CN116060325A - Method for rapidly sorting consistency of power batteries - Google Patents
Method for rapidly sorting consistency of power batteries Download PDFInfo
- Publication number
- CN116060325A CN116060325A CN202310121286.8A CN202310121286A CN116060325A CN 116060325 A CN116060325 A CN 116060325A CN 202310121286 A CN202310121286 A CN 202310121286A CN 116060325 A CN116060325 A CN 116060325A
- Authority
- CN
- China
- Prior art keywords
- battery
- sorting
- data
- points
- center
- 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
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B07—SEPARATING SOLIDS FROM SOLIDS; SORTING
- B07C—POSTAL SORTING; SORTING INDIVIDUAL ARTICLES, OR BULK MATERIAL FIT TO BE SORTED PIECE-MEAL, e.g. BY PICKING
- B07C5/00—Sorting according to a characteristic or feature of the articles or material being sorted, e.g. by control effected by devices which detect or measure such characteristic or feature; Sorting by manually actuated devices, e.g. switches
- B07C5/34—Sorting according to other particular properties
- B07C5/342—Sorting according to other particular properties according to optical properties, e.g. colour
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
- Y02E60/10—Energy storage using batteries
Abstract
The invention discloses a method for rapidly sorting consistency of power batteries, which is used for detecting the power batteries by an alternating current impedance test method and reducing the dimension of battery EIS test data so as to extract characteristic quantities for reflecting battery performances and facilitate the acceleration of sorting speed. In the battery sorting problem, a self-adaptive clustering sorting method is provided, wherein the method uses the data density to acquire the clustering number and the initial value of a clustering center so as to realize battery pre-sorting, and then optimizes the sorting result according to the distance from a data point to the clustering center so as to realize further sorting. The existing method often needs a large number of battery samples for pre-training or needs to determine the number of battery clusters in advance for sorting, has low test efficiency, and is not suitable for sorting large-scale batteries. When the method is used for sorting batteries, the method has high sorting speed, the batteries in the same group have high consistency, and the method has high applicability to different types of batteries.
Description
Technical Field
The invention belongs to the technical field of electric automobiles and energy storage, and particularly relates to a power battery consistency rapid sorting method.
Background
In 2020, the accumulated decommissioning total of power cells in China reaches about 20 ten thousand tons, and by 2025, the accumulated decommissioning total is increased to about 78 ten thousand tons. Under such a background, retired batteries of electric vehicles may provide considerable economic benefits through secondary use of energy storage and the like. However, screening and reorganization of large-scale power cells faces problems of low efficiency and low accuracy.
The existing methods for detecting and sorting the consistency of the power battery mainly comprise four methods: (1) Whether the battery can be reused is determined by detecting the appearance (e.g., swelling and dripping), weight, size, and tightness of the scrapped battery. (2) The constant-current charge and discharge testing method is used for detecting and obtaining the capacity, internal resistance and other information of the power battery one by one, however, the method is long in time consumption and is not suitable for large-scale battery sorting. (3) Sorting is performed based on a battery test curve, for example, EIS, capacity Increment (IC) curve, pulse curve, etc. are used as the basis for sorting the batteries, but the test curve of the batteries has a great amount of data redundancy, which reduces the sorting speed of the data. (4) Machine learning methods, including support vector machines and Artificial Neural Networks (ANNs), can model multivariate problems of complex systems and extract implicit nonlinear relationships between variables, but neural networks require large amounts of training data and are therefore unsuitable for sample-less battery sorting. In battery sorting, the sorting number is a difficulty, and too large a sorting number results in too slow sorting speed, and too small a sorting number results in poor consistency of the sorted battery cells.
