Battery fault early warning method based on improved clustering analysis

Abstract

The application discloses a battery fault early warning method based on improved clustering analysis, finds a balance between ensuring the accuracy and efficiency of battery fault early warning by setting conditions for constraining a scanning radius and by dynamically selecting a reference cluster, realizes accurate detection of battery faults and timely early warning, and effectively ensures stable operation and safe use of a battery system.

Description

Technical Field

[0001] This invention relates to the field of battery fault detection technology, and specifically to a battery fault early warning method based on improved clustering analysis. Background Technology

[0002] Under extreme operating conditions, lithium-ion batteries may experience thermal runaway due to overcharging, short circuits, or other faults, posing serious safety risks. Accurately diagnosing abnormal battery states and potential faults is challenging due to the complex internal chemical reactions and uncertainties in the external environment. Existing methods typically involve real-time monitoring of multi-dimensional sensing parameters such as battery voltage and temperature, followed by data analysis to attempt to detect and prevent potential hazards. However, the accuracy of fault detection is insufficient, and timeliness is difficult to achieve simultaneously. Therefore, a method is needed to accurately detect battery faults and provide timely warnings to ensure the stable operation and safe use of battery systems. Summary of the Invention

[0003] This invention aims to accurately detect battery faults and provide timely warnings, thereby ensuring the stable operation and safe use of battery systems. It provides a battery fault early warning method based on improved clustering analysis.

[0004] To achieve this objective, the present invention adopts the following technical solution:

[0005] A battery fault early warning method based on improved clustering analysis is provided, including the following steps:

[0006] S1 collects real-time status data of each individual cell in the battery pack, including individual cell status data. Battery voltage ,temperature Deformation pressure Gas concentration One or more of the following; Indicates the current time;

[0007] S2, based on the state data, cluster each of the individual cells, and then dynamically select a benchmark cluster;

[0008] S3. Based on the benchmark clustering, calculate the maximum similarity distance between each cluster and compare the thresholds, then output the fault warning classification results for each cluster.

[0009] Preferably, in step S2, after normalizing the state data associated with each individual battery cell, multidimensional feature time-series data associated with each individual battery cell is obtained. The normalization method is expressed by the following formula (1):

[0010] (1)

[0011] In formula (1), This represents the normalized value;

[0012] This represents one type of the state data;

[0013] They are respectively with These are the minimum and maximum values ​​of the same state data within the acquisition time window.

[0014] Preferably, in step S2, the method for clustering the individual cells includes the following steps:

[0015] A1, Calculate the weighted similarity distance between individual cells. , Indicates a single cell and single cell battery Weighted similarity distance between them;

[0016] A2, Set the scan radius , single cell battery and single cell battery Add to the matching set;

[0017] A3, Calculation of a single cell The weighted similarity distance to each individual cell in the matching set is used to determine whether all of them are less than or equal to the scanning radius. ,

[0018] If so, then the single cell battery Add to the set of conformities;

[0019] A4. After all individual cells have undergone the above judgment, the remaining individual cells are added to the non-compliant set.

[0020] Preferably, For single cell batteries and single cell battery Voltage similarity distance Temperature similarity distance Deformation pressure similarity distance Gas concentration similarity distance One or more weighted averages of the above.

[0021] Preferably, a single cell and single cell battery Similarity distance between data of the same state The calculation method is expressed by the following formula (2):

[0022] (2)

[0023] Represents the th in the minimum path The value of each node;

[0024] This represents the total number of nodes in the minimum path;

[0025] The minimum path is found through the following steps:

[0026] C1, for calculation Construct the corresponding matrix;

[0027] C2, find the path with the minimum sum of element values ​​from the top right element to the bottom left element of the matrix, and use this path as the found minimum path.

[0028] Preferably, step C1 involves calculation. The method for constructing the corresponding matrix is ​​expressed by the following formula (3):

[0029] (3)

[0030] Represents the first in the matrix Liede The value of the element in the row;

[0031] This indicates the individual cell at each moment within the acquisition time window. The same type of state data was collected in the first state data time series. A set of state data values, wherein each element in the first state data time series is arranged in chronological order of collection time;

[0032] This indicates that in terms of individual cells Within the same acquisition time window for the same type of status data, for individual battery cells... The second state data time series collected There are several state data values, and the elements in the second state data time series are arranged in chronological order of collection time.

[0033] Represents the first in the matrix Column, number The element value of the row;

[0034] Represents the first in the matrix Column, number The element value of the row;

[0035] Represents the first in the matrix Column, number The element value of the row.

[0036] Preferably, in step C2, the constraint condition for finding the minimum path is:

[0037] Using the top-right element of the matrix as the starting point of the path and the bottom-left element as the ending point of the path, As the current node The next node is then used to form the minimum path.

[0038] Preferably, in step S2, the method for dynamically selecting the benchmark clustering includes the following steps:

[0039] S21, determine whether the clustering result for each individual battery cell is such that the number of non-dissimilar sets is at least 1 and the number of conforming sets is 0.

[0040] If so, an abnormal alarm will be issued for the battery pack;

[0041] If not, proceed to step S22;

[0042] S22, determine whether the number of the matching sets is 1 and the number of the non-matching sets is 0.

[0043] If so, the battery pack is determined to be normal and the fault warning process is terminated;

[0044] If not, then the unique set of conforming elements or the set of conforming elements with the largest number of individual cells is selected as the baseline cluster, and the remaining sets of conforming elements or sets of non-conforming elements are each a residual cluster.

[0045] Preferably, step S3 specifically includes the following steps:

[0046] S31, Calculate the inter-class similarity distance between each remaining cluster other than the benchmark cluster and the benchmark cluster;

[0047] S32, set alarm threshold ranges corresponding to different fault warning levels;

[0048] S33, determine whether the inter-class similarity distance falls within the corresponding alarm threshold range.

[0049] If so, the remaining clusters are classified into the fault warning level corresponding to the alarm threshold range they fall into and an alarm is issued;

[0050] If not, then each individual cell in the remaining cluster is determined to be operating normally.

[0051] Preferably, in step S31, the method for calculating the inter-class similarity distance is as follows:

[0052] Calculate the similarity distance between each individual cell in the same remaining cluster and each individual cell in the benchmark cluster, and then use the calculated maximum similarity distance as the inter-cluster similarity distance between the remaining cluster and the benchmark cluster.

[0053] This application constrains the scan radius. By setting specific conditions and dynamically selecting a benchmark cluster, a balance was struck between ensuring the accuracy and efficiency of battery fault early warning. This enabled accurate detection and timely warning of battery faults, effectively ensuring the stable operation and safe use of the battery system. Attached Figure Description

[0054] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments of the present invention will be briefly described below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0055] Figure 1 This is a diagram illustrating the implementation steps of the battery fault early warning method based on improved clustering analysis provided in the embodiments of this application;

[0056] Figure 2 For single cell batteries and single cell battery An example diagram showing a matrix constructed from time series data collected at different times within the same acquisition window for the same state.

[0057] Figure 3 This is an example graph showing the minimum path found in the matrix; Detailed Implementation

[0058] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0059] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual images. They should not be construed as limiting the scope of this patent. To better illustrate the embodiments of the present invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual dimensions of the product. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0060] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "inner," and "outer" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present patent. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0061] In the description of this invention, unless otherwise explicitly specified and limited, the term "connection" or similar designation indicating a connection between components should be interpreted broadly. For example, it can refer to a fixed connection, a detachable connection, or an integral part; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can refer to the internal communication between two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0062] This application embodiment analyzes the time series data of multi-dimensional sensing parameters during the operation of lithium-ion batteries and combines it with threshold analysis of the maximum similarity distance between clusters to achieve multi-level early warning of faults, providing a basis for preventive maintenance. Simultaneously, the battery fault early warning method based on improved clustering analysis provided in this embodiment is integrated into the battery management system to improve the intelligent monitoring level of the battery, ensuring the stable operation and safe use of the battery system.

[0063] The battery fault early warning method based on improved clustering analysis provided in this embodiment, such as... Figure 1 As shown, the steps include:

[0064] S1 collects real-time status data of each individual cell in the battery pack, including battery voltage. ,temperature Deformation pressure Gas concentration , Indicates the current time, Indicates a single cell , , Indicates the length of the sampling time window, such as Each moment corresponds to 1 minute, meaning that sampling is performed once every 1 minute, for a total of 5 samplings. , This refers to the number of individual cells in the battery pack.

[0065] If the ratio of sensors (such as temperature sensors, deformation pressure sensors, and gas concentration sensors) to individual batteries cannot be 1:1, then the battery sensing data within the sensor's detection range will use the same value. For example, if the ratio of temperature sensors to batteries is 1:2, then the battery temperature readings of the two individual batteries within the temperature sensor's detection range will be the same at the same time.

[0066] S2, based on the state data of each individual cell, cluster each individual cell, and then dynamically select the benchmark cluster;

[0067] In this embodiment, considering the inconsistency of dimensions of different types of sensor data, the Min-Max method is used to normalize the state data to obtain normalized multidimensional feature time series data. The Min-Max method is expressed by the following formula (1):

[0068] (1)

[0069] In formula (1), This represents the normalized value;

[0070] This represents one type of state data;

[0071] They are respectively with These are the minimum and maximum values ​​of the same state data within the collection time window.

[0072] For example, for a single cell ,exist Individual cell voltages collected within a sampling time window of a certain length ,but , Then, the normalized expression for the single cell voltage of 3.0 using formula (1) is:

[0073] .

[0074] Single cell after state data normalization exist Multidimensional feature time series of time moments , for , , , The corresponding normalized values.

[0075] In this embodiment, the method for clustering individual battery cells includes the following steps:

[0076] A1, Calculate the weighted similarity distance between individual cells. , Indicates a single cell and single cell battery Weighted similarity distance between them;

[0077] In this embodiment, For single cell batteries and single cell battery Voltage similarity distance Temperature similarity distance Deformation pressure similarity distance Gas concentration similarity distance One or more weighted averages, preferably , , The weighted average.

[0078] Preferably, That is, the weights are equal to simplify the calculation.

[0079] Single cell battery and single cell battery Similarity distance between data of the same state (like The calculation method for ) is expressed by the following formula (2):

[0080] (2)

[0081] Represents the th in the minimum path The value of each node;

[0082] This represents the total number of nodes in the minimum path;

[0083] In this embodiment, It is a method for measuring the similarity distance between two time series. (The last part, "to calculate," appears to be incomplete and requires further context.) For example, suppose a single cell... Battery voltage time series acquired within the acquisition time window Expressed as: , This indicates the individual cell at each moment within the acquisition time window. The voltages collected separately; individual cells Battery voltage time series acquired within the same acquisition time window Expressed as: , This indicates the individual cell at each moment within the acquisition time window. The voltages were collected separately.

[0084] Assumption , Then construct a time series and time series Composed of, for example Figure 2 The matrix shown.

[0085] As explained above, the shortest path is found through the following steps:

[0086] C1, for calculation The corresponding matrix is ​​constructed, and the construction method is expressed by the following formula (3):

[0087] (3)

[0088] Represents the first in the matrix Liede The value of the element in the row;

[0089] This indicates the individual cell at each moment within the acquisition time window. The same type of state data was collected in the first state data time series. Each state data value is arranged in chronological order of its acquisition time.

[0090] This indicates that in terms of individual cells Within the same acquisition time window for collecting status data, for the same type of status data, for individual cells... The second state data time series collected The first state data value, and the elements in the second state data time series are arranged in chronological order of collection time;

[0091] Represents the first in the matrix Column, number The element value of the row;

[0092] Represents the first in the matrix Column, number The element value of the row;

[0093] Represents the first in the matrix Column, number The element value of the row.

[0094] For example, Figure 2 middle, .

[0095] C2 finds the path with the minimum sum of element values ​​from the top-right element to the bottom-left element of the matrix, and uses this path as the minimum path found.

[0096] In this embodiment, the current node The next node is As a constraint, the path starts at the top right corner of the matrix and ends at the bottom left corner, thus forming the minimum path.

[0097] For example, Figure 3 The elements marked in gray in the matrix constitute the minimum path. In this case, then... .

[0098] Complete the inter-cell battery After the calculation, the method for clustering individual cells in this embodiment proceeds to the following steps:

[0099] A2, Set the scan radius and will single cell battery and single cell battery Add to the matching set;

[0100] A3, Calculation of a single cell The weighted similarity distance to each individual cell in the matching set is used to determine whether all of them are less than or equal to the scan radius. ,

[0101] If so, then the single cell Add to the matching set;

[0102] A4. After all individual cells have undergone the above judgment, the remaining individual cells are added to the non-compliant set.

[0103] Assumption , , , All ,but Preferred Assuming the scan radius Then the single cell and single cell battery Add it to the set that does not conform to the rules.

[0104] In step S2 of this embodiment, dynamically selecting the benchmark clustering method on which step S3 depends includes the following steps:

[0105] S21, determine whether the clustering result for each individual battery cell is such that the number of non-compliant sets is at least 1 and the number of compliant sets is 0.

[0106] If so, issue an abnormal alarm for the battery pack;

[0107] If not, proceed to step S22;

[0108] S22, determine if the number of sets that satisfy the condition is 1 and the number of sets that do not satisfy the condition is 0.

[0109] If so, the battery pack is determined to be normal and the fault warning process is terminated;

[0110] If not, then select the unique set of conforming elements or the set of conforming elements with the largest number of individual cells as the baseline cluster, and the remaining sets of conforming elements or sets of non-conforming elements each become a residual set.

[0111] After dynamically selecting the benchmark cluster in step S2, as follows: Figure 1 As shown, the battery fault early warning method based on improved clustering analysis provided in this embodiment proceeds to the following steps:

[0112] S3, based on the benchmark clustering, calculates the maximum similarity distance between each cluster and compares the thresholds, then outputs the fault warning classification results for each cluster.

[0113] Specifically, step S3 includes the following steps:

[0114] S31, calculate the inter-class similarity distance between each remaining cluster (excluding the benchmark cluster) and the benchmark cluster. The calculation method is as follows:

[0115] Calculate the similarity distance between each individual cell in the same residual cluster and each individual cell in the baseline cluster, and use the maximum similarity distance as the inter-cluster similarity distance. The method for calculating the similarity distance between two individual cells is the same. The calculation method will not be elaborated here.

[0116] S32, set alarm threshold ranges corresponding to different fault warning levels;

[0117] For example, three alarm threshold ranges can be set as [0, err1), [err1, err2), and [err2, err3], with corresponding fault warning levels of minor, moderate, and severe, respectively.

[0118] S33, determine whether the inter-class similarity distance falls within the corresponding alarm threshold range.

[0119] If so, the remaining clusters will be classified into the fault warning level corresponding to the alarm threshold range they fall into and an alarm will be issued;

[0120] If not, then each individual cell in the remaining cluster is determined to be operating normally.

[0121] For example, when the inter-class similarity distance falls within the alarm threshold range of [err1, err2), it is determined that each individual cell in the remaining cluster has a moderate fault.

[0122] In summary, this application constrains the scanning radius. By setting specific conditions and dynamically selecting a benchmark cluster, a balance was struck between ensuring the accuracy and efficiency of battery fault early warning. This enabled accurate detection and timely warning of battery faults, effectively ensuring the stable operation and safe use of the battery system.

[0123] It should be stated that the above-described specific embodiments are merely preferred embodiments of the present invention and the technical principles employed. Those skilled in the art should understand that various modifications, equivalent substitutions, and variations can be made to the present invention. However, such variations, as long as they do not depart from the spirit of the present invention, should be within the scope of protection of the present invention. Furthermore, some terminology used in this specification and claims is not limiting, but merely for ease of description.

Claims

1. A battery failure early warning method based on improved cluster analysis, characterized in that, Including the following steps: S1 collects real-time status data of each individual cell in the battery pack, including individual cell status data. Battery voltage ,temperature Deformation pressure Gas concentration One or more of the following; S2, based on the state data, cluster each of the individual cells, and then dynamically select a benchmark cluster; S3, based on the benchmark clustering, calculate the maximum similarity distance between each cluster and compare the thresholds, then output the fault warning classification results for each cluster; Step S2, the method for clustering the individual cells includes the following steps: A1, Calculate the weighted similarity distance between individual cells. , Indicates a single cell and single cell battery Weighted similarity distance between them; A2, set scan radius and monocell and monocell add to the set A3, the monomer battery and the weighted similar distance of each monomer battery in the consistent set is judged whether it is less than or equal to the scanning radius , If so, the monobloc battery is added to the compliance set; A4. After all individual cells have undergone the above judgment, the remaining individual cells will be added to the non-compliant set. For single cell batteries and single cell battery Voltage similarity distance Temperature similarity distance Deformation pressure similarity distance Gas concentration similarity distance One or more weighted averages; Single cell battery and single cell battery Similarity distance between data of the same state The calculation method is expressed by the following formula (2): （2） Represents the th in the minimum path The value of each node; represents the total number of nodes in the minimum path; The minimum path is found through the following steps: C1, is the computation corresponding matrix is constructed; C2, find the path with the minimum sum of element values ​​from the top right element to the bottom left element of the matrix, and use it as the found minimum path; The calculation in Step C1 is The method of constructing the corresponding matrix is expressed by the following equation (3). （3） Represents the first in the matrix Liede The value of the element in the row; This indicates the individual cell at each moment within the acquisition time window. The same type of state data was collected in the first state data time series. A set of state data values, wherein each element in the first state data time series is arranged in chronological order of collection time; This indicates that in terms of individual cells Within the same acquisition time window for the same type of status data, for individual battery cells... The second state data time series collected There are several state data values, and the elements in the second state data time series are arranged in chronological order of collection time. denotes the value of the element in the matrix in the column, the row; Represents the first in the matrix Column, number The element value of the row; Represents the first in the matrix Column, number The element value of the row.

2. The battery failure early warning method based on improved cluster analysis according to claim 1, characterized in that, In step S2, after normalizing the state data associated with each individual battery cell, multidimensional feature time-series data associated with each individual battery cell is obtained. The normalization method is expressed by the following formula (1): （1） In Equation (1), denotes the normalized value; represents one of the state data; They are respectively with These are the minimum and maximum values ​​of the same state data within the acquisition time window.

3. The battery failure early warning method based on improved cluster analysis according to claim 1, characterized in that, In step C2, the constraint condition for finding the minimum path is: With the upper right element in the matrix as the path starting point, the lower left element as the path ending point, and the minimum path is formed by the following steps: As the next node of the current node , the minimum path is finally formed.

4. The battery failure early warning method based on improved cluster analysis according to claim 1, characterized in that, In step S2, the method for dynamically selecting the benchmark clustering includes the following steps: S21, determine whether the clustering result for each individual battery cell is such that the number of non-dissimilar sets is at least 1 and the number of conforming sets is 0. If so, an abnormal alarm will be issued for the battery pack; If not, proceed to step S22; S22, determine whether the number of the matching sets is 1 and the number of the non-matching sets is 0. If so, the battery pack is determined to be normal and the fault warning process is terminated; If not, then the unique set of conforming elements or the set of conforming elements with the largest number of individual cells is selected as the baseline cluster, and the remaining sets of conforming elements or sets of non-conforming elements are each a residual cluster.

5. The battery fault early warning method based on improved clustering analysis according to claim 1 or 4, characterized in that, Step S3 specifically includes the following steps: S31, Calculate the inter-class similarity distance between each remaining cluster other than the benchmark cluster and the benchmark cluster; S32, set alarm threshold ranges corresponding to different fault warning levels; S33, determine whether the inter-class similarity distance falls within the corresponding alarm threshold range. If so, the remaining clusters are classified into the fault warning level corresponding to the alarm threshold range they fall into and an alarm is issued; If not, then each individual cell in the remaining cluster is determined to be operating normally.

6. The battery fault early warning method based on improved clustering analysis according to claim 5, characterized in that, In step S31, the method for calculating the inter-class similarity distance is as follows: Calculate the similarity distance between each individual cell in the same remaining cluster and each individual cell in the benchmark cluster, and then use the calculated maximum similarity distance as the inter-cluster similarity distance between the remaining cluster and the benchmark cluster.

Images