Abnormality detection method and device and storage medium

Abstract

The invention discloses an anomaly detection method and device and a storage medium, and relates to the technical field of new energy batteries, and the method comprises the steps: obtaining one or more features of each battery cell for a plurality of battery cells; acquiring empirical cumulative distribution of each characteristic for the plurality of battery cells; obtaining outliers and abnormal scores of the plurality of battery cells based on the empirical cumulative distribution of each feature; and based on the outliers and the abnormal scores of the plurality of battery cells, determining whether there is an abnormality in the plurality of battery cells. The anomaly detection method provided by the invention has the technical effects that parameter adjustment is not needed, calculation is convenient and rapid, and the method is less limited by data distribution shapes.

Description

Technical Field

[0001] This application relates to the technical field of new energy batteries, and in particular, to an anomaly detection method, device, and storage medium. Background Art

[0002] In the new energy field, multiple single cells are often combined into a battery pack, and then based on the combination of multiple battery packs, electrical energy is provided for products such as electric vehicles. Therefore, anomaly detection of battery packs and cells before product shipment is of great significance for the safe use of products.

[0003] In the prior art, Six Sigma (6σ) is a technology for enterprise quality process management. According to the given upper specification limit (USL) and lower specification limit (LSL) of a product, the average value and standard deviation of product indicators are controlled, and USL and LSL are made to be at a position 6σ away from the average value as much as possible, so that the yield rate reaches about 99.99966%. This method is also often used in battery anomaly detection. For example, in the EOL (End of Line) test stage of battery production, for indicators such as cell voltage, cell internal resistance (DCR), and cell temperature, the position of 6σ is obtained by calculating the average value and variance, and then the cells outside this range are regarded as abnormal cells. Although the battery anomaly detection using the 6σ method is fast and convenient in calculation and is easy to be used as a reference for quality control, the 6σ method is based on the assumption of normal distribution, while the actual data distribution of EOL test results is usually asymmetric, and there is a quite large skewness in the distribution of each dimension of cell DCR, voltage, and temperature of a single battery pack. The number of abnormal situations given based on the fixed threshold of 6σ will not meet the expectations. Therefore, the 6σ method has limited effect in anomaly detection of EOL test results. And for the vast majority of traditional nearest neighbor-based anomaly detection algorithms (such as KNN, LOF) and machine learning model-based anomaly detection methods, due to the difficulty in obtaining labeled data and the need to train a large number of models and adjust hyperparameters, the anomaly detection effect using these methods is not very ideal.

[0004] Therefore, this application aims to propose an anomaly detection method that does not require parameter tuning, is computationally convenient, and is less restricted by the shape of the data distribution. Summary of the Invention

[0005] To overcome the above defects, the anomaly detection method, device, and storage medium proposed in this application can be used to solve or at least partially solve the technical problems in the prior art that the anomaly detection method needs to adjust hyperparameters, is computationally complex, and is restricted by the shape of the data distribution.

[0006] In a first aspect, this application provides an anomaly detection method, including:

[0007] For multiple battery cells, obtain one or more characteristics of each battery cell;

[0008] For the multiple battery cells, obtain the empirical cumulative distribution of each characteristic;

[0009] Based on the empirical cumulative distribution of each characteristic, obtain the outliers and anomaly scores of the multiple battery cells;

[0010] Based on the outliers and anomaly scores of the multiple battery cells, determine whether there are anomalies among the multiple battery cells.

[0011] Preferably, the empirical cumulative distribution includes a left-tail empirical cumulative distribution and / or a right-tail empirical cumulative distribution; correspondingly, the obtaining of the outliers and anomaly scores of the multiple battery cells based on the empirical cumulative distribution of each characteristic includes:

[0012] Based on the left-tail empirical cumulative distribution of one or more characteristics, obtain the outliers of the multiple battery cells; or based on the right-tail empirical cumulative distribution of one or more characteristics, obtain the outliers of the multiple battery cells;

[0013] Based on the left-tail empirical cumulative distribution and / or the right-tail empirical cumulative distribution of one or more characteristics, obtain the anomaly scores of the multiple battery cells.

[0014] Preferably, the outliers include a first outlier, a second outlier, and a third outlier;

[0015] The first outlier is obtained by aggregating the left-tail empirical cumulative distribution of the one or more characteristics;

[0016] The second outlier is obtained by aggregating the right-tail empirical cumulative distribution of the one or more characteristics;

[0017] The third outlier is obtained through the following steps:

[0018] Calculate the skewness corresponding to each characteristic;

[0019] Based on the skewness of each characteristic, re-determine the empirical cumulative distribution of this characteristic;

[0020] Aggregate the re-determined empirical cumulative distributions of each characteristic to obtain the third outlier of the multiple battery cells.

[0021] Preferably, the obtaining of the anomaly scores of the multiple battery cells based on the left-tail empirical cumulative distribution and / or the right-tail empirical cumulative distribution of one or more characteristics includes:

[0022] Determine a reference characteristic;

[0023] Based on the reference characteristic, re-determine the empirical cumulative distribution of each characteristic;

[0024] Obtain the anomaly scores of the multiple battery cells by aggregating the empirical cumulative distribution of the re-determined features.

[0025] Preferably, the determining of the reference feature includes: obtaining the reference feature based on the first derivative of the empirical cumulative distribution function;

[0026] The re-determining of the empirical cumulative distribution of each feature based on the reference feature includes: re-determining the empirical cumulative distribution of this feature based on the relative position of each feature and the reference feature.

[0027] Preferably, the obtaining of the reference feature based on the first derivative of the empirical cumulative distribution function includes:

[0028] Taking the first derivative of the empirical cumulative distribution function;

[0029] Taking the variable value corresponding to the maximum value of the first derivative as the reference feature.

[0030] Preferably, the re-determining of the empirical cumulative distribution of this feature based on the relative position of each feature and the reference feature includes:

[0031] In response to each feature being less than the reference feature, taking the left-tail empirical cumulative distribution of this feature as the empirical cumulative distribution of this feature;

[0032] In response to each feature being greater than or equal to the reference feature, taking the right-tail empirical cumulative distribution of this feature as the empirical cumulative distribution of this feature.

[0033] Preferably, the determining whether there is an anomaly in the multiple battery cells based on the outliers and anomaly scores of the multiple battery cells is specifically:

[0034] Obtaining the maximum value among the outliers and the anomaly scores;

[0035] When the maximum value exceeds a preset threshold, it is determined that there is an anomaly in the multiple battery cells.

[0036] In a second aspect, the present application provides a computer device, which includes a processor and a storage device. The storage device is adapted to store multiple program codes, and the program codes are adapted to be loaded and run by the processor to execute the anomaly detection method described in any one of the technical solutions of the above anomaly detection method.

[0037] In a third aspect, the present application provides a computer-readable storage medium, which stores multiple program codes therein, and the program codes are adapted to be loaded and run by a processor to execute the anomaly detection method described in any one of the technical solutions of the above anomaly detection method.

[0038] One or more of the above technical solutions of the present application have at least one or more of the following beneficial effects:

[0039] In implementing the technical solution of the present application, for a plurality of battery cells, one or more characteristics of each battery cell are obtained; for the plurality of battery cells, the empirical cumulative distribution of each characteristic is obtained; based on the empirical cumulative distribution of each characteristic, the outliers and abnormal scores of the plurality of battery cells are obtained; based on the outliers and abnormal scores of the plurality of battery cells, it is determined whether there are abnormalities among the plurality of battery cells. The abnormal detection method of the present application does not involve the adjustment of algorithm models and hyperparameters, can avoid the difficulties caused by it being difficult to obtain true labels, and is easy to use. In addition, the abnormal detection method of the present application does not require the characteristic data used for detection to be normally distributed, only assumes that the data distribution is unimodal and the data in each dimension are independent of each other, relaxes the normal distribution assumption, so that the detection of the characteristics of one dimension can simultaneously consider the tail probabilities on both the left tail and the right tail sides, thereby improving the accuracy of abnormal detection. Description of the Drawings

[0040] Referring to the accompanying drawings, the disclosure of the present application will become easier to understand. It is easy for those skilled in the art to understand that these drawings are only for illustrative purposes and are not intended to limit the protection scope of the present application. In addition, similar numbers in the drawings are used to represent similar components, where:

[0041] Figure 1 is a schematic flowchart of the main steps of the abnormal detection method according to an embodiment of the present application;

[0042] Figure 2 is an exemplary dimensional outlier graph drawn based on the sample abnormal scores obtained by the abnormal detection method according to an embodiment of the present application;

[0043] Figure 3 is a schematic block diagram of the composition of an abnormal detection device according to an embodiment of the present application. Detailed Embodiments

[0044] The following describes some embodiments of the present application with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principle of the present application and are not intended to limit the protection scope of the present application.

[0045] In the description of the present application, a "module" and a "processor" may include hardware, software, or a combination of both. A module may include a hardware circuit, various appropriate sensors, communication ports, a memory, and may also include a software part, such as program code, or may be a combination of software and hardware. A processor may be a central processing unit, a microprocessor, an image processor, a digital signal processor, or any other suitable processor. The processor has data and / or signal processing functions. The processor may be implemented in software, in hardware, or in a combination of both. A non-transitory computer-readable storage medium includes any suitable medium for storing program code, such as a magnetic disk, a hard disk, an optical disk, a flash memory, a read-only memory, a random access memory, and so on. The term "A and / or B" represents all possible combinations of A and B, such as only A, only B, or A and B. The term "at least one A or B" or "at least one of A and B" has a meaning similar to "A and / or B" and may include only A, only B, or A and B. The singular terms "a" and "the" may also include the plural form.

[0046] The preferred embodiments of the present application will be described below in conjunction with the accompanying drawings of the specification. It should be understood that the preferred embodiments described herein are only for the purpose of illustrating and explaining the present application, and are not used to limit the present application. And without conflict, the embodiments in the present application and the features in the embodiments may be combined with each other.

[0047] Embodiment 1

[0048] Refer to the attached Figure 1 , Figure 1 which is the main step flow of the anomaly detection method of an embodiment of the present application. In the anomaly detection provided by the solution of this embodiment, it is assumed that the data distribution is unimodal, and to a certain extent, the various features of the data are independent of each other, relaxing the normal distribution assumption. The anomaly detection method of this embodiment takes into account the tail probabilities on both sides of the data in each dimension when screening for abnormal data, avoiding the loss of abnormal data, and thus being able to effectively improve the accuracy of anomaly detection.

[0049] In a first aspect, the present application provides an anomaly detection method. As Figure 1 shown, the anomaly detection method of this embodiment mainly includes the following steps S11 to step S14, and the detailed introduction of each step is as follows.

[0050] Step S11: For a plurality of battery cells, obtain one or more features of each battery cell.

[0051] Optionally, the plurality of battery cells may be a plurality of battery cells from the same battery pack or a plurality of battery cells from a plurality of battery packs.

[0052] Optionally, the multiple features include at least one of a voltage value, a temperature value, and an internal resistance value.

[0053] Optionally, regarding each feature as a dimension of a single battery cell, one or more features of a battery cell can be represented as

[0054] Step S12: For the multiple battery cells, obtain the empirical cumulative distribution of each feature.

[0055] Among them, the empirical cumulative distribution of each feature is used to represent a parameter that statistically represents the distribution of any feature among the multiple features of the multiple battery cells.

[0056] In the first specific embodiment of step S12, the empirical cumulative distribution of each feature may include the left-tail empirical cumulative distribution of this feature.

[0057] In the second specific embodiment of step S12, the empirical cumulative distribution of each feature may include the right-tail empirical cumulative distribution of this feature.

[0058] In the third specific embodiment of step S12, the empirical cumulative distribution of each feature may include the left-tail empirical cumulative distribution and the right-tail empirical cumulative distribution of this feature.

[0059] It should be noted that the empirical cumulative distribution of each feature includes but is not limited to the above three specific embodiments, and any parameter that can be used to represent the statistics of the distribution of any feature among the multiple features of the multiple battery cells can be used as the empirical cumulative distribution obtained in step S12.

[0060] Step S13: Based on the empirical cumulative distribution of each feature, obtain the outlier values and anomaly scores of the multiple battery cells.

[0061] In step S13, the outlier values of the multiple battery cells are used to represent a result for predicting whether the multiple battery cells deviate from the normal battery cell distribution.

[0062] Optionally, the outlier values may include one or more values, which are respectively denoted as the first outlier value, the second outlier value, the third outlier value,....

[0063] In one embodiment, the outlier values of the multiple battery cells include three, namely the first outlier value, the second outlier value, and the third outlier value, and these three outlier values respectively reflect the outlier situation of the multiple battery cells from different dimensions.

[0064] Optionally, the first outlier value is obtained by aggregating the left-tail empirical cumulative distribution of the one or more features. For example, the left-tail empirical cumulative distributions of all features can be aggregated as the first outlier value, denoted as score_left.

[0065] Optionally, the second outlier is obtained by aggregating the right-tail empirical cumulative distribution of the one or more features. For example, the right-tail empirical cumulative distributions of all features can be aggregated as the second outlier, denoted as score_right.

[0066] Optionally, the third outlier (denoted as score_adjusted) can be obtained through the following steps: calculating the skewness corresponding to each feature; re-determining the empirical cumulative distribution of this feature based on the skewness of each feature; aggregating the re-determined empirical cumulative distributions of each feature to obtain the third outlier of the multiple battery cells.

[0067] Among them, the re-determining the empirical cumulative distribution of this feature based on the skewness of each feature specifically means: in response to the skewness value being negative, using the left-tail empirical cumulative distribution as the empirical cumulative distribution of the feature; in response to the skewness value being non-negative, using the right-tail empirical cumulative distribution as the empirical cumulative distribution of the feature.

[0068] It should be noted that the aggregation in this application generally refers to multiplying the features of the multiple battery cells in multiple empirical cumulative distributions and taking the negative logarithm.

[0069] In step S13, the outlier scores of the multiple battery cells are used to represent another result for predicting whether the multiple battery cells deviate from the normal battery cell distribution.

[0070] Optionally, the outlier scores of the multiple battery cells (denoted as score_twoside) can be obtained by aggregating the left-tail empirical cumulative distribution and / or the right-tail empirical cumulative distribution of one or more features.

[0071] Optionally, the outlier scores of the multiple battery cells are obtained through the following steps: determining a reference feature; re-determining the empirical cumulative distribution of each feature based on the reference feature; obtaining the outlier scores of the multiple battery cells by aggregating the re-determined empirical cumulative distributions of the features.

[0072] Optionally, the determining the reference feature includes: obtaining the reference feature based on the first derivative of the empirical cumulative distribution function; the re-determining the empirical cumulative distribution of each feature based on the reference feature includes: re-determining the empirical cumulative distribution of this feature based on the relative position of each feature and the reference feature.

[0073] Optionally, the obtaining the reference feature based on the first derivative of the empirical cumulative distribution function includes: taking the first derivative of the empirical cumulative distribution function; using the variable value corresponding to the maximum value of the first derivative as the reference feature.

[0074] Optionally, re - determining the empirical cumulative distribution of a feature based on the relative positions of each feature and the reference feature includes: in response to each feature being less than the reference feature, using the left - tail empirical cumulative distribution of this feature as the empirical cumulative distribution of this feature; in response to each feature being greater than or equal to the reference feature, using the right - tail empirical cumulative distribution of this feature as the empirical cumulative distribution of this feature.

[0075] Step S14: Determine whether there is an abnormality among the multiple battery cells based on the outliers and anomaly scores of the multiple battery cells.

[0076] In one embodiment, this step is specifically as follows: for multiple battery cells, obtain the maximum value among the outliers and the anomaly scores; when the maximum value exceeds a preset threshold, determine that there is an abnormality among the multiple battery cells.

[0077] Optionally, determining whether there is an abnormality among the multiple battery cells can specifically be determining whether there are abnormal battery cells.

[0078] Optionally, when the multiple battery cells come from multiple battery packs, determining whether there is an abnormality among the multiple battery cells can also specifically be determining whether there are abnormal battery packs.

[0079] Embodiment 2

[0080] In a specific implementation manner, based on the above steps S11 to S14, for an anomaly detection method provided in this embodiment, when the multiple battery cells are specifically multiple battery cells from multiple battery packs, the anomaly detection method based on this embodiment can be used to detect abnormal battery packs in multiple battery packs; when the multiple battery cells are specifically multiple battery cells from the same battery pack, the anomaly detection method based on this embodiment can be used to detect abnormal battery cells in the battery pack.

[0081] In specific implementation, taking the battery pack as a sample, the multiple battery cells can be obtained based on the sample, and multiple features of each battery cell can be obtained using the multiple performance test data of the battery cell as features. The multiple performance test data includes voltage value, temperature value, and internal resistance value (DCR). Taking each performance test data as a dimension, multiple - dimension features can be determined.

[0082] Exemplarily, taking the DCR, voltage, and temperature corresponding to each battery cell in the battery pack as columns in a data table, and each row of data corresponding to multiple features of a battery pack, the representation of multiple - dimension features with the battery pack as a sample is specifically shown in Table 1 below:

[0083]

[0084] Table 1

[0085] Exemplarily, based on the same principle, in practical applications, when taking the battery cells as samples, multiple features of each battery cell can be corresponding to each row of data, and the temperature, DCR, and voltage of the battery cells can be corresponding to each column of data respectively. Then, the representation of multiple-dimensional features with the battery cells as samples is specifically shown in Table 2 below:

[0086] Voltage value Temperature value DCR Cell #1 Cell #2 ……

[0087] Table 2

[0088] It can be understood that in the above table, each column represents a dimension, and a dimensional feature can be determined according to any row and column in the above table. For example, based on Table 1 above, the dimensional feature determined according to the first row and the first column is the voltage value of battery cell #1.

[0089] In this embodiment, when obtaining the empirical cumulative distribution of the multiple battery cells based on the features, specifically, the empirical cumulative distribution and the sample skewness of the samples can be obtained based on each dimensional feature to achieve this.

[0090] The empirical cumulative distribution function used to obtain the empirical cumulative distribution of the samples is as follows:

[0091]

[0092]

[0093] Where, is a variable, representing the feature value of the j-th dimension of the i-th sample, x is the value of a specific dimensional feature; 1(*) is an indicator function, which is 1 when its parameter is true and 0 otherwise; n is the number of samples; represents the left-tail empirical cumulative distribution function, represents the right-tail empirical cumulative distribution function.

[0094] The skewness calculation formula used to obtain the sample skewness is as follows:

[0095]

[0096] Where, skewness (j) is the sample skewness.

[0097] It can be understood that the left-tail empirical cumulative distribution of each dimensional feature x is calculated using the above formula (1), and then the left-tail empirical cumulative distribution of the samples is obtained based on the left-tail empirical cumulative distribution of each dimensional feature x; the right-tail empirical cumulative distribution of each dimensional feature x is calculated using the above formula (2), and then the right-tail empirical cumulative distribution of the samples is obtained based on the right-tail empirical cumulative distribution of each dimensional feature x; the sample skewness of each dimension j is calculated using the above formula (3).

[0098] In this embodiment, when the outliers include the first outlier, the second outlier, and the third outlier, obtaining the outliers and abnormal scores of the multiple battery cells based on the empirical cumulative distribution can be achieved in the following manner:

[0099] Obtain the first outlier by aggregating the left-tail empirical cumulative distribution of the samples;

[0100] Obtain the second outlier by aggregating the right-tail empirical cumulative distribution of the samples;

[0101] Redetermine the empirical cumulative distribution of the samples based on the sample skewness, and obtain the third outlier by aggregating the redetermined empirical cumulative distribution of the samples;

[0102] Determine the reference features based on a preset method, redetermine the empirical cumulative distribution of the samples based on the reference features, and obtain the abnormal scores by aggregating the redetermined empirical cumulative distribution of the samples.

[0103] It can be understood that each sample corresponds to the empirical cumulative distribution of multiple dimensions. In this embodiment, the aggregation refers to multiplying the empirical cumulative distributions of multiple dimensions of the sample and taking the negative logarithm.

[0104] Correspondingly, obtaining the first outlier by aggregating the left-tail empirical cumulative distribution of the samples specifically means multiplying the left-tail empirical cumulative distributions of multiple dimensions of the sample and taking the negative logarithm to obtain the first outlier; obtaining the second outlier by aggregating the right-tail empirical cumulative distribution of the samples means multiplying the right-tail empirical cumulative distributions of multiple dimensions of the sample and taking the negative logarithm to obtain the second outlier. For example, if each sample includes the dimensional features of K dimensions, the first outlier is obtained by aggregating K left-tail empirical cumulative distributions, and the second outlier is obtained by aggregating K right-tail empirical cumulative distributions.

[0105] Redetermining the empirical cumulative distribution of the samples based on the sample skewness specifically is: when the value of the sample skewness is negative, then adjust the dimensional features corresponding to the sample skewness to adopt the left-tail empirical cumulative distribution, otherwise adjust the dimensional features corresponding to the sample skewness to adopt the right-tail empirical cumulative distribution, and determine the empirical cumulative distribution of the samples based on the empirical cumulative distributions of the adjusted dimensional features. For example, if the sample skewness of dimension j calculated according to the skewness calculation formula is negative, then adjust each dimensional feature of dimension j to adopt the left-tail empirical cumulative distribution, and if the sample skewness of dimension k calculated according to the skewness calculation formula is positive, then adjust each dimensional feature of dimension k to adopt the right-tail empirical cumulative distribution. Correspondingly, obtaining the third outlier by aggregating the redetermined empirical cumulative distribution of the samples specifically means multiplying the left-tail empirical cumulative distribution or the right-tail empirical cumulative distribution of each dimensional feature of the sample and taking the negative logarithm to obtain the third outlier.

[0106] The determination of the reference feature based on a preset method and the re - determination of the empirical cumulative distribution of the sample based on the reference feature are specifically as follows: The reference feature is determined by taking the first - order derivative of the empirical cumulative distribution function. Based on the relative position of the dimensional feature and the reference feature, the empirical cumulative distribution adopted by each dimensional feature is re - determined, and the empirical cumulative distribution of the sample is determined according to the re - determined empirical cumulative distribution of each dimensional feature. For example, if the dimensional feature is less than the reference feature, it is determined that the left - tail empirical cumulative distribution is adopted for this dimensional feature; otherwise, it is determined that the right - tail empirical cumulative distribution is adopted for this dimensional feature. Correspondingly, obtaining the anomaly score by aggregating the re - determined empirical cumulative distribution of the sample specifically means multiplying the left - tail empirical cumulative distribution or the right - tail empirical cumulative distribution of each dimensional feature of the sample and taking the negative logarithm to obtain the anomaly score.

[0107] In this embodiment, the specific implementation manner of determining whether there is an anomaly based on the outlier value and the anomaly score of multiple battery cells is as follows:

[0108] Based on the first outlier value, the second outlier value, the third outlier value, and the anomaly score of each sample, the outlier samples in the sample are determined.

[0109] Specifically in this embodiment, for each sample, the maximum value is selected from the first outlier value, the second outlier value, the third outlier value, and the anomaly score as the anomaly score of the sample; a threshold is set based on the anomaly scores of all samples, and the outlier samples are screened based on the threshold.

[0110] For example, if the anomaly score is denoted as score_twoside, then the anomaly score outlier score of the sample is expressed as follows: outlier score = max(score_left, score_right, score_adjusted, score_twoside). It can be understood that the anomaly score is not a probability but is used to measure the difference between data points. The anomaly score corresponds to a percentile in the probability distribution. The higher the anomaly score, the higher the percentile it is in, and the more likely it is to be an outlier.

[0111] Exemplarily, regarding the setting of the threshold, first, all samples are sorted based on the anomaly scores of the samples. Based on the empirical proportion of the outlier samples, the threshold of the anomaly score is set to 0.99, and the samples with a percentile exceeding the threshold of 0.99 are used as outlier samples.

[0112] It should be noted that the anomaly detection method in the embodiments of the present application makes an independence assumption for the feature of each dimension. Therefore, Formula 1 and Formula 2 used in the empirical cumulative distribution calculation are independent left / right tail empirical cumulative distribution functions. In practical applications, there may be a certain degree of correlation between the features of each dimension in the obtained battery cell performance test data. Therefore, it is necessary to find out the dimension features with strong correlation and uniformly calculate the empirical cumulative distribution of these features. Specifically, the features with strong correlation can be found by calculating linear correlation parameters, non-linear correlation parameters or according to industry experience. For example, assuming there are n samples, and each sample contains dimension features of K dimensions, for the features with strong correlation found, the empirical cumulative distribution is calculated through the combined left / right tail empirical cumulative distribution function, and for other features, the empirical cumulative distribution is calculated according to the independent left / right tail empirical cumulative distribution function.

[0113] In a corresponding specific implementation manner, the obtaining of the left tail empirical cumulative distribution of the sample and the right tail empirical cumulative distribution of the sample according to the dimension feature and the empirical cumulative distribution function in this embodiment may specifically include:

[0114] Dividing the features into relevant dimension features and non-relevant dimension features according to feature correlation;

[0115] For the non-relevant dimension features, the left tail empirical cumulative distribution of the sample and the right tail empirical cumulative distribution of the sample are calculated according to the independent left / right tail empirical cumulative distribution function;

[0116] For the relevant dimension features, the left tail empirical cumulative distribution of the sample and the right tail empirical cumulative distribution of the sample are calculated according to the combined left / right tail empirical cumulative distribution function.

[0117] The combined left / right tail empirical cumulative distribution function adopted in this embodiment is specifically as follows:

[0118]

[0119]

[0120] Wherein, and are variables, representing the feature values of the j-th dimension and the k-th dimension of the i-th sample respectively, x (j) and x (k) are respectively the values of the dimension features of the j-th dimension and the k-th dimension; 1(*) is an indicator function, which is 1 when its parameter is true and 0 otherwise; n is the number of samples; represents the left tail empirical cumulative distribution function, represents the right tail empirical cumulative distribution function.

[0121] It can be understood that when aggregating the empirical cumulative distribution of the above samples, the empirical cumulative distribution of relevant dimensional features may be included.

[0122] Furthermore, based on the obtaining of the above-mentioned anomaly score, specifically, first determine the reference feature by taking the first derivative of the empirical cumulative distribution function, then re-determine the empirical cumulative distribution of the sample based on the reference feature, and obtain the anomaly score by aggregating the re-determined empirical cumulative distribution of the sample. It should be noted here that when taking the first derivative of the empirical cumulative distribution function, since the reference feature is determined based on the variable corresponding to the maximum value of the first derivative result, and in practical applications, there may be a situation where the variable corresponding to the maximum value of the first derivative result is not unique. For example, if the empirical cumulative distribution function is expressed as f(x), and the approximate method is used to obtain the first derivative of f(x), the first derivative can be expressed as: (f(x') - f(x)) / (x' - x), where x' is the value of the smallest data greater than x, and the value of the variable x when finding the maximum of the first derivative is marked as the reference feature x0. If there are multiple values of the variable x found, then determine x0 according to the sample skewness. Therefore, in this embodiment, the following situations are processed when determining the reference feature: if the variable value corresponding to the maximum value of the first derivative result is unique, then record this variable value as the reference feature; if the variable value corresponding to the maximum value of the first derivative result is not unique, then select one of the multiple variable values as the reference feature based on the sample skewness. For example, if the sample skewness is negative, take the largest x value among the multiple variable x values found and record it as x0, otherwise take the smallest x value among them and record it as x0.

[0123] In the embodiment of the present application, the outlier score of the sample obtained by the above anomaly detection method is at least one of the four anomaly scores. The outlier score can be split in each dimension. In practical applications, by drawing a dimensional outlier graph, it is possible to more directly understand the anomaly scores of the dimensional features of each dimension of the data point, the influence of each dimensional feature on the final outlier score, and whether the anomaly score exceeds the percentile, thereby giving some indications for the data point to be judged as an anomaly, and having good interpretability. An exemplary dimensional outlier graph is as Figure 2 shown. In the figure, the solid line is the percentile of the anomaly score of the data point in a certain dimension, and the dashed line is the outlier score of the cell sample. The dimensional features of the cell sample include voltage value, temperature, and DCR. If the solid line is specifically the percentile of the anomaly score of the data point in the voltage value dimension, then it can be intuitively analyzed from the figure that there are anomalies in the voltages of cell 1 (Volt_#1) and cell 2 (Volt_#2).

[0124] Embodiment 3

[0125] Refer to the appendixFigure 3 , Figure 3 is the main structural block diagram of an anomaly detection device according to an embodiment of the present application. The device mainly includes:

[0126] A data processing module 201, configured to obtain one or more features of each of a plurality of battery cells.

[0127] A data calculation module 202, configured to obtain the empirical cumulative distribution of each feature of the plurality of battery cells.

[0128] An anomaly detection module 203, configured to obtain the outlier values and anomaly scores of the plurality of battery cells based on the empirical cumulative distribution of each feature.

[0129] An anomaly determination module 204, configured to determine whether there is an anomaly among the plurality of battery cells based on the outlier values and anomaly scores of the plurality of battery cells.

[0130] For ease of description, the introduction of the above anomaly detection device only shows the parts related to the embodiments of the present application. For the specific technical details not disclosed, please refer to the method part of the embodiments of the present application.

[0131] It should be understood that since the setting of each module is only for explaining the functional units of the present application, the physical devices corresponding to these modules can be the processor itself, or a part of the software in the processor, a part of the hardware, or a part of the combination of software and hardware. Therefore, the number of each module in the figure is only illustrative.

[0132] Those skilled in the art can understand that the various modules in the device can be adaptively split or combined. Such splitting or combination of specific modules will not cause the technical solution to deviate from the principle of the present application. Therefore, the technical solutions after splitting or combination will all fall within the protection scope of the present application.

[0133] Those skilled in the art can understand that all or part of the processes in the method of implementing the above embodiments of the present application can also be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file or some intermediate form, etc. The computer-readable storage medium can include: any entity or device, medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier signal, telecommunication signal, and software distribution medium, etc., that can carry the computer program code.

[0134] Furthermore, the present application also provides a computer device.

[0135] In an embodiment of a computer device according to the present application, the computer device mainly includes a processor and a storage device. The storage device can be configured to store a program for executing the anomaly detection method in the above method embodiments. The processor can be configured to execute the program in the storage device, and the program includes, but is not limited to, the program for executing the anomaly detection method in the above method embodiments. For ease of illustration, only the parts related to the embodiments of the present application are shown. For specific technical details not disclosed, please refer to the method part of the embodiments of the present application.

[0136] In the embodiments of the present application, the computer device can be a control device formed by various electronic devices. In some possible implementation manners, the computer device can include multiple storage devices and multiple processors. The program for executing the anomaly detection method in the above method embodiments can be divided into multiple sub-programs, and each sub-program can be loaded and run by the processor to execute different steps of the anomaly detection method in the above method embodiments. Specifically, each sub-program can be stored in different storage devices respectively, and each processor can be configured to execute the program in one or more storage devices to jointly implement the anomaly detection method in the above method embodiments, that is, each processor respectively executes different steps of the anomaly detection method in the above method embodiments to jointly implement the anomaly detection method in the above method embodiments.

[0137] The above multiple processors can be processors deployed on the same device. For example, the above computer device can be a high-performance device composed of multiple processors, and the above multiple processors can be the processors configured on the high-performance device. In addition, the above multiple processors can also be processors deployed on different devices. For example, the above computer device can be a server cluster, and the above multiple processors can be the processors on different servers in the server cluster.

[0138] Furthermore, the present application also provides a computer-readable storage medium.

[0139] In an embodiment of a computer-readable storage medium according to the present application, the computer-readable storage medium can be configured to store a program for executing the anomaly detection method in the above method embodiments. The program can be loaded and run by the processor to implement the above anomaly detection method. For ease of illustration, only the parts related to the embodiments of the present application are shown. For specific technical details not disclosed, please refer to the method part of the embodiments of the present application. The computer-readable storage medium can be a storage device formed by various electronic devices. Optionally, the computer-readable storage medium in the embodiments of the present application is a non-transitory computer-readable storage medium.

[0140] So far, the technical solution of the present application has been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present application is obviously not limited to these specific embodiments. Without departing from the principle of the present application, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present application.

Claims

1. An anomaly detection method, characterized in that, The method includes: For a plurality of battery cells, obtaining one or more characteristics of each battery cell; For the plurality of battery cells, obtaining the empirical cumulative distribution of each characteristic; Based on the empirical cumulative distribution of each characteristic, obtaining the outliers and anomaly scores of the plurality of battery cells; Based on the outliers and anomaly scores of the plurality of battery cells, determining whether there are anomalies among the plurality of battery cells.

2. The method according to claim 1, characterized in that, The empirical cumulative distribution includes a left-tail empirical cumulative distribution and / or a right-tail empirical cumulative distribution; The obtaining the outliers and anomaly scores of the plurality of battery cells based on the empirical cumulative distribution of each characteristic includes: Based on the left-tail empirical cumulative distribution of one or more characteristics, obtaining the outliers of the plurality of battery cells; or based on the right-tail empirical cumulative distribution of one or more characteristics, obtaining the outliers of the plurality of battery cells; Based on the left-tail empirical cumulative distribution and / or the right-tail empirical cumulative distribution of one or more characteristics, obtaining the anomaly scores of the plurality of battery cells.

3. The method according to claim 2, characterized in that, The outliers include a first outlier, a second outlier, and a third outlier; The first outlier is obtained by aggregating the left-tail empirical cumulative distribution of the one or more characteristics; The second outlier is obtained by aggregating the right-tail empirical cumulative distribution of the one or more characteristics; The third outlier is obtained through the following steps: Calculating the skewness corresponding to each characteristic; Based on the skewness of each characteristic, re-determining the empirical cumulative distribution of this characteristic; Aggregating the re-determined empirical cumulative distributions of each characteristic to obtain the third outlier of the plurality of battery cells.

4. The method according to claim 2, characterized in that, The obtaining the anomaly scores of the plurality of battery cells based on the left-tail empirical cumulative distribution and / or the right-tail empirical cumulative distribution of one or more characteristics includes: Determining a reference characteristic; Based on the reference characteristic, re-determining the empirical cumulative distribution of each characteristic; Obtaining the anomaly scores of the plurality of battery cells by aggregating the re-determined empirical cumulative distributions of the characteristics.

5. The method according to claim 4, characterized in that, The determining the reference characteristic includes: obtaining the reference characteristic based on the first derivative of the empirical cumulative distribution function; The re-determining the empirical cumulative distribution of each characteristic based on the reference characteristic includes: re-determining the empirical cumulative distribution of this characteristic based on the relative position of each characteristic and the reference characteristic.

6. The method according to claim 5, characterized in that, The obtaining the reference characteristic based on the first derivative of the empirical cumulative distribution function includes: Taking the first derivative of the empirical cumulative distribution function; Taking the variable value corresponding to the maximum value of the first derivative as the reference characteristic.

7. The method according to claim 5, characterized in that, The re-determining the empirical cumulative distribution of this characteristic based on the relative position of each characteristic and the reference characteristic includes: In response to each characteristic being less than the reference characteristic, taking the left-tail empirical cumulative distribution of this characteristic as the empirical cumulative distribution of this characteristic; In response to each characteristic being greater than or equal to the reference characteristic, taking the right-tail empirical cumulative distribution of this characteristic as the empirical cumulative distribution of this characteristic.

8. The method according to claim 1, characterized in that, The determining whether there are anomalies among the plurality of battery cells based on the outliers and anomaly scores of the plurality of battery cells is specifically: Obtaining the maximum value among the outliers and the anomaly scores; When the maximum value exceeds a preset threshold, determining that there are anomalies among the plurality of battery cells.

9. A computer device, comprising a processor and a storage device, the storage device being adapted to store multiple program codes, characterized in that, The program code is adapted to be loaded and run by the processor to execute the anomaly detection method according to any one of claims 1 to 8.

10. A computer-readable storage medium, in which multiple program codes are stored, characterized in that, The program code is adapted to be loaded and run by a processor to execute the anomaly detection method according to any one of claims 1 to 8.

Images