A lithium battery voltage abnormality detection method and system

Abstract

The present application relates to the field of data processing, and more particularly to a lithium battery voltage abnormality detection method and system, the method comprising: collecting reference measurement values and reference true values of voltages of each cell of a historical standard battery under three states of charging, standing and discharging; collecting measured values of voltages of each cell of a battery to be measured under the three states of charging, standing and discharging; calculating consistency scores of the battery to be measured under the three states of charging, standing and discharging; and determining that the voltage of the battery to be measured is abnormal when the minimum consistency score is less than a preset value. The present application significantly reduces the negative impact of system measurement accuracy on the detection result, and comprehensively utilizes multi-state information to improve the accuracy and reliability of the detection result.

Description

Technical Field

[0001] This invention relates to the field of data processing. In particular, it relates to a method and system for detecting abnormal lithium battery voltage. Background Technology

[0002] In the quality inspection process of lithium battery pack production, the mainstream methods for identifying abnormal cells based on voltage detection mainly fall into two categories: the absolute voltage threshold method and the static consistency comparison method. The absolute voltage threshold method sets a fixed normal range for the cell voltage (e.g., 3.0V to 4.2V). During testing, if the voltage measurement of a cell exceeds this range, the cell is considered abnormal. This method is simple to implement but relies too heavily on a single or a few static measurements. The static consistency comparison method does not rely on absolute voltage values ​​but compares the voltage measurements of all cells in the battery pack at the same time (usually in a static state). By calculating the statistical characteristics of these voltage values ​​(such as range and standard deviation) and setting a threshold, the consistency of cells within the pack is judged. If the voltage of a cell deviates significantly from the average level within the pack, it is considered abnormal.

[0003] The existing methods mentioned above directly use the raw measurements from the battery management system for threshold comparison or consistency calculation. This makes the detection results heavily dependent on the absolute accuracy of the battery management system. A normal cell may be misjudged as having too low a voltage due to negative measurement errors, while an abnormal cell may be missed due to positive measurement errors. The accuracy and robustness of this method, which heavily relies on the absolute accuracy of the battery management system, are difficult to guarantee. Moreover, most of the judgments are based on voltage data at a certain static moment, lacking analysis of the cell's behavior under complete working dynamics (such as charging and discharging), resulting in insufficient comprehensiveness and reliability of the diagnostic conclusions.

[0004] Therefore, there is a need in this field for a lithium battery voltage anomaly detection method and system to solve the problem of poor accuracy in the diagnostic conclusions of the above-mentioned detection methods. Summary of the Invention

[0005] To address the technical problem of poor accuracy in diagnostic conclusions of the aforementioned detection methods, the present invention provides solutions in the following aspects.

[0006] In the first aspect, a method for detecting abnormal lithium battery voltage includes:

[0007] Collect reference measured values ​​and reference true values ​​of voltage for each cell of a historical standard battery under three states: charging, resting, and discharging. Collect the measured values ​​of voltage for each cell of the battery under test under three states: charging, resting, and discharging.

[0008] Calculate the consistency score of the battery under test in three states: charging, resting, and discharging. When the lowest consistency score is less than a preset value, the voltage of the battery under test is determined to be abnormal.

[0009] The methods for calculating the consistency score in a single state include:

[0010] Based on the reference measured values ​​and reference true values ​​of all cells in a historical standard battery under the same conditions, the error value set of each measured value of each cell in the battery under test under this condition is obtained. The error accuracy range of the battery management system is divided into several error regions. Based on the error value set, the error representative value and error probability of the measured value corresponding to the error value set in each error region are calculated. Based on all measured values ​​of a single cell and the corresponding error representative value and error probability, all estimated true value sequences of the cell and the probability of each estimated true value sequence are generated. Based on the estimated true value sequence of the cell and its probability, the matching degree between a single cell and each adjacent cell is calculated. When the matching degree is greater than a preset threshold, the consistency matching count of the cell is updated. Based on the consistency matching count of each cell and the total number of adjacent cells, the consistency score of the battery under test is calculated.

