CN115219902A - Method and system for rapidly testing service life of power battery - Google Patents
Method and system for rapidly testing service life of power battery Download PDFInfo
- Publication number
- CN115219902A CN115219902A CN202210789672.XA CN202210789672A CN115219902A CN 115219902 A CN115219902 A CN 115219902A CN 202210789672 A CN202210789672 A CN 202210789672A CN 115219902 A CN115219902 A CN 115219902A
- Authority
- CN
- China
- Prior art keywords
- battery
- variable
- power battery
- cycle life
- life
- 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
Links
- 238000012360 testing method Methods 0.000 title claims abstract description 105
- 238000000034 method Methods 0.000 title claims abstract description 93
- 238000010219 correlation analysis Methods 0.000 claims abstract description 35
- 238000010801 machine learning Methods 0.000 claims abstract description 19
- 239000011159 matrix material Substances 0.000 claims description 30
- 239000013598 vector Substances 0.000 claims description 28
- 238000004590 computer program Methods 0.000 claims description 14
- 230000006870 function Effects 0.000 claims description 11
- 238000003860 storage Methods 0.000 claims description 10
- 238000005457 optimization Methods 0.000 claims description 9
- 238000013507 mapping Methods 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000010998 test method Methods 0.000 abstract description 7
- 230000001419 dependent effect Effects 0.000 description 11
- 238000010586 diagram Methods 0.000 description 9
- 238000007599 discharging Methods 0.000 description 8
- 238000012549 training Methods 0.000 description 8
- 238000004422 calculation algorithm Methods 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000007637 random forest analysis Methods 0.000 description 5
- HBBGRARXTFLTSG-UHFFFAOYSA-N Lithium ion Chemical compound [Li+] HBBGRARXTFLTSG-UHFFFAOYSA-N 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 4
- 229910001416 lithium ion Inorganic materials 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000010200 validation analysis Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 230000018109 developmental process Effects 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 230000032683 aging Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 125000004122 cyclic group Chemical group 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000005087 graphitization Methods 0.000 description 2
- 230000014759 maintenance of location Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 206010011906 Death Diseases 0.000 description 1
- WHXSMMKQMYFTQS-UHFFFAOYSA-N Lithium Chemical compound [Li] WHXSMMKQMYFTQS-UHFFFAOYSA-N 0.000 description 1
- 238000002441 X-ray diffraction Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 239000010406 cathode material Substances 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 229910002804 graphite Inorganic materials 0.000 description 1
- 239000010439 graphite Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 229910052744 lithium Inorganic materials 0.000 description 1
- 238000010946 mechanistic model Methods 0.000 description 1
- 238000011056 performance test Methods 0.000 description 1
- BULVZWIRKLYCBC-UHFFFAOYSA-N phorate Chemical compound CCOP(=S)(OCC)SCSCC BULVZWIRKLYCBC-UHFFFAOYSA-N 0.000 description 1
- 238000004904 shortening Methods 0.000 description 1
- 238000007086 side reaction Methods 0.000 description 1
Images
Classifications
-
- 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]
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Tests Of Electric Status Of Batteries (AREA)
- Secondary Cells (AREA)
Abstract
The invention belongs to the field of power batteries, and discloses a method and a system for quickly testing the service life of a power battery, which comprise the following steps: acquiring a variable curve of a power battery to be tested, and acquiring a transverse variable sequence and a longitudinal variable sequence based on the acquired variable curve; fitting the independent variables to the battery cycle life based on a kernel canonical correlation analysis method by using the transverse and longitudinal variable sequences as the independent variables, and introducing a recursive kernel canonical correlation analysis method to obtain the maximum correlation coefficient between the independent variables and the battery cycle life; and testing the sequence corresponding to the battery cycle life with the maximum correlation coefficient higher than the threshold value by combining a machine learning model to obtain a power battery life test result. The invention avoids the requirement on prior knowledge, shows good effectiveness in the battery cycle life test and improves robustness compared with the traditional test method, and greatly shortens the test time.
Description
Technical Field
The invention belongs to the field of power batteries, and particularly relates to a method and a system for quickly testing the service life of a power battery.
Background
The battery life refers to the number of charge and discharge times when the available capacity falls to 80% of the initial value. The battery is subjected to cycle life testing, so that the characteristics of the battery can be further known, whether the battery reaches a design target or not is demonstrated, better management and control are realized in the process of using the battery, and the capacity of meeting the requirements of different application scenes of the battery is evaluated at the same time.
At present, the cycle life test of the lithium ion battery mainly adopts the national standard GB/T31484-2015 'requirement and test method for cycle life of power storage batteries for electric vehicles'. In this standard, when a standard cycle life test of a battery is performed, the battery needs to be continuously charged and discharged to the end-of-life condition, i.e., "once-test-to-end". The cycle times of charging and discharging of the lithium ion battery can reach thousands of times or even more than ten thousand times, the test time can reach 1 year or even longer, and the test usually consumes very much time. The prior art discloses a battery testing method: controlling the battery to perform N times of charging and discharging operations, wherein each time of charging and discharging operation comprises a plurality of times of charging operation and a plurality of times of discharging operation, the plurality of times of charging operation is performed based on at least two current specifications, and the plurality of times of discharging operation is performed based on at least two current specifications; after controlling the battery to carry out N times of charging and discharging operations, adding 1 to the test frequency record; and re-executing the step of determining the battery capacity value of the battery until the battery capacity value is smaller than the preset capacity value. The prior art discloses a battery testing method: acquiring a current charging and discharging curve, and dividing a voltage interval based on the current charging and discharging curve based on the upper and lower voltage limits of the battery; performing cycle test on the same cell sample in parallel in different voltage intervals, and correspondingly obtaining cycle attenuation of the different voltage intervals; and fusing the cyclic attenuation of the different voltage intervals to obtain the cyclic total attenuation. However, the above methods require passing full cycle, long, high cost tests to achieve battery cycle life.
