CN117102082A - Sorting method and system for liquid metal batteries - Google Patents

Sorting method and system for liquid metal batteries Download PDF

Info

Publication number
CN117102082A
CN117102082A CN202310793055.1A CN202310793055A CN117102082A CN 117102082 A CN117102082 A CN 117102082A CN 202310793055 A CN202310793055 A CN 202310793055A CN 117102082 A CN117102082 A CN 117102082A
Authority
CN
China
Prior art keywords
sorting
sample
point
battery
curve
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202310793055.1A
Other languages
Chinese (zh)
Inventor
王康丽
张娥
蒋凯
李浩秒
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huazhong University of Science and Technology
Original Assignee
Huazhong University of Science and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Huazhong University of Science and Technology filed Critical Huazhong University of Science and Technology
Priority to CN202310793055.1A priority Critical patent/CN117102082A/en
Publication of CN117102082A publication Critical patent/CN117102082A/en
Pending legal-status Critical Current

Links

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
    • B07C5/344—Sorting according to other particular properties according to electric or electromagnetic properties
    • 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

Landscapes

  • Secondary Cells (AREA)

Abstract

The invention discloses a method and a system for sorting liquid metal batteries, and belongs to the technical field of secondary battery application. Reconstructing the sampled discharge voltage data of the liquid metal battery to obtain a smooth voltage curve; identifying the inflection point of the voltage curve to generate a curve characterization index; repeated mapping screening is carried out on the curve characterization indexes to form sorting indexes; and filtering out sample outliers in the sorting indexes by using an improved DBSCAN clustering algorithm, and optimally dividing the rest sample space by using a Mean Shift algorithm to obtain a battery sorting result. The improved DBSCAN clustering algorithm is utilized to filter out the sample outliers, and then the Mean Shift algorithm is adopted to optimally divide the rest sample space, so that the combination algorithm can achieve the purposes of outlier detection and clustering at the same time, and the sorting precision is improved.

