Battery abnormal point identification method, system and device based on dot product similarity and medium

Abstract

The invention belongs to the technical field of power battery fault monitoring, and particularly relates to a battery abnormal point identification method, system and equipment based on dot product similarity and a medium. The method comprises the steps of obtaining power battery voltage and temperature data conforming to a GBT32960 protocol, dividing charging and discharging segments by using a sliding window technology, and constructing a battery cell voltage matrix; and calculating the dot product similarity between the battery cells to represent the consistency characteristics of the battery cells. And by performing scaling processing on the dot product similarity matrix, the calculation precision of the consistency characteristic of the battery cell is improved. The characteristic value of the battery cell consistency is analyzed in combination with a box plot method, the battery cell numbers with inconsistency are effectively recognized, and efficient and accurate early warning and abnormal point recognition of power battery faults are achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power battery fault monitoring, and particularly relates to a method, system, device and medium for identifying battery abnormal points based on dot product similarity. Background Art

[0002] As a core component of new energy vehicles, the performance state of power batteries is directly related to the safe and stable driving of vehicles. With the booming development of the new energy vehicle industry, the data monitoring and management of power batteries have become particularly important. According to national requirements, new energy vehicles need to send data to the national big data platform in accordance with the protocol format of national standard GBT32960. These data provide an important basis for the fault detection of power batteries. Existing technologies monitor these data in real time through big data technologies, aiming to timely discover potential safety hazards existing in power batteries. Generally, a power battery pack is composed of multiple battery groups, and each battery group contains multiple battery cells. During the use of the battery pack, performance differences may occur between different battery groups or battery cells due to factors such as actual temperature, aging rate, internal resistance, etc.

[0003] Although existing technologies have been able to use big data technologies to monitor power battery data in real time, there are still some defects in actual applications. On the one hand, due to the complexity of the power battery system, it is difficult to accurately identify the performance differences between different battery cells through traditional methods. On the other hand, sensors may be interfered by various noises when collecting data, resulting in inaccurate data, which further increases the difficulty of identifying the performance differences of battery cells. In addition, existing technologies may have problems of low computing efficiency when dealing with large-scale data, and it is difficult to respond to the requirements of battery fault warnings in a timely manner. Therefore, there is an urgent need for a method that can efficiently and accurately identify the performance differences of battery cells to improve the power battery fault warning ability of new energy vehicles. Summary of the Invention

[0004] The purpose of the present invention is to provide a method, system, device and medium for identifying battery abnormal points based on dot product similarity, so as to solve the problem of how to accurately and quickly identify the fault warnings and abnormal points of power batteries in large-scale data.

[0005] The present invention achieves the above purpose through the following technical solutions:

[0006] In a first aspect, the present invention proposes a method for identifying battery abnormal points based on dot product similarity, and the method includes:

[0007] Obtain the original battery data and process it into a voltage time series array;

[0008] Process the voltage time series array into charge and discharge segments;

[0009] Extract the segment to be recognized from the charge and discharge segments using a sliding window;

[0010] Construct a cell voltage matrix based on the sliding window data in each segment to be recognized; wherein, the sliding window data includes a preset data volume and a preset sliding window time interval;

[0011] Determine the cell consistency eigenvalue at the set timestamp of a single sliding window based on the cell voltage matrix, and the eigenvalue is used to characterize the dot product similarity between cells;

[0012] Analyze the eigenvalues of the cell consistency based on the box plot method to identify the inconsistent cell numbers.

[0013] Further, the obtaining of the original battery data and processing it into a voltage time series array includes:

[0014] Obtain the original battery data according to the data format of the GBT32960 protocol and real-time parsing of the message;

[0015] Sort the original battery data based on the data acquisition time, and clean the duplicate data and null data to obtain a voltage time series array.

[0016] Further, the processing of the voltage time series array into charge and discharge segments includes:

[0017] Distinguish whether the battery is currently in a charging or discharging state based on the charging status code;

[0018] And / or, distinguish whether the battery is currently in a charging or discharging state based on the current direction;

[0019] Obtain the charge and discharge segments of the battery when it is currently in a charging or discharging state.

[0020] Further, the constructing of the cell voltage matrix according to the data in the sliding window of each segment to be recognized includes:

[0021] Construct a cell voltage matrix X according to the data in each sliding window, where the data volume in the sliding window is d, the sliding window time interval is ΔT, and the total number of segments to be recognized is S;

[0022] X=(x1,x2,…,x n ) T

[0023] where, n represents the number of cells, and x i represents the time series data of the i-th cell, and the matrix dimension is n×d.

[0024] Further, the determining of the cell consistency eigenvalue at the set timestamp of a single sliding window based on the cell voltage matrix includes:

[0025] Calculate the dot product similarity between the row vectors in the cell voltage matrix X to obtain a matrix XX representing the similarity between cells T ;

[0026] Scale the dot product similarity XX T to obtain the scaled XX T′ , and the scaling method is as follows:

[0027] Based on the amount of data in the sliding window, the time interval of the sliding window, and the total number S of segments to be recognized, determine the timestamp of the last frame of data in a single sliding window;

[0028] According to the formula where S is the total number of segments and s represents the segment number, calculate the cell consistency eigenvalue I corresponding to the timestamp s , where I s is an n-dimensional column vector used to represent the consistency relationship between cells, and Min-max Norm is a regularization function.

[0029] Furthermore, analyzing the eigenvalue of the cell consistency based on the box plot method to identify the inconsistent cell numbers includes:

[0030] For a column of numbers X = (x1, x2,..., x n ), calculate the upper and lower quartiles Q1 and Q3 of the data set; T Calculate the interquartile range IQR, IQR = Q3 - Q1;

[0031] Then for the data point x

[0032] , if x i satisfies x i < Q1 - 1.5 × IQR or x i > Q3 + 1.5 × IQR, it is determined as an outlier, and its cell number is recorded.

[0033] In a second aspect, the present invention proposes a battery outlier recognition system based on dot product similarity for implementing the battery outlier recognition method described in any one of the above, and the system includes:

[0034] A first processing module for acquiring battery raw data and processing it into a voltage time series array;

[0035] A second processing module for processing the voltage time series array into charge and discharge segments;

[0036] A segment extraction module for extracting segments to be recognized from the charge and discharge segments using a sliding window;

[0037] A matrix construction module, configured to construct a battery cell voltage matrix based on the sliding window data in each fragment to be recognized; wherein, the sliding window data includes a preset data volume and a preset sliding window time interval;

[0038] A similarity calculation module, configured to determine the battery cell consistency eigenvalue at the set timestamp of a single sliding window based on the battery cell voltage matrix, and the eigenvalue is used to characterize the dot product similarity between battery cells;

[0039] An identification module, configured to analyze the eigenvalue of the battery cell consistency based on the box plot method to identify the inconsistent battery cell numbers.

[0040] In a third aspect, the present invention provides an electronic device, including:

[0041] A processor; a memory for storing executable instructions of the processor;

[0042] Wherein, the processor is configured to execute the instructions to implement the battery anomaly point identification method as described in any one of the above.

[0043] In a fourth aspect, the present invention provides a computer-readable storage medium, when the instructions in the computer-readable storage medium are executed by the processor of the electronic device, enabling the electronic device to execute the battery anomaly point identification method as described in any one of the above.

[0044] The beneficial effects of the present invention are as follows:

[0045] By acquiring the voltage and temperature data of the power battery that conforms to the national standard GBT32960 protocol and using the sliding window technology to divide the data into charge and discharge fragments, the present invention effectively enhances the flexibility and adaptability of data processing. Constructing a battery cell voltage matrix and using matrix multiplication to calculate the dot product similarity between battery cells, this innovative method not only simplifies the calculation process but also significantly improves the running efficiency of the program. At the same time, scaling the dot product similarity matrix effectively reduces the influence of noise during sensor data acquisition and improves the calculation accuracy of the battery cell consistency eigenvalue. Combining the box plot method to analyze the eigenvalue of the battery cell consistency can accurately identify the inconsistent battery cell numbers, providing strong support for the early warning and repair of battery failures. Description of the Drawings

[0046] Figure 1 It is a schematic flowchart of a battery anomaly point identification method based on dot product similarity provided by Embodiment 1 of the present application;

[0047] Figure 2 It is another schematic flowchart of a battery anomaly point identification method based on dot product similarity provided by Embodiment 1 of the present application;

[0048] Figure 3Schematic diagram of the cell voltage curve of normal inbound vehicles in the case part of the specific implementation manner of this application;

[0049] Figure 4 Schematic diagram of the cell consistency characteristic values of normal inbound vehicles in the case part of the specific implementation manner of this application;

[0050] Figure 5 Schematic diagram of the cell voltage curve of normal non - inbound vehicles in the case part of the specific implementation manner of this application;

[0051] Figure 6 Schematic diagram of the cell consistency characteristic values of normal non - inbound vehicles in the case part of the specific implementation manner of this application. Specific implementation manner

[0052] The following further describes this application in detail with reference to the accompanying drawings. It is necessary to point out here that the following specific implementation manner is only used to further illustrate this application and cannot be understood as a limitation on the protection scope of this application. Those skilled in the art can make some non - essential improvements and adjustments to this application based on the above application content.

[0053] Example 1

[0054] As Figure 1-2 shown, this example proposes a method for identifying battery abnormal points based on dot - product similarity. The method includes the following steps:

[0055] S1. Obtain the original battery data according to the data format of the GBT32960 protocol and real - time parsing of the message; sort the original battery data based on the data acquisition time, and clean the duplicate data and null data to obtain the voltage time - series array.

[0056] S2. Distinguish whether the battery is in the charging or discharging state based on the charging status code; and / or, distinguish whether the battery is in the charging or discharging state based on the current direction; obtain the charge - discharge segments of the battery in the current charging or discharging state.

[0057] S3. Construct the cell voltage matrix X according to the data in each sliding window, where the amount of data in the sliding window is d, the time interval of the sliding window is ΔT, and the total number of segments to be identified is S;

[0058] X=(x1,x2,…,x n ) T

[0059] where n represents the number of cells, x i represents the time - series data of the i - th cell, and the matrix dimension is n×d.

[0060] S4. Construct a battery cell voltage matrix based on the sliding window data in each fragment to be recognized; wherein, the sliding window data includes a preset data volume and a preset sliding window time interval;

[0061] S5. Determine the battery cell consistency eigenvalue at the set timestamp of a single sliding window based on the battery cell voltage matrix, and the eigenvalue is used to characterize the dot product similarity between battery cells;

[0062] S6. Analyze the eigenvalues of the battery cell consistency based on the box plot method to identify the inconsistent battery cell numbers.

[0063] In specific implementation, the present invention accurately identifies abnormal points in the power battery through dot product similarity calculation, and applies these abnormal points to the early warning system of the battery management system. In specific implementation, once an abnormal point in the battery cell voltage or consistency eigenvalue is identified, the system immediately triggers an early warning and timely notifies the vehicle owner or maintenance personnel. This early warning mechanism not only improves the accuracy of battery fault detection, but also provides timely maintenance suggestions for vehicle owners, effectively avoids the occurrence of battery pack failures, extends the battery service life, and at the same time improves the safety performance and driving stability of new energy vehicles.

[0064] Further preferably, in step S5, determining the battery cell consistency eigenvalue at the set timestamp of a single sliding window based on the battery cell voltage matrix includes:

[0065] S5.1. Calculate the dot product similarity between the row vectors in the battery cell voltage matrix X to obtain a matrix XX representing the similarity between battery cells T ; The dot product similarity measures the similarity degree between the battery cell voltage time series data.

[0066] S5.2. Scale the dot product similarity XX T to obtain the scaled XX T′ , and the scaling method is: Scaling is used to reduce the noise influence when collecting data by sensors. The scaling method can be selected according to actual needs, such as scaling using the Gauss-Markov inequality and the central limit theorem.

[0067] S5.3. Determine the timestamp of the last frame of data of a single sliding window based on the sliding window data volume, the sliding window time interval, and the total number S of fragments to be recognized.

[0068] S5.4. According to the formula S is the total number of fragments, and s represents the fragment number, calculate the battery cell consistency eigenvalue I at the corresponding timestamp s , where I s is an n-dimensional column vector used to characterize the consistency relationship of each battery cell, and Min-max Norm is a regularization function.

[0069] Exemplarily, for the timestamp of the last frame of data in the above single sliding window, assume the original data sequence is as follows:

[0070] [Data 1, Data 2, Data 3, Data 4, Data 5, Data 6, Data 7, Data 8, Data 9, Data 10].

[0071] If the sliding window size d is 3 and the sliding interval ΔT is 1, the following sliding windows can be obtained:

[0072] Window 1: [Data 1, Data 2, Data 3], with the timestamp being the time of Data 3;

[0073] Window 2: [Data 2, Data 3, Data 4], with the timestamp being the time of Data 4;

[0074] Window 3: [Data 3, Data 4, Data 5], with the timestamp being the time of Data 5;

[0075] …

[0076] Window 7: [Data 7, Data 8, Data 9], with the timestamp being the time of Data 9;

[0077] Window 8: [Data 8, Data 9, Data 10], with the timestamp being the time of Data 10.

[0078] It should be noted that in this embodiment, in step S5.1, the dot product similarity between the row vectors in the cell voltage matrix X is calculated to obtain the matrix XX representing the similarity between cells T , specifically, the similarity between two vectors is measured by calculating their dot product. Exemplarily, given two vectors A = (a1, …, a n ) and B = (b1, …, b n ), their dot product similarity is defined as follows:

[0079]

[0080] where a i and b i are the i-th elements of vectors A and B.

[0081] Properties of the dot product:

[0082] ① Positive value: If the directions of two vectors are the same, the dot product will be positive.

[0083] ② Zero value: If two vectors are orthogonal (perpendicular), the dot product is zero.

[0084] ③ Negative value: If the directions of two vectors are opposite, the dot product is negative.

[0085] The result calculated by the dot product represents the similarity between two vectors, and it measures the projection size of one vector in the direction of the other vector.

[0086] Further preferably, in step S6, based on the box plot method to analyze the characteristic values of the cell consistency, the inconsistent cell numbers are identified, including:

[0087] S6.1. For a column of numbers X = (x1, x2, …, x n ) T , calculate the upper and lower quartiles Q1 and Q3 of the data set;

[0088] S6.2. Calculate the interquartile range IQR, IQR = Q3 - Q1;

[0089] S6.3. Then for the data point x i , if it satisfies x i < Q1 - 1.5×IQR or x i > Q3 + 1.5×IQR, it is determined as an outlier, and its cell number is recorded.

[0090] According to the above embodiments of the present invention, the identification method obtains the battery voltage and temperature data that conform to the GBT32960 protocol, uses the sliding window technology to divide the data into charge and discharge segments, and constructs a cell voltage matrix. Subsequently, matrix multiplication is used to calculate the dot product similarity between cells to characterize the consistency characteristics of cells. By scaling the dot product similarity matrix, the calculation accuracy of the cell consistency characteristics is improved. At the same time, combined with the box plot method to analyze the characteristic values of the cell consistency, the cell numbers with inconsistencies can be effectively identified.

[0091] In order to more clearly illustrate the present invention and its advantages, the following will further explain the battery outlier identification method provided by the present invention in combination with specific cases and relevant parts of the drawings.

[0092] Exemplarily, combining Figure 3 and Figure 5 are respectively the cell voltage curves of the inbound vehicle (the battery has a fault and needs to be repaired in the factory) and the normal non-inbound vehicle. Figure 4 and Figure 6 are respectively the schematic diagrams of the cell consistency characteristic values of the inbound vehicle and the normal non-inbound vehicle. It can be detected from Figure 4 that the #12 cell has a consistency problem, corresponding to Figure 3 where the cell voltage drops significantly. And this phenomenon can be detected since the first sliding window of this discharge segment, with obvious early warning and identification ability. While Figure 3 , Figure 4 of the normal vehicles do not have similar phenomena.

[0093] Embodiment 2

[0094] This embodiment proposes a battery anomaly point recognition system based on dot product similarity for implementing the above-mentioned battery anomaly point recognition method. This system can be pre-configured in the new energy vehicle battery management system for early warning of new energy vehicle battery anomaly monitoring. The system includes:

[0095] A first processing module for acquiring battery raw data and processing it into a voltage time series array;

[0096] A second processing module for processing the voltage time series array into charge and discharge segments;

[0097] A segment extraction module for extracting segments to be recognized from the charge and discharge segments using a sliding window;

[0098] A matrix construction module for constructing a cell voltage matrix based on the sliding window data in each segment to be recognized; where the sliding window data includes a preset data volume and a preset sliding window time interval;

[0099] A similarity calculation module for determining the cell consistency eigenvalue of a single sliding window set timestamp based on the cell voltage matrix, and the eigenvalue is used to characterize the dot product similarity between cells;

[0100] An identification module for analyzing the eigenvalues of cell consistency based on the box plot method to identify inconsistent cell numbers.

[0101] It should be noted here that each module in the above battery anomaly point recognition system corresponds to steps S1 to S6 in implementing the above battery anomaly point recognition method. The examples and application scenarios implemented by multiple modules and the corresponding steps are the same, but are not limited to the content disclosed in the above Embodiment 1.

[0102] Embodiment 3

[0103] This embodiment proposes an electronic device, including:

[0104] A processor; a memory for storing processor-executable instructions;

[0105] Wherein, the processor is configured to execute instructions to implement the battery anomaly point recognition method as described above.

[0106] Embodiment 4

[0107] This embodiment proposes a computer-readable storage medium. When the instructions in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device can execute the battery anomaly point recognition method as described above.

[0108] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part.

[0109] The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more integrated available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)).

[0110] Those of ordinary skill in the art will realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. A professional technician can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of this application.

[0111] In addition, in each embodiment of the present application, the functional modules can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.

[0112] The above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for identifying battery outliers based on dot product similarity, characterized in that The method includes: Obtaining the original battery data and processing it into a voltage time series array; Processing the voltage time series array into charge and discharge segments; Using a sliding window to extract segments to be recognized from the charge and discharge segments; Constructing a cell voltage matrix based on the sliding window data in each segment to be recognized; wherein, the sliding window data includes a preset data volume and a preset sliding window time interval; Determining the cell consistency eigenvalue at the set timestamp of a single sliding window based on the cell voltage matrix, and the eigenvalue is used to characterize the dot product similarity between cells; Analyzing the eigenvalues of the cell consistency based on the box plot method to identify inconsistent cell numbers.

2. The battery anomaly point recognition method based on dot product similarity according to claim 1, characterized in that The obtaining the original battery data and processing it into a voltage time series array includes: Obtaining the original battery data according to the data format of the GBT32960 protocol and real-time parsing of the message; Sorting the original battery data based on the data acquisition time, and cleaning duplicate data and null data to obtain a voltage time series array.

3. The method for identifying battery anomaly points based on dot product similarity according to claim 1, wherein, The processing the voltage time series array into charge and discharge segments includes: Distinguishing whether the battery is currently in a charging or discharging state based on the charging status code; And / or, distinguishing whether the battery is currently in a charging or discharging state based on the current direction; Obtaining the charge and discharge segments of the battery in the current charging or discharging state.

4. A method for identifying battery outliers based on dot product similarity according to claim 1, characterized in that The constructing a cell voltage matrix based on the data of the sliding window in each segment to be recognized includes: Constructing a cell voltage matrix X according to the data in each sliding window, wherein the data volume in the sliding window is d, the sliding window time interval is ΔT, and the total number of segments to be recognized is S; X = (x1, x2, …, x n ) T where n represents the number of battery cells, and x i represents the timing data of the i-th battery cell, and the matrix dimension is n×d.

5. The method for identifying battery anomaly points based on dot product similarity according to claim 4, wherein, The determining the cell consistency eigenvalue at the set timestamp of a single sliding window based on the cell voltage matrix includes: Calculate the dot product similarity between the row vectors in the cell voltage matrix X to obtain a matrix XX that characterizes the similarity between cells T ; Dot product similarity XX T Scale it to obtain the scaled XX T′ , and the scaling method is as follows: Determining the timestamp of the last frame of data of a single sliding window based on the sliding window data volume, the sliding window time interval, and the total number of segments to be recognized S; According to the formula where S is the total number of segments, s represents the segment number, and the cell consistency characteristic value I corresponding to the time stamp is calculated s , where I s is an n-dimensional column vector used to characterize the consistency relationship of each cell, and Min-max Norm is a regularization function.

6. The method for identifying battery outliers based on dot product similarity according to claim 5, wherein The analyzing the eigenvalues of the cell consistency based on the box plot method to identify inconsistent cell numbers includes: For a sequence of numbers \(X=(x_1,x_2,\ldots,x\) n ) T , calculate the lower and upper quartiles \(Q_1\), \(Q_3\) of the data set; Calculating the interquartile range IQR, IQR = Q3 - Q1; For the data point x i , if x i < Q1 - 1.5 × IQR or x i > Q3 + 1.5 × IQR, it is determined as an outlier, and its cell number is recorded.

7. A battery outlier recognition system based on dot product similarity, which is used to implement the battery outlier recognition method according to any one of claims 1-6, and is characterized in that, The system includes: A first processing module for obtaining the original battery data and processing it into a voltage time series array; A second processing module for processing the voltage time series array into charge and discharge segments; A segment extraction module for using a sliding window to extract segments to be recognized from the charge and discharge segments; A matrix construction module for constructing a cell voltage matrix based on the sliding window data in each segment to be recognized; wherein, the sliding window data includes a preset data volume and a preset sliding window time interval; A similarity calculation module for determining the cell consistency eigenvalue at the set timestamp of a single sliding window based on the cell voltage matrix, and the eigenvalue is used to characterize the dot product similarity between cells; An identification module for analyzing the eigenvalues of the cell consistency based on the box plot method to identify inconsistent cell numbers.

8. An electronic device, characterized in that, Includes: A processor; A memory for storing executable instructions of the processor; Wherein, the processor is configured to execute the instructions to implement the battery anomaly point identification method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the battery anomaly point recognition method according to any one of claims 1 to 6.

Images