A prediction-based method for rapidly testing battery life is an excellent solution to the above-mentioned problem, i.e. "to estimate generation testing". By replacing the full-period test with prediction, a large amount of time can be saved, and the efficient and accurate evaluation of the cycle life of the battery can be realized. The prior art discloses a prediction method, which comprises the steps of carrying out short-term cycle performance tests on batteries with different cycle times, recording the cycle times and the capacity retention rate, then disassembling the batteries with different cycle times, testing the graphitization degree of a graphite cathode material by using an X-ray diffraction method, and carrying out tests according to three data of the cycle times, the capacity retention rate and the graphitization degree. The method needs to disassemble and destroy the battery, and has certain limitation. The prior art discloses a method for testing the cycle life of a lithium ion battery, which comprises the steps of installing a pressure sensor on the surface of the battery, recording capacity information of the battery within a certain cycle number, and fitting according to the cycle number, the discharge capacity and voltage data in the pressure sensor to test the service life of the battery. The method needs an additional pressure sensor, actually increases the cost and has certain limitations. The prior art also discloses a method for rapidly testing the cycle life of lithium ions, which realizes the life test of a battery by placing the battery under different multiplying powers to carry out 500 times of cycle test and fitting an equation to obtain a test equation. The method does not need precise test equipment and complex theoretical calculation, but still needs long-time cycle test, and has low practicability. The battery life test method is beneficial to the production, optimization and development of the battery, but the power battery is a strong nonlinear system and is highly sensitive to the environment, and the cycle life test is extremely difficult. The new energy automobile industry requires that the power battery can quickly perform technical iteration and product upgrading, but the existing method for quickly testing the cycle life of the battery is lacked, so that the quick and high-quality development of the new energy automobile industry is severely restricted. Therefore, how to shorten the life test time, test and evaluate the cycle life of the battery quickly, accurately and efficiently becomes an important means for breaking through the bottleneck of the key technology for the quick development of the power battery and the related industries.
At present, many model test methods for testing the cycle life of the lithium battery exist, and the model test methods can be generally divided into a mechanism model, an experience model and a machine learning model. Many of the mechanistic models currently used for power battery life testing are pseudo-two-dimensional based electrochemical models that account for capacity degradation caused by side reactions, but they fail to account for battery inconsistencies. Unlike the mechanism model, the empirical model directly fits the capacity drop trajectory and ignores the internal mechanism of the battery; in order to update the model parameters according to the latest battery state and operating conditions, kalman Filter (KF) and local filter (PF) are widely used; however, since these empirical models fit the collected aging curves, they can usually only be applied under similar conditions, which means poor generalization ability. The machine learning model is tested by mapping a set of extracted battery features to battery cycle life based on a large amount of data. However, most machine learning models generally input fewer features, which may result in poor robustness in practical applications; furthermore, these features are typically concentrated at several points throughout the data set, with the remainder of the data set being underutilized, resulting in reduced accuracy; furthermore, manually finding a suitable set of data features often requires a great deal of effort.
Disclosure of Invention
In order to solve the technical problems of long period and high cost of the traditional testing method, the invention provides a method and a system for quickly testing the service life of a power battery, wherein a transverse variable sequence and a longitudinal variable sequence are obtained based on an obtained variable curve; the transverse variable sequence and the longitudinal variable sequence are used as independent variables, the independent variables are fitted to the cycle life of the battery based on a kernel canonical correlation analysis method, a recursive kernel canonical correlation analysis method is introduced to obtain a maximum correlation coefficient between the independent variables and the cycle life of the battery and a corresponding kernel correlation analysis solution, good effectiveness is shown in the test of the cycle life of the battery, and the test time is greatly shortened.
In order to achieve the purpose, the invention adopts the following technical scheme:
the invention provides a method for quickly testing the service life of a power battery, which adopts estimation to replace a method to be tested and comprises the following steps:
acquiring a variable curve of a power battery to be tested;
obtaining transverse and longitudinal variable sequences based on the obtained variable curves;
fitting the independent variable to the battery cycle life based on a kernel canonical correlation analysis method by taking the transverse and longitudinal variable sequences as the independent variable, and introducing a recursive kernel canonical correlation analysis method to obtain the maximum correlation coefficient between the independent variable and the battery cycle life;
and testing the sequence corresponding to the battery cycle life with the maximum correlation coefficient higher than the threshold value by combining a machine learning model to obtain a power battery life test result.