Description

Sorting method and system for liquid metal batteries
Technical Field
The invention belongs to the technical field of secondary battery application, and particularly relates to a method and a system for sorting liquid metal batteries.
Background
The liquid metal battery is used as a novel electrochemical energy storage technology, and the three-layer liquid structure has the advantages of large capacity, high efficiency, long service life, low cost and the like, and can effectively solve the problems of short service life and low reliability faced by the current electrochemical energy storage technology. In practical application, a large number of single batteries need to be used in series-parallel connection in groups so as to meet the voltage and capacity grade requirements of a large-scale energy storage system. However, due to the difference of the production and manufacturing processes, inconsistency exists among the single batteries inevitably, and specifically refers to inconsistency of parameters such as internal resistance, capacity and the like of the batteries. In addition, in the use process of the battery pack, the inconsistency of each single battery can be further amplified due to the difference of application environments, so that battery aging is caused, the capacity utilization rate and the service life of the whole battery pack can be limited by the single battery with quicker performance degradation, even the single battery is more likely to fail, the whole battery pack cannot work normally, and safety accidents can be caused under extreme conditions. In theory, although the liquid metal battery cannot have a malignant safety accident caused by the fault of the lithium ion battery due to the special structural design and material selection, the failure of the single battery can affect the utilization rate of the whole battery pack system. Therefore, a battery sorting method facing system application needs to be developed, scientific and efficient integration of batteries before grouping is realized, and long-term guarantee of economy of a battery system is facilitated.
In chinese patent specification CN202110581420.3, a sorting method for monitoring consistency of discharge end capacity of a battery is disclosed, and battery sorting is performed on the battery according to a characteristic element, so that the sorting speed is fast, but the accuracy is low. A retired battery sorting method is disclosed in chinese patent specification CN 201811475223.8. According to the method, open-circuit voltage, fixed-frequency internal resistance, discharge energy, direct-current internal resistance and charge-discharge energy efficiency parameters of retired batteries are respectively obtained through a plurality of test working conditions, and then are respectively compared with corresponding thresholds, so that batteries which are not in the threshold range are removed, and battery sorting is achieved. According to the method, 5 battery parameters are comprehensively considered, the reliability of the sorting result is high, but the battery test working conditions are more, and the time consumption is long. In the method, three battery sorting parameters are extracted through electrochemical impedance spectrum test, relaxation time analysis and grey correlation reconstruction of the lithium ion battery, and the parameters can reflect dynamic change of the battery, so that the battery sorting accuracy is further improved, but the process of extracting the sorting parameters is complex, and the battery sorting efficiency is affected. Therefore, a method that can perform simple and easy operation, rapid measurement or estimation on the dynamic parameters is needed to improve the classification efficiency while further constructing the dynamic parameters that can reflect the battery state to improve the classification accuracy. In addition, the method is complete clustering, and when obvious outliers exist in battery sorting parameters, the anti-interference capability of a sorting algorithm on the outliers is weak.
Because of the defects and shortcomings, the field needs to be further improved and perfected, a targeted screening method for battery parameter consistency is developed according to the characteristics of the liquid metal battery, a rapid screening method based on a liquid metal battery voltage curve is designed, the problem of low battery sorting efficiency is effectively solved, meanwhile, the battery sorting reliability is improved, and therefore reliable and safe operation of the liquid metal battery pack is guaranteed.
Disclosure of Invention
Aiming at the defects of the prior art, the invention aims to provide a method and a system for sorting liquid metal batteries, which aim to solve the problem of low efficiency of the conventional multi-parameter-based battery sorting technology.
In order to achieve the above purpose, the invention provides a rapid battery sorting method based on a liquid metal battery discharge curve, which is characterized in that characteristic parameters representing curve changes are extracted from the battery discharge curve to form battery sorting indexes, and then an improved Density-based noise application spatial clustering (Density-Based Spatial Clustering ofApplications with Noise, DBSCAN) +mean Shift (Mean Shift) combined clustering optimization algorithm is utilized to simultaneously realize battery outlier detection and self-adaptive spatial clustering division of sorting indexes, so that battery sorting efficiency and accuracy are improved.
In order to achieve the above object, according to one aspect of the present invention, there is provided a sorting method of liquid metal batteries, comprising the steps of:
reconstructing the sampled discharge voltage data of the liquid metal battery to obtain a smooth voltage curve;
identifying the inflection point of the voltage curve to generate a curve characterization index;
repeated mapping screening is carried out on the curve characterization indexes to form sorting indexes;
and filtering out sample outliers in the sorting indexes by using an improved DBSCAN clustering algorithm, and optimally dividing the rest sample space by using a Mean Shift algorithm to obtain a battery sorting result.
The invention firstly provides the battery discharging curve characteristic extraction frame, and characteristic parameters which can effectively reflect curve shape differences are extracted on the discharging curve, so that the test process of obtaining battery parameters is avoided, and the test time is saved. The characteristic parameters can provide reference basis for the subsequent battery sorting.
Preferably, the feature extraction framework comprises four phases: data acquisition, data preprocessing, curve characterization index generation and battery sorting index generation.
The specific method is realized by the following steps:
(1) Respectively collecting discharge data of the liquid metal battery to be selected under the multiplying power of 0.2C, wherein the sampling time interval is 30s, and forming an original data set;
(2) Smoothing each group of data in the original data set;
preferably, in the constant current discharge mode, the terminal voltage of the liquid metal battery should theoretically satisfy the following conditions:
wherein the method comprises the steps ofI.e. the battery is at t k The terminal voltage at the moment in time, where k represents the kth sample point.
Under the influence of noise collected by measuring equipment, the actually measured terminal voltage possibly does not meet the condition of the formula (1), and the curve needs to be subjected to smoothing treatment by reconstructing discharge voltage data, preferably, the invention adopts a linear interpolation method to reconstruct the voltage curve, and the steps are as follows:
a. markingSample points: the detection is performed sequentially from the first sampling point. If a sample point meets the condition of equation (1), the sample point will be marked as normal (noted as) Otherwise, will be marked as an outlier (noted +.>) And the outlier must conform to formula (2);
b. reconstructing outliers: if the outlier satisfies equation (3), reconstruct the outlier according to equation (4) (denoted as) The method comprises the steps of carrying out a first treatment on the surface of the If the abnormal point satisfies the formula (5), reconstructing j-1 points after the point is started according to the formula (6), wherein the j-th point satisfies the formula (7):
wherein the method comprises the steps ofNamely, the battery terminal voltage is +.>Corresponding time.
c. And after reconstruction, the abnormal points are updated into normal values in time, all sampling points are traversed, and reconstruction is completed.
(3) Identifying curve inflection points and generating curve characterization indexes;
the liquid metal battery discharge curve has two obvious voltage inflection points, which are one of important characteristics reflecting the internal state of the battery, and therefore can be used for representing the battery curve. Preferably, the invention adopts a sliding window algorithm to respectively identify two curve inflection points, and the voltage signal can be approximately seen as linear change in a smaller time range, so that the invention uses an autoregressive model to model the voltage signal, and the related loss function is based on least squares residual error:
wherein the method comprises the steps ofThe predicted value of the voltage at time t. Assuming a window width of 2ω, the loss function of the first window signal can be expressed as:
the loss function for the second window is:
the window total loss function is:
the window signal difference can be obtained according to equations (9), (10), (11):
Z(V tk-ω...tk+ω )＝c(V tk-ω...tk+ω )-c(V tk-ω...tk )-c(V tk...tk+ω ) (12)
preferably, the difference curve obtained by identifying the first inflection point is defined as Z high The difference curve obtained by identifying the inflection point of the second curve is defined as Z low The moment when the peak value of the difference curve appears corresponds to the moment when the voltage appears at the inflection point, so that the moment corresponding to the two inflection points can be obtained:
wherein t is high Indicating the moment corresponding to the first inflection point, t low The time corresponding to the second inflection point is represented, and then the inflection point voltage value can be obtained:
according to the characteristics of the liquid metal battery voltage curve, preferably, the invention extracts the following characteristic values as curve characterization indexes:
wherein F represents a characteristic index of the first x, wherein F1-F4 characterize two voltage inflection point locations and F5-F7 can be used to characterize the local shape of the curve.
(4) Generating a sorting index.
A total of 7 feature indices were extracted herein. However, there may be duplicate mapping information between different features, resulting in redundancy of information, which in turn affects the effectiveness of the sorting algorithm. Therefore, it is desirable to optimize important features among the above-described multi-dimensional features so that both feature dimensions are reduced and most of the useful information is retained. Preferably, the invention adopts the Spearman correlation coefficient matrix to preferably select a new characteristic matrix with less characteristics representing most of information of the original characteristic matrix, thereby forming a sorting index. The method comprises the following steps:
a. constructing a multidimensional feature quantity matrix A, A= [ F ] 1 ,F 2 ,…F n ]N is the number of features, n=7. Wherein F is n The characteristic vector matrix is m rows and 1 columns, and m is the number of battery samples;
b. the data in the matrix A are ordered according to columns, and the data positions after the ordering are recorded by the matrix P, which is expressed as follows:
c. calculating the correlation coefficient according to the formula (18) to obtain the correlation between the feature vectors:
wherein ρ is ij Is the correlation coefficient between the ith feature vector and the jth feature vector.
d. And (3) comparing the correlation among the feature vectors, and taking the feature vector with lower correlation as a sorting index (marked as S), wherein the dimension of the sorting index is less than or equal to 3.
The invention also provides a sorting algorithm based on the combination algorithm.
Preferably, the invention adopts an improved DBSCAN+mean Shift algorithm to sort batteries, and the specific implementation steps are as follows:
(5) Standardized treatment sorting indexes: in order to avoid the influence of the difference on the clustering result, the sample space of the sorting index is firstly standardized when the sample space is constructed, and preferably, zero-mean normalization is adopted, and the formula is as follows:
where x represents the sorting index and μ and σ are the mean and variance of each sorting index.
(6) Determining optimal DBSCAN algorithm neighborhood parameters (Eps, minPts) by using an elbow detection method, wherein Eps is a neighborhood distance threshold, minPts is a neighborhood sample number threshold, and an initial value is taken as a sorting index dimension plus 1;
1) The k-distance is calculated. Firstly, calculating the distance between each sample point and the rest sample points, wherein the calculation formula is as follows:
wherein r is the sorting index sequence number, and q is the sorting index dimension. And then will beSequencing from small to large, wherein the sequenced distance set is as follows
D(i)＝{d(i,1),d(i,2),…,d(i,k),…d(i,m)} (21)
D (i, k) is calledWhere k=minpts, the initial value of MinPts is set to h+1 in the present invention.
2) Obtaining a radius Eps: drawing the k-distances of all points on a k-distance graph in ascending order, and taking the maximum curvature point of Eps;
(7) DBSCAN algorithm clustering
1) Initializing parameters: core object collectionCluster number a=0; the set of unvisited samples h=e; cocooning frame partition set->
2) Traversing all sample points in the set E to obtain sample pointsIs-domain subsampled set:
the number of elements in the sub-sample set is denoted as b, if b is more than or equal to MinPts, the number of elements in the sub-sample set is more than or equal to MinPtsAdding the core object set omega;
3) In the core object set Ω, a set element o is randomly selected, and parameters are updated:
based on the core object point, finding an Eps-neighborhood sub-sample set N through a neighborhood distance threshold Eps (o) put it into the temporary point set Z;
4) Taking out a core object o' from the temporary point set Z, and finding out all Eps-neighborhood sub-sample sets N through a neighborhood distance threshold value Eps (o') updating
5) Repeating the above 4) until the temporary point set z=Φ, and then clustering the cluster P a After the generation is finished, updating cluster division groups:
P＝(P 1 ，P 2 ，…，P a ) (25)
marking non-core object points in the temporary point set Z as boundary points;
6) The above 3) -5) process is repeated until the core object set is an empty set. Wherein, the element in the set (E-P) is marked as noise point, and the cluster number a is output;
7) If a is 1, skipping to the step 8), otherwise updating MinPts, returning to the step (2), wherein the updating formula of MinPts is as follows:
MinPts＝MinPts+1 (26)
8) Outputting a first clustering result, wherein the number of samples in the set P is recorded as N;
(8) Mean Shift algorithm clustering
1) Calculating the mean shift M of each sample point in the set P according to equation (21) h (x)：
Wherein the method comprises the steps ofH is the bandwidth, which is the kernel function;
2) Each sample point is translated, i.e. along M h (x) And (3) direction movement:
x i ＝x i +M h (x i ) (28)
3) Repeating steps 1) -2) until the sample point converges, namely:
M h (x i )＝0 (29)
4) Sample points that converge to the same point are considered to be the same cluster class.
5) And outputting a second clustering result, namely a battery sorting result.
The invention also provides a sorting system of the liquid metal battery, which comprises the following steps: a computer readable storage medium and a processor;
the computer-readable storage medium is for storing executable instructions;
the processor is used for reading executable instructions stored in the computer readable storage medium and executing the sorting method of the liquid metal battery.
Compared with the prior art, the technical scheme of the invention can achieve the following beneficial results:
1. the invention provides a method for extracting characteristics of a discharge curve of a liquid metal battery, which is characterized in that voltage data is subjected to smoothing treatment by a linear difference method to avoid the influence of acquisition noise on the later characteristic extraction, the obvious voltage inflection point position of the liquid metal battery is identified by a sliding window algorithm, 7 characteristic indexes capable of representing the discharge curve of the battery are extracted based on the voltage inflection point, 3 key indexes are extracted by correlation analysis and are used for battery sorting indexes, and the index set can be directly applied to the later battery sorting work.
2. The invention provides a combined sorting algorithm capable of simultaneously realizing outlier detection and battery clustering, which utilizes an improved DBSCAN clustering algorithm to filter sample outliers, and then adopts a Mean Shift algorithm to optimally divide the rest sample space, so that the combined algorithm can simultaneously achieve the purposes of outlier detection and battery clustering, thereby improving sorting precision.
3. The rapid battery screening method based on the voltage discharge curve does not need to accurately measure battery parameters, can realize rapid separation of liquid metal batteries only by means of the battery discharge curve, and is simple in operation, short in testing time and convenient in data processing. The method is a rapid and highly reliable battery sorting method by combining the characteristics of the liquid metal battery and the characteristics of the application environment.
Drawings
FIG. 1 is a flow chart of a construction method for battery sorting based on a discharge voltage curve provided by the invention;
FIG. 2 is a graph of the voltage of a liquid metal cell according to an embodiment of the present invention; wherein, (a) is an original voltage curve of the liquid metal battery and a voltage curve of the liquid metal battery after data reconstruction; (b) 212 voltage curves after data reconstruction;
FIG. 3 is a graph of the identification result of inflection points provided by the embodiment of the present invention; (a) is the first inflection point identification result of the discharge curve; (b) a second inflection point identification of the discharge curve;
FIG. 4 is a matrix of characteristic index Spearman correlation coefficients according to an embodiment of the present invention;
FIG. 5 is a sample space diagram of each sorting index distribution diagram and its constitution provided by the embodiment of the present invention; wherein (a), (b) and (c) are scatter distribution diagrams after standardized treatment of the sorting index S1, the sorting index S2 and the sorting index S3 respectively; (d) a sample space formed by the three sorting indexes;
FIG. 6 is a graph showing the result of the algorithm clustering provided by the embodiment of the invention; wherein (a) is a DBSCAN algorithm output cluster result; (b) outputting a clustering result by a Mean Shift algorithm;
fig. 7 is a diagram of a sorting result of a liquid metal battery according to an embodiment of the present invention; wherein, (a) is a discharge curve graph of the separated group single battery; (b) Is cluster C 1 A corresponding battery voltage curve; (c) Is cluster C 2 A corresponding battery voltage curve; (d) Is cluster C 3 A corresponding battery voltage curve; (e) Is cluster C 4 A corresponding battery voltage curve; (f) Is cluster C 5 Corresponding battery voltage curves.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. 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 addition, the technical features of the embodiments of the present invention described below may be combined with each other as long as they do not interfere with each other.
The invention provides a sorting method of liquid metal batteries, which comprises the following steps:
reconstructing the sampled discharge voltage data of the liquid metal battery to obtain a smooth voltage curve;
identifying the inflection point of the voltage curve to generate a curve characterization index;
repeated mapping screening is carried out on the curve characterization indexes to form sorting indexes;
and filtering out sample outliers in the sorting indexes by using an improved DBSCAN clustering algorithm, and optimally dividing the rest sample space by using a Mean Shift algorithm to obtain a battery sorting result.
In one embodiment of the invention, the battery used is a laboratory-prepared 200Ah grade Li Bi liquid metal battery.
Fig. 1 is a flowchart of a battery sorting method provided by the invention, which mainly comprises the following steps:
(1) Obtaining battery data;
after the activation of the liquid metal batteries is completed, the original data set can be obtained by only testing the discharge curve of each liquid metal battery under the multiplying power of 0.2C. In the embodiment, 212 200 Ah-level Li Bi liquid metal battery voltage curves are measured in total;
(2) Smoothing the discharge data and identifying the curve inflection point of the discharge data;
(3) Generating a sorting index set;
(4) Optimizing neighborhood parameters of a DBSCAN algorithm, clustering a sample space formed by a sorting index set by using the DBSCAN algorithm, and marking outlier sample points;
(5) And removing outlier sample points, and re-clustering the sample space by using a Mean Shift algorithm to obtain a battery sorting result.
Fig. 2 (a) is an original voltage curve of a liquid metal battery and a voltage curve of the liquid metal battery after data reconstruction, where table 1 shows related information of the first 6 sampling points in the enlarged view, including 3 normal sampling points, 3 abnormal sampling points and corresponding reconstruction values. The first two outliers meet the condition of equation (5), the reconstructed value is processed according to equation (7), and the third outlier meets the condition of equation (3), the reconstructed value is processed according to equation (4). Fig. 2 (b) is a graph of discharge voltage of 212 liquid metal batteries subjected to smoothing treatment according to the present embodiment. As can be seen from the graph, the difference in the curve shape between the batteries is remarkable, reflecting that the difference in the battery parameters is large.
TABLE 1 sample Point voltage reconstruction values
Fig. 3 is a graph of the identification result of the inflection point provided in the present embodiment, and it can be seen from the graph that the sliding window algorithm used can accurately identify the location of the inflection point of the voltage. Based on the voltage inflection point locations, a characteristic index characterizing the curve shape may be generated.
Fig. 4 is a characteristic index Spearman correlation coefficient matrix provided in this embodiment. The correlation coefficient matrix is a symmetric matrix, the rows and columns describe the same series of feature quantities, and a higher correlation of one or more features indicates that the features have similar variation trends and possibly contain similar information. In general, when the correlation coefficient is smaller than 0.20, the variables are considered to have very weak correlation or no correlation. If the correlation coefficient of less than 0.20 in a certain column exceeds 3 (50% of the number of feature amounts), the feature is considered to have low correlation and is an independent feature. As can be seen from the figure, F3 and F4 are independent features, with which both F1 and F5 exhibit low correlation, wherein F5 correlation is lower, whereas F1 has a strong correlation with F5 itself. According to the analysis, the final feature dimension is determined to be 3 dimensions, so that the dimension reduction selection of the multi-dimensional feature quantity is completed, and the corresponding features are F3, F4 and F5. The features F3, F4, F5 are also selected as final sorting indicators, labeled S1, S2, S3, respectively.
Fig. 5 is a sample space diagram of each sorting index distribution diagram and its constitution provided by the present invention. In fig. 5, (a), (b) and (c) are distribution plots of the sorting index S1, the sorting index S2 and the sorting index S3 after normalization, respectively, and (d) is a sample space constituted by the three sorting indexes, and a total of 212 sample points are obtained. The sample distribution has obvious high-density clustering centers, and the density distribution accords with the rules of large middle and small edges.
Fig. 6 is an algorithmic clustering result provided by the present invention. The optimized neighborhood parameter is (1.152,7), the final DBSCAN algorithm output clustering result is shown in (a) of FIG. 6, and the statistical result is shown in Table 2. As can be seen from fig. 6 (a) and table 2, the DBSCAN algorithm generates a cluster, which contains 190 sample points in total, and indicates that 190 single cells are sorted out, which indicates that the single cells have relatively good consistency, wherein the core sample points have 137, and the single cells have better consistency. The algorithm marks 22 outliers as noise points, indicating that this example has 22 cells out of population. As shown in fig. 6 (b), the Mean Shift algorithm divides the sample space formed by 190 sample points into 5 clusters, and the number of samples in the clusters is 101, 41, 20, 15 and 13. Table 3 shows the cluster center of each cluster, cluster C 1 The index of the group is closest to the average level, and the number of samples of the group is the largest, which indicates that the samples have obvious aggregation forms; cluster C 3 Is the second inflection point of (2)The pressure (index S2) is lower than cluster C 2 The other two indexes are similar and the value of the index S3 is relatively higher; cluster C 4 The second inflection voltage of (1) is highest and the corresponding time (index S1) is longest, and the value of the index S3 is also lowest. Cluster C 5 The second corner voltage value of (c) is the lowest and the corresponding time is the shortest.
Table 2DBSCAN clustering algorithm output results
TABLE 3 Cluster center
Fig. 7 is a schematic diagram of the battery sorting result provided by the invention. Fig. 7 (a) shows voltage curves corresponding to all the sample cells, wherein the dotted line represents a curve marked as a noise point corresponding to a cell, and it can be seen from the figure that the noise point corresponding to a cell discharge curve is distributed at the edge position of the whole curve cluster. Further, the battery voltage curves corresponding to the respective clusters are shown in (b), (c), (d), (e), and (f) of fig. 7, respectively. As can be seen from the figure, the voltage curves within each cluster overlap higher. Therefore, according to the method provided by the invention, according to the inconsistency information reflected by 212 battery monomers on the discharge curve, 22 batteries with larger difference are removed based on the feature extraction and cluster analysis method, and then the consistency of the batteries is obviously improved by clustering. The result shows that the method has strong feasibility, and abnormal batteries can be rapidly and reliably screened out in practical application, so that consistency of groups of batteries is improved.
It will be readily appreciated by those skilled in the art that the foregoing description is merely a preferred embodiment of the invention and is not intended to limit the invention, but any modifications, equivalents, improvements or alternatives falling within the spirit and principles of the invention are intended to be included within the scope of the invention.

Claims (9)

1. A method for sorting liquid metal cells, comprising the steps of:
reconstructing the sampled discharge voltage data of the liquid metal battery to obtain a smooth voltage curve;
identifying the inflection point of the voltage curve to generate a curve characterization index;
repeated mapping screening is carried out on the curve characterization indexes to form sorting indexes;
and filtering out sample outliers in the sorting indexes by using an improved DBSCAN clustering algorithm, and optimally dividing the rest sample space by using a Mean Shift algorithm to obtain a battery sorting result.
2. The method according to claim 1, wherein in the constant current discharge mode, the terminal voltage of the liquid metal battery satisfies the following condition:
wherein the method comprises the steps ofI.e. the battery is at t k The terminal voltage at the moment in time, where k represents the kth sample point.
3. The method of claim 2, wherein the step of reconstructing the voltage curve using a linear interpolation method comprises:
sequentially from the first sampling point, if one sample point is coincident withThe sample point will be marked as normal value, marked as +.>Otherwise, it will be marked as an outlier, noted +.>And the outliers must be in line with
If the abnormal point satisfiesPress->Reconstructing the point, noted asIf the outlier satisfies +.>Then pressReconstructing j-1 points after the start of the point, wherein the j-th point satisfies the condition +.> Namely, the battery terminal voltage is +.>Corresponding time;
and after reconstruction, the abnormal points are updated into normal values in time, all sampling points are traversed, and reconstruction is completed.
4. The method of claim 1, wherein the step of identifying two inflection points using a sliding window algorithm comprises:
using self-returningModeling the voltage with a model, the associated loss function being based on a least squares residual:wherein->The voltage predicted value at the time t; let the window width be 2ω, the loss function of the first window be denoted +.>The loss function of the second window is expressed asThe window total loss function is +.>Obtaining a window difference Z (V) tk-ω...tk+ω )＝c(V tk-ω...tk+ω )-c(V tk-ω...tk )-c(V tk...tk+ω )；
The moments corresponding to the two inflection points are:wherein Z is high To identify the difference curve obtained at the first curve inflection point, Z low A difference curve obtained for identifying the inflection point of the second curve;
two corner voltage values are obtained:wherein t is high Indicating the moment corresponding to the first inflection point, t low Indicating the moment corresponding to the second inflection point.
5. The method of claim 4, wherein the curve characterization indicator comprises:
wherein F represents a characteristic index of the first x, wherein F1-F4 characterize two voltage inflection point locations and F5-F7 can be used to characterize the local shape of the curve.
6. The method of claim 5, wherein the forming a sorting index specifically comprises:
constructing a multidimensional feature quantity matrix A, A= [ F ] 1 ,F 2 ,…F 7 ]Wherein F n The characteristic vector matrix is m rows and 1 columns, m is the number of battery samples, and n is the characteristic number;
the data in the matrix A are ordered according to columns, and the data positions after the ordering are recorded by the matrix P, which is expressed as follows:
calculating the correlation coefficient to obtain the correlation between the feature vectors:rho in ij Is the correlation coefficient between the ith feature vector and the jth feature vector;
and (3) comparing the correlation among the feature vectors, and taking the feature vector with lower correlation as a sorting index, wherein the dimension of the sorting index is less than or equal to 3.
7. The method of claim 1, wherein the filtering the sample outliers in the sorting index by using the improved DBSCAN clustering algorithm, and then optimally dividing the remaining sample space by using the Mean Shift algorithm, to obtain the battery sorting result, comprises:
(1) Using zero-mean normalizationSorting indexes:wherein x represents sorting indexes, and mu and sigma are the mean value and variance of each sorting index;
(2) Determining optimal DBSCAN algorithm neighborhood parameters Eps and MinPts by using an elbow detection method, wherein Eps is a neighborhood distance threshold, minPts is a neighborhood sample number threshold, and an initial value is taken as a sorting index dimension plus 1;
(3) DBSCAN algorithm clustering
(31) Initializing parameters: core object collectionCluster number a=0; the set of unvisited samples h=e; cocooning frame partition set->
(32) Traversing all sample points in the set E to obtain sample pointsIs-domain subsampled set:the number of elements in the sub-sample set is denoted by b, and if b is greater than or equal to MinPts, the number of elements in the sub-sample set is greater than or equal to ∈>Adding the core object set omega;
(33) In the core object set Ω, a set element o is randomly selected, and parameters are updated:based on the core object point, finding an Eps-neighborhood sub-sample set N through a neighborhood distance threshold Eps (o) put it into the temporary point set Z;
(34) Taking one from the temporary point set ZFinding out all Eps-neighborhood sub-sample sets N through neighborhood distance threshold values of core objects o Eps (o') updating parameters:
(35) Repeating the process (34) until the temporary point set Z=phi, and then clustering the cluster P a After the generation is finished, updating cluster division groups: p= (P) 1 ，P 2 ，…，P a ) Marking non-core object points in the temporary point set Z as boundary points;
(36) Repeating the processes (33) - (35) until the core object set is an empty set, wherein the elements in the set (E-P) are marked as noise points, and outputting a cluster number a;
(37) If a is 1, skipping to step (38), otherwise updating MinPts, returning to step (32), wherein the updating formula of MinPts is as follows: minpts=minpts+1;
(38) Outputting a first clustering result, wherein the number of samples in the set P is recorded as N;
(4) Mean Shift algorithm clustering
(41) Calculating the mean shift M of each sample point in the set P h (x)：Wherein the method comprises the steps ofH is the bandwidth, which is the kernel function;
(42) Each sample point is translated, i.e. along M h (x) And (3) direction movement: x is x i ＝x i +M h (x i )；
(43) Repeating steps (41) - (42) until the sample point converges, i.e.: m is M h (x i )＝0；
(44) Sample points converging to the same point are considered to be the same cluster class;
(45) And outputting a second clustering result, namely a battery sorting result.
8. The method of claim 7, wherein determining optimal DBSCAN algorithm neighborhood parameters using an elbow detection method comprises:
calculating k-distance: firstly, calculating the distance between each sample point and the rest sample points, wherein the calculation formula is as follows:and then->Sequencing according to the sequence from small to large, wherein the sequenced distance set is as follows: d (i) = { D (i, 1), D (i, 2),., D (i, k), … D (i, m) }, then D (i, k) is referred to as +.>Wherein r is a sorting index number, q is a sorting index dimension, k=minpts;
obtaining a radius Eps: and (3) drawing the k-distances of all points on a k-distance graph in ascending order, and taking the maximum curvature point of Eps.
9. A sorting system for liquid metal cells, comprising: a computer readable storage medium and a processor;
the computer-readable storage medium is for storing executable instructions;
the processor is configured to read executable instructions stored in the computer readable storage medium and execute the method of sorting liquid metal cells according to any one of claims 1 to 8.
CN202310793055.1A 2023-06-29 2023-06-29 Sorting method and system for liquid metal batteries Pending CN117102082A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310793055.1A CN117102082A (en) 2023-06-29 2023-06-29 Sorting method and system for liquid metal batteries

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310793055.1A CN117102082A (en) 2023-06-29 2023-06-29 Sorting method and system for liquid metal batteries

Publications (1)

Publication Number Publication Date
CN117102082A true CN117102082A (en) 2023-11-24

Family

ID=88793672

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310793055.1A Pending CN117102082A (en) 2023-06-29 2023-06-29 Sorting method and system for liquid metal batteries

Country Status (1)

Country Link
CN (1) CN117102082A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
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

Cited By (2)

* Cited by examiner, † Cited by third party
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
CN116502112B (en) New energy power supply test data management method and system
CN107422266B (en) Fault diagnosis method and device for high-capacity battery energy storage system
CN111398833A (en) Battery health state evaluation method and evaluation system
CN113962259B (en) Multi-mode double-layer fault diagnosis method for fuel cell system
CN115511013B (en) Large-scale energy storage power station abnormal battery identification method, device and storage medium
CN115166563B (en) Power battery aging state evaluation and retirement screening method and system
CN113866642A (en) Lithium ion battery fault diagnosis method based on gradient lifting tree
CN110187281B (en) Lithium battery health state estimation method based on charging stage health characteristics
CN114818831B (en) Bidirectional lithium ion battery fault detection method and system based on multi-source perception
CN117102082A (en) Sorting method and system for liquid metal batteries
CN114024328A (en) Comprehensive evaluation method and device for energy storage power station system and readable medium
CN115366683A (en) Fault diagnosis strategy for new energy automobile power battery multi-dimensional model fusion
CN117273489A (en) Photovoltaic state evaluation method and device
CN116150572A (en) Automobile battery monomer consistency defect evaluation method based on cluster analysis
CN115015789A (en) Lithium ion battery health state estimation method and device based on incremental capacitance curve
CN113406524A (en) Inconsistent fault diagnosis method and system for power battery system
CN113765179A (en) Cloud battery pack consistency analysis method based on hierarchical clustering and class center distance
CN112287980A (en) Power battery screening method based on typical feature vector
CN113533995B (en) Consistency detection method for power battery
CN114325433A (en) Lithium ion battery fault detection method and system based on electrochemical impedance spectrum test
CN113391214A (en) Battery micro-fault diagnosis method based on battery charging voltage ranking change
CN117171588A (en) Method for detecting gradient utilization faults of power battery
CN116400244A (en) Abnormality detection method and device for energy storage battery
CN115128468A (en) Chemical energy storage battery PHM undervoltage fault prediction method
CN112485694B (en) Battery pack detection method and device

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