Battery system consistency evaluation method, electronic equipment and medium

Abstract

The invention provides a battery system consistency evaluation method, electronic equipment and a medium. The battery system consistency evaluation method comprises the following steps: acquiring voltage data of each battery monomer in a battery system; calculating a characteristic value of the voltage data of each battery monomer by adopting a fast weighted Shannon entropy method; and processing the characteristic value of the voltage data of each single battery by adopting a clustering method, and screening out abnormal single batteries. According to the method, the normal distribution-based fast weighted Shannon entropy method is adopted to process the voltage data, the normal distribution is combined to divide the probability matrix interval, and the probability matrix is weighted through the weight factor, so that the method is more sensitive to the abnormal voltage data, the identification of the outlier degree of the battery cell monomers is improved, and the battery consistency detection precision and accuracy are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of battery management, and in particular relates to a battery system consistency evaluation method, electronic equipment and medium. Background Art

[0002] During the use of power battery packs, there will be pressure or temperature differences between battery cells. These inconsistencies reflect problems inside the battery. At the least, they will lead to capacity reduction, abnormal battery changes, etc. At the worst, they will cause battery overvoltage and lead to battery safety problems. Research on battery consistency mainly judges the inconsistency of battery voltage, etc., which can reflect the outliers of battery cells to a certain extent and warn of battery safety problems.

[0003] Patent application CN117370758A discloses a battery pack consistency assessment method, including: extracting the voltage data and temperature data of each battery in the energy storage power station battery pack, using the voltage data and temperature data to obtain the variance, Shannon entropy and range characteristics of the voltage and temperature, and adopting a hybrid clustering algorithm for the consistency assessment of the energy storage power station battery pack. This method uses the traditional Shannon entropy and calculates the probability through uniform distribution, which has a low degree of discrimination for abnormal battery data and low battery consistency detection precision and accuracy. Summary of the invention

[0004] The purpose of the present invention is to address the deficiencies in the prior art and provide a battery system consistency evaluation method, electronic equipment and medium to improve the accuracy and precision of battery consistency detection.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] A battery system consistency assessment method comprises the following steps:

[0007] S1: Obtain voltage data of each battery cell in the battery system;

[0008] S2: The fast weighted Shannon entropy method is used to calculate the characteristic value of the voltage data of each battery cell. The specific process is as follows:

[0009] Set the length of the sliding window and extract the voltage data X in the current sliding window t ;

[0010] Calculate X t The Z value distribution, if -2σ<z ij <2σ, then in X t Remove the voltage data of the corresponding row and get X t ′;

[0011] Based on the normal distribution (-∞, -3σ), (-3σ, -2σ), (-2σ, 2σ), (2σ, 3σ), (3σ, ∞), calculate X t 'The probability matrix P t ;

[0012] Use weight W to calculate the probability matrix P t For weighted processing, the expression of W is as follows:

[0013]

[0014] Calculate P t The Shannon entropy h of each row in i , and obtain the Shannon entropy H of the voltage data in the sliding window t ,h i The expression is as follows:

[0015]

[0016] Calculate the characteristic value S of the voltage data in the sliding window t , S t =H t T X t ′;

[0017] Move the sliding window and repeat step S2 to obtain the characteristic value of the voltage data of each battery cell;

[0018] S3: using a clustering method to process the characteristic value of the voltage data of each battery cell and screen out abnormal battery cells;

[0019] Where σ is the voltage data X t The variance of x ij is the voltage of the jth battery cell at time i, For X t The average value of the voltage data in the i-th row, s i For X t The standard deviation of the voltage data in the i-th row, W0 is the initial weight, λ0 and λ1 are weight coefficients, and p ij is the probability, H t T H t The transpose of .

[0020] The present invention adopts a fast weighted Shannon entropy method based on normal distribution to process voltage data, divides the probability matrix interval in combination with normal distribution, and weights the probability matrix by weight factors, which is more sensitive to abnormal voltage data, improves the recognition of the degree of outliers of battery cells, and improves the accuracy and precision of battery consistency detection.

[0021] Further: W0=[w1,w2,w3,w4,w5], where w1=w5, w2=w4, w5>w2, w4>w3.

[0022] Furthermore, λ0 and λ1 are calculated by the following formula:

[0023]

[0024] Among them, x min 、x max They are voltage data X t The minimum and maximum values ​​in is the voltage data X t The average value of .

[0025] Furthermore, the clustering methods include DBSCAN clustering, hierarchical clustering, Gaussian mixture clustering, fuzzy G clustering, and K-means.

[0026] Preferably, the charging state of the battery system is selected to obtain voltage data of each battery cell in the battery system.