The second aspect of the present invention provides a system for rapidly testing the service life of a power battery, including:
the data acquisition module is used for acquiring a variable curve of the power battery to be tested;
the sequence construction module is used for obtaining transverse and longitudinal variable sequences based on the obtained variable curves;
the data fitting module is used for fitting the independent variables to the battery cycle life based on a kernel canonical correlation analysis method by taking the transverse and longitudinal variable sequences as the independent variables, and introducing a recursive kernel canonical correlation analysis method to obtain the maximum correlation coefficient between the independent variables and the battery cycle life;
and the battery life testing module is used for testing the sequence corresponding to the battery cycle life with the maximum correlation coefficient higher than the threshold value by combining the machine learning model to obtain a power battery life testing result.
A third aspect of the invention provides a computer-readable storage medium.
A computer-readable storage medium, on which a computer program is stored, which when executed by a processor implements the steps in a method for fast testing of power battery life as described above.
A fourth aspect of the invention provides a computer apparatus.
A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method for rapidly testing the service life of a power battery.
Compared with the prior art, the invention has the beneficial effects that:
(1) Compared with the traditional test method, the method of replacing the test by estimation is adopted, the method has good effectiveness in the battery cycle life test, the test time is greatly shortened, the life test can be completed by depending on accurate prediction, the life test time is greatly shortened, and the upgrading and updating of the battery are accelerated.
(2) The invention tests the cycle life by comparing the states of different batteries in the same cycle and predicts the cycle life by the change trend of different batteries at the same point, introduces a kernel typical correlation analysis method, solves the problem of overfitting by fitting a variable sequence to almost perfect cycle life, can be combined with the traditional characteristics, and shows good effectiveness and improves robustness in the battery cycle life test compared with the traditional test method.
(3) The method is based on a kernel canonical correlation analysis method to fit the independent variable to the battery cycle life, introduces the kernel canonical correlation analysis method to obtain the maximum correlation coefficient between the independent variable and the battery cycle life, and provides a recursive canonical correlation analysis solving algorithm to improve the solving operation speed; and testing the sequence corresponding to the battery cycle life with the maximum correlation coefficient higher than the threshold value by combining a machine learning model to obtain a power battery life test result, so that the requirement on prior knowledge is avoided.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.
Fig. 1 is a schematic overall flow chart of a method for rapidly testing the service life of a power battery according to an embodiment of the present invention;
fig. 2 is a schematic diagram of the transverse and longitudinal variable sequences of the battery according to the embodiment of the invention.
FIG. 3 is a test result obtained using the first 50 cycles of data for an embodiment of the present invention.
FIG. 4 is a test result obtained using the first 100 cycles of data for an embodiment of the present invention.
Detailed Description
The invention is further described with reference to the following figures and examples.
It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
The overall thought of the invention is as follows: first, a set of transverse sequences and a set of longitudinal sequences are derived from each selected battery variable, and these sequences are fitted to the battery cycle life by kernel canonical correlation analysis. The calculation load is reduced by using the typical correlation analysis of the recursive kernel; in addition, a recursive canonical correlation analysis solving algorithm is proposed to improve the solving operation speed. By means of the above analysis method, appropriate fitting results are generated as optional inputs to the machine learning model. And finally, a threshold value of the correlation coefficient is appointed, and high-quality machine learning algorithm input is selected.
Example one
As shown in fig. 1, the present embodiment provides a method for quickly testing the lifetime of a power battery, which uses estimation instead of a method to be tested, and includes the following steps:
s1: acquiring a variable curve of a power battery to be tested;
as one or more embodiments, in S1, the present embodiment selects a data set collected by the ministry of labor (MIT).
It should be noted that the number of batteries in the data set and the variable curve of the batteries selected in this embodiment may be set by those skilled in the art.
For example, the present embodiment selects 123 batteries in total, and 7 battery variable curves;
the variable curve of the power battery to be tested comprises: a discharge capacity-voltage curve (Q (V)), a discharge temperature-voltage curve (T (V)), a voltage increment-capacity (IC), a discharge voltage-time curve (V (T)), a discharge temperature-time curve (T))), a discharge voltage/current-time curve (V/I (T)), and a charging current interpolation curve (ICC) during the entire CV (abbreviation of constant voltage, i.e. constant voltage phase of charging) step.
Wherein the ICC uses linear interpolation for normalization.
S2: obtaining transverse and longitudinal variable sequences based on the obtained variable curves;
as shown in fig. 2, a sequence of battery variables that change in one period in the variable curve is used as a transverse variable sequence, and a sequence of battery variables that change at the same sampling point in different periods is used as a longitudinal variable sequence.
Different transverse variable sequence curves are more pronounced in later cycles and therefore generally perform better.
In contrast, in early cycles, the differences between cells were small and irregular, and the performance of the vertical sequence was related to the position of the sampling point, and generally performed better with more variable changes and regularity.
For example, based on the data obtained in S1, a set of sequences of horizontal and vertical variables is derived for each variable, i.e., a set of sequences of horizontal and vertical variables is derived from each selected variable.
In this way, a sequence of 14 variables can be generated as arguments for subsequent input using a recursive kernel-representative correlation analysis method.
S3: fitting the independent variables to the battery cycle life based on a kernel canonical correlation analysis method by using the transverse and longitudinal variable sequences as the independent variables, and introducing a recursive kernel canonical correlation analysis method to obtain the maximum correlation coefficient between the independent variables and the battery cycle life;
nuclear Canonical Correlation Analysis (KCCA) can test the cycle life of a battery by comparing the states of different batteries in the same cycle through a transverse variable sequence and comparing the variation trends of different batteries at the same point through a longitudinal variable sequence.
Since KCCA can fit variable sequences to almost perfect cycle life in the training step, the problem of overfitting must be addressed and hyper-parameters for which KCCA performs well on the training set may not be suitable for the test set.
For example, if a small η is simply chosen in the training step to make the correlation coefficient large, the fitting result may not be satisfactory in the test set.
To address this problem, the present embodiment creates a KCCA validation set to test and select the KCCA hyperparameters. Approximately the same number of cells in each batch were assigned to training, validation and test groups, taking into account the differences between calendar aging and rest time between batches.
As shown in fig. 2, the data set was divided into three groups, a KCCA training group, a KCCA validation group, and a KCCA test group, each of 41 cells, the KCCA training group, and the KCCA validation group constituting a machine learning training group.
The KCCA training group trains variable sequences by using different hyper-parameter sets, and the different hyper-parameter sets are effectively identified by a grid search method.
The embodiment proposes a Recursive KCCA (RKCCA) solving algorithm based on Kernel Canonical Correlation Analysis (KCCA) to facilitate the calculation.
Conventional Canonical Correlation Analysis (CCA) aims to find linear relationships between two sets of multidimensional variables.
Two sets of random variables, of the form (x, y),
mean value zero, assuming there is one power cell data instance S = ((x) 1 ，y 1 )，…，(x n ，y n ) X, y), wherein x represents a sequence consisting of a sequence of transverse and longitudinal variables of the power cell, i.e. an independent variable, and y represents the lifetime of the corresponding power cell, i.e. a dependent variable.
Wherein, X is represented as (X) 1 ，…，x n ) T And Y is represented by (Y) 1 ，…，y n ) T 。
Assume that n ≧ rank (X), rank (Y), and both X and Y have zero mean. A new vector w is defined, named load vector x and projects x into that direction, and the projection of y is obtained by selecting load vector v to do the same for y.
Projections t = Xw and u = Yv are defined as the score vectors of X and Y, respectively. CCA determines w and v to maximize the correlation coefficient between t = Xw and u = Yv.
In this embodiment, all | · | | | represent euclidean norms.
The method for fitting the independent variables to the battery cycle life based on the kernel canonical correlation analysis method by taking the transverse variable sequence and the longitudinal variable sequence as the independent variables specifically comprises the following steps:
s301: mapping the arguments to the high dimensional space using a kernel-functionalized solution includes:
each x is i Mapping to a high-dimensional vector space F with infinite dimensions:
wherein x is i Representing each element in the sequence of arguments, Φ (x) i ) Representing the result of mapping each element in the sequence of arguments to a high dimensional space;
Φ(X)＝(Φ(x 1 )，Φ(x 2 )，…，Φ(x n )) T (2)
where the mean value of Φ (X) should also be zero:
avoiding solving specific Φ (x) by applying kernel techniques i )：
K(X) i，j ＝Φ(x i )Φ(x j ) T ＝k(x i ，x j ) (4)
K(X)＝Φ(X)Φ(X) T (5)
Wherein the kernel function k (x) i ，x j ) Is a symmetric function. Thus, K (X) is a symmetric matrix, representing the argument kernel matrix.
S302: centralization
Usually k (x) i ，x j ) Cannot ensure phi (X)The mean is zero, so centering is required:
s303: calculating a projection of a sequence mapped to the high-dimensional space independent variable and the cycle life of the battery; and (4) constructing an optimization function by using the maximum correlation coefficient among the projections as an object.
Note that, for convenience, Φ and K are used instead of Φ (X) and K (X), respectively, in the rest of the present embodiment.
The optimization problem becomes to maximize the correlation coefficient ρ 1 ：
Where Φ and Φ (X) have the same meaning and represent an independent variable matrix mapped to a high-dimensional space, Y represents a dependent variable matrix, and w 1 Projection vector, v, representing an argument matrix mapped to a high dimensional space 1 A projection vector representing the dependent variable matrix.
Φ is usually full rank, so w i Is always phi (x) i ) Linear combination of (a):
w i ＝Φ T α i (8)
similar to CCA, one can obtain:
(K T K) -1 K T Y(Y T Y) -1 Y T Kα 1 ＝ρ 1 2 α 1 (9)
(Y T Y) -1 Y T K(K T K) -1 K T Yv 1 ＝ρ 1 2 v 1 (10)
however, the symmetry of K, which results in:
(K T K) -1 K T Y(Y T Y) -1 Y T K＝I (11)
where I is the identity matrix and all eigenvalues are equal to 1. At arbitrary selection of alpha i Time | ρ i Always, i | =1 holds, and perfect correlation can always be formed.
Regularization, eta alpha, is required in the kernel method 1 T Kα 1 The term is added to the optimization equation:
where η is a manually selected hyper-parameter, and is usually small. Alpha is alpha T K α is a term in the optimization problem of the Kernel Partial Least Squares (KPLS) method.
The regularization problem can be viewed as a transitional form of the non-regularization KCCA and KPLS.
After this step, it can be ensured that the equations are solvable and that the matrix eigenvalues are not always 1:
(K T K+η 1 I) -1 K T Y(Y T Y) -1 Y T Kα 1 ＝λ 1 2 α 1 ： (13)
(Y T Y) -1 Y T K(K T K+η 1 I) -1 K T Yv 1 ＝λ 1 2 v 1 (14)
it is emphasized that λ is changed since the constraint has been changed 1 2 And rho 1 2 Are not exactly the same, but their differences should be small.
Wherein, the step of obtaining the maximum correlation coefficient between the independent variable and the battery cycle life by introducing the recursive kernel canonical correlation analysis method comprises the following steps:
based on a power method, based on a given diagonalizable matrix and a random non-zero vector, an iterative solution method is adopted to find out the maximum eigenvalue of the diagonalizable matrix, and the maximum correlation coefficient corresponding to the maximum eigenvalue is obtained according to the relation between the eigenvalue and the correlation coefficient.
Similar to CCA, α and v are (K) T K+ηI) -1 K T Y(Y T Y) -1 Y T Kα 1 And (Y) T Y) -1 Y T K(K T K+η 1 I) -1 K T Yv 1 Of the corresponding feature vector.
Using the power-law to accomplish this task is a better approach because only the maximum eigenvalue needs to be found.
Given a diagonalizable matrix D and a random non-zero vector b 0 When k → ∞ is satisfied, the eigenvector b corresponding to the maximum eigenvalue of D k This can be obtained by the following recursive relationship:
the relationship between α and v can be found:
v∝(YY T ) -1 YK T α (16)
α∝(KK T +ηI) -1 KY T v (17)
in this embodiment, according to the property of a typical correlation analysis CCA, the relationship between the characteristic value and the correlation coefficient is: the square of the eigenvalues is the corresponding correlation coefficient.
For this feature, the present embodiment finds the maximum eigenvalue of the diagonalizable matrix by using an iterative solution method based on the given diagonalizable matrix and the random non-zero vector, including:
(1) A random non-zero vector is selected as the initial value of alpha, usually the first column of K
(normalization to the weight vector alpha of the kernel matrix);
the following iterations are then performed:
(2) Score vector for Φ: t = K α, and t = K α,
(score vector of argument matrix mapped to high dimensional space at convergence timet is the product of the independent variable kernel matrix K and the weight vector alpha, normalized to t)
(3) Y load vector: v = (Y) T Y) -1 Y T t，
(the relationship between the load vector v of the dependent variable at the time of convergence, the dependent variable matrix Y, and the score vector t of the independent variable matrix mapped to the high-dimensional space, v is normalized)
(4) Score vector for Y: u = Yv; (the score vector u of the dependent variable at convergence is the product of the dependent variable matrix Y and the load vector v of the dependent variable)
(5) Weight vector: α = (KK) T +ηI) -1 KY T v，
(the relationship between the weight vector alpha of the kernel matrix at the time of convergence and the load vectors v of the kernel matrix, the dependent variable matrix Y and the dependent variable is normalized by the weight vector alpha of the kernel matrix)
Through the method, the embodiment can find the maximum eigenvalue and the corresponding eigenvector in a more convenient way at the cost of small precision loss instead of performing complete eigenvector decomposition on the two matrixes.
In the test algorithm of KCCA, x is considered as an independent variable and y is taken as a dependent variable. The goal is to use the new given argument X according to the previously captured relationships new Test Y new 。X new Kernel matrix K of new The pretreatment was as follows:
K new ＝Φ(X new )Φ(X) T (18)
then centralizing K new ：
By assuming X new And Y new Is the same to test Y new 。X new The scoring matrix of (a) is:
T new ＝K new A (20)
will T new Is regarded as Y new The score matrix of (c):
estimated Y new ：
S4: and testing the sequence corresponding to the battery cycle life with the maximum correlation coefficient higher than the threshold value by combining a machine learning model to obtain a power battery life test result.
The embodiment considers that the quality of the data input by the test model is more important than the quantity, and the performance of the deep learning method is superior to that of the traditional machine learning methods (GPR and Elastic net).
In this embodiment, the sequences corresponding to the fitting results of these correlation coefficients above the threshold were tested using 4 representative machine learning models, which include Artificial Neural Network (ANN), random Forest (RF), gaussian Process Regression (GPR), and Elastic network (Elastic net).
Using the first 100 cycles of data, the best Mean Absolute Percent Error (MAPE) for ANN, RF, GPR, and Elastic net were 8.4%, 8.6%, 11.1%, and 14.0%, respectively, and using the first 50 cycles of data, the best results ranged from 9.5% to 14.9%, with RF achieving the best performance. The performance of ANN and RF was verified to be superior to the traditional machine learning algorithms (GPR and Elastic networks) by the above experiments.
For the setting of the threshold, it is not always better that the higher correlation coefficient threshold is, especially for the deep learning method, the lower threshold means that more fitting results can be input into the machine learning model, which also facilitates the testing.
The setting of the threshold value is therefore set according to the actually selected machine learning model.
For example, in general for ANN, MAPE decreases first and then increases as the threshold decreases, with a minimum occurring when the threshold is set to 0.8, and a balance exists between the number and quality of these inputs. For GPR and Elastic net, their best performance occurs when the threshold is high.
To verify the effectiveness of the method of this example, experiments were performed:
in table 1, experimental results based on the MIT data set are shown, and fig. 3 is a test result obtained using the first 50-cycle data according to an embodiment of the present invention; FIG. 4 is a test result obtained using the first 100 cycles of data for an embodiment of the present invention.
The results of the experiments were evaluated by introducing MAPE, which is the mean absolute percent error, and RMSE, which is the root mean square error.
The final experimental result shows that compared with the traditional full life cycle test which needs 2000 times, the invention can complete the life test by the accurate test with little precision, only 50 to 300 times of circulation is needed to complete the estimation, the comprehensive test is not needed, 1950 test circulation is shortened, the speed is improved by 40 times, the life test time can be greatly reduced, and the upgrading and updating of the battery are accelerated.
TABLE 1 results of the experiment
| Number of cycles | MAPE(％) | RMSE (circulation) | Shortening the number of cycles |
| 50(2.5％) | 9.1 | 122 | 1950 |
| 100(5％) | 8.2 | 98 | 1900 |
| 200(10％) | 8.0 | 93 | 1800 |
| 300(15％) | 7.7 | 89 | 1700 |
Example two
The embodiment provides a quick test system of power battery life, includes:
the data acquisition module is used for acquiring a variable curve of the power battery to be tested;
the sequence construction module is used for obtaining transverse and longitudinal variable sequences based on the obtained variable curves;
the data fitting module is used for fitting the independent variables to the battery cycle life based on a kernel canonical correlation analysis method by taking the transverse and longitudinal variable sequences as the independent variables, and introducing a recursive kernel canonical correlation analysis method to obtain the maximum correlation coefficient between the independent variables and the battery cycle life;
and the battery life testing module is used for testing the sequence corresponding to the battery cycle life with the maximum correlation coefficient higher than the threshold value by combining the machine learning model to obtain a power battery life testing result.
EXAMPLE III
The embodiment provides a computer readable storage medium, on which a computer program is stored, which when executed by a processor implements the steps in the method for rapidly testing the service life of a power battery as described above.
Example four
The embodiment provides a computer device, which includes a memory, a processor and a computer program stored in the memory and capable of running on the processor, and when the processor executes the program, the steps in the method for rapidly testing the service life of a power battery are implemented.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The present invention has been described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A method for rapidly testing the service life of a power battery is characterized in that a method for replacing a method to be tested by estimation comprises the following steps:
acquiring a variable curve of a power battery to be tested;
obtaining transverse and longitudinal variable sequences based on the obtained variable curves;
fitting the independent variables to the battery cycle life based on a kernel canonical correlation analysis method by using the transverse and longitudinal variable sequences as the independent variables, and introducing a recursive kernel canonical correlation analysis method to obtain the maximum correlation coefficient between the independent variables and the battery cycle life;
and testing the sequence corresponding to the battery cycle life with the maximum correlation coefficient higher than the threshold value by combining a machine learning model to obtain a power battery life test result.
2. The method for rapidly testing the service life of the power battery as claimed in claim 1, wherein the variable curves of the power battery to be tested comprise a discharge capacity-voltage curve, a discharge temperature-voltage curve, a voltage increment-capacity, a discharge voltage-time curve, a discharge temperature-time curve, a discharge voltage/current-time curve and a charge current interpolation curve during the whole constant voltage stage step of charging.
3. The method for rapidly testing the service life of the power battery as claimed in claim 1, wherein the obtaining of the transverse and longitudinal variable sequences based on the obtained variable curve is as follows:
and taking a sequence formed by the battery variables changing in one period in the variable curve as a transverse variable sequence, and taking a sequence formed by the battery variables changing at the same sampling point in different periods as a longitudinal variable sequence.
4. The method for rapidly testing the service life of the power battery as claimed in claim 1, wherein the step of fitting the independent variables to the cycle life of the power battery based on a kernel canonical correlation analysis method by using the transverse and longitudinal variable sequences as the independent variables comprises the following steps:
mapping a sequence corresponding to the independent variable to a high-dimensional space by adopting a kernel functional solution method;
calculating a projection of a sequence mapped to the high-dimensional space independent variable and the cycle life of the battery;
and (4) constructing an optimization function by using the maximum correlation coefficient among the projections as an object.
5. The method for rapidly testing the service life of the power battery as claimed in claim 1, wherein when the optimization function is constructed, terms in the optimization problem of the hyperparameter and the kernel partial least squares method are added to the objective function through regularization.
6. The method for rapidly testing the service life of the power battery as claimed in claim 1, wherein the step of introducing a recursive kernel canonical correlation analysis method to obtain the maximum correlation coefficient between the independent variable and the battery cycle life comprises the following steps: based on a power method, based on a given diagonalizable matrix and a random non-zero vector, an iterative solution method is adopted to find out the maximum eigenvalue of the diagonalizable matrix, and the maximum correlation coefficient corresponding to the maximum eigenvalue is obtained according to the relation between the eigenvalue and the correlation coefficient.
7. A power battery life rapid test system is characterized by comprising:
the data acquisition module is used for acquiring a variable curve of the power battery to be tested;
the sequence construction module is used for obtaining transverse and longitudinal variable sequences based on the obtained variable curves;
the data fitting module is used for fitting the independent variables to the cycle life of the battery based on a kernel canonical correlation analysis method by taking the transverse and longitudinal variable sequences as the independent variables, and introducing a recursive kernel canonical correlation analysis method to obtain the maximum correlation coefficient between the independent variables and the cycle life of the battery;
and the battery life testing module is used for testing the sequence corresponding to the battery cycle life with the maximum correlation coefficient higher than the threshold value by combining the machine learning model to obtain a power battery life testing result.
8. The method for rapidly testing the service life of the power battery as claimed in claim 7, wherein the step of fitting the independent variables to the cycle life of the power battery based on a kernel canonical correlation analysis method by using the transverse and longitudinal variable sequences as the independent variables comprises the following steps:
mapping a sequence corresponding to the independent variable to a high-dimensional space by adopting a kernel functional solution method;
calculating a projection of a sequence mapped to the high-dimensional space independent variable and the cycle life of the battery;
and (4) constructing an optimization function by using the maximum correlation coefficient among the projections as an object.
Or the like, or, alternatively,
the method for obtaining the maximum correlation coefficient between the independent variable and the battery cycle life by introducing the recursive kernel canonical correlation analysis method comprises the following steps: based on a power method, based on a given diagonalizable matrix and a random non-zero vector, an iterative solution method is adopted to find out the maximum eigenvalue of the diagonalizable matrix, and the maximum correlation coefficient corresponding to the maximum eigenvalue is obtained according to the relation between the eigenvalue and the correlation coefficient.
9. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the steps of a method for rapid testing of the lifetime of a power battery according to any one of claims 1 to 7.
10. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to implement the steps of a method for rapid testing of power battery life as claimed in any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210789672.XA CN115219902B (en) | 2022-07-06 | 2022-07-06 | Rapid power battery life test method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210789672.XA CN115219902B (en) | 2022-07-06 | 2022-07-06 | Rapid power battery life test method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115219902A true CN115219902A (en) | 2022-10-21 |
| CN115219902B CN115219902B (en) | 2024-08-16 |
Family
ID=83610385
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210789672.XA Active CN115219902B (en) | 2022-07-06 | 2022-07-06 | Rapid power battery life test method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115219902B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115825755A (en) * | 2022-12-30 | 2023-03-21 | 哈尔滨光宇新能源有限公司 | Method for evaluating voltage consistency of battery core of energy storage battery |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050134282A1 (en) * | 2003-12-19 | 2005-06-23 | Moshe Averbuch | [method and apparatus for battery testing and measuring] |
| US20050288666A1 (en) * | 2001-12-04 | 2005-12-29 | Bertolero Arthur A | Cardiac ablation devices and methods |
| CN102778653A (en) * | 2012-06-20 | 2012-11-14 | 哈尔滨工业大学 | Data-driven lithium ion battery cycle life prediction method based on AR (Autoregressive) model and RPF (Regularized Particle Filtering) algorithm |
| CN103954913A (en) * | 2014-05-05 | 2014-07-30 | 哈尔滨工业大学深圳研究生院 | Predication method of electric vehicle power battery service life |
| US20150066406A1 (en) * | 2013-08-27 | 2015-03-05 | The Regents Of The University Of Michigan | On-board state of health monitoring of batteries using incremental capacity analysis |
| CN107024663A (en) * | 2017-04-01 | 2017-08-08 | 湖南银杏数据科技有限公司 | The lithium battery screening technique clustered based on charging curve feature KPCA |
| CN107728072A (en) * | 2017-10-10 | 2018-02-23 | 合肥国轩高科动力能源有限公司 | Method for rapidly predicting cycle life of lithium ion battery |
| CN107942261A (en) * | 2017-12-29 | 2018-04-20 | 上海电气集团股份有限公司 | The method of estimation and system of battery charge state |
| US20200064406A1 (en) * | 2016-11-04 | 2020-02-27 | Avl List Gmbh | Method for diagnosing a technical system |
| GB202017682D0 (en) * | 2020-11-09 | 2020-12-23 | Horiba Mira Ltd | Battery performance optimisation |
| CN113533988A (en) * | 2021-06-04 | 2021-10-22 | 上海空间电源研究所 | Long-term circulation capacity attenuation analysis method for lithium ion battery |
| WO2022019607A1 (en) * | 2020-07-21 | 2022-01-27 | 삼성전자 주식회사 | Method for identifying battery state, electronic device therefor, and storage medium therefor |
-
2022
- 2022-07-06 CN CN202210789672.XA patent/CN115219902B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050288666A1 (en) * | 2001-12-04 | 2005-12-29 | Bertolero Arthur A | Cardiac ablation devices and methods |
| US20050134282A1 (en) * | 2003-12-19 | 2005-06-23 | Moshe Averbuch | [method and apparatus for battery testing and measuring] |
| CN102778653A (en) * | 2012-06-20 | 2012-11-14 | 哈尔滨工业大学 | Data-driven lithium ion battery cycle life prediction method based on AR (Autoregressive) model and RPF (Regularized Particle Filtering) algorithm |
| US20150066406A1 (en) * | 2013-08-27 | 2015-03-05 | The Regents Of The University Of Michigan | On-board state of health monitoring of batteries using incremental capacity analysis |
| CN103954913A (en) * | 2014-05-05 | 2014-07-30 | 哈尔滨工业大学深圳研究生院 | Predication method of electric vehicle power battery service life |
| US20200064406A1 (en) * | 2016-11-04 | 2020-02-27 | Avl List Gmbh | Method for diagnosing a technical system |
| CN107024663A (en) * | 2017-04-01 | 2017-08-08 | 湖南银杏数据科技有限公司 | The lithium battery screening technique clustered based on charging curve feature KPCA |
| CN107728072A (en) * | 2017-10-10 | 2018-02-23 | 合肥国轩高科动力能源有限公司 | Method for rapidly predicting cycle life of lithium ion battery |
| CN107942261A (en) * | 2017-12-29 | 2018-04-20 | 上海电气集团股份有限公司 | The method of estimation and system of battery charge state |
| WO2022019607A1 (en) * | 2020-07-21 | 2022-01-27 | 삼성전자 주식회사 | Method for identifying battery state, electronic device therefor, and storage medium therefor |
| GB202017682D0 (en) * | 2020-11-09 | 2020-12-23 | Horiba Mira Ltd | Battery performance optimisation |
| CN113533988A (en) * | 2021-06-04 | 2021-10-22 | 上海空间电源研究所 | Long-term circulation capacity attenuation analysis method for lithium ion battery |
Non-Patent Citations (3)
| Title |
|---|
| GUANGCAI ZHAO: "Capacity Prediction and Remaining Useful Life Diagnosis of Lithium-ion Batteries Using CNN-LSTM Hybrid Neural Network", 《2021 CHINA AUTOMATION CONGRESS (CAC)》, 24 October 2021 (2021-10-24) * |
| XUEBIN CAO: "A Fast Capacity Estimation Approach for Retired Lithium-ion Batteries", 《2021 40TH CHINESE CONTROL CONFERENCE (CCC)》, 28 July 2021 (2021-07-28) * |
| 王竹晴: "基于KPCA-ANFIS 的锂离子电池ＲUL 预测", 《电子测量与仪器学报》, 31 October 2018 (2018-10-31) * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115825755A (en) * | 2022-12-30 | 2023-03-21 | 哈尔滨光宇新能源有限公司 | Method for evaluating voltage consistency of battery core of energy storage battery |
| CN115825755B (en) * | 2022-12-30 | 2023-09-19 | 哈尔滨昆宇新能源有限公司 | Method for evaluating consistency of voltages of battery cells of energy storage battery |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115219902B (en) | 2024-08-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109991542B (en) | Lithium ion battery residual life prediction method based on WDE optimization LSTM network | |
| Zhang et al. | Long short-term memory recurrent neural network for remaining useful life prediction of lithium-ion batteries | |
| Oji et al. | Data-driven methods for battery soh estimation: Survey and a critical analysis | |
| Zhou et al. | A fast screening framework for second-life batteries based on an improved bisecting K-means algorithm combined with fast pulse test | |
| CN113361197B (en) | Method and system for predicting remaining service life of lithium battery | |
| CN106446940A (en) | Support vector machine (SVM) based prediction method for the degradation tendency of super-capacitor capacitance value | |
| JP2023501995A (en) | Battery performance prediction | |
| CN113189490B (en) | Lithium battery health state estimation method based on feature screening and Gaussian process regression | |
| CN109917292A (en) | A kind of lithium ion battery life-span prediction method based on DAUPF | |
| CN113985294A (en) | Method and device for estimating remaining life of battery | |
| Wu et al. | Optimized multi-source fusion based state of health estimation for lithium-ion battery in fast charge applications | |
| CN115201686B (en) | Lithium ion battery health state assessment method under incomplete charge and discharge data | |
| Wei et al. | State of health and remaining useful life prediction of lithium-ion batteries with conditional graph convolutional network | |
| CN113721149A (en) | Lithium battery capacity prediction method based on semi-supervised transfer learning | |
| CN115219902B (en) | Rapid power battery life test method and system | |
| CN115356647A (en) | Method for early predicting service life of lithium ion battery | |
| CN116401770A (en) | Quick charge strategy design method based on battery digital twin model and machine learning | |
| Cai et al. | Prediction of lithium-ion battery remaining useful life based on hybrid data-driven method with optimized parameter | |
| Schmitt et al. | Effective estimation of battery state-of-health by virtual experiments via transfer-and meta-learning | |
| CN117688364A (en) | Battery remaining service life prediction method, system, equipment and medium | |
| Abdelghafar et al. | Satellite lithium-ion battery remaining useful life estimation by coyote optimization algorithm | |
| CN116224070A (en) | OCV (open circuit voltage) -SOC (system on a chip) model-based full-temperature lithium ion battery OCV (open circuit voltage) evaluation method and system | |
| Liu et al. | An end-to-end learning framework for early prediction of battery capacity trajectory | |
| Zhao et al. | Lithium‐Ion Battery State‐of‐Health Estimation Method Using Isobaric Energy Analysis and PSO‐LSTM | |
| CN114594397A (en) | Lithium ion battery health state estimation method and system based on generalization model |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |