Battery cell outlier state detection method and device, electronic equipment and storage medium

Abstract

The invention discloses a battery cell outlier state detection method and device, electronic equipment and a storage medium, and relates to the technical field of battery detection. The method comprises the following steps: acquiring a multi-source sensor parameter and a frequency domain characteristic parameter in a battery, and acquiring a time characteristic parameter according to the multi-source sensor parameter; acquiring a feature matrix according to the multi-source sensor parameters, the frequency domain feature parameters and the time feature parameters; performing principal component analysis on the feature matrix to obtain a plurality of target principal components; calculating a score vector of the feature matrix on each target principal component; and obtaining a battery cell outlier detection result according to the score vector. Therefore, principal component analysis is carried out by utilizing the multi-dimensional battery parameters, early-stage tiny anomalies are accurately identified, and the outlier detection accuracy of the battery cell is improved.

Description

Technical Field

[0001] The present invention relates to the field of battery detection, and in particular to a method, device, electronic device and storage medium for detecting the outlier state of battery cells. Background Art

[0002] With the gradually increasing proportion of electrochemical energy storage in the energy storage field, and the optimization of materials and processes leading to a gradual reduction in the cost per kilowatt-hour, how to improve the power consumption safety throughout the life cycle has become an urgent problem to be solved in the current energy storage industry; single cells need to be connected in series and parallel to form modules or battery clusters to increase capacity and power. Any deviation in the position of one cell during operation or a reduction in its own operating range will affect the maximum available capacity and power of the entire battery pack. Moreover, the state of health (SOH) of a cell with a low SOH will decay slightly faster than other cells during operation due to more frequent deep discharges or full charges. Eventually, when a single cell decays to a certain extent, it will result in a forced reduction in the available capacity and power of the entire energy storage system. How to identify minor changes in a certain cell within the battery pack in advance has become a key problem to be solved.

[0003] In the current technology, during operation, the problem of inconsistent state of charge (SOC) and inconsistent SOH is basically identified only by collecting voltage, and in most cases, an alarm is given only when the pressure difference between the maximum and minimum voltages exceeds a simple voltage threshold set, resulting in false alarms in judgment due to insufficient input data. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to overcome the deficiencies in the prior art, and provide a method, device, electronic device and storage medium for detecting the outlier state of battery cells, which are used for battery safety monitoring in scenarios such as electric vehicles and energy storage systems, and improve the reliability and safety of the battery pack throughout the life cycle.

[0005] The present invention provides the following technical solutions:

[0006] In a first aspect, the present invention proposes a method for detecting the outlier state of battery cells, including:

[0007] Obtain multi-source sensor parameters and frequency-domain characteristic parameters in the battery, and obtain time characteristic parameters according to the multi-source sensor parameters;

[0008] Obtain a feature matrix according to the multi-source sensor parameters, the frequency-domain characteristic parameters and the time characteristic parameters;

[0009] Perform principal component analysis on the feature matrix to obtain a plurality of target principal components;

[0010] Calculating the score vector of the feature matrix on each of the target principal components;

[0011] Outlier identification is performed according to the score vector to obtain a battery cell outlier detection result.

[0012] In one embodiment, the acquiring a feature matrix according to the multi-source sensor parameters, the frequency domain feature parameters, and the time feature parameters includes:

[0013] performing standardization processing on the multi-source sensor parameters to obtain multi-source sensor input quantities;

[0014] Performing standardization on the time characteristic parameters to obtain a time characteristic input;

[0015] Performing dimensionality reduction processing on the frequency domain feature parameters to obtain a frequency domain feature input;

[0016] The characteristic matrix is determined according to the multi-source sensing input, the time characteristic input and the frequency domain characteristic input.

[0017] In one embodiment, the performing dimensionality reduction processing on the frequency domain characteristic parameters to obtain the frequency domain characteristic parameter input includes:

[0018] Dividing the frequency domain characteristic parameters into a first frequency region, a second frequency region, and a third frequency region according to a preset frequency threshold;

[0019] Calculating the mean of all real part values and imaginary part values corresponding to each frequency point in the first frequency region to obtain a first real part mean value and a first imaginary part mean value;

[0020] Calculating the mean of all real part values and imaginary part values corresponding to each frequency point in the second frequency region to obtain a second real part mean and a second imaginary part mean;

[0021] The first real part mean, the first imaginary part mean, the second real part mean, the second imaginary part mean, and the real part values and imaginary part values corresponding to the frequency points in the third frequency zone are determined as the frequency domain feature parameter inputs.

[0022] In one embodiment, the principal component analysis is performed on the feature matrix to obtain multiple target principal components, including:

[0023] Obtaining the contribution of the multi-source sensor parameters, the frequency domain characteristic parameters, and the time characteristic parameters to battery failure respectively;

[0024] Determining a weight matrix according to the contribution;

[0025] Calculating a weighted covariance matrix based on the feature matrix and the weight matrix;

[0026] Determine each of the target principal components according to a preset contribution rate threshold and the weighted covariance matrix.

[0027] In one embodiment, the determining each of the target principal components according to a preset contribution rate threshold and the weighted covariance matrix includes:

[0028] Perform singular value decomposition on the weighted covariance matrix to obtain eigenvalues and eigenvectors;

[0029] Determine a principal component set according to the eigenvalues and the eigenvectors;

[0030] Determine each of the target principal components from the principal component set according to the preset contribution rate threshold.

[0031] In one embodiment, the calculating a score vector of the feature matrix on each of the target principal components includes:

[0032] Construct a PCA plane according to each of the target principal components, and project each input quantity in the feature matrix onto the PCA plane to obtain a plurality of reconstructed values;

[0033] Calculate the mean of all the reconstructed values, and subtract the mean from each of the reconstructed values respectively to obtain the score vector of the feature matrix on each of the target principal components.

[0034] In one embodiment, the battery cell outlier detection result includes a global outlier detection result, and the obtaining the battery cell outlier detection result by performing outlier identification according to the score vector includes:

[0035] Calculate a reconstructed covariance matrix according to each of the reconstructed values;

[0036] Calculate a T 2 statistic according to the score vector and the reconstructed covariance matrix;

[0037] If the T 2 statistic is greater than a preset statistic threshold, determine the global outlier detection result, where the global outlier detection result is used to indicate that the battery has a battery cell outlier.

[0038] In one embodiment, the battery cell outlier detection result includes a local outlier detection result, and the obtaining the battery cell outlier detection result by performing outlier identification according to the score vector further includes:

[0039] Calculate a reconstruction error according to the feature matrix and the reconstructed values;

[0040] If the reconstruction error is greater than a preset error threshold, identify the reconstructed values based on the local outlier factor algorithm to obtain abnormal sample points;

[0041] Obtain the local outlier detection result according to the abnormal sample points, and the local outlier detection result is used to indicate the abnormal battery cells of the battery.

[0042] In one embodiment, the obtaining the local outlier detection result according to the abnormal sample points includes:

[0043] Obtain the relaxation voltage decay rate of the abnormal sample points;

[0044] If the relaxation voltage decay rate is within a preset abnormal range, determine the outlier cells in the battery according to the position information of the abnormal sample points, and use the outlier cells as the local outlier detection result.

[0045] In a second aspect, the present invention provides a device for detecting the outlier state of battery cells, including:

[0046] A first acquisition module, configured to acquire multi-source sensor parameters and frequency-domain characteristic parameters in the battery, and acquire time-domain characteristic parameters according to the multi-source sensor parameters;

[0047] A second acquisition module, configured to acquire a feature matrix according to the multi-source sensor parameters, the frequency-domain characteristic parameters, and the time-domain characteristic parameters;

[0048] An analysis module, configured to perform principal component analysis on the feature matrix to obtain a plurality of target principal components;

[0049] A calculation module, configured to calculate a score vector of the feature matrix on each of the target principal components;

[0050] A third acquisition module, configured to perform outlier recognition according to the score vector to obtain a detection result of the outlier state of the battery cells.

[0051] In a third aspect, the present invention provides an electronic device, including a memory and a processor, where the memory stores a computer program, and when the computer program is executed by the processor, the method for detecting the outlier state of battery cells as described in the first aspect is implemented.

[0052] In a fourth aspect, the present invention provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the method for detecting the outlier state of battery cells as described in the first aspect is implemented.

[0053] The battery cell outlier detection method, device, electronic device, and storage medium disclosed in the present invention obtain multi-source sensor parameters and frequency domain characteristic parameters in the battery, and obtain time characteristic parameters based on the multi-source sensor parameters; obtain a feature matrix based on the multi-source sensor parameters, the frequency domain characteristic parameters, and the time characteristic parameters; perform principal component analysis on the feature matrix to obtain multiple target principal components; calculate score vectors of the feature matrix on each of the target principal components; and perform outlier identification based on the score vectors to obtain battery cell outlier detection results. In this way, principal component analysis using multi-dimensional battery parameters can accurately identify early-stage minor anomalies and improve the accuracy of battery cell outlier detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope of protection of the present invention. In each of the drawings, similar components are numbered similarly.

[0055] Figure 1 A first flow chart of the battery cell outlier state detection method proposed in this embodiment is shown;

[0056] Figure 2 A second flow chart of the battery cell outlier state detection method proposed in this embodiment is shown;

[0057] Figure 3 A third flow chart of the battery cell outlier state detection method proposed in this embodiment is shown;

[0058] Figure 4 A fourth flow chart of the battery cell outlier state detection method proposed in this embodiment is shown;

[0059] Figure 5 FIG4 shows a fifth flow chart of the battery cell outlier state detection method proposed in this embodiment;

[0060] Figure 6 1 shows a sixth flow chart of the battery cell outlier state detection method proposed in this embodiment;

[0061] Figure 7 shows a seventh flow chart of the battery cell outlier state detection method proposed in this embodiment;

[0062] Figure 8 A structural schematic diagram of the battery cell outlier state detection device proposed in this embodiment is shown.

[0063] Description of the accompanying drawings:

[0064] 800 - Battery cell outlier state detection device; 801 - First acquisition module; 802 - Second acquisition module; 803 - Analysis module; 804 - Calculation module; 805 - Third acquisition module. Detailed implementation manners

[0065] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments.

[0066] Generally, the components of the embodiments of the present invention described and illustrated herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present invention.

[0067] In the following text, the terms "including", "having" and their cognates that can be used in various embodiments of the present invention are only intended to represent specific features, numbers, steps, operations, elements, components or combinations of the foregoing items, and should not be construed as first excluding the existence of one or more other features, numbers, steps, operations, elements, components or combinations of the foregoing items or increasing the possibility of one or more features, numbers, steps, operations, elements, components or combinations of the foregoing items.

[0068] In addition, the terms "first", "second", "third", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance.

[0069] Unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art to which the various embodiments of the present invention belong. The terms (such as those defined in a commonly used dictionary) will be interpreted as having the same meaning as the contextual meaning in the relevant technical field and will not be interpreted as having an idealized meaning or an overly formal meaning unless clearly defined in the various embodiments of the present invention.

[0070] The embodiments of the present disclosure provide a method for detecting the outlier state of battery cells. In the case of high-rate charge and discharge or floating charge, outlier cells are likely to cause local overheating or overvoltage, and parameters such as voltage, internal resistance, capacity, or temperature deviate significantly from other cells. Multi-source data can be monitored by multiple sensors for outlier identification, so as to be used for battery safety monitoring in scenarios such as electric vehicles and energy storage systems, and improve the reliability and safety of the battery pack throughout its life cycle. Please refer to Figure 1, the method for detecting the outlier state of the battery cell includes steps S101 to S105, and the following is a detailed description of each step.

[0071] Step S101, obtain multi-source sensor parameters and frequency-domain characteristic parameters in the battery, and obtain time characteristic parameters according to the multi-source sensor parameters.

[0072] In this embodiment, the battery includes multiple battery cells. Obtain the multi-source sensor parameters and frequency-domain characteristic parameters of each battery cell in the battery, and calculate the time characteristic parameters according to the multi-source sensor parameters. In other embodiments, parameters such as lithium ion concentration and positive / negative electrode potential difference can also be collected.

[0073] Among them, the multi-source sensor parameters are physical quantity parameters that can be directly collected, such as battery voltage, battery positive temperature, battery negative temperature, microstrain and other data, where microstrain is a mechanical parameter; the time characteristic parameters are calculated time-related characteristic parameters, such as the change of the applied voltage and relaxation voltage over time, the open circuit voltage (Open Circuit Voltage, OCV) data obtained after standing for a sufficient time, the charge and discharge plateau time, etc.; the frequency-domain characteristic parameters are, for example, Electrochemical Impedance Spectroscopy (EIS).

[0074] Step S102, obtain a feature matrix according to the multi-source sensor parameters, the frequency-domain characteristic parameters, and the time characteristic parameters.

[0075] In this embodiment, a feature matrix is obtained by using the multi-source sensor parameters, the frequency-domain characteristic parameters, and the time characteristic parameters. Specifically, the multi-source sensor parameters, the frequency-domain characteristic parameters, and the time characteristic parameters of each battery cell are used as the label data of the corresponding battery cell. This feature matrix includes the label data of each battery cell in the battery. Among them, for example, if the battery has N battery cells, the feature matrix can be expressed as X = [x1…x N T , x N = [x N1 …x Nn , N represents the number of battery cells, and n represents the number of label input quantities.

[0076] Please refer to Figure 2 , in a specific embodiment, step S102 includes steps S1021 to S1024, and the following is a detailed description of each step.

[0077] Step S1021, perform normalization processing on the multi-source sensor parameters to obtain multi-source sensing input quantities.

[0078] ​In this embodiment, the multi-source sensor parameters are standardized to obtain the standardized data, that is, the multi-source sensing input quantity with a mean of 0 and a standard deviation of 1.

[0079] Step S1022: Standardize the time feature parameters to obtain the time feature input quantity.

[0080] In this embodiment, the time feature parameters are processed to obtain the standardized data, that is, the time feature input quantity with a mean of 0 and a standard deviation of 1.

[0081] It should be noted that the standardization method can be standard score (Z-score), 0-1 standardization, robust standardization, etc. The expression of the standardization process can be: In the formula, is the standardized data, x i is the data before standardization, including multi-source sensor parameters and time feature parameters, and μ and σ are the mean and standard deviation of the overall data.

[0082] Since the mean and standard deviation of the overall population cannot be accurately obtained, it is advisable to use the mean and standard deviation of the sample during calculation, and the formula is changed to: In the formula, is the mean of the sample, and S is the standard deviation of the sample.

[0083] It should be added that for the missing values in the multi-source sensor parameters and frequency domain feature parameters, the mean filling method can be used to fill the missing data in the multi-source sensor parameters and frequency domain feature parameters.

[0084] Step S1023: Perform dimensionality reduction on the frequency domain feature parameters to obtain the frequency domain feature input quantity.

[0085] In this embodiment, the original frequency domain feature parameters usually have a high dimension. Direct use may lead to problems such as increased computational complexity and model overfitting. Therefore, it is necessary to perform dimensionality reduction on the frequency domain feature parameters to obtain the frequency domain feature input quantity after dimensionality reduction. In addition, the characteristic frequency band can also be dynamically adjusted according to the failure mode.

[0086] In a specific embodiment, step S1023 includes: dividing the frequency domain feature parameters into a first frequency region, a second frequency region, and a third frequency region according to a preset frequency threshold; calculating the mean values of all real part values and imaginary part values corresponding to each frequency point in the first frequency region to obtain a first real part mean value and a first imaginary part mean value; calculating the mean values of all real part values and imaginary part values corresponding to each frequency point in the second frequency region to obtain a second real part mean value and a second imaginary part mean value; determining the first real part mean value, the first imaginary part mean value, the second real part mean value, the second imaginary part mean value, and the real part value and imaginary part value corresponding to the frequency point in the third frequency region as the input quantity of the frequency domain feature parameters.

[0087] In this embodiment, the frequency domain feature parameters are divided into a first frequency region, a second frequency region, and a third frequency region according to a preset frequency threshold. For example, the preset frequency threshold is 1 kHz, and the frequency domain feature parameters with a frequency of 0.1 kHz to 20 kHz are obtained. Now, the frequency domain feature parameters are divided into a first frequency region [0.1 kHz, 1 kHz), a second frequency region (1 kHz, 20 kHz], and a third frequency region (1 kHz). In other embodiments, the frequency domain feature parameters can also be divided into four or more frequency regions according to data compression requirements or dimensionality reduction requirements.

[0088] For the first frequency region, calculate the mean values of all real part values and imaginary part values corresponding to each frequency point in the first frequency region to obtain a first real part mean value and a first imaginary part mean value. It is expressed as: z′1 = SUM(Z′ lowFreq ) / N1, z″1 = SUM(Z″ lowFreq ) / N1. In the formula, z′1 is the first real part mean value, z″1 is the first imaginary part mean value, Z′ lowFreq is the real part value corresponding to a frequency point in the first frequency region, Z″ lowFreq is the imaginary part value corresponding to a frequency point in the first frequency region, and N1 is the number of frequencies in the first frequency region.

[0089] For the second frequency region, calculate the mean values of all real part values and imaginary part values corresponding to each frequency point in the second frequency region to obtain a second real part mean value and a second imaginary part mean value. It is expressed as: z′2 = SUM(Z′ highFreq ) / N2, z″2 = SUM(Z″ high2req ) / N2. In the formula, z′2 is the second real part mean value, z″2 is the second imaginary part mean value, Z′ highFreq is the real part value corresponding to a frequency point in the second frequency region, Z″ highFreq is the imaginary part value corresponding to a frequency point in the second frequency region, and N2 is the number of frequencies in the second frequency region.

[0090] For the third frequency region, extract the frequency of 1 kHz as a separate feature: z′ 1kHz , z″1kHz 。

[0091] Further, the first real - part mean value, the first imaginary - part mean value, the second real - part mean value, the second imaginary - part mean value, and the real - part values and imaginary - part values corresponding to the frequency points in the third frequency region are determined as the input quantities of the frequency - domain characteristic parameters. By performing feature compression on the EIS high - dimensional data, the running time and resource occupancy of the algorithm are reduced, making the algorithm easier to be deployed on embedded devices for real - time monitoring.

[0092] Step S1024, determine the feature matrix according to the multi - source sensing input quantity, the time - feature input quantity, and the frequency - domain feature input quantity.

[0093] In this embodiment, the feature matrix is determined according to the multi - source sensing input quantity, the time - feature input quantity, and the frequency - domain feature input quantity of each battery cell for subsequent data calculation.

[0094] Step S103, perform principal component analysis on the feature matrix to obtain multiple target principal components.

[0095] In this embodiment, by performing principal component analysis on the feature matrix, multiple target principal components are obtained. The target principal components retain most of the information of the original data, thus retaining the key features of the battery data as much as possible while reducing the dimension.

[0096] Please refer to Figure 3 , in a specific embodiment, step S103 includes steps S1031 to S1034, and the following is a detailed description of each step.

[0097] Step S1031, obtain the contribution degrees of the multi - source sensor parameters, the frequency - domain feature parameters, and the time - feature parameters to battery failure respectively.

[0098] In this embodiment, the contribution degrees of the multi - source sensor parameters, the frequency - domain feature parameters, and the time - feature parameters to battery failure are obtained respectively. It can be obtained from the results of correlation analysis on a cluster of battery samples with known different aging degrees measured.

[0099] Step S1032, determine the weight matrix according to the contribution degrees.

[0100] In this embodiment, by calculating the weights through the contribution degrees, the influence degree of the battery input parameters on failure can be quantified. The weight matrix can be expressed as W = [W1…W N T 。

[0101] Step S1033, calculate the weighted covariance matrix according to the feature matrix and the weight matrix.

[0102] ​In this embodiment, a weighted covariance matrix C is calculated according to the feature matrix and the weight matrix wighted , which is expressed as:

[0103] Step S1034: Determine each of the target principal components according to a preset contribution rate threshold and the weighted covariance matrix.

[0104] In this embodiment, by decomposing the weighted covariance matrix and using the preset contribution rate threshold to select the first k principal components as the target principal components. For example, k can be 3, 5, or other values.

[0105] In a specific embodiment, step S1034 includes: performing singular value decomposition on the weighted covariance matrix to obtain eigenvalues and eigenvectors; determining a set of principal components according to the eigenvalues and the eigenvectors; and determining each of the target principal components from the set of principal components according to the preset contribution rate threshold.

[0106] In this embodiment, the weighted covariance matrix is subjected to singular value decomposition to obtain eigenvalues and eigenvectors. A set of vectors obtained by sorting the eigenvectors in descending order of eigenvalues is the set of principal components. The contribution rate of the first principal component is the largest, and the contribution rates of subsequent principal components decrease in turn. The individual contribution rate represents the proportion of the variance of this principal component in the total variance of all principal components, reflecting the amount of information it carries from the original data.

[0107] Further, determine the target principal components from the set of principal components according to a preset contribution rate threshold (such as 95%), that is, the k principal components whose cumulative contribution rate is greater than or equal to the preset contribution rate threshold in the order of arrangement in the set of principal components are used as the k target principal components.

[0108] Step S104: Calculate the score vector of the feature matrix on each of the target principal components.

[0109] In this embodiment, each input quantity in the original feature matrix is projected onto the selected k target principal components to obtain the scores of each input quantity on the target principal components, that is, the score vector of the feature matrix on each target principal component is obtained.

[0110] Please refer to Figure 4 , in a specific embodiment, step S104 includes steps S1041 to S1042, and the following will explain each step in detail.

[0111] Step S1041: Construct a PCA plane according to each of the target principal components, and project each input quantity in the feature matrix onto the PCA plane to obtain a plurality of reconstructed values.

[0112] In this embodiment, a principal component analysis (PCA) plane is constructed from k target principal components, and the PCA plane is a k-dimensional subspace. Each input quantity in the feature matrix is projected onto the PCA plane to obtain a reconstructed feature matrix, which includes multiple reconstructed values.

[0113] Step S1042: Calculate the mean of all the reconstructed values, and subtract the mean from each of the reconstructed values respectively to obtain the score vectors of the feature matrix on each of the target principal components.

[0114] In this embodiment, calculate the mean of all the reconstructed values as the benchmark for centralization. Subtract the mean from each reconstructed value to obtain the centralized reconstructed values. The centralized reconstructed values are the score vectors of the feature matrix on each target principal component.

[0115] Step S105: Perform outlier identification based on the score vectors to obtain the battery cell outlier detection result.

[0116] In this embodiment, the score vectors represent the relative positions of each sample on the PCA plane, reflecting the contribution degree of the samples on each principal component to battery failure. Thus, the battery cell outlier detection result is obtained based on the score vectors, thereby improving the detection accuracy, being able to detect early tiny multi-parameter coupling anomalies, intervening in maintaining the battery pack in advance, and reducing the maintenance cost when problems occur. Meanwhile, with lightweight algorithm design, it supports online monitoring and fast response.

[0117] Please refer to Figure 5 , in a specific embodiment, the battery cell outlier detection result includes a global outlier detection result, and step S105 includes steps S1051 to S1053. The following will elaborate on each step in detail.

[0118] Step S1051: Calculate the reconstructed covariance matrix according to each of the reconstructed values.

[0119] In this embodiment, calculate the reconstructed covariance matrix according to each of the reconstructed values in the reconstructed feature matrix, that is, the covariance matrix in the principal component space.

[0120] Step S1052: Calculate the T 2 statistic according to the score vectors and the reconstructed covariance matrix.

[0121] In this embodiment, the T 2 statistic can be called Hotelling's T 2 statistic. Hotelling's T 2 statistic is a hypothesis testing method in multivariate statistical analysis used to test whether a multivariate mean vector is equal to a specified value.

[0122] Calculate T according to the score vector and the reconstructed covariance matrix 2 statistic, specifically expressed as: In the formula, x new is the score vector, and C -1 is the inverse matrix of the reconstructed covariance matrix. By retaining sufficient principal components to capture the main variations of the data, the T 2 statistic measures the deviation of the sample from the standardized principal component scores and is used for outlier detection.

[0123] Step S1053, if the T 2 statistic is greater than the preset statistic threshold, then determine the global outlier detection result, and the global outlier detection result is used to indicate that the battery has a battery cell outlier.

[0124] In this embodiment, if the T 2 statistic is greater than the preset statistic threshold, it can be known that the battery cell may have a battery cell outlier, and at this time, an outlier warning is output.

[0125] Please refer to Figure 6 , in a specific embodiment, the battery cell outlier detection result includes a local outlier detection result, and step S105 further includes steps S601 to S603, and the following details each step.

[0126] Step S601, calculate the reconstruction error according to the feature matrix and the reconstructed value.

[0127] In this embodiment, calculate the reconstruction error SPE according to each input quantity in the feature matrix and each reconstructed value in the reconstructed feature matrix, specifically expressed as In the formula, is the reconstructed value.

[0128] Step S602, if the reconstruction error is greater than the preset error threshold, then identify the reconstructed value based on the local outlier factor algorithm to obtain abnormal sample points.

[0129] In this embodiment, if the reconstruction error is greater than the preset error threshold, the battery cell may show an outlier performance, and an outlier warning is output; and identify the reconstructed value based on the local outlier factor algorithm to obtain the samples in the sparse density region in the principal component space, that is, abnormal sample points.

[0130] Step S603, obtain the local outlier detection result according to the abnormal sample points, and the local outlier detection result is used to indicate the abnormal battery cell of the battery.

[0131] In this embodiment, obtain the position information of the abnormal battery cell according to the abnormal sample points, and output the outlier warning of the battery cell at the corresponding position.

[0132] Please refer to Figure 7 , in a specific embodiment, step S603 includes steps S6031 to S6032, and the following will elaborate on each step in detail.

[0133] Step S6031, obtain the relaxation voltage decay rate of the abnormal sample point.

[0134] In this embodiment, the relaxation voltage decay rate of the abnormal sample point is obtained, and the relaxation voltage decay rate is expressed as dV / dt.

[0135] Step S6032, if the relaxation voltage decay rate is within a preset abnormal range, determine the outlier cell in the battery according to the position information of the abnormal sample point, and use the outlier cell as the local outlier detection result.

[0136] In this embodiment, if the relaxation voltage decay rate is within a preset abnormal range, that is, deviates from the normal range (±10%), it is confirmed as a real outlier, and the outlier cell in the battery is determined according to the position information of the abnormal sample point, and the outlier cell is used as the local outlier detection result to output an early warning.

[0137] It should be noted that the global analysis of the T 2 statistic determines the cell abnormality to trigger the local analysis calculation based on the reconstruction error. When the local is still abnormal, the outlier battery samples are calculated and screened through the local outlier factor algorithm, and the screened cells are further verified by time series to obtain the final outlier detection result, so as to accurately identify early minor abnormalities.

[0138] The method for detecting the outlier state of battery cells proposed in this embodiment obtains multi-source sensor parameters and frequency-domain characteristic parameters in the battery, and obtains time characteristic parameters according to the multi-source sensor parameters; obtains a feature matrix according to the multi-source sensor parameters, the frequency-domain characteristic parameters and the time characteristic parameters; performs principal component analysis on the feature matrix to obtain multiple target principal components; calculates the score vectors of the feature matrix on each of the target principal components; performs outlier recognition according to the score vectors to obtain the outlier detection result of the battery cells. In this way, principal component analysis is performed using multi-dimensional battery parameters to accurately identify early minor abnormalities and improve the detection accuracy of outlier battery cells.

[0139] In addition, the embodiments of the present disclosure provide a device 800 for detecting the outlier state of battery cells. Please refer to Figure 8 , including:

[0140] The first acquisition module 801 is configured to acquire multi-source sensor parameters and frequency-domain characteristic parameters in the battery, and acquire time characteristic parameters according to the multi-source sensor parameters;

[0141] The second acquisition module 802 is configured to obtain a feature matrix according to the multi-source sensor parameters, the frequency-domain feature parameters, and the time-domain feature parameters;

[0142] The analysis module 803 is configured to perform principal component analysis on the feature matrix to obtain a plurality of target principal components;

[0143] The calculation module 804 is configured to calculate a score vector of the feature matrix on each of the target principal components;

[0144] The third acquisition module 805 is configured to perform outlier identification according to the score vector to obtain a battery cell outlier detection result.

[0145] Optionally, the second acquisition module 802 is further configured to perform standardization processing on the multi-source sensor parameters to obtain a multi-source sensing input quantity; perform standardization processing on the time-domain feature parameters to obtain a time-domain feature input quantity; perform dimensionality reduction processing on the frequency-domain feature parameters to obtain a frequency-domain feature input quantity; and determine the feature matrix according to the multi-source sensing input quantity, the time-domain feature input quantity, and the frequency-domain feature input quantity.

[0146] Optionally, the second acquisition module 802 is further configured to divide the frequency-domain feature parameters into a first frequency region, a second frequency region, and a third frequency region according to a preset frequency threshold; calculate the mean value of all real part values and imaginary part values corresponding to each frequency point in the first frequency region to obtain a first real part mean value and a first imaginary part mean value; calculate the mean value of all real part values and imaginary part values corresponding to each frequency point in the second frequency region to obtain a second real part mean value and a second imaginary part mean value; and determine the first real part mean value, the first imaginary part mean value, the second real part mean value, the second imaginary part mean value, and the real part values and imaginary part values corresponding to the frequency points in the third frequency region as the frequency-domain feature parameter input quantity.

[0147] Optionally, the analysis module 803 is further configured to obtain the contribution degrees of the multi-source sensor parameters, the frequency-domain feature parameters, and the time-domain feature parameters to battery failure respectively; determine a weight matrix according to the contribution degrees; calculate a weighted covariance matrix according to the feature matrix and the weight matrix; and determine each of the target principal components according to a preset contribution rate threshold and the weighted covariance matrix.

[0148] Optionally, the analysis module 803 is further configured to perform singular value decomposition on the weighted covariance matrix to obtain eigenvalues and eigenvectors; determine a principal component set according to the eigenvalues and the eigenvectors; and determine each of the target principal components from the principal component set according to the preset contribution rate threshold.

[0149] Optionally, the calculation module 804 is further configured to construct a PCA plane based on each of the target principal components, project each input quantity in the feature matrix onto the PCA plane to obtain a plurality of reconstruction values; calculate the mean value of all the reconstruction values, and subtract the mean value from each of the reconstruction values respectively to obtain a score vector of the feature matrix on each of the target principal components.

[0150] Optionally, the third acquisition module 805 is further configured to calculate a reconstruction covariance matrix according to each of the reconstruction values; calculate a T 2 statistic according to the score vector and the reconstruction covariance matrix; if the T 2 statistic is greater than a preset statistic threshold, determine the global outlier detection result, and the global outlier detection result is used to indicate that a battery cell in the battery is outlier.

[0151] Optionally, the battery cell outlier detection result includes a local outlier detection result, and the third acquisition module 805 is further configured to calculate a reconstruction error according to the feature matrix and the reconstruction values; if the reconstruction error is greater than a preset error threshold, identify the reconstruction values based on the local outlier factor algorithm to obtain abnormal sample points; obtain the local outlier detection result according to the abnormal sample points, and the local outlier detection result is used to indicate the abnormal battery cells in the battery.

[0152] Optionally, the third acquisition module 805 is further configured to obtain a relaxation voltage decay rate of the abnormal sample points; if the relaxation voltage decay rate is within a preset abnormal range, determine the outlier cell in the battery according to the position information of the abnormal sample points, and use the outlier cell as the local outlier detection result.

[0153] The device provided in the embodiments of the present disclosure can execute the steps of the battery cell outlier state detection method provided in the embodiments. To avoid repetition, details are not described herein again.

[0154] The battery cell outlier state detection device proposed in this embodiment acquires multi-source sensor parameters and frequency domain feature parameters in the battery, and obtains time feature parameters according to the multi-source sensor parameters; obtains a feature matrix according to the multi-source sensor parameters, the frequency domain feature parameters and the time feature parameters; performs principal component analysis on the feature matrix to obtain a plurality of target principal components; calculates a score vector of the feature matrix on each of the target principal components; performs outlier identification according to the score vector to obtain a battery cell outlier detection result. In this way, principal component analysis is performed using multi-dimensional battery parameters to accurately identify early tiny anomalies and improve the detection accuracy of battery cell outliers.

[0155] In addition, an embodiment of the present disclosure provides a computer device, including a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, the battery cell outlier state detection method described in the embodiment is implemented.

[0156] The device provided by the embodiment of the present disclosure can execute the steps of the battery cell outlier state detection method provided in the embodiment. To avoid repetition, it will not be elaborated here.

[0157] An embodiment of the present disclosure proposes a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the battery cell outlier state detection method described in this embodiment is implemented.

[0158] In this embodiment, the computer-readable storage medium may be a read-only memory (ROM for short), a random access memory (RAM for short), a magnetic disk, or an optical disc, etc.

[0159] The computer-readable storage medium provided in this embodiment can implement the battery cell outlier state detection method provided in the embodiment. To avoid repetition, it will not be elaborated here.

[0160] In all the examples shown and described here, any specific value should be construed as merely exemplary, not as a limitation. Therefore, other examples of the exemplary embodiments may have different values.

[0161] It should be noted that similar reference numerals and letters denote similar items in the following figures. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0162] The above-described embodiments merely represent several implementation manners of the present invention. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention.

Claims

1. A method for detecting the outlier state of a battery cell, characterized in that, Including: Obtain multi-source sensor parameters and frequency-domain characteristic parameters in the battery, and obtain time characteristic parameters according to the multi-source sensor parameters; Obtain a feature matrix according to the multi-source sensor parameters, the frequency-domain characteristic parameters, and the time characteristic parameters; Perform principal component analysis on the feature matrix to obtain a plurality of target principal components; Calculate the score vectors of the feature matrix on each of the target principal components; Perform outlier identification according to the score vectors to obtain the outlier detection result of the battery cell.

2. The method for detecting the outlier state of a battery cell according to claim 1, wherein The obtaining a feature matrix according to the multi-source sensor parameters, the frequency-domain characteristic parameters, and the time characteristic parameters includes: Perform normalization processing on the multi-source sensor parameters to obtain multi-source sensing input quantities; Perform normalization processing on the time characteristic parameters to obtain time characteristic input quantities; Perform dimensionality reduction processing on the frequency-domain characteristic parameters to obtain frequency-domain characteristic input quantities; Determine the feature matrix according to the multi-source sensing input quantities, the time characteristic input quantities, and the frequency-domain characteristic input quantities.

3. The method for detecting the outlier state of a battery cell according to claim 2, wherein, The performing dimensionality reduction processing on the frequency-domain characteristic parameters to obtain frequency-domain characteristic parameter input quantities includes: Divide the frequency-domain characteristic parameters into a first frequency region, a second frequency region, and a third frequency region according to a preset frequency threshold; Calculate the mean values of all real part values and imaginary part values corresponding to each frequency point in the first frequency region to obtain a first real part mean value and a first imaginary part mean value; Calculate the mean values of all real part values and imaginary part values corresponding to each frequency point in the second frequency region to obtain a second real part mean value and a second imaginary part mean value; Determine the first real part mean value, the first imaginary part mean value, the second real part mean value, the second imaginary part mean value, and the real part values and imaginary part values corresponding to the frequency points in the third frequency region as the frequency-domain characteristic parameter input quantities.

4. The method for detecting the outlier state of a battery cell according to claim 1, wherein The performing principal component analysis on the feature matrix to obtain a plurality of target principal components includes: Obtain the contribution degrees of the multi-source sensor parameters, the frequency-domain characteristic parameters, and the time characteristic parameters to battery failure respectively; Determine a weight matrix according to the contribution degrees; Calculate a weighted covariance matrix according to the feature matrix and the weight matrix; Determine each of the target principal components according to a preset contribution rate threshold and the weighted covariance matrix.

5. The method for detecting the outlier state of a battery cell according to claim 4, wherein The determining each of the target principal components according to a preset contribution rate threshold and the weighted covariance matrix includes: Perform singular value decomposition on the weighted covariance matrix to obtain eigenvalues and eigenvectors; Determine a principal component set according to the eigenvalues and the eigenvectors; Determine each of the target principal components from the principal component set according to the preset contribution rate threshold.

6. The method for detecting the outlier state of a battery cell according to claim 1, wherein, The calculating the score vectors of the feature matrix on each of the target principal components includes: Construct a PCA plane according to each of the target principal components, and project each input quantity in the feature matrix onto the PCA plane to obtain a plurality of reconstructed values; Calculate the mean value of all the reconstructed values, and subtract the mean value from each of the reconstructed values respectively to obtain the score vectors of the feature matrix on each of the target principal components.

7. The method for detecting the outlier state of a battery cell according to claim 6, wherein The outlier detection result of the battery cell includes a global outlier detection result, and the obtaining the outlier detection result of the battery cell according to the score vectors includes: Calculate the reconstructed covariance matrix according to each of the reconstructed values; Calculate T according to the score vector and the reconstructed covariance matrix 2 statistic; If the described T 2 statistic is greater than a preset statistic threshold, then determine the global outlier detection result, and the global outlier detection result is used to indicate that a battery cell of the battery is outlier.

8. The method for detecting the outlier state of a battery cell according to claim 7, wherein The battery cell outlier detection result includes a local outlier detection result. The outlier identification based on the score vector to obtain the battery cell outlier detection result further includes: Calculate the reconstruction error according to the feature matrix and the reconstructed value; If the reconstruction error is greater than a preset error threshold, identify the reconstructed value based on the local outlier factor algorithm to obtain abnormal sample points; Obtain the local outlier detection result according to the abnormal sample points, and the local outlier detection result is used to indicate the abnormal battery cells of the battery.

9. The method for detecting the outlier state of a battery cell according to claim 8, wherein The obtaining the local outlier detection result according to the abnormal sample points includes: Obtain the relaxation voltage decay rate of the abnormal sample points; If the relaxation voltage decay rate is within a preset abnormal range, determine the outlier cells in the battery according to the position information of the abnormal sample points, and use the outlier cells as the local outlier detection result.

10. A battery cell outlier state detection device, characterized in that, Includes: A first acquisition module, configured to acquire multi-source sensor parameters and frequency-domain feature parameters in the battery, and acquire time feature parameters according to the multi-source sensor parameters; A second acquisition module, configured to acquire a feature matrix according to the multi-source sensor parameters, the frequency-domain feature parameters, and the time feature parameters; An analysis module, configured to perform principal component analysis on the feature matrix to obtain a plurality of target principal components; A calculation module, configured to calculate the score vector of the feature matrix on each of the target principal components; A third acquisition module, configured to perform outlier identification according to the score vector to obtain a battery cell outlier detection result.

11. An electronic device, characterized in that, Includes a memory and a processor, where the memory stores a computer program, and when the computer program is executed by the processor, it implements the battery cell outlier state detection method according to any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that, It stores a computer program, and when the computer program is executed by a processor, it implements the battery cell outlier state detection method according to any one of claims 1 to 9.

Images