Battery fault recognition model training method and device

Abstract

The invention relates to the technical field of battery fault recognition, and discloses a battery fault recognition model training method and device. The method comprises the following steps: respectively carrying out serialization tests related to battery characteristics on a preset reference battery and each target to-be-tested battery to obtain first time sequence test data and second time sequence test data; and performing dynamic time warping and average differential voltage difference calculation on each second time sequence test data and the first time sequence test data one by one, and finally inputting each dynamic time warping value and each average differential voltage difference value as sample feature data into a preset battery fault identification model. And training the battery fault identification model according to the sample feature data. According to the invention, the trained battery fault identification model can identify the fault of the target to-be-detected battery more accurately.

Description

Technical Field

[0001] The present application relates to the technical field of battery fault identification, and in particular to a battery fault identification model training method and device. Background Art

[0002] In the prior art, the safety management of battery systems faces many technical problems, especially in the diagnosis of multiple faults. As the core component of electric vehicles, the performance of batteries directly affects the safety and reliability of the entire vehicle. However, common faults in battery systems are hidden, diverse and complex. These faults are often difficult to identify individually, especially during actual operation, when minor faults and multiple faults may occur at the same time. Existing methods cannot effectively and accurately identify these faults, so the prior art has the problem of inaccurate fault identification. Summary of the invention

[0003] The present application provides a battery fault identification model training method and device thereof to solve one or more technical problems existing in the prior art and at least provide a beneficial option or create conditions.

[0004] Other features and advantages of the present application will become apparent from the following detailed description, or may be learned in part by the practice of the present application.

[0005] According to one aspect of an embodiment of the present application, a battery fault identification model training method is provided, the method comprising: Performing serial tests related to battery characteristics on a preset reference battery and each target battery to be tested, respectively, to obtain first time series test data of the reference battery and second time series test data of each target battery to be tested; Performing dynamic time warping on each of the second time series test data and the first time series test data one by one to obtain dynamic time warping values ​​of each of the target battery to be tested and the reference battery; Calculate the difference of average differential voltage between each of the second time series test data and the first time series test data one by one, to obtain the average differential voltage difference between each of the target test batteries and the reference battery; Each of the dynamic time warping values ​​and each of the average differential voltage difference values ​​are input into a preset battery fault identification model as sample feature data of each of the target batteries to be tested, and the battery fault identification model is trained according to each of the sample feature data.

[0006] In one embodiment of the present application, based on the above solution, the dynamic time warping value of a single target battery to be tested and the reference battery is obtained by the following steps: Constructing a distance matrix according to the first time series test data of the reference battery and the second time series test data of the target battery to be tested; Calculate the best bending distance between any point of the first time series test data and any point of the second time series test data in the distance matrix according to the dynamic time warping method, and generate a best bending distance set; A target optimal bending distance with the smallest bending distance is searched in the optimal bending distance set as the dynamic time warping value.

[0007] In one embodiment of the present application, based on the above solution, the average differential voltage difference between the single target battery to be tested and the reference battery is obtained by the following steps: Determine first open circuit voltage sequence test data of the reference battery according to the first time series test data, and determine second open circuit voltage sequence test data of the target battery to be tested according to the second time series test data; Determining a first average differential voltage value of the reference battery according to each open circuit voltage value and the number of each open circuit voltage value in the first open circuit voltage sequence test data; Determining a second average differential voltage value of the target battery to be tested according to each open circuit voltage value and the number of each open circuit voltage value in the second open circuit voltage sequence test data; The average differential voltage difference value is determined based on the first average differential voltage value and the second average differential voltage value.

[0008] In one embodiment of the present application, based on the above solution, the battery fault identification model includes a first-layer Gaussian mixture model and a second-layer Gaussian mixture model, and the training of the battery fault identification model according to each of the sample feature data includes: Acquire a dynamic time warping value and an average differential voltage difference between each of the target batteries to be tested and the reference battery according to each of the sample characteristic data; Inputting each of the dynamic time warping values ​​and each of the average differential voltage difference values ​​into the first-layer Gaussian mixture model to obtain a first fault analysis result of each of the target batteries to be tested; Inputting each of the dynamic time warping values ​​and each of the average differential voltage difference values ​​into the second-layer Gaussian mixture model to obtain a second fault analysis result of each of the target batteries to be tested; The battery fault identification model is trained according to the first fault analysis result and the second fault analysis result.

[0009] In one embodiment of the present application, based on the aforementioned scheme, the first-layer Gaussian mixture model includes a first probability density function, which is used to determine the first probability distribution point of each of the target batteries to be tested, and to determine whether the target batteries to be tested have a short circuit fault based on each of the first probability distribution points; the second-layer Gaussian mixture model includes a second probability density function, which is used to determine the second probability distribution point of each of the target batteries to be tested, and to determine whether the target batteries to be tested have an abnormal degradation fault based on each of the second probability distribution points.

[0010] In one embodiment of the present application, based on the aforementioned scheme, the horizontal coordinates and vertical coordinates of the first probability distribution point and the second probability distribution point are respectively the dynamic time warping value and the average differential voltage difference.

[0011] In one embodiment of the present application, based on the above solution, whether the target battery to be tested has a short circuit fault can be determined by the following steps: Acquire a horizontal coordinate value and a vertical coordinate value of a first probability distribution point of the target battery to be tested, wherein the horizontal coordinate value is a dynamic time warping value of the target battery to be tested and the reference battery, and the vertical coordinate value is an average differential voltage difference between the target battery to be tested and the reference battery; If the horizontal coordinate value is greater than a first preset threshold value and / or the vertical coordinate value is greater than a second preset threshold value, it is determined that a short circuit fault occurs in the target battery to be tested.

[0012] In one embodiment of the present application, based on the above solution, whether the target battery to be tested has an abnormal degradation fault can be determined by the following steps: Obtaining a horizontal coordinate value and a vertical coordinate value of a second probability distribution point of the target battery to be tested, wherein the horizontal coordinate value is a dynamic time warping value of the target battery to be tested and the reference battery, and the vertical coordinate value is an average differential voltage difference between the target battery to be tested and the reference battery; If the horizontal coordinate value is greater than the third preset threshold or the vertical coordinate value is greater than the fourth preset threshold, it is determined that the target battery to be tested has an abnormal degradation fault.

[0013] In one embodiment of the present application, based on the above-mentioned solution, the serialized test related to the battery characteristics is to test the voltage characteristics of the battery in sequence according to the time sequence in the preset time period.

[0014] According to one aspect of an embodiment of the present application, a battery fault identification model training device is provided, the device comprising: A testing unit, used to perform serial tests related to battery characteristics on a preset reference battery and each target battery to be tested, respectively, to obtain first time series test data of the reference battery and second time series test data of each target battery to be tested; A first calculation unit, configured to perform dynamic time warping on each of the second time series test data and the first time series test data one by one, to obtain a dynamic time warping value of each of the target battery to be tested and the reference battery; A second calculation unit, configured to calculate the difference of average differential voltage between each of the second time series test data and the first time series test data one by one, to obtain the average differential voltage difference between each of the target test cells and the reference cell; A training unit is used to input each of the dynamic time warping values ​​and each of the average differential voltage difference values ​​as sample feature data of each of the target batteries to be tested into a preset battery fault identification model, and train the battery fault identification model according to each of the sample feature data.

[0015] The beneficial effect of the present application is that by performing serialized tests related to battery characteristics on a preset reference battery and each target battery to be tested, respectively, the obtained first time series test data of the reference battery and the second time series test data of each target battery to be tested can be used to characterize the battery characteristics of the reference battery and the target battery to be tested that change over time.

[0016] Furthermore, the dynamic time warping values ​​of each of the target battery to be tested and the reference battery can be used to reflect the difference between the target battery to be tested and the reference battery (normal battery), and the average differential voltage difference between each of the target battery to be tested and the reference battery can reflect the inconsistency between the target battery to be tested and the reference battery (normal battery). The differences and inconsistencies obtained through analysis can be used as sample feature data of the target battery to be tested and input into the battery fault identification model to train the model, so that the trained battery fault identification model can accurately identify whether the difference between the target battery to be tested and the normal reference battery is too large or there is inconsistency, and then analyze and identify whether the target battery to be tested has a fault, thereby improving the accuracy of the model's fault identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following is a brief description of the drawings required for use in the description of the embodiments. Obviously, the drawings described are only part of the embodiments of the present application, not all of the embodiments, and those skilled in the art can also obtain other design solutions and drawings based on these drawings without creative work.

[0018] Figure 1 It is a flowchart of a battery fault identification model training method according to an embodiment of the present application; Figure 2 is a circuit structure diagram of an equivalent circuit model according to an embodiment of the present application; Figure 3 is a block diagram of a battery fault identification model training device according to an embodiment of the present application; Figure 4 A probability density distribution diagram of fault identification using a double-layer Gaussian mixture model according to an embodiment of the present application. DETAILED DESCRIPTION

[0019] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this application will be more comprehensive and complete and fully convey the concept of the example embodiments to those skilled in the art.

[0020] In addition, described feature, structure or characteristic can be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present application. However, those skilled in the art will appreciate that the technical scheme of the present application can be put into practice without one or more of the specific details, or other methods, components, devices, steps, etc. can be adopted. In other cases, known methods, devices, realizations or operations are not shown or described in detail to avoid blurring the various aspects of the application.

[0021] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities may be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or micro-control node devices.

[0022] The flowcharts shown in the accompanying drawings are only exemplary and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps can be decomposed, and some operations / steps can be combined or partially combined, so the actual execution order may change according to actual conditions.

[0023] It should be noted that the "multiple" mentioned in this article refers to two or more. "And / or" describes the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0024] The implementation details of the technical solution of the embodiment of the present application are described in detail below: According to one aspect of an embodiment of the present application, a battery fault identification model training method is provided. Figure 1 This is a flow chart of a battery thermal runaway state warning method according to an embodiment of the present application. The method includes at least steps 110 to 140, which are described in detail as follows: In step 110, a preset reference battery and each target battery to be tested are subjected to serial tests related to battery characteristics to obtain first time series test data of the reference battery and second time series test data of each target battery to be tested.

[0025] Specifically, the input data of the target battery to be tested or the reference battery cannot directly form sequence test data to provide a data basis for subsequent fault analysis. The input data input to the target battery to be tested or the reference battery needs to be transformed by an equivalent circuit model to form time series test data. Therefore, the preset reference battery and each target battery to be tested are subjected to serial tests related to battery characteristics, that is, the equivalent circuit model is used to form sequence data with spatiotemporal characteristics according to the changes in the internal battery parameters of the target battery to be tested or the reference battery in time sequence, that is, the first time series test data of the reference battery and the second time series test data of each of the target batteries to be tested.

[0026] In one embodiment of the present application, the serialized test related to the battery characteristics is to test the voltage characteristics of the battery in sequence according to the time sequence in a preset time period.

[0027] Specifically, the preset time period can be a time period selected according to actual needs, such as a time period during the charging process of the battery, or a time period during the discharging process of the battery. The characteristics of the battery, especially the voltage characteristics, can be analyzed during the selected preset time period, and the data related to the voltage characteristics actually collected can be converted into the time series test data in the embodiment of this application through the equivalent circuit model.

[0028] In one embodiment of the present application, the equivalent circuit model can be specifically as follows: Figure 2 As shown in Figure 1, the equivalent circuit model consists of an ideal voltage source, a series resistor, and two resistor-capacitor (RC) loops. R0 is an ohmic resistor, R1 and C1 represent the activation polarization resistor and capacitor, respectively, and R2 and C2 represent the concentration polarization resistor and capacitor, respectively. According to Kirchhoff's law, the state equation of the second-order RC model can be described by equations (1) and (2): (1) (2) Where U t and U OCV They represent the terminal voltage and open circuit voltage of the battery respectively; I refers to the current; U1 and U2 are the polarization voltages of the two RC circuits.

[0029] In addition, when the method is actually applied to the time domain series, discretization processing is required, which is described in (3) and (4). Next, the present invention applies RLS to online parameter identification. The specific process is as follows: (3) (4) (5) Where k is the sampling time, K k is the gain matrix; P k-1 represents the error covariance matrix; is the data matrix; is the parameter matrix; E is the identity matrix; y k is the system observation; µ is the forgetting factor, which is 1 here.

[0030] Expand the data matrix to: (6) Therefore, the above formulas (1)-(6) can be used to generate time series test data based on the actual internal parameter changes of the equivalent simulated battery.

[0031] In step 120, each of the second time series test data is dynamically time-aligned with the first time series test data one by one to obtain a dynamic time-aligned value of each of the target battery to be tested and the reference battery.

[0032] Specifically, a battery pack is usually composed of multiple single cells connected in series or in parallel. When the battery is operating normally, each battery exhibits similar characteristics. Therefore, the differences between the series-connected batteries are regarded as fault features for diagnosing battery faults. The Dynamic Time Warping (DTW) algorithm is a method for measuring the similarity of two time series. In this application, DTW is used to quantify the degree of fault, that is, the fault difference between the target battery to be tested and the normal reference battery, that is, calculated by the first time series test data and the second time series test data. In addition, the smaller the DTW value, the higher the similarity between the curves.

[0033] DTW is used to accurately calculate the difference between each independent series battery. Specifically, it accurately calculates the error by comparing and analyzing the battery's time series. Battery operation data (P and Q), where P and Q represent the first time series test data and the second time series test data, respectively, and the difference is calculated based on the Euclidean distance of the time series (such as the difference between voltage points). The DTW rule is to minimize their differences by dynamically adjusting the time series and finding the optimal matching path. In one embodiment of the present application, the dynamic time warping value of a single target battery to be tested and the reference battery is obtained by the following steps: Constructing a distance matrix according to the first time series test data of the reference battery and the second time series test data of the target battery to be tested; Calculate the best bending distance between any point of the first time series test data and any point of the second time series test data in the distance matrix according to the dynamic time warping method, and generate a best bending distance set; A target optimal bending distance with the smallest bending distance is searched in the optimal bending distance set as the dynamic time warping value.

[0034] Specifically, the calculation process of the dynamic time warping value is as follows: Construct an n×m matrix, namely the distance matrix described in the embodiment of the present application, which is defined as the distance matrix D base : ; In the formula, dbase represents the basic distance, which specifically means the midpoint p in the time series P. i The midpoint q of the time series Q j The Euclidean distance between them, time series P and time series Q represent the first time series test data and the second time series test data respectively, and the database is: ; Therefore, the optimal curved path can be defined as: ; Where F is the optimal bending path length; W F =(i, j) means p i With q j The corresponding relationship.

[0035] Specifically, the role of the optimal path bending is to dynamically adjust the alignment of the time series on the time axis to find the optimal matching path between the two time series. This can help simplify the Euclidean distance of the two time series at a specific point when the time is not synchronized, and form a dbase curved path to accumulate these Euclidean distances, thereby minimizing the overall difference.

[0036] The DTW distance between time series P and Q is calculated as: ; The specific recursive definition is as follows: ; where DTW(i, j) represents the optimal warping distance between the first i points of the time series P and the first j points of the time series Q; i∈(1,n], j∈(1,m), DTW(1,1)=dbase(p1,q1), and DTW(n,m) is the optimal warping distance between two DTW distances between two time series P and Q.

[0037] Specifically, the DTW(i, j) set, that is, the optimal bending distance set, is formed by calculating the optimal bending distance between any point of the first time series test data and any point of the second time series test data, and the dynamic time warping value is obtained by searching for DTW(n, m), that is, searching for the target optimal bending distance with the minimum bending distance.

[0038] In step 130, the average differential voltage difference between each of the second time series test data and the first time series test data is calculated one by one to obtain the average differential voltage difference between each of the target cells to be tested and the reference cell.

[0039] Specifically, the mean differential voltage, denoted by MDV, is used to determine the voltage difference between batteries in series. The judgment principle is usually to determine the inconsistency by comparing the open circuit voltage of each target battery to be tested with the voltage difference of the reference battery. Inconsistency refers to the performance difference exhibited by the battery under working conditions, which may affect the safety and reliability of the entire battery pack. Inconsistency emphasizes that the target battery to be tested may cause potential failure risks, so the purpose is to ensure the overall performance and safety of the battery pack.

[0040] In one embodiment of the present application, the average differential voltage difference between a single target battery to be tested and the reference battery is obtained by the following steps: Determine first open circuit voltage sequence test data of the reference battery according to the first time series test data, and determine second open circuit voltage sequence test data of the target battery to be tested according to the second time series test data; Determining a first average differential voltage value of the reference battery according to each open circuit voltage value and the number of each open circuit voltage value in the first open circuit voltage sequence test data; Determining a second average differential voltage value of the target battery to be tested according to each open circuit voltage value and the number of each open circuit voltage value in the second open circuit voltage sequence test data; The average differential voltage difference value is determined based on the first average differential voltage value and the second average differential voltage value.

[0041] The embodiment of the present application uses the mean differential voltage (MDV) to represent the inconsistency between the batteries in series. 0~ t e The detection method is described as an example. The open circuit voltage OCV of each target battery to be tested in the target battery pack to be tested is expressed as: ; Wherein, n is the number of target batteries to be tested in the battery system; i represents the i-th target battery unit to be tested.

[0042] ; Where DVOCV,i is the differential OCV of the i-th battery, and MDVOCV,i is the average value of DVOCV,i.

[0043] Specifically, by comparing the average differential voltage difference between the reference battery and the target battery to be tested, the inconsistency between the reference battery and the target battery to be tested can be analyzed and obtained.

[0044] In step 140, each of the dynamic time warping values ​​and each of the average differential voltage difference values ​​are input into a preset battery fault identification model as sample feature data of each of the target batteries to be tested, and the battery fault identification model is trained according to each of the sample feature data.

[0045] Specifically, the present application inputs each of the dynamic time warping values ​​and each of the average differential voltage difference values ​​as sample feature data of each of the target batteries to be tested into a preset battery fault recognition model. The input sample feature data can be used as training samples of the battery fault recognition model, so that the battery fault recognition model can quickly identify whether the target battery to be tested has a fault and what faults have occurred according to the input sample labels during subsequent recognition, and can accurately identify multiple faults. In one embodiment of the present application, the battery fault identification model includes a first-layer Gaussian mixture model and a second-layer Gaussian mixture model, and the training of the battery fault identification model according to each of the sample feature data includes: Acquire a dynamic time warping value and an average differential voltage difference between each of the target batteries to be tested and the reference battery according to each of the sample characteristic data; Inputting each of the dynamic time warping values ​​and each of the average differential voltage difference values ​​into the first-layer Gaussian mixture model to obtain a first fault analysis result of each of the target batteries to be tested; Inputting each of the dynamic time warping values ​​and each of the average differential voltage difference values ​​into the second-layer Gaussian mixture model to obtain a second fault analysis result of each of the target batteries to be tested; The battery fault identification model is trained according to the first fault analysis result and the second fault analysis result.

[0046] In one embodiment of the present application, the first-layer Gaussian mixture model includes a first probability density function, which is used to determine the first probability distribution point of each of the target batteries to be tested, and to determine whether the target batteries to be tested have a short circuit fault based on each of the first probability distribution points; the second-layer Gaussian mixture model includes a second probability density function, which is used to determine the second probability distribution point of each of the target batteries to be tested, and to determine whether the target batteries to be tested have an abnormal degradation fault based on each of the second probability distribution points. The horizontal coordinates and vertical coordinates of the first probability distribution point and the second probability distribution point are respectively the dynamic time warping value and the average differential voltage difference.

[0047] Specifically, Figure 4 As shown, Figure 4 (a) shows the first probability distribution points obtained by the first probability density function used in the first-layer Gaussian mixture model, where the first probability distribution points represented by the blue dots are normal target batteries under test without any faults, and the first probability distribution points represented by the yellow dots are target batteries under test with SC faults (short circuit faults).

[0048] Figure 4 (b) The second probability distribution points obtained by the second probability density function used by the second-layer Gaussian mixture model, where the second probability distribution points represented by the blue triangle points are normal target batteries under test without faults, and the second probability distribution points represented by the yellow triangle points are target batteries under test with abnormal degradation faults.

[0049] It should be noted that the embodiment of the present application can also identify SC faults and abnormal degradation faults by a double-layer clustering method. First, the normal battery and the battery with SC fault are distinguished by the first-layer Gaussian mixture model, and then the normal battery identified by the first-layer Gaussian mixture model is further identified by abnormal degradation faults. Figure 4 It can be seen that when the DTW value (dynamic time warping value) is high, SC faults and abnormal degradation faults are prone to occur. When the MDV value (mean differential voltage value) is high, abnormal degradation faults are prone to occur, but SC faults may not necessarily occur.

[0050] Further, assuming that the Gaussian mixture model is composed of G Gaussian models (in the embodiment of the present application, the value of G is 2), the probability density function of the Gaussian mixture model is as follows: ; Where p(g) =πg is the weight of the g-th Gaussian model; p(x|g) = N(x|ug, ∑g|) is the probability density function of the g-th Gaussian model. The first-layer probability density function can be obtained by the above formula, and the formula of the second-layer probability density function is specifically expanded as follows: ; where D represents the number of variables in each Gaussian model.

[0051] Whether the target battery to be tested has a short circuit fault can be determined by the following steps: Acquire a horizontal coordinate value and a vertical coordinate value of a first probability distribution point of the target battery to be tested, wherein the horizontal coordinate value is a dynamic time warping value of the target battery to be tested and the reference battery, and the vertical coordinate value is an average differential voltage difference between the target battery to be tested and the reference battery; If the horizontal coordinate value is greater than a first preset threshold value and / or the vertical coordinate value is greater than a second preset threshold value, it is determined that a short circuit fault occurs in the target battery to be tested.

[0052] Specifically, the first preset threshold, the second preset threshold, the third preset threshold and the fourth preset threshold can be set according to actual needs and are not limited here. In the embodiment of the present application, the first preset threshold, the second preset threshold, the third preset threshold and the fourth preset threshold can be specifically 50, 0.05, 8, and 0.02, respectively. The units of the second preset threshold and the fourth preset threshold are volts.

[0053] In one embodiment of the present application, whether the target battery to be tested has an abnormal degradation fault can be determined by the following steps: Obtaining a horizontal coordinate value and a vertical coordinate value of a second probability distribution point of the target battery to be tested, wherein the horizontal coordinate value is a dynamic time warping value of the target battery to be tested and the reference battery, and the vertical coordinate value is an average differential voltage difference between the target battery to be tested and the reference battery; If the horizontal coordinate value is greater than the third preset threshold or the vertical coordinate value is greater than the fourth preset threshold, it is determined that the target battery to be tested has an abnormal degradation fault.

[0054] Specifically, if the dynamic time warping value of the target battery to be tested and the reference battery is greater than 8 or the average differential voltage value of the target battery to be tested exceeds 0.02, it means that the target battery to be tested has an abnormal degradation fault. Therefore, the target battery to be tested can be attached with an abnormal degradation fault label, and the dynamic time warping value and the average differential voltage difference of the target battery to be tested can be combined as training samples and input into the battery fault identification model. In this way, the battery fault identification model can directly output the analysis result that the current battery has an abnormal degradation fault when it identifies a battery with battery parameters within this range next time.

[0055] Based on the selected fault features, the present application uses a double-layer Gaussian mixture model clustering algorithm to perform multi-fault diagnosis on the battery system. The clustering algorithm is an unsupervised learning method, which means that the applied data is usually unlabeled. It should be emphasized that although unlabeled data is used for clustering, due to previous experiments, the label of each sample is known, that is, the label is assigned by the sample feature data given in the embodiment of the present application. In particular, the clustering of the first layer of Gaussian mixture model is used to separate short circuit (SC) faults from other faults. The results are shown in Figure 2. Figure 4 As shown in (a). Based on the first-layer clustering results, the samples considered to be SC faults are removed, and the remaining samples are clustered for the second time. Figure 4 As shown in (b), abnormally degraded batteries can be accurately identified.

[0056] This application designs a cycle experiment by connecting normal batteries and degraded batteries, and in the process, an equivalent circuit model is used to simulate the artificial triggering of short circuit (SC) faults. The technology is developed based on equivalent circuit model (ECM), similarity metric and unsupervised learning framework. First, an equivalent circuit model is constructed to model the battery and identify the open circuit voltage (OCV) in combination with recursive least squares. Then, dynamic time warping (DTW) is used to measure the similarity between OCVs. In addition, the mean differential voltage (MDV) is defined to evaluate the inconsistency of the battery system. Both DTW and MDV are extracted as fault features for subsequent diagnosis. Finally, a dual Gaussian mixture model is applied to cluster multiple faults, with the first step used to detect short circuit faults and the second step used to distinguish abnormally degraded batteries from normal batteries. Throughout the process, short circuit faults and abnormal degradation faults can be quantified, classified and located in a timely and effective manner.

[0057] In summary, the present application performs serialized tests related to battery characteristics on a preset reference battery and each target battery to be tested, respectively, so that the obtained first time series test data of the reference battery and the second time series test data of each target battery to be tested can be used to characterize the battery characteristics of the reference battery and the target battery to be tested that change over time.

[0058] Furthermore, the dynamic time warping values ​​of each of the target battery to be tested and the reference battery can be used to reflect the difference between the target battery to be tested and the reference battery (normal battery), and the average differential voltage difference between each of the target battery to be tested and the reference battery can reflect the inconsistency between the target battery to be tested and the reference battery (normal battery). The differences and inconsistencies obtained through analysis can be used as sample feature data of the target battery to be tested and input into the battery fault identification model to train the model, so that the trained battery fault identification model can accurately identify whether the difference between the target battery to be tested and the normal reference battery is too large or there is inconsistency, and then analyze and identify whether the target battery to be tested has a fault, thereby improving the accuracy of the model's fault identification.

[0059] Figure 3 1 is a block diagram of a battery thermal runaway state warning device 300 according to an embodiment of the present application. According to a battery thermal runaway state warning device 300 according to an embodiment of the present application, the device 300 includes: a testing unit 301, a first computing unit 302, a second computing unit 303, and a training unit 304.

[0060] The test unit 301 is used to perform serial tests related to battery characteristics on a preset reference battery and each target battery to be tested, respectively, to obtain first time series test data of the reference battery and second time series test data of each target battery to be tested; A first calculation unit 302 is used to perform dynamic time warping on each of the second time series test data and the first time series test data one by one to obtain a dynamic time warping value of each of the target battery to be tested and the reference battery; A second calculation unit 303 is used to calculate the difference of the average differential voltage between each of the second time series test data and the first time series test data one by one, so as to obtain the average differential voltage difference between each of the target test cells and the reference cell; The training unit 304 is used to input each of the dynamic time warping values ​​and each of the average differential voltage difference values ​​as sample feature data of each of the target batteries to be tested into a preset battery fault identification model, and train the battery fault identification model according to each of the sample feature data.

[0061] As another aspect, the present application further provides a computer-readable storage medium on which a program product capable of implementing the method provided above in this specification is stored. In some possible implementations, various aspects of the present application may also be implemented in the form of a program product, which includes a program code, and when the program product is run on a terminal device, the program code is used to cause the terminal device to execute the steps according to various exemplary implementations of the present application described in the above "Embodiment Method" section of this specification.

[0062] According to the program product for implementing the above method in the embodiment of the present application, it can adopt a portable compact disk read-only memory (CD-ROM) and include program code, and can be run on a terminal device, such as a personal computer. However, the program product of the present application is not limited thereto. In this document, a readable storage medium can be any tangible medium containing or storing a program, which can be used by or in combination with an instruction execution system, an apparatus or a device.

[0063] The program product may be any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0064] Computer readable signal media may include data signals propagated in baseband or as part of a carrier wave, in which readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Readable signal media may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0065] The program code embodied on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the foregoing.

[0066] Program code for performing the operations of the present application may be written in any combination of one or more programming languages, including object-oriented programming languages, such as Java, C++, etc., and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., using an Internet service provider to connect through the Internet).

[0067] In addition, the above-mentioned figures are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present application, and are not intended to be limiting. It is easy to understand that the processes shown in the above-mentioned figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0068] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be performed without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A battery fault identification model training method, characterized in that: The method comprises: Performing serial tests related to battery characteristics on a preset reference battery and each target battery to be tested respectively, to obtain first time series test data of the reference battery and second time series test data of each target battery to be tested; Performing dynamic time warping on each of the second time series test data and the first time series test data one by one to obtain dynamic time warping values ​​of each of the target battery to be tested and the reference battery; Calculate the difference of average differential voltage between each of the second time series test data and the first time series test data one by one, to obtain the average differential voltage difference between each of the target test batteries and the reference battery; Each of the dynamic time warping values ​​and each of the average differential voltage difference values ​​are input into a preset battery fault identification model as sample feature data of each of the target batteries to be tested, and the battery fault identification model is trained according to each of the sample feature data.

2. The battery fault identification model training method according to claim 1, characterized in that: The dynamic time warping values ​​of the single target battery to be tested and the reference battery are obtained by the following steps: Constructing a distance matrix according to the first time series test data of the reference battery and the second time series test data of the target battery to be tested; Calculate the best bending distance between any point of the first time series test data and any point of the second time series test data in the distance matrix according to the dynamic time warping method, and generate a best bending distance set; A target optimal bending distance with the smallest bending distance is searched in the optimal bending distance set as the dynamic time warping value.

3. The battery fault identification model training method according to claim 1, characterized in that: The average differential voltage difference between the single target battery to be tested and the reference battery is obtained by the following steps: Determine first open circuit voltage sequence test data of the reference battery according to the first time series test data, and determine second open circuit voltage sequence test data of the target battery to be tested according to the second time series test data; Determining a first average differential voltage value of the reference battery according to each open circuit voltage value and the number of each open circuit voltage value in the first open circuit voltage sequence test data; Determining a second average differential voltage value of the target battery to be tested according to each open circuit voltage value and the number of each open circuit voltage value in the second open circuit voltage sequence test data; The average differential voltage difference value is determined based on the first average differential voltage value and the second average differential voltage value.

4. The battery fault identification model training method according to claim 1, characterized in that: The battery fault identification model includes a first-layer Gaussian mixture model and a second-layer Gaussian mixture model, and the training of the battery fault identification model according to each of the sample feature data includes: Acquire a dynamic time warping value and an average differential voltage difference between each of the target batteries to be tested and the reference battery according to each of the sample characteristic data; Inputting each of the dynamic time warping values ​​and each of the average differential voltage difference values ​​into the first-layer Gaussian mixture model to obtain a first fault analysis result of each of the target batteries to be tested; Inputting each of the dynamic time warping values ​​and each of the average differential voltage difference values ​​into the second-layer Gaussian mixture model to obtain a second fault analysis result of each of the target batteries to be tested; The battery fault identification model is trained according to the first fault analysis result and the second fault analysis result.

5. The battery fault identification model training method according to claim 4, characterized in that: The first-layer Gaussian mixture model includes a first probability density function, which is used to determine the first probability distribution point of each target battery to be tested, and determine whether the target battery to be tested has a short circuit fault based on each of the first probability distribution points; the second-layer Gaussian mixture model includes a second probability density function, which is used to determine the second probability distribution point of each target battery to be tested, and determine whether the target battery to be tested has an abnormal degradation fault based on each of the second probability distribution points.

6. The battery fault identification model training method according to claim 5, characterized in that: The horizontal coordinates and vertical coordinates of the first probability distribution point and the second probability distribution point are respectively the dynamic time warping value and the average differential voltage difference.

7. The battery fault identification model training method according to claim 6, characterized in that: Whether the target battery to be tested has a short circuit fault can be determined by the following steps: Acquire a horizontal coordinate value and a vertical coordinate value of a first probability distribution point of the target battery to be tested, wherein the horizontal coordinate value is a dynamic time warping value of the target battery to be tested and the reference battery, and the vertical coordinate value is an average differential voltage difference between the target battery to be tested and the reference battery; If the horizontal coordinate value is greater than a first preset threshold value and / or the vertical coordinate value is greater than a second preset threshold value, it is determined that a short circuit fault occurs in the target battery to be tested.

8. The battery fault identification model training method according to claim 6, characterized in that: Whether the target battery to be tested has an abnormal degradation fault can be determined by the following steps: Obtaining a horizontal coordinate value and a vertical coordinate value of a second probability distribution point of the target battery to be tested, wherein the horizontal coordinate value is a dynamic time warping value of the target battery to be tested and the reference battery, and the vertical coordinate value is an average differential voltage difference between the target battery to be tested and the reference battery; If the horizontal coordinate value is greater than the third preset threshold or the vertical coordinate value is greater than the fourth preset threshold, it is determined that the target battery to be tested has an abnormal degradation fault.

9. The battery fault identification model training method according to claim 1, characterized in that: The serialized test related to the battery characteristics is to test the voltage characteristics of the battery in sequence according to the time sequence in the preset time period.

10. A battery fault identification model training device, characterized in that: The device comprises: A testing unit, used to perform serial tests related to battery characteristics on a preset reference battery and each target battery to be tested, respectively, to obtain first time series test data of the reference battery and second time series test data of each target battery to be tested; A first calculation unit, configured to perform dynamic time warping on each of the second time series test data and the first time series test data one by one, to obtain a dynamic time warping value of each of the target battery to be tested and the reference battery; A second calculation unit, configured to calculate the difference of average differential voltage between each of the second time series test data and the first time series test data one by one, to obtain the average differential voltage difference between each of the target test cells and the reference cell; A training unit is used to input each of the dynamic time warping values ​​and each of the average differential voltage difference values ​​as sample feature data of each of the target batteries to be tested into a preset battery fault identification model, and train the battery fault identification model according to each of the sample feature data.

Images