CN114118290A - Electric vehicle battery fault diagnosis method based on signal processing - Google Patents

Electric vehicle battery fault diagnosis method based on signal processing Download PDF

Info

Publication number
CN114118290A
CN114118290A CN202111465737.7A CN202111465737A CN114118290A CN 114118290 A CN114118290 A CN 114118290A CN 202111465737 A CN202111465737 A CN 202111465737A CN 114118290 A CN114118290 A CN 114118290A
Authority
CN
China
Prior art keywords
battery
electric vehicle
data
sample
support vector
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
CN202111465737.7A
Other languages
Chinese (zh)
Inventor
赵妙颖
焦智
王宏宇
任思源
刘志宾
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.)
North China Institute of Aerospace Engineering
Original Assignee
North China Institute of Aerospace Engineering
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 North China Institute of Aerospace Engineering filed Critical North China Institute of Aerospace Engineering
Priority to CN202111465737.7A priority Critical patent/CN114118290A/en
Publication of CN114118290A publication Critical patent/CN114118290A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
    • G01R31/392—Determining battery ageing or deterioration, e.g. state of health
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
    • G01R31/396—Acquisition or processing of data for testing or for monitoring individual cells or groups of cells within a battery
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Secondary Cells (AREA)
  • Tests Of Electric Status Of Batteries (AREA)

Abstract

The invention discloses a method for diagnosing battery faults of an electric vehicle based on signal processing, which adopts an empirical wavelet transform and sample entropy method to extract the characteristics of historical data of the battery of the electric vehicle, establishes a battery state characteristic space, takes the battery state characteristic space as a training sample, and adopts a hierarchical support vector machine with multi-classification performance to establish a battery fault diagnosis model; the test sample input by the model is a battery online data state characteristic space, and the model output is a fault type, so that online real-time fault diagnosis of the electric vehicle battery is realized. According to the electric vehicle battery fault diagnosis method based on signal processing, the method based on signal processing is combined with machine learning, and battery use data is fully utilized to realize electric vehicle battery fault diagnosis research, so that on one hand, a complex electrochemical analytic mathematical model can be avoided being established, the fault diagnosis efficiency is improved, and on the other hand, the fault diagnosis accuracy can be improved through effective feature extraction and accurate modeling.

