Method and device for detecting and positioning short-circuit fault in lithium ion battery pack and medium

Abstract

The invention discloses a lithium ion battery pack internal short circuit fault detection and positioning method and device and a medium, and the method comprises the steps: designing a battery monomer voltage abnormity index based on the correlation analysis of the local resultant force of a lithium ion battery pack monomer voltage signal; battery pack temperature feature extraction based on kernel principal component analysis is carried out to obtain a single battery temperature square prediction error; and constructing a multi-scale fault detection statistic based on the voltage anomaly index and the temperature square prediction error to realize fault detection, and constructing a fault contribution degree function based on the temperature square prediction error to realize fault positioning. According to the method, the voltage anomaly index of the single battery is designed, the square prediction error of the temperature of the single battery is obtained through extraction, finally, the multi-scale fault detection statistics and the fault contribution degree function are constructed to achieve fault detection and positioning, and the method is a pure data driving method and is low in equipment cost, easy to operate and high in reliability. And the temperature and voltage signal changes of the battery monomers in the early stage of internal short circuit are sensitive.

Description

Technical Field

[0001] The present invention belongs to the technical field of battery pack fault diagnosis, and specifically relates to a method, device and medium for detecting and locating short-circuit faults in a lithium-ion battery pack based on multi-scale variable analysis. Background Art

[0002] As a new type of green energy, lithium-ion batteries are widely used in electric vehicles, large-scale energy storage systems and various electronic devices due to their excellent properties such as high energy density, long cycle life, low self-discharge rate and environmental friendliness. However, safety accidents caused by lithium-ion battery systems occur from time to time. For example, internal short circuit is one of the most common lithium-ion battery failures, which can cause the battery system to generate a lot of heat and cause thermal runaway, which not only causes economic losses but also affects people's life safety. Therefore, it is crucial to develop effective methods to quickly detect and accurately locate early internal short circuit failures in lithium-ion battery packs.

[0003] At present, some data-based and model-based methods have been proposed for detecting and locating short-circuit faults in battery packs. For example, a fault diagnosis method based on the correlation coefficient of battery cell voltage signals, a short-circuit fault detection and location method for lithium-ion battery packs based on voltage signal kernel principal component analysis, and a model method that uses the measured values ​​of battery cell temperature and voltage to estimate the current electrochemical state inside the battery. However, in the early stage of internal short circuit, the changes in signals such as battery cell voltage and temperature are not obvious, which limits the performance of traditional data- and model-based detection and location methods.

[0004] In order to improve the accuracy of diagnosis, many advanced technologies are introduced into the detection and positioning of short circuits in lithium-ion battery packs. For example, electrochemical impedance spectroscopy, infrared thermal imaging, and ultrasonic detection methods are used in the diagnosis of short circuit faults in lithium battery packs. These methods have good diagnostic effects, but the high equipment cost and complexity limit their promotion in practical applications. At the same time, some current diagnostic methods can only detect faults but cannot accurately locate them, making it inconvenient to carry out targeted maintenance of battery packs. Therefore, it is necessary to propose a method based on multi-scale variable analysis to quickly detect and accurately locate the early internal short circuit phenomenon of battery packs. Summary of the invention

[0005] In order to solve the problems that the existing data and model-based battery pack internal short circuit fault diagnosis methods are insensitive to changes in voltage and temperature signals in the early internal short circuit stage, and the methods based on advanced technologies are costly and complicated, the present invention provides a lithium-ion battery pack internal short circuit fault detection and positioning method, equipment and medium, which can quickly detect and accurately locate the early internal short circuit phenomenon of the battery pack based on a multi-scale variable analysis method.

[0006] In order to solve the above technical problems, the technical means adopted by the present invention are as follows:

[0007] In the first aspect of the present invention, a method for detecting and locating short circuit faults in a lithium-ion battery pack based on multi-scale variable analysis is provided, the method comprising:

[0008] S1. Design the battery cell voltage abnormality index based on the local resultant force correlation analysis of the lithium-ion battery pack cell voltage signal;

[0009] S2, battery pack temperature feature extraction based on kernel principal component analysis to obtain the square prediction error of battery cell temperature;

[0010] S3. A multi-scale fault detection statistic is constructed based on the voltage anomaly index and the temperature square prediction error to achieve fault detection, and a fault contribution function is constructed based on the temperature square prediction error to achieve fault location.