The existing battery sorting method is low in test speed and not suitable for sorting large-scale batteries, and the method is beneficial to accelerating the sorting speed by extracting characteristic quantities related to battery performance in an EIS curve as sorting basis of the batteries based on battery EIS test data; when in cluster sorting, the data point density is utilized to pre-sort, the clustering number and the center of the batteries are preliminarily determined, the optimization of the battery sorting result is realized by adjusting the distance between the data point and the center point, the self-adaptive selection of the battery sorting number of different types is realized, the pre-training is not required for the test data of the battery sample to be sorted, the sorting speed is increased, the applicability of the sorting method to various batteries is improved, and the consistency of the batteries in the sorted group is ensured.
Disclosure of Invention
In order to solve the technical problems, the invention provides a power battery consistency rapid sorting method, which has the advantages that due to the differences in battery production process or use environment after assembly and the like, performance indexes such as battery internal resistance, capacity, open-circuit voltage, coulomb efficiency and the like are different, and by utilizing the power battery sorting method, battery groups with similar performances can be obtained through screening, and the service life maximization and use safety of the battery during reapplication are ensured.
The invention adopts the following technical scheme for solving the technical problems:
a power battery consistency rapid sorting method comprises the following steps:
s1, testing a battery sample to be tested by using an alternating current impedance method to obtain an electrochemical impedance spectrum data set of a battery of the battery sample to be tested, and obtaining an electrochemical impedance spectrum curve of the battery;
s2, performing data dimension reduction on an electrochemical impedance spectrum data set of a battery sample to be tested, wherein the dimension-reduced data represents the battery to be tested;
s3, pre-sorting the batteries based on the dimension reduced data, acquiring the clustering number and the clustering center by using the data intensity as an index according to a self-adaptive clustering algorithm, and carrying out self-adaptive updating on the intensity along with iteration;
s4, further sorting the batteries, and adjusting the clustering center by calculating the distance between the data point and the clustering center.
Further, the specific process of obtaining the electrochemical impedance spectrum curve of the battery in S1 is as follows:
step (1), applying excitation current signals with different frequencies to the battery, wherein the excitation current value is 1/20 of the rated capacity value of the battery, and synchronously sampling a voltage signal U (t) and a current signal I (t) of the battery;
step (2), extracting the amplitude values of a voltage signal U (t) and a current signal I (t) of the battery under different frequencies through Fourier decomposition;
wherein ω is an angular frequency corresponding to the frequency f,
the phase is corresponding to the voltage and the current; n represents the frequency belonging to the nth implant, t represents the time interval;
step (3) calculating the amplitude Z (n) and phase of the battery impedance under the excitation signal according to the voltage signal U (t) and the current signal I (t) of the battery
And (4) changing the frequency of the excitation signal, repeating the step (2) and the step (3), and fitting an electrochemical impedance spectrum curve of the battery through the impedance of the battery under different frequencies.
Further, in the step S2, the data dimension reduction is to sort and weight the feature values based on feature value decomposition to reduce the number of feature values, which specifically includes:
the number of samples of the battery is m, each battery is tested for n frequencies, and the corresponding electrochemical impedance spectrum curve correspondingly contains n points, so that the original data of the battery is expressed as:
X n,m ＝(x 1 ,…,x m ) (16)
wherein X represents the set of raw data of the whole battery, X m Representing the column vector formed by the test data of the mth cell.
Construction of X n,m Is a covariance matrix B of (1):
performing feature decomposition on the covariance matrix B to obtain a feature value lambda of the covariance matrix i And corresponding feature vector v i The method comprises the steps of carrying out a first treatment on the surface of the By combining the feature vectors v i Arranging the characteristic values into a matrix according to the size of the corresponding characteristic values from left to right in descending order of columns, and taking the first q columns to form a matrix T q，m ；
Contribution degree R of eigenvalue i The definition adopted is:
wherein lambda is i For matrix T q，m Is a characteristic value of (2);
in order to determine the selected feature quantity, the generation of high-dimensional data is prevented from affecting the subsequent sorting, and the processing is performed in the following manner:
when the contribution degree of the characteristic values exceeds 1%, selecting the current characteristic values as characteristic values of subsequent sorting;
for eigenvalues with contribution between 0.5% and 1%, assuming that there are e numbers, the weighted eigenvalues θ are calculated as:
reducing the characteristic value of low contribution degree to 1;
for the characteristic value with the contribution degree lower than 0.5%, discarding the corresponding characteristic quantity, and considering the characteristic value as noise;
through the processing, if the number of the selected characteristic quantities is p, most of information of the original data set can be considered to be covered;
data set X p，m Is used for sorting of subsequent batteries.
Further, the pre-selection of S3 obtains the number of clusters and the cluster center by using the data density, which specifically includes:
step (1) first, a data set X is formed based on the extracted feature quantity p，m Calculating a data intensity range r for subsequent data intensity calculation, and calculating the maximum value of the distance between any two points:
wherein x is i And x j Representing vectors corresponding to data points corresponding to the ith and j batteries 2 Representing the 2 norms of the vector, namely, the sum of squares of all elements in the vector is further rooted, and the superscript 2 represents squaring the calculated result.
To calculate the intensity of these points, the influence range r of each point is defined as:
r＝δd max (21)
wherein δ represents a desired degree of consistency, ranging from 0 to 1;
step (2) calculating an ith data point intensity level:
step (3) finding the data point with the highest data point density as the first clustering center c 1 This data point is then removedAnd (3) influencing the range, namely, calculating the points in the influence range by the formula (21), repeating the step (1) and the step (2), and dynamically adjusting the data range while searching for a new cluster;
step (4) sequentially circulating until the new density degree is delta of the previous density degree 2 At times delta 2 The value is 0 to 1;
and (5) obtaining an initial value of the clustering number and an initial position of the clustering center according to the classification condition at the termination.
Further, the further sorting in S4 is achieved by adjusting the distance between the data point and the clustering center, and the specific process is as follows:
step (1), initializing: taking the number of clusters and the cluster center obtained in the pre-sorting process as initialization parameters;
clustering samples: the distance of each sample to each cluster center is calculated, each sample is categorized into the class in which the cluster center closest to it is located:
D ki ＝||c i -x j || 2 (23)
wherein D is ki Representing the distance between the data point and the center point, c i Represents the vector corresponding to the ith center point, x j Representing the vector corresponding to the data point corresponding to the jth cell, I 2 Representing the 2-norm of the vector, i.e., the sum of the squares of the individual elements in the vector, and the root number.
Step (3), when the distances from the data point to the two center points are equal, distributing the points to clusters with more current clusters; distance d from the point to each center point ci When the following conditions are satisfied:
i.e. d ci The method is characterized in that the method is larger than half of the maximum distance between any two points in a data set, the points are considered to be outlier points, and the outlier points are used as independent clustering centers so as to accelerate the sorting speed; in which x is i And x j Representing the i and j-th battery data correspondence vectors。
Step (4) calculating a new cluster center: calculating the central positions of all data points in each cluster as a new cluster center c newi ：
Wherein k is the number of data points contained in the cluster;
and (5) when the change of the clustering center is small, the distance between two points is as follows:
when the distance between the two points is smaller than the minimum distance between any two points in the data set, the clustering process is considered to be ended, otherwise, the step (2) and the step (3) are repeated;
and (6) sorting the m batteries into g clusters, and finishing sorting.
According to the rapid evaluation method for the consistency of the power battery, the feature value is obtained by reducing the dimension of EIS curve data of the power battery to be evaluated, the battery is pre-sorted by utilizing the data density degree, then the clustering center is adjusted by adjusting the distance between the data point and the clustering center, the battery is further sorted, and the consistency of the battery in the sorted group is improved.
Compared with the prior art, the invention has the beneficial effects that:
according to the technical scheme, the battery is tested by using the alternating current impedance spectrum, and the characteristic quantity of the battery is extracted from the EIS curve, so that the battery is more rapid compared with a conventional testing method.
According to the technical scheme, the EIS data is subjected to data dimension reduction, consistency evaluation is converted into the clustering problem of the dimension reduced data, and the influence caused by data redundancy is reduced.
According to the technical scheme, sorting is performed according to the data characteristics of the test data, the method is not dependent on a specific battery model, and is suitable for sorting of various types of batteries, the test period is short, and the working efficiency is high.
According to the technical scheme, the batteries are pre-sorted by adopting the data density degree, and then the distance between the data points and the center point is adjusted according to the obtained battery clustering number and the center point, so that the clustering number can be selected in a self-adaptive mode, and the consistency sorting of power batteries with different number scales is realized.
Drawings
FIG. 1 is a flow chart of a method of rapid assessment of power cell consistency in accordance with the present invention;
FIG. 2 is a flow chart of a data dimension reduction method of an embodiment;
FIG. 3 is a flow chart of a battery cluster sorting algorithm of an embodiment;
FIG. 4 is the battery sorting result of the example; wherein, (a) shows the center points obtained by pre-sorting, 5 center points are obtained by calculation, (b) shows the optimization of the positions of the center points after further sorting, (c) shows the pre-sorting result, and (d) shows the result after further sorting.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
In order to improve reliability and rapidity of power battery consistency sorting, the invention provides a rapid power battery consistency assessment method, which is a power battery consistency assessment method based on a principal component analysis method and a clustering algorithm, and the assessment method of the embodiment of the invention is described below with reference to the accompanying drawings:
as shown in fig. 1, the method for quickly sorting consistency of power batteries according to the embodiment of the invention comprises the following steps:
s1, carrying out alternating current impedance test on a battery pack to be tested to obtain a battery Electrochemical Impedance Spectroscopy (EIS) data set, wherein the acquisition mode specifically comprises the following steps:
step (1) applying excitation current signals with different frequencies to the battery, wherein the excitation current value is 1/20 of the rated capacity value of the battery, and synchronously sampling a voltage signal U (t) and a current signal I (t) of the battery;
step (2) extracting to obtain the amplitude values of a battery voltage signal U (t) and a current signal I (t) at the frequency f through Fourier decomposition;
wherein ω is an angular frequency corresponding to the frequency f,
is the phase corresponding to the voltage and the current. n represents the frequency belonging to the nth implant, t represents the time interval;
step (3) according to the voltage signal U (t) and the current signal I (t) of the battery, calculating to obtain the amplitude Z (n) and the phase of the battery impedance under the excitation signal
And (4) changing the frequency of the excitation signal, repeating the step (2) and the step (3), and fitting an electrochemical impedance spectrum curve of the battery through the impedance of the battery at different frequencies.
S2, performing data dimension reduction processing on sample data of the battery pack to be tested to obtain a new battery characteristic set;
because the data obtained by the test is high-dimensional data and contains a large amount of redundant information, the sorting algorithm is not easy to carry out, and the data is subjected to characteristic extraction by adopting a data dimension reduction method, and fig. 2 is a flow chart of the data dimension reduction method in the embodiment, and the specific process is as follows:
the number of samples of the battery composition is m, and the EIS curve of each battery correspondingly contains n points, so the raw data of the battery is expressed as:
X n,m ＝(x 1 ,…,x m ) (16)
where X represents the set of raw data of the entire battery, and xm represents the column vector formed by the test data of the mth battery.
Construction of X n,m Is a covariance matrix B of (1):
performing feature decomposition on the covariance matrix B to obtain a feature value lambda of the covariance matrix i And corresponding feature vector v i . By combining the feature vectors v i Arranging the characteristic values into a matrix according to the size of the corresponding characteristic values from left to right in descending order of columns, and taking the first q columns to form a matrix T q，m 。
The contribution degree of the feature value is generally defined as follows:
lambda in i For matrix T q，m Is a characteristic value of (a).
In order to determine the number of selected feature quantities, to prevent the generation of high-dimensional data affecting the subsequent sorting, the processing is performed in the following manner:
when the contribution degree of the characteristic values exceeds 1%, selecting the current characteristic values as characteristic values of subsequent sorting;
for eigenvalues with contribution between 0.5% and 1%, assuming the number of them is e, calculating weighted eigenvalues:
the feature value of low contribution is reduced to 1.
For feature values with a contribution of less than 0.5%, the corresponding feature values are discarded, and are considered to be noise.
Through the above processing, the number of selected feature amounts is p, and it is considered that most of the information of the original data set can be covered.
Data set X p，m Will be used for sorting of subsequent cells.
S3, calculating the data density of each point through the dimension-reduced data set, selecting a center point with high density as a center point, and obtaining a cluster center initial point and a cluster number of sorting, wherein the method comprises the following steps:
the battery to be measured is represented by the data after dimension reduction, and sorting is performed, and fig. 3 is a flowchart of a battery clustering sorting algorithm in an embodiment, wherein the clustering algorithm is used for pre-sorting the battery based on the data point density, then further sorting, and the power battery sorting method specifically comprises the following steps:
because the data belongs to the unmarked state when the clustering is carried out, the battery sorting is converted into the unmarked clustering, each data point is used as a potential clustering center, the function of the completed clustering center is subtracted, the clustering center is searched again, and the clustering number and the initial value of the clustering center are provided for the subsequent further sorting, and the specific process is as follows:
first, a data set X composed according to the extracted feature quantity p，m Calculating a data intensity range r for subsequent data intensity calculation, and calculating the maximum value of the distance between any two points:
wherein x is i And x j Representing vectors corresponding to data points corresponding to the ith and j batteries 2 Representing the 2 norms of the vector, namely, the sum of squares of all elements in the vector is further rooted, and the superscript 2 represents squaring the calculated result.
To calculate the intensity of these points, define the impact range of each point:
r＝δd max (21)
where δ represents the desired degree of consistency, ranging from 0 to 1, with smaller values being the higher the degree of consistency, but with a corresponding slower sorting speed.
Next, the i-th data point intensity level is calculated:
finding the data point with the highest data point density as the first clustering center c 1 And then removing the data points and the influence range, calculating the points of the influence range by the formula (21), calculating a new influence range and the degree of density, and dynamically adjusting the data range while searching for a new cluster.
Then, the steps are sequentially circulated until the new density degree is delta of the last density degree 2 At times delta 2 The smaller the value of 0 to 1, the more classification numbers will be generated, but the sorting speed will be lowered, typically 0.5.
And finally, obtaining an initial value of the clustering number and an initial position of a clustering center according to the classification condition at the termination.
S4, calculating the distance between each point and the central point, optimizing the sorting result according to the distance, and adjusting the clustering central position and the clustered batteries, wherein the further sorting shown in the figure 3 comprises the following steps:
because the center obtained by sorting is always the data point in the original data set and is not necessarily the most suitable clustering center, a new clustering center is generated by adjusting according to the distance from the data point to the clustering center, and the battery is further sorted, and the specific process is as follows:
initializing: according to the number of clusters obtained in the pre-sorting and the initial center point set of the clusters, the number of clusters and the initial center point set of the clusters are used as initialization parameters;
clustering samples: the distance from each sample to each cluster center is calculated, and each sample is classified into the class in which the cluster center closest to the sample is located.
D ki ＝|x i -c i | (23)
Wherein D is ki Representing the distance between the data point and the center point, c i Represents the vector corresponding to the ith center point, x j Representing the vector corresponding to the data point corresponding to the jth cell, I 2 Representing the 2-norm of the vector, i.e., the sum of the squares of the individual elements in the vector, and the root number.
Data point allocation: when the distances from the data point to the two center points are equal, the points are distributed to clusters with more current clusters; distance d from the point to each center point ci The following conditions are satisfied:
i.e. d ci And the point is considered to be an outlier point which is more than half of the maximum distance between any two points in the data set, and the outlier point is taken as a clustering center so as to accelerate the sorting speed. In which x is i And x j Representing the i and j-th battery data correspondence vectors.
Calculating a new cluster center: calculating the central positions of all data points in each cluster as a new cluster center c newi 。
Where k is the number of data points contained in the cluster.
When the distance change between the new cluster center and the old cluster center is smaller:
i.e. when the distance between the two is smaller than the minimum distance between any two points in the data set, the clustering algorithm is stopped, otherwise, the sample is continuously clustered and a new clustering center is calculated. In c newi And c oldi Respectively represent new and old central points, D li Represents the distance between the new and old center points, d limit Representing the minimum distance between any two points within the dataset.
Based on the method, the sorting result is adjusted, so that the clustering accuracy is improved, as shown in fig. 4, which is a battery sorting result of a certain group of battery data, 3 feature quantities are obtained through data dimension reduction extraction, corresponding 3-dimensional coordinates are drawn, the center points obtained through pre-sorting are shown in fig. 4 (a), 5 center points are obtained through calculation, and the optimization of the positions of the center points after further sorting is shown in fig. 4 (b). Fig. 4 (c) and fig. 4 (d) are the results after pre-sorting and further sorting, respectively, the differently shaped points represent the clusters belonging to different groups, the outliers are effectively identified and individually sorted, and the sorting result of the groups is optimized, as compared to fig. 4 (c).
After the battery sorting is performed by using the clustering algorithm, the performance difference of the batteries sorted into the same cluster is small, and the batteries in the group have good consistency.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the invention.
Claims (5)
1. The power battery consistency rapid sorting method is characterized by comprising the following steps of:
s1, testing a battery sample to be tested by using an alternating current impedance method to obtain an electrochemical impedance spectrum data set of a battery of the battery sample to be tested, and obtaining an electrochemical impedance spectrum curve of the battery;
s2, performing data dimension reduction on an electrochemical impedance spectrum data set of a battery sample to be tested, wherein the dimension-reduced data represents the battery to be tested;
s3, pre-sorting the batteries based on the dimension reduced data, acquiring the clustering number and the clustering center by using the data intensity as an index according to a self-adaptive clustering algorithm, and carrying out self-adaptive updating on the intensity along with iteration;
s4, further sorting the batteries, and adjusting the clustering center by calculating the distance between the data point and the clustering center.
2. The method for uniform and rapid sorting of power cells according to claim 1, wherein in S1, the specific process of obtaining the electrochemical impedance spectrum curve of the cells is as follows:
step (1), applying excitation current signals with different frequencies to the battery, wherein the excitation current value is 1/20 of the rated capacity value of the battery, and synchronously sampling a voltage signal U (t) and a current signal I (t) of the battery;
step (2), extracting the amplitude values of a voltage signal U (t) and a current signal I (t) of the battery under different frequencies through Fourier decomposition;
wherein ω is an angular frequency corresponding to the frequency f,
the phase is corresponding to the voltage and the current; n represents the frequency belonging to the nth implant, t represents the time interval;
step (3) calculating the amplitude Z (n) and phase of the battery impedance under the excitation signal according to the voltage signal U (t) and the current signal I (t) of the battery
And (4) changing the frequency of the excitation signal, repeating the step (2) and the step (3), and fitting an electrochemical impedance spectrum curve of the battery through the impedance of the battery under different frequencies.
3. The method for quickly sorting the consistency of the power battery according to claim 2, wherein in the step S2, the data dimension reduction is to sort and weight the feature values based on feature value decomposition so as to reduce the number of the feature values, and the specific process is as follows:
the number of samples of the battery is m, each battery is tested for n frequencies, and the corresponding electrochemical impedance spectrum curve correspondingly contains n points, so that the original data of the battery is expressed as:
X n,m ＝(x 1 ,…,x m ) (16)
wherein X represents the set of raw data of the whole battery, X m A column vector formed by test data representing the mth cell;
construction of X n,m Is a covariance matrix B of (1):
performing feature decomposition on the covariance matrix B to obtain a feature value lambda of the covariance matrix i And corresponding feature vector v i The method comprises the steps of carrying out a first treatment on the surface of the By combining the feature vectors v i Arranging the characteristic values into a matrix according to the size of the corresponding characteristic values from left to right in descending order of columns, and taking the first q columns to form a matrix T q，m ；
Contribution degree R of eigenvalue i The definition adopted is:
lambda in i For matrix T q，m Is a characteristic value of (2);
in order to determine the selected feature quantity, the generation of high-dimensional data is prevented from affecting the subsequent sorting, and the processing is performed in the following manner:
when the contribution degree of the characteristic values exceeds 1%, selecting the current characteristic values as characteristic values of subsequent sorting;
for eigenvalues with contribution between 0.5% and 1%, assuming that there are e numbers, the weighted eigenvalues θ are calculated as:
reducing the characteristic value of low contribution degree to 1;
for the characteristic value with the contribution degree lower than 0.5%, discarding the corresponding characteristic quantity, and considering the characteristic value as noise;
through the processing, if the number of the selected characteristic quantities is p, most of information of the original data set can be considered to be covered;
data set X p，m Is used for sorting of subsequent batteries.
4. The rapid sorting method for consistency of power batteries according to claim 3, wherein the pre-sorting of S3 uses data intensity to obtain the number of clusters and the cluster center, and the specific process is as follows:
step (1) first, a data set X is formed based on the extracted feature quantity p，m Calculating a data intensity range r for subsequent data intensity calculation, and calculating the maximum value of the distance between any two points:
wherein x is i And x j Representing vectors corresponding to data points corresponding to the ith and j batteries 2 Representing 2 norms of the vector, namely, summing squares of all elements in the vector and then opening root numbers, wherein 2 is the square of the calculated result;
to calculate the intensity of these points, the influence range r of each point is defined as:
r＝δd max (21)
wherein δ represents a desired degree of consistency, ranging from 0 to 1;
step (2) calculating an ith data point intensity level:
step (3) finding the data point with the highest data point density as the first clustering center c 1 This data point is then removed andthe influence range, the point in the influence range calculated by the formula (21), repeating the step (1) and the step (2), and dynamically adjusting the data range while searching for a new cluster;
step (4) sequentially circulating until the new density degree is delta of the previous density degree 2 At times delta 2 The value is 0 to 1;
and (5) obtaining an initial value of the clustering number and an initial position of the clustering center according to the classification condition at the termination.
5. The method for rapid sorting of consistency of power cells according to claim 4, wherein the further sorting in S4 is achieved by adjusting the distance between the data points and the clustering center, specifically comprising the following steps:
step (1), initializing: taking the number of clusters and the cluster center obtained in the pre-sorting process as initialization parameters;
clustering samples: calculating the distance from each sample to each cluster center, and classifying each sample into the class of the cluster center closest to the sample:
D ki ＝||c i -x j || 2 (23)
d in ki Representing the distance between the data point and the center point, c i Represents the vector corresponding to the ith center point, x j Representing the vector corresponding to the data point corresponding to the jth cell, I 2 Representing the 2 norms of the vector, namely, the sum of squares of all elements in the vector is further root-signed;
step (3), when the distances from the data point to the two center points are equal, distributing the points to clusters with more current clusters; distance d from the point to each center point ci When the following conditions are satisfied:
i.e. d ci Greater than half the maximum distance between any two points in the dataset, which are considered outliers, which are taken asIs an independent clustering center to accelerate the sorting speed; in which x is i And x j Representing the i and j-th battery data corresponding vectors;
step (4) calculating a new cluster center: calculating the central positions of all data points in each cluster as a new cluster center c newi ：
Wherein k is the number of data points contained in the cluster;
and (5) when the change of the clustering center is small, the distance between two points is as follows:
when the distance between the two points is smaller than the minimum distance between any two points in the data set, the clustering process is considered to be ended, otherwise, the step (2) and the step (3) are repeated; in c newi And c oldi Respectively represent new and old central points, D li Represents the distance between the new and old center points, d limit Representing a minimum distance between any two points within the dataset;
and (6) sorting the m batteries into g clusters, and finishing sorting.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310121286.8A CN116060325A (en) | 2023-02-16 | 2023-02-16 | Method for rapidly sorting consistency of power batteries |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310121286.8A CN116060325A (en) | 2023-02-16 | 2023-02-16 | Method for rapidly sorting consistency of power batteries |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116060325A true CN116060325A (en) | 2023-05-05 |
Family
ID=86176722
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310121286.8A Pending CN116060325A (en) | 2023-02-16 | 2023-02-16 | Method for rapidly sorting consistency of power batteries |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116060325A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117706377A (en) * | 2024-02-05 | 2024-03-15 | 国网上海能源互联网研究院有限公司 | Battery inconsistency identification method and device based on self-adaptive clustering |
| CN117706377B (en) * | 2024-02-05 | 2024-05-14 | 国网上海能源互联网研究院有限公司 | Battery inconsistency identification method and device based on self-adaptive clustering |
-
2023
- 2023-02-16 CN CN202310121286.8A patent/CN116060325A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117706377A (en) * | 2024-02-05 | 2024-03-15 | 国网上海能源互联网研究院有限公司 | Battery inconsistency identification method and device based on self-adaptive clustering |
| CN117706377B (en) * | 2024-02-05 | 2024-05-14 | 国网上海能源互联网研究院有限公司 | Battery inconsistency identification method and device based on self-adaptive clustering |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110752410B (en) | Method for rapidly sorting and recombining retired lithium battery | |
| CN112465040B (en) | Software defect prediction method based on class unbalance learning algorithm | |
| CN108655028A (en) | A kind of method and system classified to battery based on Fuzzy Mean Clustering Algorithm | |
| CN108363810A (en) | A kind of file classification method and device | |
| CN110503245A (en) | A kind of prediction technique of air station flight large area risk of time delay | |
| CN112287980B (en) | Power battery screening method based on typical feature vector | |
| CN112791997B (en) | Method for cascade utilization and screening of retired battery | |
| CN110610225A (en) | Multi-objective particle swarm optimization algorithm based on kriging proxy model plus-point strategy | |
| CN106778846A (en) | A kind of method for forecasting based on SVMs | |
| CN110738232A (en) | grid voltage out-of-limit cause diagnosis method based on data mining technology | |
| CN114564982A (en) | Automatic identification method for radar signal modulation type | |
| CN112305441B (en) | Power battery health state assessment method under integrated clustering | |
| CN112633337A (en) | Unbalanced data processing method based on clustering and boundary points | |
| CN111814826B (en) | Rapid detection and rating method for residual energy of retired power battery | |
| CN115829105A (en) | Photovoltaic power prediction method based on historical data feature search | |
| CN110659682A (en) | Data classification method based on MCWD-KSMOTE-AdaBoost-DenseNet algorithm | |
| CN110263834A (en) | A kind of detection method of new energy power quality exceptional value | |
| Wang et al. | Otsu multi-threshold image segmentation algorithm based on improved particle swarm optimization | |
| CN116060325A (en) | Method for rapidly sorting consistency of power batteries | |
| CN111797899A (en) | Low-voltage transformer area kmeans clustering method and system | |
| CN115454990A (en) | Oil paper insulation data cleaning method based on improved KNN | |
| CN113033683B (en) | Industrial system working condition monitoring method and system based on static and dynamic joint analysis | |
| CN112485694B (en) | Battery pack detection method and device | |
| CN112348700B (en) | Line capacity prediction method combining SOM clustering and IFOU equation | |
| CN110647671A (en) | Data stream classification algorithm based on AAE-DWMIL-LearnNSE |
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 |