Description

Electric vehicle battery fault diagnosis method based on signal processing
Technical Field
The invention relates to the technical field of electric vehicle battery fault diagnosis, in particular to a signal processing-based electric vehicle battery fault diagnosis method.
Background
The power of the electric vehicle is mainly provided by an internally installed battery. Lithium batteries are widely used in electric vehicles because of their advantages of long service life, high energy density, controllable cost, etc. Along with the accumulation of service time and the increase of the number of charging and discharging cycles, the battery capacity is reduced, aging faults occur, and meanwhile, the small difference between single lithium batteries is gradually increased. When the single lithium batteries form the battery pack in a series-parallel connection mode, the difference among different monomers easily causes the lithium batteries to have faults of overcharge, overdischarge, thermal runaway and the like, reduces the thermal stability of the batteries, shortens the service life of the batteries and increases the potential safety hazard of electric vehicles. Therefore, in order to prolong the service life of the battery and ensure the use safety of the electric automobile, research needs to be carried out on the battery fault diagnosis technology of the electric automobile.
Currently, common fault diagnosis methods can be divided into three categories: mathematical model-based methods, signal processing-based methods, and knowledge-based methods. The method based on the mathematical model can realize real-time fault detection, but the establishment of the mathematical model has certain difficulty, the fault detection precision is too dependent on the precision degree of the mathematical model, and the requirement on the accuracy of the mathematical model is high. Although the knowledge-based method does not need to establish an accurate mathematical model, the precondition is that a large amount of prior knowledge is obtained in advance, and certain difficulty is brought to some systems. The basic idea of the signal processing-based method is to perform signal processing on a measurable signal output by a system to obtain the output signal characteristics of a dynamic system, and accordingly, the system state and the fault type are judged. The method is easy to realize, an accurate mathematical model does not need to be established for a complex system, and the fault diagnosis rate is improved. Therefore, for the electric vehicle battery fault diagnosis, application research of a signal processing method needs to be enhanced, and the vehicle battery fault diagnosis efficiency needs to be further improved.
Disclosure of Invention
The invention aims to provide a method for diagnosing the battery fault of an electric automobile based on signal processing.
In order to achieve the purpose, the invention provides the following scheme:
a method for diagnosing battery faults of an electric automobile based on signal processing comprises the following steps:
s1, obtaining historical use data of the electric vehicle battery, extracting features of the historical use data of the electric vehicle battery by adopting an empirical wavelet transform and sample entropy method, and establishing a battery state feature space based on the extracted feature vectors;
s2, establishing an electric vehicle battery fault diagnosis model by using a hierarchical support vector machine with the battery state feature space as a training sample;
s3, collecting the online data of the electric vehicle battery, and extracting the characteristics of the online data of the electric vehicle battery by the same method as the step S1 to obtain a battery online data state characteristic space;
and S4, inputting the battery online data state characteristic space as a test sample into the electric vehicle battery fault diagnosis model, and outputting a corresponding fault type.
Further, in step S1, the historical usage data of the electric vehicle battery includes: and historical data of the electric automobile battery under four different states of a normal state, overcharge, overdischarge and aging fault.
Further, in step S1, performing feature extraction on historical usage data of the electric vehicle battery by using an empirical wavelet transform and a sample entropy method, and establishing a battery state feature space based on the extracted feature vectors, specifically including:
performing adaptive decomposition on historical data of the electric vehicle battery in four different states of a normal state, overcharge, overdischarge and aging faults by adopting empirical wavelet transform to generate an empirical wavelet function component;