[0011] In the above scheme, step S1 comprises:

[0012] S11, calculating the local resultant voltage of the battery cell;

[0013] S12, performing a local resultant force correlation analysis on the local resultant force of the voltage of the battery cell according to the correlation coefficient to obtain a local resultant force correlation coefficient of the voltage;

[0014] S13. Design voltage anomaly index based on local resultant force, including:

[0015] Calculate the average value of the correlation coefficient between each monomer and the local resultant voltage of all monomers;

[0016] The absolute value of the maximum value of the average value of the local resultant voltage correlation coefficients of all monomers at each moment is taken as the voltage anomaly index.

[0017] In the above solution, step S11 includes:

[0018] Standardize the voltage signal of the battery cell to obtain the standard score of the voltage:

[0019]

[0020] Among them, V i ′ (t) is the voltage standard fraction of battery cell i at the tth second, V i (t) is the voltage signal of battery cell i at the tth second, and σ(t) represent the mean and standard deviation of the voltage signal of battery cell i in a given time window with t as the end point, respectively;

[0021] Based on this, the local voltage force of the battery cell is calculated:

[0022]

[0023] Among them, F i (t) is the local resultant force generated by the battery cell i due to the influence of n′ other cells, V j ′ (t) and V i ′ (t) represents the voltage standard fraction of monomer j and monomer i at time t, n ′ is the number of monomers other than monomer i;

[0024] The step S12 comprises:

[0025] According to the correlation coefficient, the local resultant force correlation analysis of the voltage local resultant force of the battery cell is performed:

[0026]

[0027] Among them, F i and F j are the local resultant voltage forces on monomer i and monomer j, ρ i,j (t) represents the time at t 0 The correlation coefficient of the local resultant voltage of monomer i and monomer j in the time window from t to t, cov i,j (t) represents the F in the time window i and F j The covariance of i (t) and σ j (t) represents the F i and F j The variance of i (t) and μ j (t) represents the F i and F j The mean of

[0028] The step S13 comprises:

[0029] Calculate the average value of the correlation coefficient between each monomer and the local resultant voltage of all monomers:

[0030]

[0031] in, is the average value of the correlation coefficient between each monomer and the local resultant voltage of all monomers, n is the number of monomers, ρ i,j (t) is the correlation coefficient of the local resultant voltage between monomer i and monomer j;

[0032] Take n monomers at each moment The absolute value of the maximum value is taken as the voltage anomaly index:

[0033]

[0034] Wherein, h(t) represents the voltage anomaly index.

[0035] In the above scheme, step S2 includes:

[0036] S21, calculating the temperature signal kernel matrix;

[0037] S22, extracting nonlinear features based on the temperature signal kernel matrix to obtain temperature data after feature extraction;

[0038] S23. Based on the obtained temperature data after feature extraction, derive the temperature square prediction error.

[0039] In the above solution, step S21 includes:

[0040] According to the battery cell temperature data measured by the sensor, the temperature signal kernel matrix is ​​constructed:

[0041]

[0042] Where K is the temperature signal kernel matrix, l represents the number of training samples, x(t) = [x 1 (t),…,x m (t)] T is the temperature data of each battery cell at the tth second, m is the number of cells, φ(x(t)) represents a mapping from m dimensions to higher dimensions, and k(x, x′) is a radial basis function, which is specifically defined as follows:

[0043]

[0044] The radial basis function is used to measure the relationship between two vectors x and x. ′ The similarity between them, where σ is called the bandwidth of the function;

[0045] The step S22 comprises:

[0046] Solve the characteristic roots of the temperature signal kernel matrix K:

[0047] λv=Kv

[0048] Among them, λ is the characteristic root of the temperature signal kernel matrix K, and v is the characteristic vector of the matrix K;

[0049] According to the order of the characteristic roots from large to small, the corresponding eigenvectors are organized into a characteristic matrix V by column:

[0050] V=[v 1 ,v 2 ,…,vl ]

[0051] Among them, v 1 ,v 2 ,…,v l The corresponding eigenvalue is λ 1 ,λ 2 ,…,λ n , and λ 1 >λ 2 >…>λ n ;

[0052] Based on the temperature signal kernel matrix K, nonlinear feature extraction of temperature data is performed:

[0053] G d =KV d

[0054] Among them, G d =[g′(1),…,g′(l)] T is the temperature data after feature extraction, V d is the first d columns of V, g′(t)=[g′ 1 (t),…,g′ d (t)] T ;

[0055] The step S23 comprises:

[0056] Derive the temperature squared prediction error:

[0057]

[0058] Where SPE(t) is the squared prediction error of temperature at time t, g(t) = [g 1 (t),…,g l (t)] T It is the tth row of G=KV.

[0059] In the above scheme, step S3 includes:

[0060] S31, constructing a multi-scale fault detection statistic based on the voltage anomaly index and the temperature square prediction error, if the multi-scale fault detection statistic is greater than or equal to a threshold value of the multi-scale statistic, a short circuit fault occurs in the lithium-ion battery pack;

[0061] S32. A fault contribution function is constructed based on the voltage local resultant force correlation coefficient and the temperature square prediction error. The battery cell with the largest fault contribution is considered to be a faulty cell.

[0062] In the above solution, step S31 includes:

[0063] Based on the voltage anomaly index h(t) and the temperature square prediction error SPE(t), a multi-scale fault detection statistic is designed:

[0064]

[0065] Where M(t) is the multi-scale fault detection statistic, t is time, h r and SPE r are the reference signals of voltage anomaly index and temperature square prediction error, respectively, and α is the weight parameter;

[0066] Based on the weight parameter α, the threshold of the multi-scale statistic is calculated:

[0067] M r =αh r +(1-α)SPE r

[0068] Among them, M r is the threshold of the multi-scale statistic;

[0069] The internal short circuit fault in the lithium-ion battery pack is detected by multi-scale fault detection statistics and multi-scale statistical thresholds:

[0070]

[0071] Among them, the first time M(t)≥M r The time is recorded as t d ;

[0072] The step S32 comprises:

[0073] Based on the local resultant voltage correlation coefficient and the temperature square prediction error, the fault contribution function is designed:

[0074]

[0075] Among them, C j (t) is the internal short-circuit fault contribution function of battery cell j, β is the weight parameter, is the average value of the correlation coefficient between each monomer and the local resultant voltage of all monomers, sgn() is the sign function, x j (t) is the temperature data of battery cell j at the tth second; d At this moment, the battery cell with the largest fault contribution is considered to be the faulty cell:

[0076] z f = arg maxC z (t d )

[0077] Among them, z fIt is the serial number of the faulty cell.

[0078] In the above scheme, the reference signals of voltage anomaly index and temperature square prediction error are calculated by kernel probability density estimation method:

[0079]

[0080] Where p is the confidence level, f(s) is the probability density function of variable s, s∈{h(t),SPE(t)}, β r ∈{h r ,SPE r}.

[0081] According to a second aspect of the present invention, a wind turbine bearing service life prediction device is provided, comprising: a processor and a memory, the memory storing programs or instructions that can be run on the processor, and the programs or instructions, when executed by the processor, implement the steps of any one of the above-mentioned methods for detecting and locating short-circuit faults in lithium-ion battery packs based on multi-scale variable analysis.

[0082] According to a third aspect of the present invention, a readable storage medium is provided, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the method for detecting and locating short-circuit faults in a lithium-ion battery pack based on multi-scale variable analysis are implemented.

[0083] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:

[0084] The present invention provides a method for detecting and locating short-circuit faults in lithium-ion battery packs based on multi-scale variable analysis. The method designs a battery cell voltage anomaly index based on the local resultant correlation analysis of the lithium-ion battery cell voltage signal, and obtains the square prediction error of the battery cell temperature based on the battery pack temperature feature extraction based on the kernel principal component analysis. Finally, a multi-scale fault detection statistic and a fault contribution function are constructed to achieve fault detection and location. This method is a purely data-driven method with low equipment cost and simple operation, so it is suitable for internal short-circuit fault detection and location in actual battery packs. At the same time, this method is more sensitive to changes in temperature and voltage signals of battery cells in the early stage of internal short circuit. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 It is a schematic flow chart of a method for detecting and locating short-circuit faults in a lithium-ion battery pack based on multi-scale variable analysis of the present invention;

[0086] Figure 2 It is a schematic diagram of the structure of the battery pack used in the experiment of the present invention;

[0087] Figure 3is a waveform diagram of the UDDS of an embodiment of the present invention;

[0088] Figure 4 is a waveform diagram of the voltage and temperature signals of each battery cell in Fault 1 of the embodiment of the present invention; wherein, Figure 4 (a) is the voltage waveform. Figure 4 (b) is the temperature waveform diagram;

[0089] Figure 5 is a diagram of the internal short circuit fault detection result in an embodiment of the present invention; wherein, Figure 5 (a), (b), (c) and (d) are the internal short circuit fault detection result diagrams under the conditions of fault 6, fault 8, fault 9 and fault 10 respectively;

[0090] Figure 6 is a diagram showing the internal short circuit fault location result in an embodiment of the present invention; wherein, Figure 6 (a), (b), (c) and (d) are the internal short circuit fault location result diagrams under fault 1, fault 2, fault 3 and fault 4 conditions respectively. DETAILED DESCRIPTION

[0091] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0092] The present invention provides a method, device and medium for detecting and locating short-circuit faults in lithium-ion battery packs based on multi-scale variable analysis. The battery cell voltage anomaly index is designed based on the local resultant correlation analysis of the lithium-ion battery cell voltage signal, and the square prediction error of the battery cell temperature is obtained according to the battery pack temperature feature extraction based on the kernel principal component analysis. Finally, a multi-scale fault detection statistic and a fault contribution function are constructed to achieve fault detection and location. This method is a purely data-driven method with low equipment cost and simple operation, so it is suitable for internal short-circuit fault detection and location in actual battery packs. At the same time, this method is more sensitive to the temperature and voltage signal changes of the battery cell in the early stage of internal short circuit.

[0093] like Figure 1 As shown, according to one embodiment of the present invention, a method for detecting and locating short-circuit faults in a lithium-ion battery pack based on multi-scale variable analysis is proposed, comprising the following steps:

[0094] S1. Design the battery cell voltage abnormality index based on the local resultant force correlation analysis of the lithium-ion battery pack cell voltage signal;

[0095] S2, battery pack temperature feature extraction based on kernel principal component analysis to obtain the square prediction error of battery cell temperature;

[0096] S3. Construct multi-scale fault detection statistics and fault contribution functions to achieve fault detection and location.

[0097] The specific process of step S1 includes:

[0098] S11, calculation of local resultant force of voltage signal;

[0099] S12, local resultant force correlation analysis;

[0100] S13. Design of voltage anomaly index based on local resultant force.

[0101] Wherein, step S11 is specifically as follows:

[0102] Standardize the voltage signal of the battery cell to obtain the standard score of the voltage:

[0103]

[0104] Among them, V i ′ (t) is the voltage standard fraction of battery cell i at the tth second, V i (t) is the voltage signal of battery cell i measured by the sensor at the tth second, and σ(t) represent the mean and standard deviation of the voltage signal of monomer i in a given time window with t as the end point, respectively.

[0105] Based on this, the local voltage force of the battery cell is calculated:

[0106]

[0107] Among them, F i (t) is the local resultant force of monomer i due to the influence of n′ other monomers, V j ′ (t) and V i ′ (t) represents the standard fraction of monomer j and monomer i at time t, n ′ is the number of monomers other than monomer i.

[0108] Wherein, step S12 is specifically as follows:

[0109] According to the correlation coefficient, the local resultant force correlation analysis of the voltage local resultant force of the battery cell is performed:

[0110]

[0111] Among them, F i and F j are the local forces received by monomer i and monomer j, respectively, and ρ i,j (t) represents the time at t 0 The correlation coefficient of the local resultant voltage of monomer i and monomer j in the time window from t to t, cov i,j (t) represents the F in the time window i and F j The covariance of i (t) and σ j (t) represents the F i and F j The variance of i (t) and μ j (t) represents the F i and F j The mean of .

