoring method

Country Status (1)

Country Link
CN (1) CN102749585B (en)

Cited By (25)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103529397A (en) * 2013-10-24 2014-01-22 华南师范大学 Method for estimating electric quantity of battery and system for managing electric quantity of battery
CN103745229A (en) * 2013-12-31 2014-04-23 北京泰乐德信息技术有限公司 Method and system of fault diagnosis of rail transit based on SVM (Support Vector Machine)
CN105277891A (en) * 2014-07-18 2016-01-27 三星电子株式会社 Method and apparatus for estimating state of battery
CN105319512A (en) * 2014-07-30 2016-02-10 三星电子株式会社 Method and apparatus for estimating state of battery
JP2017026616A (en) * 2015-07-21 2017-02-02 三星電子株式会社Ｓａｍｓｕｎｇ Ｅｌｅｃｔｒｏｎｉｃｓ Ｃｏ．，Ｌｔｄ． Method and apparatus for estimating state of battery
CN106610478A (en) * 2017-01-10 2017-05-03 中国电力科学研究院 Energy storage battery characteristic estimation method and system based on mass data
CN106772099A (en) * 2017-01-23 2017-05-31 合肥国轩高科动力能源有限公司 A kind of dynamic lithium battery degradation evaluation method
CN107290679A (en) * 2017-07-03 2017-10-24 南京能瑞电力科技有限公司 The Intelligentized battery method for detecting health status of charging pile is shared for electric automobile
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
CN107436409A (en) * 2017-07-07 2017-12-05 淮阴工学院 A kind of electric automobile power battery SOC intelligent predicting devices
CN108805217A (en) * 2018-06-20 2018-11-13 山东大学 A kind of health state of lithium ion battery method of estimation and system based on support vector machines
CN108872863A (en) * 2018-05-02 2018-11-23 广东工业大学 A kind of electric car charged state monitoring method of Optimum Classification
CN108885241A (en) * 2016-03-28 2018-11-23 Ntn株式会社 The degradation determination device of secondary cell
CN108985336A (en) * 2018-06-19 2018-12-11 三峡大学 A kind of fruit quality classification method based on multi-layer SVM
CN109143085A (en) * 2018-07-18 2019-01-04 中国电力科学研究院有限公司 A kind of method and system carrying out early warning to lithium ion battery based on intelligent algorithm
CN110146817A (en) * 2019-05-13 2019-08-20 上海博强微电子有限公司 The diagnostic method of lithium battery failure
CN111178383A (en) * 2019-11-29 2020-05-19 合肥国轩高科动力能源有限公司 Battery monomer quality classification method based on multi-core support vector machine
CN111308353A (en) * 2020-01-07 2020-06-19 天津市中力神盾电子科技有限公司 Storage battery abnormity judgment method and system
CN112630660A (en) * 2020-12-14 2021-04-09 湖北工业大学 Battery fault identification method based on support vector machine
CN113159790A (en) * 2021-05-19 2021-07-23 中国银行股份有限公司 Abnormal transaction identification method and device
CN113391209A (en) * 2021-05-26 2021-09-14 江苏小牛电动科技有限公司 Method, device and system for predicting health state of battery and battery
CN114127570A (en) * 2020-01-22 2022-03-01 株式会社Lg新能源 Battery diagnosis system, power system, and battery diagnosis method
CN114118290A (en) * 2021-12-03 2022-03-01 北华航天工业学院 Electric vehicle battery fault diagnosis method based on signal processing
CN114646880A (en) * 2022-04-13 2022-06-21 中国铁塔股份有限公司江西省分公司 Intelligent diagnosis method and system for lead-acid battery
CN116609676A (en) * 2023-07-14 2023-08-18 深圳先进储能材料国家工程研究中心有限公司 Method and system for monitoring state of hybrid energy storage battery based on big data processing

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070031041A1 (en) * 2005-08-02 2007-02-08 Samsung Electronics Co., Ltd. Apparatus and method for detecting a face
CN101540048A (en) * 2009-04-21 2009-09-23 北京航空航天大学 Image quality evaluating method based on support vector machine

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070031041A1 (en) * 2005-08-02 2007-02-08 Samsung Electronics Co., Ltd. Apparatus and method for detecting a face
CN101540048A (en) * 2009-04-21 2009-09-23 北京航空航天大学 Image quality evaluating method based on support vector machine

Non-Patent Citations (6)

* Cited by examiner, † Cited by third party
Title
占勇等: "基于S变换的电能质量扰动支持向量机分类识别", 《中国电机工程学报》 *
吕干云等: "基于S变换和多级SVM的电能质量扰动检测识别", 《电工技术学报》 *
吕干云等: "基于多级支持向量机分类器的电力变压器故障识别", 《电力系统及其自动化学报》 *
姜慧研等: "基于多级SVM的回转窑烧结状态识别方法", 《东北大学学报（自然科学版）》 *
戴洪德等: "基于多层子支持向量机的多传感器故障诊断", 《仪器仪表学报》 *
陈艳等: "基于SVM的蓄电池容量智能在线检测技术研究", 《压电与声光》 *

Cited By (37)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103529397B (en) * 2013-10-24 2016-06-22 华南师范大学 A kind of method estimating battery electric quantity and battery electric quantity management system
CN103529397A (en) * 2013-10-24 2014-01-22 华南师范大学 Method for estimating electric quantity of battery and system for managing electric quantity of battery
CN103745229A (en) * 2013-12-31 2014-04-23 北京泰乐德信息技术有限公司 Method and system of fault diagnosis of rail transit based on SVM (Support Vector Machine)
CN105277891A (en) * 2014-07-18 2016-01-27 三星电子株式会社 Method and apparatus for estimating state of battery
EP2975421B1 (en) * 2014-07-18 2024-05-22 Samsung Electronics Co., Ltd. Method and apparatus for estimating state of battery
US10295601B2 (en) 2014-07-18 2019-05-21 Samsung Electronics Co., Ltd. Method and apparatus for estimating state of battery
CN105319512A (en) * 2014-07-30 2016-02-10 三星电子株式会社 Method and apparatus for estimating state of battery
CN105319512B (en) * 2014-07-30 2020-06-23 三星电子株式会社 Method and apparatus for estimating state of battery
JP2017026616A (en) * 2015-07-21 2017-02-02 三星電子株式会社Ｓａｍｓｕｎｇ Ｅｌｅｃｔｒｏｎｉｃｓ Ｃｏ．，Ｌｔｄ． Method and apparatus for estimating state of battery
CN108885241A (en) * 2016-03-28 2018-11-23 Ntn株式会社 The degradation determination device of secondary cell
CN107340475B (en) * 2016-04-29 2021-03-16 株式会社日立制作所 Battery fault detection method and battery fault detection device
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
CN106610478A (en) * 2017-01-10 2017-05-03 中国电力科学研究院 Energy storage battery characteristic estimation method and system based on mass data
CN106772099A (en) * 2017-01-23 2017-05-31 合肥国轩高科动力能源有限公司 A kind of dynamic lithium battery degradation evaluation method
CN107290679B (en) * 2017-07-03 2019-11-12 南京能瑞电力科技有限公司 The Intelligentized battery method for detecting health status of charging pile is shared for electric car
CN107290679A (en) * 2017-07-03 2017-10-24 南京能瑞电力科技有限公司 The Intelligentized battery method for detecting health status of charging pile is shared for electric automobile
WO2019006995A1 (en) * 2017-07-07 2019-01-10 淮阴工学院 Intelligent prediction system for power battery soc of electric vehicle
CN107436409A (en) * 2017-07-07 2017-12-05 淮阴工学院 A kind of electric automobile power battery SOC intelligent predicting devices
CN108872863A (en) * 2018-05-02 2018-11-23 广东工业大学 A kind of electric car charged state monitoring method of Optimum Classification
CN108872863B (en) * 2018-05-02 2020-09-08 广东工业大学 Optimized and classified electric vehicle charging state monitoring method
CN108985336A (en) * 2018-06-19 2018-12-11 三峡大学 A kind of fruit quality classification method based on multi-layer SVM
CN108805217A (en) * 2018-06-20 2018-11-13 山东大学 A kind of health state of lithium ion battery method of estimation and system based on support vector machines
CN108805217B (en) * 2018-06-20 2020-10-23 山东大学 Lithium ion battery health state estimation method and system based on support vector machine
CN109143085A (en) * 2018-07-18 2019-01-04 中国电力科学研究院有限公司 A kind of method and system carrying out early warning to lithium ion battery based on intelligent algorithm
CN109143085B (en) * 2018-07-18 2022-09-27 中国电力科学研究院有限公司 Method and system for early warning of lithium ion battery based on artificial intelligence algorithm
CN110146817A (en) * 2019-05-13 2019-08-20 上海博强微电子有限公司 The diagnostic method of lithium battery failure
CN111178383A (en) * 2019-11-29 2020-05-19 合肥国轩高科动力能源有限公司 Battery monomer quality classification method based on multi-core support vector machine
CN111308353A (en) * 2020-01-07 2020-06-19 天津市中力神盾电子科技有限公司 Storage battery abnormity judgment method and system
CN114127570A (en) * 2020-01-22 2022-03-01 株式会社Lg新能源 Battery diagnosis system, power system, and battery diagnosis method
CN114127570B (en) * 2020-01-22 2023-10-20 株式会社Lg新能源 Battery diagnosis system, power system, and battery diagnosis method
CN112630660A (en) * 2020-12-14 2021-04-09 湖北工业大学 Battery fault identification method based on support vector machine
CN113159790A (en) * 2021-05-19 2021-07-23 中国银行股份有限公司 Abnormal transaction identification method and device
CN113391209A (en) * 2021-05-26 2021-09-14 江苏小牛电动科技有限公司 Method, device and system for predicting health state of battery and battery
CN114118290A (en) * 2021-12-03 2022-03-01 北华航天工业学院 Electric vehicle battery fault diagnosis method based on signal processing
CN114646880A (en) * 2022-04-13 2022-06-21 中国铁塔股份有限公司江西省分公司 Intelligent diagnosis method and system for lead-acid battery
CN116609676A (en) * 2023-07-14 2023-08-18 深圳先进储能材料国家工程研究中心有限公司 Method and system for monitoring state of hybrid energy storage battery based on big data processing
CN116609676B (en) * 2023-07-14 2023-09-15 深圳先进储能材料国家工程研究中心有限公司 Method and system for monitoring state of hybrid energy storage battery based on big data processing

Also Published As

Publication number Publication date
CN102749585B (en) 2014-11-05

Similar Documents

Publication Publication Date Title
CN102749585A (en) Multi-layer SVM (support vector machine) based storage battery on-line monitoring method
Jiang et al. Data-driven fault diagnosis and thermal runaway warning for battery packs using real-world vehicle data
CN111584952B (en) Method and system for online evaluation of electrochemical cells of energy storage power station
CN110133508A (en) The safe early warning method of electric automobile power battery
CN102263303B (en) Modular lithium power battery and management systems and management method thereof
CN101067644B (en) Storage battery performance analytical expert diagnosing method
CN107656216A (en) A kind of lead-acid accumulator on-line monitoring is safeguarded and early warning system and performance estimating method
CN112186275A (en) BMS system based on high in clouds
WO2011147222A1 (en) Control method and system of intelligent detection, restoration for lead-acid battery group
CN110146817A (en) The diagnostic method of lithium battery failure
CN116455085B (en) Intelligent monitoring system of battery energy storage power station
CN109143091A (en) Battery management system failure FDIR system and method based on dual redundant
US20230059529A1 (en) Characterization of Rechargeable Batteries Using Machine-Learned Algorithms
CN116401585B (en) Energy storage battery failure risk assessment method based on big data
CN109615273A (en) A kind of electric car electrically-charging equipment method for evaluating state and system
CN103986205B (en) 16 series lithium battery group multi-mode balance control methods of dynamic adaptation
CN111509319A (en) PHM management system for energy storage power supply
Wei et al. Multi-level data-driven battery management: From internal sensing to big data utilization
He et al. Voltage abnormality-based fault diagnosis for batteries in electric buses with a self-adapting update model
CN116754967A (en) Method and system for online evaluation of electrochemical cells of an energy storage power station
Yao et al. Fault identification of lithium-ion battery pack for electric vehicle based on ga optimized ELM neural network
CN113629856B (en) Intelligent power supply management method and system for deep sea mobile platform battery
CN105873111A (en) Soft and hard fault diagnosis and self restoration method suitable for health monitoring
CN117543791A (en) Power supply detection method, device, equipment and storage medium for power supply
Wu et al. Research on short-circuit fault-diagnosis strategy of lithium-ion battery in an energy-storage system based on voltage cosine similarity

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20141105

Termination date: 20180421

CF01 Termination of patent right due to non-payment of annual fee