and calculating the sample entropy of each component, combining the feature vectors of the battery data, and establishing a battery state feature space.
Further, in step S2, the battery state feature space is used as a training sample, and a hierarchical support vector machine is used to establish a battery fault diagnosis model of the electric vehicle, which specifically includes:
the hierarchical support vector machine needs 3 support vector machine two types of classifiers, namely SVM1, SVM2 and SVM3, wherein the SVM1 realizes the classification of { normal state, aging fault } and { overcharge and overdischarge }; the SVM2 realizes the classification of normal state and aging fault; the SVM3 implements the classification of overcharge from overdischarge.
Further, in step S2, the battery state feature space is used as a training sample, and a hierarchical support vector machine is used to establish a battery fault diagnosis model of the electric vehicle, which specifically includes:
s201, performing hierarchical support vector machine training by taking the battery state feature space S in the step S1 as a training sample, wherein the battery state feature space S selects 120 groups of sample data, and 30 groups of sample data are selected for normal state, overcharge, overdischarge and aging faults respectively;
s202, establishing a first-layer support vector machine two-class classifier SVM1, and inputting a training sample vector (V)i,Yi) (i ═ 1,2, …,120, y ∈ { -1,1}), if V { [ 1,2, …,120, y { [ 1 ] is presentiE is left to { normal state, aging fault }, then yi1 is ═ 1; if ViE { overcharge, overdischarge }, then yi＝-1；
The kernel function adopted by the hierarchical support vector machine for training is a Gaussian radial basis function:
Figure BDA0003391368650000031
wherein, take σ2＝0.01，Vi,VjTwo training sample data;
solving the objective function by using a quadratic programming method:
Figure BDA0003391368650000032
obtaining an optimal Lagrange multiplier for a first layer support vector machine
Figure BDA0003391368650000033
Wherein, ai,ajLagrange multipliers, y, corresponding to the ith and jth training sample data respectivelyi,yjRespectively corresponding state values of the ith training sample data and the jth training sample data;
substituting a support vector V in the training sample into the formula:
Figure BDA0003391368650000034
wherein f (V) is the class value of the vector, i.e., -1 or 1, and the offset value of the first layer support vector machine is calculated
Figure BDA0003391368650000035
Using a trained langerhan multiplier
Figure BDA0003391368650000036
Deviation value
Figure BDA0003391368650000037
And kernel function k (V)i,Vj) Establishing a first-layer support vector machine model:
Figure BDA0003391368650000038
s203, establishing a second-layer support vector machine classifier SVM2 and SVM3 according to the method in the step S202, wherein the training sample of the SVM2 is a feature space S formed by two sample data feature vectors of a normal state and an aging fault1＝[V1，V2，…，V60]Inputting a training sample vector (V)i,Yi) (i ═ 1,2, …,60, y ∈ { -1,1}), if V { [ 1,2, …,60, y { } is presentiE is normalState }, then yi1 is ═ 1; if ViE { aging fault }, then yi＝-1；
Similarly, the training sample of the SVM3 is a feature space S composed of two sample data feature vectors of overcharge and overdischarge2＝[V61，V2，…，V120]Inputting a training sample vector (V)i,Yi) (i-61, 2, …,120, y ∈ { -2,2}), if V ∈ { -2,2})iE { overcharge }, then yi2; if ViE { overdischarge }, then yi＝-2；
Training the parameters of SVM2 according to the method in step S202
Figure BDA0003391368650000041
And
Figure BDA0003391368650000042
and parameters of SVM3
Figure BDA0003391368650000043
And
Figure BDA0003391368650000044
respectively establishing a second layer of support vector machine model
Figure BDA0003391368650000045
And third layer support vector machine model
Figure BDA0003391368650000046
Further, in step S4, the method includes using the battery online data state feature space as a test sample, inputting the test sample into a battery fault diagnosis model of the electric vehicle, and outputting a corresponding fault type, which specifically includes:
s401, dividing the acquired online data of the electric vehicle battery by taking 1000 data points as a group of sample data, selecting 200 groups of sample data, and taking the state characteristic space of the battery online data as St＝[V1，V2，…，V200]；
S402, combining the feature spaceStInputting the test sample into the battery fault diagnosis model established in step S2, and outputting f (V)i) If 1, the sample belongs to normal state; if f (V)i) If-1, the sample belongs to aging fault; if f (V)i) 2, the sample belongs to an overcharged state; if f (V)i) When the value is-2, the sample belongs to an overdischarge state.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects: the invention provides a method for diagnosing battery faults of an electric vehicle based on signal processing, which combines the method based on signal processing with machine learning, realizes the research of battery fault diagnosis of the electric vehicle by fully utilizing battery use data, specifically, adopts empirical wavelet transform and sample entropy to carry out feature extraction on the battery use data of the electric vehicle, constructs a battery state feature space, establishes a battery fault diagnosis model by utilizing a hierarchical support vector machine on the basis of the battery state feature space, inputs the online state of the battery, namely outputs the corresponding fault type, and realizes the online real-time diagnosis of the battery faults of the electric vehicle; according to the method, on one hand, the establishment of a complex electrochemical analytic mathematical model can be avoided, the fault diagnosis efficiency is improved, and on the other hand, the fault diagnosis accuracy can be improved through effective feature extraction and accurate modeling.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a schematic flow chart of a method for diagnosing a battery fault of an electric vehicle based on signal processing according to the present invention;
FIG. 2 is a schematic diagram of a hierarchical support vector machine decision according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a method for diagnosing the battery fault of an electric automobile based on signal processing, which can realize the online real-time diagnosis of the battery fault type of the electric automobile and simultaneously improve the fault diagnosis efficiency and the diagnosis accuracy.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
The invention provides a method for diagnosing the battery fault of an electric automobile based on signal processing, which comprises the following steps:
s1, obtaining historical use data of the electric vehicle battery, extracting features of the historical use data of the electric vehicle battery by adopting an empirical wavelet transform and sample entropy method, and establishing a battery state feature space based on the extracted feature vectors; the historical use data of the electric automobile battery comprises: historical data of the electric automobile battery under four different states of a normal state, overcharge, overdischarge and aging faults; in the embodiment of the invention, each type of data is divided by taking 1000 data points as a group of sample data, and 30 groups of sample data are selected;
s2, establishing an electric vehicle battery fault diagnosis model by using a hierarchical support vector machine with the battery state feature space as a training sample;
s3, collecting the online data of the electric vehicle battery, and extracting the characteristics of the online data of the electric vehicle battery by the same method as the step S1 to obtain a battery online data state characteristic space;
and S4, inputting the battery online data state characteristic space as a test sample into the electric vehicle battery fault diagnosis model, and outputting a corresponding fault type.
In step S1, performing feature extraction on historical use data of the electric vehicle battery by using an empirical wavelet transform and a sample entropy method, and establishing a battery state feature space based on the extracted feature vectors, specifically including:
s101, performing adaptive decomposition on historical data of the electric vehicle battery in a normal state, an overcharge state, an overdischarge state and an aging fault by adopting empirical wavelet transform to generate an empirical wavelet function component, which specifically comprises the following steps:
1) taking a certain sample data as an example, the time dispersion sequence X ═ X is used1,x2,…,x1000]Representing the sample, and carrying out Fourier transform on X to obtain a Fourier spectrum F (omega);
2) f (ω) is adaptively divided according to the following spectral division principle: at omeganIs bounded by [0, π]If F (ω) is divided into N consecutive paragraphs, there are N +1 boundaries, where ω0＝0,ωNPi, the remaining N-1 boundaries are determined as follows: detecting local maximum values of F (omega) and arranging the local maximum values in a descending order, assuming that M maximum values exist, if M is larger than or equal to N, keeping the first N-1 maximum values; if M is less than or equal to N, retaining all maximum values and resetting N value, and finally taking the intermediate frequency of two continuous maximum values as omegan；
3) Constructing a wavelet filter bank based on the spectral segmentation of 2), determining the components of empirical wavelet function, and establishing an empirical wavelet basis, wherein the empirical scale function
Figure BDA0003391368650000061
And empirical wavelet mother function
Figure BDA0003391368650000062
Respectively as follows:
Figure BDA0003391368650000063
Figure BDA0003391368650000071
in the formula, ωn,ωn+1Respectively representing the nth and n +1 th spectral boundaries, τn,τn+1Half the length of the n-th and n + 1-th conversion periods, respectively, and is usually taken as τn＝γωnGamma is more than 0 and less than 1; beta (x) satisfies
Figure BDA0003391368650000072
And is
Figure BDA0003391368650000073
Empirical wavelet transform:
Figure BDA0003391368650000074
the approximation coefficients are:
Figure BDA0003391368650000075
wherein psinPhi and phi1Respectively an empirical wavelet function and an empirical scale function,
Figure BDA0003391368650000076
and
Figure BDA0003391368650000077
are respectively psinAnd phi1Complex conjugation of (a).
Recovering the original sample sequence from the empirical wavelet transform to obtain:
Figure BDA0003391368650000078
the empirical wavelet function components are:
Figure BDA0003391368650000079
the empirical wavelet function component resulting from the empirical wavelet transform decomposition is denoted as F1,F2,…,FnThe method comprises different amplitude-frequency characteristic information of the battery sample data.
S102, calculating sample entropies of all components, combining feature vectors of battery data, and establishing a battery state feature space, specifically:
1) with an empirical wavelet function component F1To illustrate the sample entropy calculation procedure, a data sequence F is calculated1Standard deviation of (2)
Figure BDA00033913686500000710
Wherein f isiFor the ith element in the data sequence,
Figure BDA00033913686500000711
taking the similarity tolerance r as 0.2 sigma, and initializing a basic parameter embedding dimension m for the average value of the data sequence elements, wherein N is 1000;
2) defining a sequence of vectors Y of dimension mm＝[Ym(1),Ym(2),…,Ym(N-m+1)]Wherein Y ism(i)＝[f1(i),f1(i+1),…,f1(i+m-1)]I is more than or equal to 1 and less than or equal to N-m +1, and calculating vector Ym(i) And Ym(j) The distance between, we get:
Figure BDA0003391368650000081
3) for a given Ym(i) Statistics of Ym(i) And Ym(j) The number of j (j is more than or equal to 1 and less than or equal to N-m +1, j is not equal to i) with the distance between the two is less than or equal to r, and the following results are obtained: b isi＝num(d[Ym(i),Ym(j)]R is less than or equal to r); calculating the ratio of the distance to the total distance to obtain:
Figure BDA0003391368650000082
on this basis, average all i to get:
Figure BDA0003391368650000083
4) increasing the dimensionality to m +1, repeating steps 2) -3) to obtain Bm+1(r), then the empirical wavelet function component F1The sample entropy of (d) is:
Figure BDA0003391368650000084
5) calculating an empirical wavelet function F according to steps 1) -4)1,F2,…,FnThe characteristic vector V of the battery sample data is combined and constructed: v ═ SampEn1, SampEn2, …, SampEn]。
According to the steps S101 and S102, feature extraction is carried out on battery history data (120 groups of sample data in total) of the electric automobile in four different states such as a normal state, an aging fault, overcharge and overdischarge, feature vectors of each group of sample data are constructed, and a battery history state feature space S is formed by 120 sample data feature vectors, wherein S is [ V ═ V [ V ] ]1，V2，…，V120]。
In step S2, the hierarchical support vector machine needs 3 support vector machine classifiers, which are SVM1, SVM2 and SVM3, respectively, and the SVM1 realizes the classification of { normal state, aging fault } and { overcharge, overdischarge }; the SVM2 realizes the classification of normal state and aging fault; the SVM3 implements the classification of overcharge from overdischarge.
In the step S2, the battery state feature space is used as a training sample, and a hierarchical support vector machine is used to establish a battery fault diagnosis model of the electric vehicle, which specifically includes:
s201, performing hierarchical support vector machine training by taking the battery state feature space S in the step S1 as a training sample, wherein the battery state feature space S selects 120 groups of sample data, and 30 groups of sample data are selected for normal state, overcharge, overdischarge and aging faults respectively;
s202, establishing a first-layer support vector machine two-class classifier SVM1, and inputting a training sample vector (V)i,Yi) (i ═ 1,2, …,120, y ∈ { -1,1}), if V { [ 1,2, …,120, y { [ 1 ] is presentiE is left to { normal state, aging fault }, then yi1 is ═ 1; if ViE { overcharge, overdischarge }, then yi＝-1；
The kernel function adopted by the hierarchical support vector machine for training is a Gaussian radial basis function:
Figure BDA0003391368650000091
wherein, take σ2＝0.01，Vi,VjTwo training sample data;
solving the objective function by using a quadratic programming method:
Figure BDA0003391368650000092
obtaining an optimal Lagrange multiplier for a first layer support vector machine
Figure BDA0003391368650000093
Wherein, ai,ajLagrange multipliers, y, corresponding to the ith and jth training sample data respectivelyi,yjRespectively corresponding state values of the ith training sample data and the jth training sample data;
substituting a support vector V in the training sample into the formula:
Figure BDA0003391368650000094
wherein f (V) is the class value of the vector, i.e., -1 or 1, and the offset value of the first layer support vector machine is calculated
Figure BDA0003391368650000095
Using a trained langerhan multiplier
Figure BDA0003391368650000096
Deviation value
Figure BDA0003391368650000097
And kernel function k (V)i,Vj) Establishing a first-layer support vector machine model:
Figure BDA0003391368650000098
s203, establishing a second-layer support vector machine classifier SVM2 and SVM3 according to the method in the step S202, wherein the training sample of the SVM2 is a feature space S formed by two sample data feature vectors of a normal state and an aging fault1＝[V1，V2，…，V60]Inputting a training sample vector (V)i,Yi) (i ═ 1,2, …,60, y ∈ { -1,1}), if V { [ 1,2, …,60, y { } is presentiE is left to the normal state, then yi1 is ═ 1; if ViE { aging fault }, then yi＝-1；
Similarly, the training sample of the SVM3 is a feature space S composed of two sample data feature vectors of overcharge and overdischarge2＝[V61，V2，…，V120]Inputting a training sample vector (V)i,Yi) (i-61, 2, …,120, y ∈ { -2,2}), if V ∈ { -2,2})iE { overcharge }, then yi2; if ViE { overdischarge }, then yi＝-2；
Training the parameters of SVM2 according to the method in step S202
Figure BDA0003391368650000101
And
Figure BDA0003391368650000102
and parameters of SVM3
Figure BDA0003391368650000103
And
Figure BDA0003391368650000104
respectively establishing a second layer of support vector machine model
Figure BDA0003391368650000105
And third layer support vector machine model
Figure BDA0003391368650000106
Step S4, inputting the battery online data state feature space as a test sample into the electric vehicle battery fault diagnosis model, and outputting a corresponding fault type, specifically including:
s401, dividing the acquired online data of the electric vehicle battery by taking 1000 data points as a group of sample data, selecting 200 groups of sample data, and taking the state characteristic space of the battery online data as St＝[V1，V2，…，V200]；
S402, converting the feature space StInputting the test sample into the battery fault diagnosis model established in step S2, and outputting f (V)i) If 1, the sample belongs to normal state; if f (V)i) If-1, the sample belongs to aging fault; if f (V)i) 2, the sample belongs to an overcharged state; if f (V)i) When the value is-2, the sample belongs to an overdischarge state.
The invention provides a method for diagnosing battery faults of an electric vehicle based on signal processing, which combines the method based on signal processing with machine learning, realizes the research of battery fault diagnosis of the electric vehicle by fully utilizing battery use data, specifically, adopts empirical wavelet transformation and sample entropy to carry out feature extraction on the battery use data of the electric vehicle, constructs a battery state feature space, and establishes a battery fault diagnosis model by utilizing a hierarchical support vector machine on the basis of the battery state feature space to realize the online real-time diagnosis of the battery faults of the electric vehicle; according to the method, on one hand, the establishment of a complex electrochemical analytic mathematical model can be avoided, the fault diagnosis efficiency is improved, and on the other hand, the fault diagnosis accuracy can be improved through effective feature extraction and accurate modeling.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.

Claims (6)

1. A method for diagnosing battery faults of an electric automobile based on signal processing is characterized by comprising the following steps:
s1, obtaining historical use data of the electric vehicle battery, extracting features of the historical use data of the electric vehicle battery by adopting an empirical wavelet transform and sample entropy method, and establishing a battery state feature space based on the extracted feature vectors;
s2, establishing an electric vehicle battery fault diagnosis model by using a hierarchical support vector machine with the battery state feature space as a training sample;
s3, collecting the online data of the electric vehicle battery, and extracting the characteristics of the online data of the electric vehicle battery by the same method as the step S1 to obtain a battery online data state characteristic space;
and S4, inputting the battery online data state characteristic space as a test sample into the electric vehicle battery fault diagnosis model, and outputting a corresponding fault type.
2. The method for diagnosing the battery fault of the electric vehicle based on the signal processing as claimed in claim 1, wherein in the step S1, the historical usage data of the battery of the electric vehicle includes: and historical data of the electric automobile battery under four different states of a normal state, overcharge, overdischarge and aging fault.
3. The method for diagnosing battery faults of an electric vehicle based on signal processing as claimed in claim 2, wherein in step S1, the method for extracting the features of the historical use data of the battery of the electric vehicle by using empirical wavelet transform and sample entropy method, and establishing the battery state feature space based on the extracted feature vectors specifically includes:
performing adaptive decomposition on historical data of the electric vehicle battery in four different states of a normal state, overcharge, overdischarge and aging faults by adopting empirical wavelet transform to generate an empirical wavelet function component;
and calculating the sample entropy of each component, combining the feature vectors of the battery data, and establishing a battery state feature space.
4. The method for battery fault diagnosis of electric vehicle based on signal processing as claimed in claim 3, wherein in step S2, the hierarchical support vector machine requires 3 support vector machine classifiers, which are SVM1, SVM2 and SVM3, respectively, SVM1 implements the classification of { normal state, aging fault } and { overcharge, overdischarge }; the SVM2 realizes the classification of normal state and aging fault; the SVM3 implements the classification of overcharge from overdischarge.
5. The method for diagnosing battery faults of an electric vehicle based on signal processing according to claim 4, wherein the step S2 is implemented by taking a battery state feature space as a training sample and establishing a battery fault diagnosis model of the electric vehicle by using a hierarchical support vector machine, and specifically comprises:
s201, performing hierarchical support vector machine training by taking the battery state feature space S in the step S1 as a training sample, wherein the battery state feature space S selects 120 groups of sample data, and 30 groups of sample data are selected for normal state, overcharge, overdischarge and aging faults respectively;
s202, establishing a first-layer support vector machine two-class classifier SVM1, and inputting a training sample vector (V)i,Yi) (i ═ 1,2, …,120, y ∈ { -1,1}), if V { [ 1,2, …,120, y { [ 1 ] is presentiE is left to { normal state, aging fault }, then yi1 is ═ 1; if ViE { overcharge, overdischarge }, then yi＝-1；
The kernel function adopted by the hierarchical support vector machine for training is a Gaussian radial basis function:
Figure FDA0003391368640000021
wherein, take σ2＝0.01，Vi,VjTwo training sample data;
solving the objective function by using a quadratic programming method:
Figure FDA0003391368640000022
obtaining an optimal Lagrange multiplier for a first layer support vector machine
Figure FDA0003391368640000023
Wherein, ai,ajLagrange multipliers, y, corresponding to the ith and jth training sample data respectivelyi,yjRespectively corresponding state values of the ith training sample data and the jth training sample data;
substituting a support vector V in the training sample into the formula:
Figure FDA0003391368640000024
wherein f (V) is the class value of the vector, i.e., -1 or 1, and the offset value of the first layer support vector machine is calculated
Figure FDA0003391368640000025
Using a trained langerhan multiplier
Figure FDA0003391368640000026
Deviation value
Figure FDA0003391368640000027
And kernel function k (V)i,Vj) Establishing a first-layer support vector machine model:
Figure FDA0003391368640000031
s203, establishing a second-layer support vector machine classifier SVM2 and SVM3 according to the method in the step S202, wherein the training sample of the SVM2 is a feature space S formed by two sample data feature vectors of a normal state and an aging fault1＝[V1，V2，…，V60]Inputting a training sample vector (V)i,Yi) (i ═ 1,2, …,60, y ∈ { -1,1}), if V { [ 1,2, …,60, y { } is presentiE is left to the normal state, then yi1 is ═ 1; if ViE { aging fault }, then yi＝-1；
Similarly, the training sample of the SVM3 is a feature space S composed of two sample data feature vectors of overcharge and overdischarge2＝[V61，V2，…，V120]Inputting a training sample vector (V)i,Yi) (i-61, 2, …,120, y ∈ { -2,2}), if V ∈ { -2,2})iE { overcharge }, then yi2; if ViE { overdischarge }, then yi＝-2；
Training the parameters of SVM2 according to the method in step S202
Figure FDA0003391368640000032
And
Figure FDA0003391368640000033
and parameters of SVM3
Figure FDA0003391368640000034
And
Figure FDA0003391368640000035
respectively establishing a second layer of support vector machine model
Figure FDA0003391368640000036
And third layer support vector machine model
Figure FDA0003391368640000037
6. The method for diagnosing the battery fault of the electric vehicle based on the signal processing as claimed in claim 5, wherein the step S4, taking the battery online data state feature space as a test sample, inputs the test sample into a battery fault diagnosis model of the electric vehicle, and outputs a corresponding fault type, specifically comprises:
s401, dividing the acquired online data of the electric vehicle battery by taking 1000 data points as a group of sample data, selecting 200 groups of sample data, and taking the state characteristic space of the battery online data as St＝[V1，V2，…，V200]；
S402, converting the feature space StInputting the test sample into the battery fault diagnosis model established in step S2, and outputting f (V)i) If 1, the sample belongs to normal state; if f (V)i) If-1, the sample belongs to aging fault; if f (V)i) 2, the sample belongs to an overcharged state; if f (V)i) When the value is-2, the sample belongs to an overdischarge state.
CN202111465737.7A 2021-12-03 2021-12-03 Electric vehicle battery fault diagnosis method based on signal processing Pending CN114118290A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111465737.7A CN114118290A (en) 2021-12-03 2021-12-03 Electric vehicle battery fault diagnosis method based on signal processing

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111465737.7A CN114118290A (en) 2021-12-03 2021-12-03 Electric vehicle battery fault diagnosis method based on signal processing

Publications (1)

Publication Number Publication Date
CN114118290A true CN114118290A (en) 2022-03-01

Family

ID=80366079

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111465737.7A Pending CN114118290A (en) 2021-12-03 2021-12-03 Electric vehicle battery fault diagnosis method based on signal processing

Country Status (1)

Country Link
CN (1) CN114118290A (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102749585A (en) * 2011-04-21 2012-10-24 李昌 Multi-layer SVM (support vector machine) based storage battery on-line monitoring method
CN110703109A (en) * 2019-10-25 2020-01-17 山东大学 Battery string multi-fault diagnosis method and system based on correction sample entropy
CN111090050A (en) * 2020-01-21 2020-05-01 合肥工业大学 Lithium battery fault diagnosis method based on support vector machine and K mean value
CN113111618A (en) * 2021-03-09 2021-07-13 电子科技大学 Analog circuit fault diagnosis method based on improved empirical wavelet transform

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102749585A (en) * 2011-04-21 2012-10-24 李昌 Multi-layer SVM (support vector machine) based storage battery on-line monitoring method
CN110703109A (en) * 2019-10-25 2020-01-17 山东大学 Battery string multi-fault diagnosis method and system based on correction sample entropy
CN111090050A (en) * 2020-01-21 2020-05-01 合肥工业大学 Lithium battery fault diagnosis method based on support vector machine and K mean value
CN113111618A (en) * 2021-03-09 2021-07-13 电子科技大学 Analog circuit fault diagnosis method based on improved empirical wavelet transform

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
万书亭等: "基于EWT和多尺度熵的高压断路器故障诊断", 振动.测试与诊断, vol. 38, no. 04, pages 89 - 90 *
燕海霞等: "中医脉诊研究与临床应用", 上海科学技术出版社, pages: 114 - 115 *

Similar Documents

Publication Publication Date Title
Modi et al. Estimation of energy consumption of electric vehicles using deep convolutional neural network to reduce driver’s range anxiety
CN112379269B (en) Battery abnormality detection model training and detection method and device thereof
CN112858919A (en) Battery system online fault diagnosis method and system based on cluster analysis
CN111027625B (en) Battery screening method based on SAE and K-means clustering algorithm
CN104459560A (en) Method for predicting remaining service life of lithium battery based on wavelet denoising and relevant vector machine
Li et al. Fast battery capacity estimation using convolutional neural networks
Li et al. Remaining useful life prediction for lithium-ion batteries with a hybrid model based on TCN-GRU-DNN and dual attention mechanism
Su et al. Prediction of remaining useful life for lithium-ion battery with multiple health indicators
CN112305441B (en) Power battery health state assessment method under integrated clustering
CN113569485A (en) Short-time abnormal battery identification method based on variational modal decomposition
WO2023226355A1 (en) Dual-ion battery fault detection method and system based on multi-source perception
Park et al. Capacity estimation of Li-ion batteries using constant current charging voltage with multilayer perceptron
CN104090238A (en) Lithium ion battery remaining service life prediction method based on AR ensemble learning model
CN114966436A (en) Lithium battery state of charge prediction method, device, equipment and readable storage medium
CN113361692A (en) Lithium battery residual life combined prediction method
CN116400244B (en) Abnormality detection method and device for energy storage battery
CN114118290A (en) Electric vehicle battery fault diagnosis method based on signal processing
Alam Recurrent neural networks in electricity load forecasting
CN113721151B (en) Battery capacity estimation model and method based on double-tower deep learning network
CN116307097A (en) Energy storage system residual electric quantity prediction method based on deep learning
Wang et al. An inconsistency fault diagnosis method for lithium-ion cells in the battery pack based on piecewise dimensionality reduction and outlier identification
Wu et al. Network traffic prediction based on a CNN-LSTM with attention mechanism
CN113884903A (en) Battery fault diagnosis method based on multilayer perceptron neural network
CN115545111A (en) Network intrusion detection method and system based on clustering self-adaptive mixed sampling
CN114545279A (en) Lithium battery health state estimation method based on neural network ordinary differential equation

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination