Abnormal battery detection method based on voltage change characteristics and local abnormal factors

Abstract

The invention discloses an abnormal battery detection method based on voltage change characteristics and local abnormal factors, relates to the technical field of battery detection, and solves the technical problems that only abnormity of a battery pack can be detected and abnormal batteries in the battery pack cannot be determined in the prior art. The method comprises the following steps: acquiring real-time operation data of a lithium ion battery pack from a cloud database, and preprocessing the real-time operation data; extracting standard voltage data from the preprocessed real-time operation data, and screening out suspected abnormal batteries according to the standard voltage data by using a 3 sigma criterion; according to the standard voltage data, extracting feature data of a plurality of batteries in a PCA mode; calculating the internal resistance of the battery according to the real-time operation data; screening real abnormal batteries from the suspected abnormal batteries according to the characteristic data and the internal resistance of the plurality of batteries; the accuracy of an abnormal battery detection result can be improved, and thermal runaway early warning can be made in advance.

Description

Technical Field

[0001] The present invention belongs to the field of battery detection and relates to abnormal battery detection technology, in particular to an abnormal battery detection method based on voltage variation characteristics and local abnormal factors. Background Art

[0002] With the booming electric vehicle industry, the high popularity of electronic products and the continuous innovation of energy storage system technology, lithium-ion batteries have become the mainstream battery technology widely used in many application scenarios due to their significant advantages such as high energy density and long cycle life. Abnormal detection of lithium-ion batteries can timely detect defects inside the battery such as internal short circuits and electrolyte leakage, thereby effectively preventing the occurrence of these safety accidents. Abnormal detection can measure the key performance parameters of the battery, which can help users understand the health status and usage of the battery, timely detect the trend of battery performance degradation, and adopt corresponding maintenance measures, thereby extending the battery life.

[0003] The prior art (invention patent with application number: 2017101671789) discloses a battery pack abnormality monitoring method, which includes an air pressure sensor for monitoring the air pressure inside the battery pack; a strain type pressure sensor for detecting the inner wall pressure of the battery pack; a processor for determining whether the battery pack has an abnormality based on a first pressure value collected by the air pressure sensor and a second pressure value collected by the strain type pressure sensor, and generating a first alarm message under the condition that the battery pack has an abnormality; the prior art can analyze the abnormality of the battery pack by analyzing the first pressure value and the second pressure value in the battery pack; however, the battery pack is composed of multiple batteries. When a battery in the battery pack is abnormal, the entire battery pack will cause an abnormality; the prior art can only detect that the battery pack has an abnormality, but cannot determine the abnormal battery in the battery pack.

[0004] The present invention provides an abnormal battery detection method based on voltage variation characteristics and local abnormal factors to solve the above technical problems. Summary of the invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes an abnormal battery detection method based on voltage change characteristics and local abnormal factors, which is used to solve the technical problem that a battery pack is composed of multiple batteries, and when a battery in the battery pack is abnormal, it will cause the entire battery pack to be abnormal; the prior art can only detect the abnormality of the battery pack, but cannot identify the abnormal battery in the battery pack.

[0006] To achieve the above object, a first aspect of the present invention provides an abnormal battery detection method based on voltage variation characteristics and local abnormal factors, comprising: Step S1: obtaining real-time operation data of the lithium-ion battery pack from a cloud database, and preprocessing some real-time operation data; Step S2: extracting standard voltage data from the pre-processed real-time operation data, and screening out suspected abnormal batteries according to the standard voltage data using the 3σ criterion; Step S3: extracting characteristic data of several batteries by PCA according to the standard voltage data; Step S4: Calculating the internal resistance of the battery according to the real-time operation data; Step S5: Screening out truly abnormal batteries from suspected abnormal batteries based on the characteristic data and internal resistance of the batteries.

[0007] Preferably, the preprocessing of the plurality of real-time operation data includes: Retrieve real-time operation data; the real-time operation data includes: operation time, operation status, total battery voltage, battery current, battery voltage; the operation status includes: charging status, discharging status and static status; The real-time operation data is cleaned, and the battery voltages of several batteries are extracted from the cleaned real-time operation data; the median of several battery voltages is calculated to obtain the median voltage; the difference between several battery voltages and the median voltage is calculated, and the formula is expressed as follows: ; By formula Calculate the voltage data of the decentralized battery; By formula Standardize the decentralized voltage data to obtain standard voltage data; Among them, n represents the number of collection points, and m represents the number of batteries; represents the battery voltage of the mth battery in the nth acquisition point; Indicates the median voltage of the nth acquisition point; is the battery voltage data after Yeo-Johnson standardization; i represents the i-th collection point; j represents the j-th battery; represents the transformation coefficient; It represents the difference between the battery voltage of the jth battery in the i-th collection point and the median voltage.

[0008] It should be noted that Yeo-Johnson, or Yeo-Johnson transform, is a set of piecewise functions composed of power functions, which are used to transform data of random variables to reduce their heteroscedasticity and amplify their normality, so that the shape of their probability density function is closer to the normal distribution; the transformation coefficient is calculated by calling the Python function; the collection point refers to the collection time point, and the real-time operating data of the battery is collected according to the artificially set frequency.

[0009] The present invention pre-processes the real-time operation data and converts the battery voltage into a standard voltage using the Yeo-Johnson transformation, which lays a foundation for subsequent screening of suspected abnormal batteries and facilitates analysis and processing of the battery voltage.

[0010] Preferably, screening out abnormal batteries according to standard voltage data includes: Retrieve standard voltage data; calculate the standard score of the battery through the formula ; ; in, Indicates the standard voltage data of the jth battery; represents the standard score of the jth battery; Obtain a score threshold; determine whether the absolute value of the standard score of the battery is greater than the score threshold; if yes, mark the corresponding battery as an abnormal battery; if no, continue to screen the abnormal batteries.

[0011] It should be noted that in statistics and mathematical analysis, standard scores are used to evaluate the degree of deviation of a data point from the mean of its data set; the setting of the score threshold is based on conventional statistical practices and experience in practical applications; when the standard voltage after the Yeo-Johnson change processing presents a normal distribution, the score threshold is set to 2; in a normal distribution, about 95% of the data points will be within the standard deviation range on both sides of the mean, that is, between -2σ and 2σ; therefore, when the absolute value of the standard score is greater than 2, it is judged to be abnormal.

[0012] The present invention analyzes the standard voltage of the battery and uses the 3σ criterion to screen out suspected abnormal batteries; it is capable of performing preliminary detection on the abnormal batteries; and the preliminary screening of the batteries is conducive to improving the accuracy of the abnormal battery detection results and making a thermal runaway warning in advance.

[0013] Preferably, the extracting characteristic data of a plurality of batteries by PCA according to the standard voltage data includes: By formula The charging time of several batteries is divided into q time windows; wherein n represents the length of the charging segment, i.e., the charging time; and p represents the size of the time window; in each time window, the voltage variation coefficient of the battery and the distance between the voltage curve and the voltage median line are calculated respectively; The voltage variation coefficient of the battery and the average, standard deviation, range, maximum increment and minimum increment of the distance between the voltage curve and the voltage median line are calculated respectively, and integrated into the initial characteristic data of the battery; the PCA method is used to reduce the dimension of the initial characteristics of the battery to obtain the characteristic data of the battery.

[0014] It should be noted that PCA, or principal component analysis, is a common data analysis that uses the idea of ​​dimensionality reduction to transform multiple indicators into a few comprehensive indicators using a linear transformation technique.

[0015] The present invention divides the charging time into several time windows, calculates the voltage variation coefficient of the battery and the distance between the voltage curve and the voltage median line in the time windows; and respectively calculates the average value, standard deviation, range, maximum increment and minimum increment of the two to obtain initial feature data, and uses PCA to reduce the dimension of the initial feature data of the battery to obtain the feature data of the battery; the battery features can be extracted, laying a foundation for subsequent abnormal analysis of the battery.

[0016] Preferably, calculating the voltage variation coefficient of the battery in each time window includes: The coefficient of variation of the battery voltage is calculated by the formula: ; in, represents the coefficient of variation of the cell voltage of the jth cell in the qth time window; represents the average voltage of the jth battery in the qth time window; Indicates the voltage dispersion of the jth battery in the qth time window; Represents the battery voltage collected by the jth battery in the rth row.

[0017] Preferably, the step of calculating the voltage variation coefficient of the battery and the distance between the voltage curve and the voltage median line in each time window includes: The distance between the battery voltage curve and the voltage midline is calculated using the formula: ; in, represents the distance between the battery curve of the jth battery and the median voltage curve in the qth time window; Represents the median voltage curve formed by the median voltage of each record in the qth window; represents the voltage curve of the jth battery in the qth window; yes and The inverse covariance matrix of .

[0018] Preferably, the calculation of the mean value, standard deviation, range, maximum increment and minimum increment of the voltage coefficient of variation of the battery includes: Get the voltage variation coefficient of several batteries; through the formula Calculate the average value of the voltage coefficient of variation of the battery; By formula Calculate the standard deviation of the battery's voltage coefficient of variation; By formula Calculate the range of the battery's voltage coefficient of variation; By formula Calculate the maximum increment of the voltage variation coefficient of the battery; By formula Calculate the minimum increment of the voltage variation coefficient of the battery; where abs represents the absolute value function; represents the coefficient of variation of the cell voltage of the jth cell in the kth time window.

[0019] Preferably, the calculation of the average value, standard deviation, range, maximum increment and minimum increment of the distance between the voltage curve of the battery and the voltage median line includes: Get the voltage curves of several batteries and the distance between the voltage midline; through the formula Calculate the average value of the distance between the voltage curve and the voltage median line; By formula Calculate the standard deviation of the distance between the voltage curve and the voltage median line; By formula Calculate the range of the distance between the voltage curve and the voltage median line; By formula Calculate the maximum increment of the distance between the voltage curve and the voltage neutral line; By formula Calculate the minimum increment of the distance between the voltage curve and the voltage neutral line; where, Represents the distance between the battery curve of the jth battery in the kth time window and the median voltage curve.

[0020] Preferably, the calculating the internal resistance of the battery according to the real-time operating data includes: Retrieve the real-time operating data of the battery during the charging process, and extract the real-time operating data of the variable current stage from the real-time operating data of the charging process; By formula Calculate the internal resistance of the battery; where, represents the voltage of the jth battery before current reduction, represents the current of the jth battery before current reduction; represents the voltage of the jth battery after current reduction, Represents the current of the jth battery after current reduction.

[0021] Preferably, the method of screening out truly abnormal batteries from suspected abnormal batteries according to the characteristic data and internal resistance of a plurality of batteries includes: Retrieve characteristic data and internal resistance of several batteries, integrate the characteristic data and internal resistance of the batteries into analysis data; input the analysis data into the LOF algorithm to obtain the LOF score of the battery; Determine whether the battery with the largest LOF score is a suspected abnormal battery; if yes, mark the corresponding suspected abnormal battery as a real abnormal battery; if no, mark the suspected abnormal battery as a normal battery.

[0022] It should be noted that the LOF algorithm, namely the local outlier factor algorithm, is an unsupervised learning algorithm for anomaly detection. It identifies outliers by comparing the local density difference between a data point and its neighbors.

[0023] The present invention integrates the characteristic data and internal resistance of the battery into analysis data, uses the LOF algorithm to analyze the analysis data of the battery, and obtains the LOF score of the battery; and screens out the real abnormal batteries according to the LOF score and the suspected abnormal batteries; can perform a second screening on the abnormal batteries, which is beneficial to avoid false alarms; and reduces the recall cost of the battery.

[0024] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention pre-processes the real-time operation data and converts the battery voltage into a standard voltage using the Yeo-Johnson transformation, which lays a foundation for the subsequent screening of suspected abnormal batteries and facilitates the analysis and processing of the battery voltage; analyzes the standard voltage of the battery to screen out suspected abnormal batteries; can perform preliminary detection of abnormal batteries; and performs preliminary screening of batteries, which is conducive to improving the accuracy of abnormal battery detection results and making thermal runaway warnings in advance.

[0025] 2. The present invention divides the charging time into several time windows, calculates the voltage variation coefficient of the battery and the distance between the voltage curve and the voltage median line in the time window; and respectively calculates the average value, standard deviation, range, maximum increment and minimum increment of the two to obtain initial feature data, and uses PCA to reduce the dimension of the initial feature data of the battery to obtain the feature data of the battery; the battery features can be extracted, laying the foundation for subsequent abnormal analysis of the battery; the battery feature data and internal resistance are integrated into analysis data, and the battery analysis data is analyzed using the LOF algorithm to obtain the battery LOF score; and the real abnormal batteries are screened out according to the LOF score and suspected abnormal batteries; the abnormal batteries can be screened for a second time, which is conducive to avoiding false alarms; and the battery recovery cost is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0027] Figure 1 It is a schematic diagram of the overall steps of the present invention; Figure 2 A schematic diagram of the suspected abnormal battery detection steps of the present invention; Figure 3 It is a schematic diagram of the actual abnormal battery detection steps of the present invention. DETAILED DESCRIPTION

[0028] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] See also Figure 1 The first embodiment of the present invention provides an abnormal battery detection method based on voltage change characteristics and local abnormal factors, including: Step S1: obtaining real-time operation data of the lithium-ion battery pack from a cloud database, and preprocessing some real-time operation data; Step S2: extracting standard voltage data from the pre-processed real-time operation data, and screening out suspected abnormal batteries according to the standard voltage data using the 3σ criterion; Step S3: extracting characteristic data of several batteries by PCA according to the standard voltage data; Step S4: Calculating the internal resistance of the battery according to the real-time operation data; Step S5: Screening out truly abnormal batteries from suspected abnormal batteries based on the characteristic data and internal resistance of the batteries.

[0030] It should be noted that a lithium-ion battery pack is a battery group composed of multiple lithium-ion battery cells connected in series or in parallel; it can provide higher voltage and larger capacity to meet the power needs of different devices.

[0031] See also Figure 2 , obtain real-time operating data of several batteries in the lithium-ion battery pack, wherein the real-time operating data includes: operating time, operating status, total battery voltage, battery current, and battery voltage; the operating status includes: charging state, discharging state, and static state; The real-time operation data is cleaned, and the battery voltages of several batteries are extracted from the cleaned real-time operation data; the median of several battery voltages is calculated to obtain the median voltage; the difference between several battery voltages and the median voltage is calculated, and the formula is expressed as follows: ; By formula Calculate the voltage data of the battery after decentralization; through the formula The decentralized voltage data is standardized to obtain standard voltage data; where n represents the number of sampling points and m represents the number of batteries; represents the battery voltage of the mth battery in the nth acquisition point; Indicates the median voltage of the nth acquisition point; is the battery voltage data after Yeo-Johnson standardization; i represents the i-th collection point; j represents the j-th battery; Represents the transform coefficients.

[0032] Retrieve standard voltage data; calculate the standard score of the battery through the formula ; ; in, Indicates the standard voltage data of the jth battery; Indicates the standard score of the j-th battery; obtains the score threshold; determines whether the absolute value of the standard score of the battery is greater than the score threshold; if yes, marks the corresponding battery as an abnormal battery; if no, continues to screen the abnormal batteries.

[0033] It should be noted that the first layer of anomaly detection cleans the real-time operation data of several batteries in the lithium-ion battery pack and standardizes the battery voltage to obtain the standard voltage; the standard score of the battery is calculated according to the standard voltage, and the battery is screened for the first time based on the standard score to obtain suspected abnormal batteries; this is beneficial to improving the accuracy of faulty battery detection and making thermal runaway warnings in advance, and has good generalization.

[0034] For example: suppose that several batteries in a lithium-ion battery pack are tested, the real-time operation data of the lithium-ion battery pack is obtained, the real-time operation data is cleaned, and the decentralized battery voltage is standardized using the Yeo-Johnson method to obtain the standard voltage of several batteries; the standard scores of several batteries are calculated according to the standard voltage; the score threshold is set to 2; the standard scores of several batteries are compared with the score threshold to obtain the suspected abnormal batteries as battery 5, battery 12, and battery 23.

[0035] See also Figure 3 , through the formula The charging time of several batteries is divided into q time windows; wherein n represents the length of the charging segment, i.e., the charging time; and p represents the size of the time window; in each time window, the voltage variation coefficient of the battery and the distance between the voltage curve and the voltage median line are calculated respectively; The coefficient of variation of the battery voltage is calculated by the formula: ; in, represents the coefficient of variation of the cell voltage of the jth cell in the qth time window; represents the average voltage of the jth battery in the qth time window; Indicates the voltage dispersion of the jth battery in the qth time window; represents the battery voltage collected by the jth battery in the kth row; The distance between the battery voltage curve and the voltage midline is calculated using the formula: ; in, represents the distance between the battery curve of the jth battery and the median voltage curve in the qth time window; Represents the median voltage curve formed by the median voltage of each record in the qth window; represents the voltage curve of the jth battery in the qth window; yes and The inverse matrix of the covariance of ; Get the voltage variation coefficient of several batteries; through the formula Calculate the average value of the voltage coefficient of variation of the battery; By formula Calculate the standard deviation of the battery's voltage coefficient of variation; by the formula Calculate the range of the battery voltage variation coefficient; by the formula Calculate the maximum increment of the battery voltage variation coefficient; by the formula Calculate the minimum increment of the voltage variation coefficient of the battery; retrieve the distance between the voltage curves and the voltage median line of several batteries; By formula Calculate the average value of the distance between the voltage curve and the voltage midline; by the formula Calculate the standard deviation of the distance between the voltage curve and the voltage median line; by the formula Calculate the range of the distance between the voltage curve and the voltage midline; by the formula Calculate the maximum increment of the distance between the voltage curve and the voltage neutral line; by the formula Calculate the minimum increment of the distance between the voltage curve and the voltage median line; integrate the voltage coefficient of variation of the battery and the average value, standard deviation, range, maximum increment and minimum increment of the distance between the voltage curve and the voltage median line into the initial characteristic data of the battery; use the PCA method to reduce the dimension of the initial characteristics of the battery to obtain the characteristic data of the battery.

[0036] Retrieve the real-time operating data of the battery during the charging process, and extract the real-time operating data of the variable current stage from the real-time operating data of the charging process; through the formula Calculate the internal resistance of the battery; where, represents the voltage of the jth battery before current reduction, represents the current of the jth battery before current reduction; represents the voltage of the jth battery after current reduction, represents the current of the jth battery after current reduction; Retrieve characteristic data and internal resistance of several batteries, and integrate the characteristic data and internal resistance of the batteries into analysis data; input the analysis data into the LOF algorithm to obtain the LOF score of the battery; determine whether the battery with the largest LOF score is a suspected abnormal battery; if yes, mark the corresponding suspected abnormal battery as a real abnormal battery; if not, mark the suspected abnormal battery as a normal battery.

[0037] It should be noted that the second-level abnormality detection detects the batteries that have been marked as suspected abnormal batteries based on the first-level abnormality detection.

[0038] It should be noted that the second-level anomaly detection uses the time window processing mechanism to calculate the voltage variation coefficient and the median voltage distance within the window, and extracts the statistical characteristics of the voltage variation coefficient and the median voltage distance between windows, such as the mean, standard deviation, range, maximum increment, and minimum increment. PCA is then used to extract the main features as voltage change characteristics. Based on the changes in the charging current drop section, the internal resistance of the battery is calculated to measure the degree of battery aging. Finally, LOF is used to score the battery for abnormality. If the battery with the largest score is the suspected abnormal battery detected in the first layer, it will be determined as a real abnormal battery. This is helpful to avoid false alarms and reduce the cost of battery recovery.

[0039] For example: suppose a second-level test is performed on a lithium-ion battery pack; the standard voltage of the suspected abnormal battery in the lithium-ion battery pack is obtained; the voltage variation coefficient and median voltage distance within the window are calculated based on the standard voltage, and the mean, standard deviation, range, maximum increment, and minimum increment of the voltage variation coefficient and median voltage distance between windows are extracted, and then PCA is used for dimensionality reduction to obtain the characteristic data of the battery; The internal resistance of the battery is calculated based on the change in the charging current reduction stage; the characteristic data and internal resistance of the battery are integrated into analysis data, and the analysis data is input into the LOF algorithm to generate the LOF score of the battery; the battery with the largest LOF score is extracted as battery 12; then battery 12 is marked as a real abnormal battery, and battery 5 and battery 23 are marked as normal batteries.

[0040] Part of the data in the above formula is calculated by removing the dimension and taking its numerical value. The formula is a formula closest to the actual situation obtained by software simulation of a large amount of collected data; the preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained through simulation of a large amount of data.

[0041] The working principle of the present invention is as follows: the present invention obtains the real-time operation data of the lithium-ion battery pack from the cloud database, and pre-processes some of the real-time operation data; extracts the standard voltage data from the pre-processed real-time operation data, and screens out suspected abnormal batteries according to the standard voltage data using the 3σ criterion; extracts the characteristic data of some batteries according to the standard voltage data using the PCA method; calculates the internal resistance of the battery according to the real-time operation data; and screens out the real abnormal batteries from the suspected abnormal batteries according to the characteristic data and internal resistance of some batteries.

[0042] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. An abnormal battery detection method based on voltage variation characteristics and local abnormal factors, characterized in that: include: Step S1: obtaining real-time operation data of the lithium-ion battery pack from a cloud database, and preprocessing some real-time operation data; Step S2: extracting standard voltage data from the pre-processed real-time operation data, and screening out suspected abnormal batteries according to the standard voltage data using the 3σ criterion; Step S3: extracting characteristic data of several batteries by PCA according to the standard voltage data; Step S4: Calculating the internal resistance of the battery according to the real-time operation data; Step S5: Screening out truly abnormal batteries from suspected abnormal batteries based on the characteristic data and internal resistance of the batteries.

2. The abnormal battery detection method based on voltage variation characteristics and local abnormal factors according to claim 1 is characterized in that: The preprocessing of the real-time operation data includes: Retrieve real-time operation data; the real-time operation data includes: operation time, operation status, total battery voltage, battery current, battery voltage; the operation status includes: charging status, discharging status and static status; The real-time operation data is cleaned, and the battery voltages of several batteries are extracted from the cleaned real-time operation data; the median of several battery voltages is calculated to obtain the median voltage; the difference between several battery voltages and the median voltage is calculated, and the formula is expressed as follows: ； By formula Calculate the voltage data of the decentralized battery; By formula Standardize the decentralized voltage data to obtain standard voltage data; Among them, n represents the number of collection points, and m represents the number of batteries; represents the battery voltage of the mth battery in the nth acquisition point; Indicates the median voltage of the nth acquisition point; is the battery voltage data after Yeo-Johnson standardization; i represents the i-th collection point; j represents the j-th battery; represents the transformation coefficient; It represents the difference between the battery voltage of the jth battery in the i-th collection point and the median voltage.

3. The abnormal battery detection method based on voltage variation characteristics and local abnormal factors according to claim 1 is characterized in that: The method of screening out abnormal batteries according to standard voltage data includes: Retrieve standard voltage data; calculate the standard score of the battery through the formula ; ； in, Indicates the standard voltage data of the jth battery; represents the standard score of the jth battery; Obtain a score threshold; determine whether the absolute value of the standard score of the battery is greater than the score threshold; if yes, mark the corresponding battery as an abnormal battery; if no, continue to screen the abnormal batteries.

4. The abnormal battery detection method based on voltage variation characteristics and local abnormal factors according to claim 1 is characterized in that: The method of extracting characteristic data of several batteries by PCA according to the standard voltage data includes: By formula The charging time of several batteries is divided into q time windows; wherein n represents the length of the charging segment, i.e., the charging time; and p represents the size of the time window; in each time window, the voltage variation coefficient of the battery and the distance between the voltage curve and the voltage median line are calculated respectively; The voltage variation coefficient of the battery and the average, standard deviation, range, maximum increment and minimum increment of the distance between the voltage curve and the voltage median line are calculated respectively, and integrated into the initial characteristic data of the battery; the PCA method is used to reduce the dimension of the initial characteristics of the battery to obtain the characteristic data of the battery.

5. The abnormal battery detection method based on voltage variation characteristics and local abnormal factors according to claim 4 is characterized in that: The step of calculating the voltage variation coefficient of the battery in each time window includes: The coefficient of variation of the battery voltage is calculated by the formula: ； in, represents the coefficient of variation of the cell voltage of the jth cell in the qth time window; represents the average voltage of the jth battery in the qth time window; Indicates the voltage dispersion of the jth battery in the qth time window; Represents the battery voltage collected by the jth battery in the rth row.

6. The abnormal battery detection method based on voltage variation characteristics and local abnormal factors according to claim 4 is characterized in that: The step of calculating the voltage variation coefficient of the battery and the distance between the voltage curve and the voltage median line in each time window includes: The distance between the battery voltage curve and the voltage midline is calculated using the formula: ； in, represents the distance between the battery curve of the jth battery and the median voltage curve in the qth time window; Represents the median voltage curve formed by the median voltage of each record in the qth window; represents the voltage curve of the jth battery in the qth window; yes and The inverse covariance matrix of .

7. The abnormal battery detection method based on voltage variation characteristics and local abnormal factors according to claim 5, characterized in that: The calculation of the mean value, standard deviation, range, maximum increment and minimum increment of the voltage coefficient of variation of the battery includes: Get the voltage variation coefficient of several batteries; through the formula Calculate the average value of the voltage coefficient of variation of the battery; By formula Calculate the standard deviation of the battery's voltage coefficient of variation; By formula Calculate the range of the battery's voltage coefficient of variation; By formula Calculate the maximum increment of the voltage variation coefficient of the battery; By formula Calculate the minimum increment of the voltage variation coefficient of the battery; where abs represents the absolute value function; represents the coefficient of variation of the cell voltage of the jth cell in the kth time window.

8. The abnormal battery detection method based on voltage variation characteristics and local abnormal factors according to claim 6, characterized in that: The method of calculating the average value, standard deviation, range, maximum increment and minimum increment of the distance between the voltage curve of the battery and the voltage median line includes: Get the voltage curves of several batteries and the distance between the voltage midline; through the formula Calculate the average value of the distance between the voltage curve and the voltage median line; By formula Calculate the standard deviation of the distance between the voltage curve and the voltage median line; By formula Calculate the range of the distance between the voltage curve and the voltage median line; By formula Calculate the maximum increment of the distance between the voltage curve and the voltage neutral line; By formula Calculate the minimum increment of the distance between the voltage curve and the voltage neutral line; where, Represents the distance between the battery curve of the jth battery in the kth time window and the median voltage curve.

9. The abnormal battery detection method based on voltage variation characteristics and local abnormal factors according to claim 1, characterized in that: The calculating the internal resistance of the battery according to the real-time operation data includes: Retrieve the real-time operating data of the battery during the charging process, and extract the real-time operating data of the variable current stage from the real-time operating data of the charging process; By formula Calculate the internal resistance of the battery; where, represents the voltage of the jth battery before current reduction, represents the current of the jth battery before current reduction; represents the voltage of the jth battery after current reduction, Represents the current of the jth battery after current reduction.

10. The abnormal battery detection method based on voltage variation characteristics and local abnormal factors according to claim 1, characterized in that: The method of screening out truly abnormal batteries from suspected abnormal batteries based on characteristic data and internal resistance of a plurality of batteries includes: Retrieve characteristic data and internal resistance of several batteries, integrate the characteristic data and internal resistance of the batteries into analysis data; input the analysis data into the LOF algorithm to obtain the LOF score of the battery; Determine whether the battery with the largest LOF score is a suspected abnormal battery; if yes, mark the corresponding suspected abnormal battery as a real abnormal battery; if no, mark the suspected abnormal battery as a normal battery.

Images