[0027] Based on the same inventive concept, the present invention also provides an electronic device, including:

[0028] one or more processors;

[0029] A memory having one or more programs stored thereon, which, when executed by the one or more processors, enables the one or more processors to implement the steps of the battery system consistency assessment method.

[0030] Based on the same inventive concept, the present invention also provides a computer-readable storage medium storing a computer program, which implements the steps of the battery system consistency assessment method when executed by a processor.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] The present invention adopts a fast weighted Shannon entropy method based on normal distribution to process voltage data, divides the probability matrix interval in combination with normal distribution, and weights the probability matrix by weight factors, which is more sensitive to abnormal voltage data, improves the recognition of the degree of outliers of battery cells, and improves the accuracy and precision of battery consistency detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is a schematic diagram of a battery system consistency assessment method of the present invention;

[0034] Figure 2 It is a schematic diagram of the DBSCAN clustering algorithm;

[0035] Figure 3 This is a voltage diagram of the lithium-ion battery charging process;

[0036] Figure 4 Schematic diagram of generating characteristic curves for sliding window calculations;

[0037] Figure 5 This is a schematic diagram of noise points obtained by the DBSCAN clustering algorithm;

[0038] Figure 6 Schematic diagram of cluster distribution obtained by DBSCAN clustering algorithm. DETAILED DESCRIPTION

[0039] The present invention will be described in detail below in conjunction with the embodiments. It should be noted that the embodiments and features of the embodiments of the present invention can be combined with each other without conflict. For the convenience of description, if the words "upper", "lower", "left" and "right" appear in the following, they only indicate that the upper, lower, left and right directions are consistent with the drawings themselves, and do not limit the structure.

[0040] Example

[0041] like Figure 1-Figure 6 In this embodiment, a battery system consistency assessment method is used to obtain characteristic values ​​for the national standard voltage data of the battery system stored on the platform (the national standard voltage data refers to GB32960 "Technical Specifications for Electric Vehicle Remote Service and Management System") through fast 3sigma (normal distribution) adaptive weighted Shannon entropy, and then use clustering methods to identify abnormal self-discharge of the battery cell, abnormal whole package, etc., so as to provide timely warning and processing. Clustering methods include DBSCAN clustering, hierarchical clustering, Gaussian mixture clustering, fuzzy G clustering, and K-means. The DBSCAN clustering method is used in this embodiment.

[0042] 1. Fast weighted Shannon entropy based on normal distribution

[0043] This technology introduces weighted Shannon entropy considering the practical significance of data, considers normal distribution, and weights the probability matrix by adaptively determining the weight factor in combination with the battery cell charging characteristic curve, thereby improving the recognition of the degree of battery cell outliers, as follows.

[0044] 1.1 Shannon entropy

[0045] During the use of power batteries, pressure differences or temperature differences will occur. These inconsistencies reflect problems inside the battery. At the least, they will lead to a decrease in capacity, abnormal changes in the battery, etc. At the worst, they will cause battery overvoltage and lead to battery safety problems. Information entropy, also known as Shannon entropy, is an effective means of judging battery inconsistency. Its basic definition is as follows:

[0046]

[0047] Among them, p i is the probability of the ith event occurring among d events, V i is the ith event.

[0048] Applying information entropy to battery data means statistically analyzing the interval distribution of battery data according to certain rules to obtain the probability distribution P, and then calculating the Shannon entropy. The improved Shannon entropy based on the 3sigma distribution divides the intervals by normal distribution, which is more sensitive to abnormal data. The specific method is as follows: For the following battery data,

[0049]

[0050] Among them, x ij is the voltage data of the jth battery cell at time i, the number of battery cells is m, and the total recording time is n. For time t, select the sliding window length k;

[0051]

[0052] Among them, u = t + 1, v = t + k - 1. Through standardization, the Z value distribution is obtained.

[0053]

[0054] in, For X t The mean value of the voltage data in the i-th row, s i For X t The standard deviation of the voltage data in the i-th row.

[0055] Based on the 3sigma distribution (-∞, -3σ), (-3σ, -2σ), (-2σ, 2σ), (2σ, 3σ), (3σ, ∞), we can get the following probability matrix:

[0056]

[0057] Where σ is the voltage data X t The variance of .

[0058] Through formula (1), the Shannon entropy of each row is calculated as follows:

[0059]

[0060] The Shannon entropy at this moment is:

[0061] H t =[h1,h2,...,h k ] T (7)

[0062] The voltage data is processed by Shannon entropy to obtain its eigenvalue. The eigenvalue calculation method at this moment is as follows:

[0063] S t =H t T X t =[s 11 ,s 12 ,...,s 1m ](8)

[0064] By moving the sliding window, we can get the Shannonen eigenvalue of each monomer at all times, such as Figure 4 , Different colors in the picture represent different battery cells.

[0065]

[0066] 1.2 Fast 3sigma Shannon entropy

[0067] In actual battery systems, the consistency between cells is good in most cases. For such data, the Z value in equation (4) after normalization falls within the ±2σ range, which is calculated as follows:

[0068] P i =[0,0,1,0,0] (10)

[0069] Calculated by formula (6):

[0070] h i =0 (11)

[0071] The Shannon entropy in formula (7) is obtained as follows:

[0072] H t =[h1,...,0,...,h k ] T (12)

[0073] The eigenvalues ​​calculated by formula (8) are as follows:

[0074]

[0075] Right now

[0076] H(X) T X=H(X′) T X′ (14)

[0077] Where X′ is the reduced-order matrix of the original matrix, that is, the rows with good consistency between the cells are removed to obtain the reduced-order matrix. The specific method is to calculate formula (4) and then determine whether the Z value of each row meets the following conditions:

[0078] -2σ<Z ij <2σ (15)

[0079] All rows that meet the condition in the original matrix are removed to obtain the reduced-order matrix X′.

[0080] Fast Shannon entropy can reduce the amount of calculation and increase the calculation speed. For commercial vehicles that are actually equipped with hundreds of battery cells, it can increase the calculation speed by more than 30%.

[0081] 1.3 Adaptive weighted Shannon entropy based on normal distribution

[0082] For probability matrices with the same probability but distributed in different intervals, the traditional Shannon entropy results are the same. For example, the following two probability matrices have the same entropy value;

[0083]

[0084] h=-alog(a)-(1-a)log(1-a) (17)

[0085] 3sigma entropy determines the probability matrix P based on the 3sigma distribution. In the case of normal distribution, the distribution probabilities of data in the five intervals (-∞, -3σ), (-3σ, -2σ), (-2σ, 2σ), (2σ, 3σ), (3σ, ∞) are:

[0086] p=[0.135,2.14,95.45,2.14,0.135] (18)

[0087] In actual situations, the probability of P1 occurring is much lower than that of P2, and the abnormal probability of P1 is also much greater than that of P2, but the Shannon entropy h values ​​of the two are the same, indicating that the 3sigma Shannon entropy cannot well characterize the abnormality of the data. In view of the fact that the traditional Shannon entropy cannot reflect the importance of each interval, this embodiment adopts the weighted Shannon entropy as follows:

[0088]

[0089] The traditional 3sigma Shannon entropy is improved to 3sigma weighted Shannon entropy, and the weight matrix is ​​determined by adaptively based on the normal distribution and the voltage change during battery charging. The specific method is as follows:

[0090] With reference to the normal distribution probability, the initial weight W0 is given according to the rule of w1=w5>w2=w4>w3.

[0091] W0=[w1,w2,w3,w4,w5] (20)

[0092] At the front end of charging, due to the inconsistency of single-cell SOC, the battery voltage will rise sharply one after another, resulting in a large voltage difference. This situation also occurs at the end of charging, such as Figure 3 .

[0093] At this time, the Z value will increase, and the h value will increase accordingly. In order to avoid interference from such situations, the weight is adjusted adaptively.

[0094] W=[W1,W2,W3,W4,W5] (21)

[0095]

[0096] Where λ0 and λ1 are weight coefficients.

[0097] The specific adaptive weight selection rules are as follows:

[0098]

[0099] where x min 、x max They are voltage data X t The minimum and maximum values ​​in is the voltage data X t The average value of .

[0100] 2. DBSCAN Clustering

[0101] DBSCAN is a density-based unsupervised clustering algorithm, which is an existing technology. When the number of samples in the sample clustering radius neighborhood meets the conditions, the sample is a core point; when the sample is in the neighborhood of other core points, it is called a boundary point; otherwise, the sample is a noise point.

[0102] The DBSCAN is applied to the battery system consistency detection as follows: 1) For the voltage data at a certain moment, its eigenvalue is obtained by fast 3sigma weighted Shannon entropy, as shown in formula (9); 2) The DBSCAN algorithm is used to process the eigenvalue, which can reflect the clustering of the battery cells and filter out abnormal battery cells, as follows:

[0103]

[0104] Among them, C is the cluster set, the number of clusters at each moment is not fixed, and O is the noise point set, that is, the abnormal battery cell.

[0105] Through the above method, the distribution of abnormal cells and cell clusters at each moment is obtained, such as Figure 2 , Figure 5 , Figure 6 .

[0106] The DBSCAN algorithm is used to obtain the distribution of battery cell clusters at each moment, and then the following algorithm is used to detect the actual battery cell clustering:

[0107] 1) Detecting the voltage difference: For the battery voltage data in formula (2), the voltage difference between the maximum value and the minimum value of each row is detected. When the voltage difference is less than a certain threshold, cluster distribution detection is performed.

[0108] 2) According to the clustering set and the clustering of each battery cell, the distribution of the whole pack is determined. Specifically, the clustering set of each battery cell is counted. If there is an inconsistent cluster number or a noise point and it meets a certain threshold, it will be eliminated and the cluster classification of the battery pack will be summarized.

[0109] For cloud platform storage of battery information, since the trend of battery cells is relatively consistent during charging, selecting charging segments can avoid the impact of timing on the results. When it is determined that the battery is in a charging state, single cell outlier detection and cluster distribution detection are performed, and the obtained single cell abnormality and whole package abnormality problems are reported.

[0110] The battery system consistency assessment method has higher warning accuracy. It can improve the accuracy of warning through adaptive weighted Shannon entropy based on normal distribution. It also increases the detection of whole pack anomalies and can identify whole pack anomalies of battery cells caused by pack replacement and other situations.

[0111] Another embodiment of the present invention provides an electronic device, including:

[0112] one or more processors;

[0113] A memory having one or more programs stored thereon, which, when executed by one or more processors, enables the one or more processors to implement the steps of the battery system consistency assessment method.

[0114] In some implementations, the memory may be a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory.

[0115] In some other implementations, the processor may be a central processing unit (CPU), a digital signal processor (DSP), or other general-purpose processors of various types, which are not limited herein.

[0116] Another embodiment of the present invention provides a computer-readable storage medium storing a computer program, which implements the steps of the battery system consistency assessment method when the computer program is executed by a processor.

[0117] The contents explained in the above embodiments should be understood as these embodiments are only used to more clearly illustrate the present invention, and are not used to limit the scope of the present invention. After reading the present invention, various equivalent forms of modifications to the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

Claims

1. A battery system consistency assessment method, characterized in that: The following steps are involved: S1: Obtain voltage data of each battery cell in the battery system; S2: The fast weighted Shannon entropy method is used to calculate the characteristic value of the voltage data of each battery cell. The specific process is as follows: Set the length of the sliding window and extract the voltage data X in the current sliding window t ; Calculate X t The Z value distribution, if -2σ<z ij <2σ, then in X t Remove the voltage data of the corresponding row and get X t ′; Based on the normal distribution (-∞, -3σ), (-3σ, -2σ), (-2σ, 2σ), (2σ, 3σ), (3σ, ∞), calculate X t The probability matrix P of t ; Use weight W to calculate the probability matrix P t For weighted processing, the expression of W is as follows: Calculate P t The Shannon entropy h of each row in i , and obtain the Shannon entropy H of the voltage data in the sliding window t ,h i The expression is as follows: Calculate the characteristic value S of the voltage data in the sliding window t , S t =H t T X t ′; Move the sliding window and repeat step S2 to obtain the characteristic value of the voltage data of each battery cell; S3: using a clustering method to process the characteristic value of the voltage data of each battery cell and screen out abnormal battery cells; Where σ is the voltage data X t The variance of x ij is the voltage of the jth battery cell at time i, For X t The average value of the voltage data in the i-th row, s i For X t The standard deviation of the voltage data in the i-th row, W0 is the initial weight, λ0 and λ1 are weight coefficients, and p ij is the probability, H t T H t The transpose of .

2. The battery system consistency assessment method according to claim 1, characterized in that: W0=[w1,w2,w3,w4,w5], where w1=w5, w2=w4, w5>w2, w4>w3.

3. The battery system consistency assessment method according to claim 1, characterized in that: λ0 and λ1 are calculated by the following formula: Among them, x min 、x max They are voltage data X t The minimum and maximum values ​​in is the voltage data X t The average value of .

4. The battery system consistency assessment method according to claim 1, characterized in that: The clustering methods include DBSCAN clustering, hierarchical clustering, Gaussian mixture clustering, fuzzy G clustering, and K-means.

5. The battery system consistency assessment method according to claim 1, characterized in that: Select the battery system charging status and obtain the voltage data of each battery cell in the battery system.

6. An electronic device, characterized in that: include: one or more processors; A memory having one or more programs stored thereon, which, when the one or more programs are executed by the one or more processors, enables the one or more processors to implement the steps of the method according to any one of claims 1 to 5.

7. A computer-readable storage medium, characterized in that: The computer program is stored therein, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images