[0011] Preferably, the method for obtaining the error value set of each measured value to be tested includes: setting a preset deviation range, traversing the reference measured values ​​of all cells of a historical standard battery under the same conditions, finding all reference measured values ​​that conform to the preset deviation range, calculating the difference between the reference measured value and the corresponding reference true value, and obtaining the error value set of the measured value to be tested.

[0012] Preferably, the step of calculating the representative error value of the measured value in each error region based on the error value set includes: for each error region, calculating the average value of the error values ​​falling into the error region in the error value set, and using it as the representative error value of the measured value in that error region.

[0013] Preferably, the step of calculating the error probability of the measured value corresponding to the error value set in each error region based on the error value set includes: for each error region, calculating the ratio of the number of error values ​​falling into the error region to the total number of error values ​​in the error value set, to obtain the error probability of the measured value in the error region.

[0014] Preferably, the method for constructing the estimated true value sequence of each cell under a single state includes: for each measured value to be measured of a single cell, the measured value to be measured is added to the error representative value of each corresponding error region to obtain multiple estimated true values, forming a set of estimated true values ​​of the measured value to be measured; each time, an estimated true value is selected from the set of estimated true values ​​corresponding to each measured value to be measured, and all estimated true value sequences are constructed according to the acquisition time sequence of the measured values ​​to be measured.

[0015] Preferably, each error region has an error probability corresponding to its error representative value. The method for calculating the probability of the estimated true value sequence includes: obtaining the error probability of each estimated true value corresponding to its error representative value in the estimated true value sequence, and multiplying all error probabilities together to obtain the probability of the estimated true value sequence.

[0016] Preferably, the step of calculating the matching degree between a single battery cell and each adjacent battery cell based on the estimated true value sequence of the battery cell and its probability includes: pairing all estimated true value sequences of a single battery cell with all estimated true value sequences of adjacent battery cells one by one; calculating the Pearson correlation coefficient between each pair of estimated true value sequences; and calculating the sum of the products of the probability corresponding to each pair of estimated true value sequences and the Pearson correlation coefficient to obtain the matching degree.

[0017] Preferably, the step of calculating the consistency score of the battery under test based on the consistency match count of each cell and the total number of adjacent cells includes: for a single cell, calculating the ratio of the consistency match count of that cell to the total number of adjacent cells; summing the ratios of all cells, and using the ratio of the sum to the total number of all cells as the consistency score of the battery under test.

[0018] In a second aspect, a lithium battery voltage anomaly detection system includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned lithium battery voltage anomaly detection method is implemented.

[0019] The present invention has the following effects:

[0020] 1. By analyzing the matching degree of the estimated true value sequence between battery cells, this invention can identify cells that are inconsistent with the behavior of other battery cells before the voltage value has significantly exceeded the absolute threshold (a fixed numerical range based on the voltage measurement value itself, which is commonly used in the prior art), thus achieving early detection of abnormal battery cells.

[0021] 2. This invention effectively overcomes system measurement errors. By incorporating system measurement errors into the analysis framework, it significantly reduces the negative impact of measurement accuracy on detection results and improves the robustness and accuracy of the method.

[0022] 3. This invention comprehensively utilizes multi-state information to improve reliability. It comprehensively examines the consistency performance of battery cells under three different states: charging, resting, and discharging. This avoids misjudgments caused by interference from specific operating conditions, making the diagnostic conclusions more comprehensive and reliable. Attached Figure Description

[0023] Figure 1 This is a flowchart of steps S1-S6 in a lithium battery voltage anomaly detection method according to an embodiment of the present invention. Detailed Implementation

[0024] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0025] A battery pack typically consists of dozens to hundreds of cells connected in series and parallel, and its overall performance is limited by the worst-performing individual cell. Therefore, quickly and accurately identifying defective cells with potential problems during the production quality inspection process is crucial to ensuring the quality of the battery pack before it leaves the factory. This invention proposes a method for detecting abnormal voltage in lithium batteries, applicable to lithium battery packs composed of multiple cells connected in series, enabling early and accurate diagnosis of voltage anomalies.

[0026] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0027] Reference Figure 1 A method for detecting abnormal lithium battery voltage includes steps S1-S6, as detailed below:

[0028] S1: Collect the reference measured value and reference true value of the voltage of each cell of the historical standard battery under the three states of charging, resting and discharging, and collect the measured value of the voltage of each cell of the battery under test under the three states of charging, resting and discharging.

[0029] The battery management system (BMS) is used to collect reference voltage measurements for each cell of multiple historical standard batteries under three states: charging, resting, and discharging. This yields the timing sequence of the reference measurement values ​​for each cell in a single state. A high-precision measuring instrument is then used to collect the reference true voltage values ​​for each cell of the same battery under the same three states, yielding the timing sequence of the reference true voltage values ​​for each cell in a single state. This process prepares for subsequent calculations of the error between the reference measurements and the reference true values.

[0030] The battery management system is used to collect the voltage measurement values ​​of each cell of the battery under test in three states: charging, resting, and discharging, so as to obtain the timing sequence of the measurement values ​​of each cell of the battery under test in a single state.

[0031] By collecting reference real values ​​and reference measured values, a data foundation was laid for the subsequent accurate establishment of an error model for the battery management system. Simultaneously, data from the battery under test was collected throughout its entire lifecycle under three typical conditions: charging, resting, and discharging. This provided comprehensive information support for the subsequent integrated analysis of the cell's behavior under different operating conditions, overcoming the limitations of single-state diagnosis.

[0032] S2: Based on the reference measured values ​​and reference true values ​​of all cells of the historical standard battery under the same condition, obtain the error value set of each measured value of each cell of the battery under test under this condition, divide the error accuracy range of the battery management system into several error regions, and calculate the error representative value and error probability of the measured value corresponding to the error value set in each error region based on the error value set.

[0033] Taking a historical standard battery and a battery under test in a charging state as an example, each cell of the battery under test corresponds to a timing sequence of the measured value under charging state. The method for obtaining the set of error values ​​of each measured value under test in the timing sequence includes: setting a preset deviation range, traversing the reference measured values ​​of all cells of the historical standard battery in a charging state, finding all reference measured values ​​that meet the preset deviation range, calculating the difference between the reference measured value and the corresponding reference true value, and obtaining the set of error values ​​of the measured value under test.

[0034] To illustrate this more clearly, let's take a specific example: For instance, the measured value is 2V (volts), and the preset deviation range is... Searching for matching values ​​among all reference measurements of historical standard batteries If there are ten compliant reference measurements within this range, then calculate the difference between each of these ten reference measurements and the corresponding true reference value to obtain ten error values. These ten error values ​​form the set of error values ​​corresponding to the measured value.

[0035] The error accuracy range of the battery management system can be directly obtained. The deviation range of the measured value can be set according to the error accuracy range. For example, the values ​​at the two ends of the deviation range can be one-tenth of the values ​​at the two ends of the error accuracy range.

[0036] The error accuracy range of the battery management system is evenly divided into ten error regions (the specific values ​​can be adjusted according to actual conditions). Based on the set of error values, the representative error value and error probability of the measured value corresponding to the set of error values ​​are calculated for each error region. The calculation method for the representative error value and error probability includes: for each error region, the average value of the error values ​​falling into that error region is calculated as the representative error value of the measured value in that error region; the ratio of the number of error values ​​falling into that error region to the total number of error values ​​in the set of error values ​​is calculated to obtain the error probability of the measured value in that error region.

[0037] This allows us to obtain the representative error value and error probability of each measured value for each cell of the battery under test in each error region during charging. This provides data support for subsequently constructing an estimated true value sequence.

[0038] The representative error value and error probability of a single measured value in each error region are expressed as follows:

[0039]

[0040] in, The first error region represents the representative error value and the representative error value. The error probability, here the error probability is ; The second error region represents the representative error value and the representative error value. The error probability, here the error probability is ; Indicates the first The representative error value and the representative error value for each error region The error probability, here the error probability is .

[0041] By constructing representative error values ​​and error probabilities for each measurement in various error regions, the impact of measurement uncertainties in the battery management system is quantified. This method formally incorporates systematic errors into the analytical framework, providing a theoretical basis for the "error-free" analysis in subsequent steps and fundamentally improving the method's robustness to systematic errors.

[0042] S3: Based on all the measured values ​​to be tested for a single cell, as well as the corresponding error representative value and error probability, generate all estimated true value sequences for that cell and the probability of each estimated true value sequence.

[0043] Still using the charging state as the background, each cell of the battery under test in the charging state corresponds to a timing sequence of the measurement value to be measured, represented as follows: ,in This represents the measured value corresponding to the first data acquisition time point. This represents the total number of collection times, then the first... The measured values ​​corresponding to each acquisition time point are represented as follows: Each measured value corresponds to a representative error value and error probability in each error region.

[0044] For each measured value to be measured in a single cell, the measured value to be measured is added to the error representative value of each corresponding error region to obtain multiple estimated true values, forming a set of estimated true values ​​of the measured value to be measured. Each time, an estimated true value is selected from the set of estimated true values ​​corresponding to each measured value to be measured, and a sequence of all estimated true values ​​is formed according to the acquisition time order of the measured values ​​to be measured.

[0045] A single estimated true value sequence is at each collection time point The measured value to be measured Select an error representative value from the set of estimated true values. and the measured value to be measured It is formed by addition. Among them, Indicates the first The data collection time point was selected at the [number]th time point. One error region, , This represents the total number of error regions. Therefore, the estimated true value sequence for a single battery cell can be expressed by the following formula:

[0046]

[0047] In the formula, Indicates the first number of the battery under test A sequence of estimated true values ​​for each battery cell; Indicates the selection vector. This is used to identify a specific combination of error representative values; Indicates the first The measured value corresponding to each acquisition time; Indicates the first The selected collection time is the first one The representative error value corresponding to each error region , This indicates the total number of error regions. This represents the total collection time.

[0048] Each error region has a corresponding error probability. The method for calculating the probability of the estimated true value sequence includes: obtaining the error probability of the error representative value corresponding to each estimated true value in the estimated true value sequence, and multiplying all error probabilities to obtain the probability of the estimated true value sequence. The specific formula is as follows:

[0049]

[0050] In the formula, Indicates the battery under test. Estimated true value sequence of individual battery cells The probability of; Indicates the time of collection Selected error representative value The probability of error; This represents the total collection time.

[0051] By traversing all vectors You can then obtain the battery cells. All estimated true value sequences and the probability of each estimated true value sequence. Theoretically, there are a total of There are several ways to estimate the true value sequence. In actual calculations, optimization methods (such as dynamic programming) can be used to avoid explicit enumeration.

[0052] Similarly, we can obtain all the estimated true value sequences for each cell of the battery under test in the charging state, as well as the probability of each estimated true value sequence.

[0053] By generating all possible estimated true value sequences and their probabilities for each cell, all possible measurement error scenarios are taken into account. This achieves a paradigm shift from pursuing a single precise value to covering probabilistic uncertainty, enabling subsequent analysis to be built on a more robust probability space that encompasses all possible real-world scenarios, thereby bypassing the dependence on the absolute precision of measurement values.

[0054] S4: Based on the estimated true value sequence of the battery cell and its probability, calculate the matching degree between a single battery cell and each adjacent battery cell. When the matching degree is greater than a preset threshold, update the consistency matching count of the battery cell.

[0055] The method for calculating the matching degree between a single battery cell and each adjacent battery cell includes: pairing all estimated true value sequences of a single battery cell with all estimated true value sequences of adjacent battery cells one by one; calculating the Pearson correlation coefficient between each pair of estimated true value sequences; and calculating the sum of the products of the probability corresponding to each pair of estimated true value sequences and the Pearson correlation coefficient to obtain the matching degree. The specific formula is as follows:

[0056]

[0057] In the formula, Indicates the first number of the battery under test The cell and the adjacent cell The degree of matching between individual battery cells; Indicates the first Estimated true value sequence of individual battery cells The probability of; Represented as the first The selection vector defined for each battery cell This is used to identify a specific combination of error representative values; Indicates the first Estimated true value sequence of individual battery cells The probability of; Represented as the first The selection vector defined for each battery cell This is used to identify a specific combination of error representative values; Indicates traversing the first... All selection vectors for each cell and the All selection vectors for each cell ; Represents the estimated true value sequence With the estimated true value sequence The Pearson correlation coefficient between them is calculated using a method known in the art and will not be elaborated here.

[0058] Thus, the first The matching degree between a cell and all its adjacent cells, the first Consistency matching count of individual cells Initially set to 0, with a preset threshold. Whenever the first Matching degree between individual cells and adjacent cells Then the first Consistency matching count of individual cells Add 1, if the first Matching degree between individual cells and adjacent cells Then keep the first Current consistency match count for each cell The consistency match count is used to count the number of times a match score greater than a threshold occurs.

[0059] Similarly, the consistency matching count of each cell of the battery under test in the charging state is obtained.

[0060] This step is crucial for achieving early and accurate diagnosis in this solution. By calculating the matching degree of two cells under all possible error scenarios, it essentially compares their consistency in dynamic voltage change trends. Even if the estimated true values ​​of two cells are similar due to errors, their matching degree will decrease if their inherent change patterns (trends) differ. This allows this solution to keenly detect early abnormal cells whose voltage values ​​are still within the normal range, but whose behavior patterns have begun to deviate from the norm.

[0061] S5: Calculate the consistency score of the battery under test based on the consistency matching count of each cell and the total number of adjacent cells.

[0062] The method for calculating the consistency score of the battery under test in the charging state includes: for a single cell, calculating the ratio of the consistency match count of the cell to the total number of adjacent cells; summing the ratios of all cells, and using the ratio of the sum to the total number of all cells as the consistency score of the battery under test.

[0063] Similarly, the consistency score of the battery under test in the static and discharged states is obtained.

[0064] By normalizing the matching count (divided by the total number of adjacent cells), the structural bias of the battery pack topology on the score is eliminated, making cells in different locations and battery packs of different sizes comparable.

[0065] S6: Obtain the consistency score of the battery under test in three states: charging, resting, and discharging. When the smallest consistency score is less than a preset value, the voltage of the battery under test is determined to be abnormal.

[0066] After obtaining the consistency scores of the battery under test in the three states of charging, resting, and discharging, the lowest consistency score is selected and compared with a preset value. When the lowest consistency score is less than the preset value, the voltage of the battery under test is determined to be abnormal.

[0067] Selecting the minimum consistency score across charging, resting, and discharging states as the final criterion reflects the "weakest link" principle. The overall health of a battery pack is determined by its weakest point. This judgment strategy is extremely conservative and reliable, ensuring that any inconsistency exhibited by the tested battery under any operating condition can be effectively detected, greatly reducing the risk of missed detections and improving the reliability of the test results.

[0068] The preset value is determined by the performance indicators of a large number of historical standard battery samples. Specifically, the consistency score of each historical standard battery sample is calculated under three states: charging, resting, and discharging. The lowest consistency score is selected, and the average of the lowest consistency scores of all historical standard battery samples is calculated as the preset value. The calculation method for the consistency score of historical standard batteries is the same as that for the battery under test, and will not be repeated here.

[0069] In addition, the threshold in step S4 Alternatively, the settings can be based on a large number of historical standard battery samples. Specifically, a large number of adjacent cell pairs are randomly selected from the historical standard battery samples, the matching degree of each pair of adjacent cells is calculated, the mean and standard deviation of all matching degrees are calculated, and the mean is subtracted from the standard deviation. The threshold is obtained by multiplying the standard deviation by a factor of 1. ,in 2 or 3 can be selected. The calculation method for the matching degree of adjacent cells in historical standard batteries is the same as that for adjacent cells in the battery under test, and will not be repeated here.

[0070] This application also discloses a lithium battery voltage anomaly detection system, the system including a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement the lithium battery voltage anomaly detection method according to the above embodiments of the present invention.

[0071] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0072] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A lithium battery voltage abnormality detection method, characterized by, The application relates to a battery voltage consistency detection method and device. The reference measurement value and the reference true value of each battery cell of a historical standard battery in charging, static and discharging states are collected, and the measurement value of each battery cell of a to-be-detected battery in the three states is collected; The consistency score of the to-be-detected battery in the three states is calculated, and when the minimum consistency score is less than a preset value, the voltage of the to-be-detected battery is determined to be abnormal; The consistency score in a single state is calculated by the following method: The error value set of each measurement value of each battery cell of the to-be-detected battery in the same state is obtained according to the reference measurement value and the reference true value of all battery cells of the historical standard battery in the same state, the error accuracy range of the battery management system is divided into a plurality of error regions, the error representative value and the error probability of the measurement value corresponding to the error value set in each error region are calculated based on the error value set, the estimated true value sequence of a single battery cell and the probability of each estimated true value sequence are generated based on all measurement values of the battery cell and the corresponding error representative value and error probability, the matching degree between the single battery cell and each adjacent battery cell is calculated according to the estimated true value sequence of the battery cell and the probability, the consistency matching count of the battery cell is updated when the matching degree is greater than a preset threshold, and the consistency score of the to-be-detected battery is calculated according to the consistency matching count of each battery cell and the total number of adjacent battery cells; The estimated true value sequence of each battery cell in a single state is formed by the following method: for each measurement value of a single battery cell, the measurement value and the error representative value of each error region corresponding to the measurement value are added respectively to obtain a plurality of estimated true values, and the estimated true value set of the measurement value is formed; one estimated true value is selected from the estimated true value set corresponding to each measurement value each time, and all estimated true value sequences are constituted in the order of the collection time of the measurement value. The error representative value of each error region corresponds to an error probability, and the probability of the estimated true value sequence is calculated by the following method: the error probability of the error representative value corresponding to each estimated true value in the estimated true value sequence is obtained, and all error probabilities are multiplied to obtain the probability of the estimated true value sequence.

2. The lithium battery voltage abnormality detection method of claim 1, wherein The error value set of each measurement value is obtained by the following method: a preset deviation range is set, all reference measurement values of all battery cells of the historical standard battery in the same state are traversed, all reference measurement values meeting the preset deviation range are found out, the difference between the reference measurement value and the corresponding reference true value is calculated, and the error value set of the measurement value is obtained.

3. The lithium battery voltage abnormality detection method of claim 1, wherein The error representative value of the measurement value in each error region is calculated based on the error value set by the following method: for each error region, the average value of the error values in the error value set falling into the error region is calculated as the error representative value of the measurement value in the error region.

4. The lithium battery voltage abnormality detection method of claim 3, wherein The error probability of the measurement value in each error region is calculated based on the error value set by the following method: for each error region, the ratio of the number of error values falling into the error region to the total number of error values in the error value set is calculated to obtain the error probability of the measurement value in the error region.

5. The lithium battery voltage abnormality detection method of claim 1, wherein The calculating the matching degree between the single battery cell and each adjacent battery cell according to the estimated real value sequence of the battery cell and the probability thereof comprises: pairing all the estimated real value sequences of the single battery cell with all the estimated real value sequences of the adjacent battery cell one by one; calculating the Pearson correlation coefficient value between each pair of estimated real value sequences; for each pair of estimated real value sequences, multiplying the probabilities of the two estimated real value sequences respectively to obtain a first product, multiplying the first product with the Pearson correlation coefficient value between the two estimated real value sequences to obtain a second product; summing up the second products corresponding to each pair of estimated real value sequences to obtain the matching degree.

6. The lithium battery voltage abnormality detection method of claim 1, wherein The calculating the consistency score of the battery under test according to the consistency matching count of each battery cell and the total number of the adjacent battery cells thereof comprises: for a single battery cell, calculating the ratio of the consistency matching count of the battery cell to the total number of the adjacent battery cells; summing up the ratios of all the battery cells, and taking the ratio of the sum value to the total number of all the battery cells as the consistency score of the battery under test.

7. A lithium battery voltage abnormality detection system characterized by comprising: The method comprises: A processor and a memory, wherein the memory stores computer program instructions which, when executed by the processor, implement the lithium battery voltage anomaly detection method according to any one of claims 1-6.

Images