[0112] Wherein, step S13 is specifically as follows:

[0113] Calculate the average value of the correlation coefficient between each monomer and the local resultant force of all monomers:

[0114]

[0115] in, is the average value of the correlation coefficient between each monomer and the local resultant force of all monomers, n is the number of monomers, ρ i,j (t) is the correlation coefficient of the local resultant voltage between monomer i and monomer j.

[0116] Take n monomers at each moment The absolute value of the maximum value is taken as the voltage anomaly index:

[0117]

[0118] Wherein, h(t) represents the voltage anomaly index.

[0119] The specific process of step S2 includes:

[0120] S21, temperature signal kernel matrix calculation;

[0121] S22, nonlinear feature extraction based on kernel matrix;

[0122] S23. Derivation of temperature squared prediction error.

[0123] Wherein, step S21 is specifically as follows:

[0124] According to the battery cell temperature data measured by the sensor, the temperature kernel matrix is ​​constructed:

[0125]

[0126] Where K is the kernel matrix of the temperature signal, l represents the number of training samples, x(t) = [x 1 (t),…,x m (t)] T is the temperature data of each battery cell at the tth second, m is the number of cells, φ(x(t)) represents a mapping from m dimensions to higher dimensions, and κ(x, x′) is a radial basis function, which is specifically defined as follows:

[0127]

[0128] The radial basis function is used to measure the relationship between two vectors x and x. ′ The similarity between them is σ, where σ is called the bandwidth of the function.

[0129] Wherein, step S22 is specifically as follows:

[0130] Solve the characteristic root problem of the temperature kernel matrix K:

[0131] λv=Kv

[0132] Among them, λ is the characteristic root of matrix K, and v is the eigenvector of matrix K.

[0133] According to the order of the characteristic roots from large to small, the corresponding eigenvectors are organized into a characteristic matrix V by column:

[0134] V=[v 1 ,v 2 ,…,v l ]

[0135] Among them, v 1 ,v 2 ,…,v l The corresponding eigenvalue is λ 1 ,λ 2 ,…,λ n , and λ 1 >λ 2 >…>λ n .

[0136] Based on the kernel matrix K, nonlinear feature extraction of temperature data is performed:

[0137] G d =KV d

[0138] Among them, G d =[g′(1),…,g′(l)] T is the temperature data after feature extraction, V d is the first d columns of V, g′(t)=[g′ 1(t),…,g′ d (t)] T .

[0139] Wherein, step S23 is specifically as follows:

[0140] Based on the above results, the temperature square prediction error is derived:

[0141]

[0142] Where SPE(t) is the squared prediction error at time t, g(t) = [g 1 (t),…,g l (t)] T It is the tth row of G=KV.

[0143] The specific process of step S3 includes:

[0144] S31. Multi-scale statistics design based on voltage anomaly index and temperature square prediction error;

[0145] S32. Design of fault contribution function based on voltage local resultant force correlation coefficient and temperature square prediction error.

[0146] Wherein, step S31 is specifically as follows:

[0147] Based on the voltage anomaly index and temperature square prediction error, multi-scale statistics are designed:

[0148]

[0149] Among them, M(t) is the multi-scale statistic, h r and SPE r are the reference signals of voltage anomaly index and temperature square prediction error respectively, and α is the weight parameter. The reference signal can be calculated by kernel probability density estimation method:

[0150]

[0151] Where p is the confidence level, usually set to 0.95 or 0.99, f(s) is the probability density function of variable s, which can be calculated by kernel density estimation, s∈{h(t),SPE(t)}, β r ∈{h r ,SPE r}.

[0152] Based on the weight parameter α, the threshold of the multi-scale statistic is calculated:

[0153] M r =αh r +(1-α)SPE r

[0154] Among them, M r is the threshold of the multi-scale statistic. Through the multi-scale statistic and its threshold, the internal short circuit fault in the lithium-ion battery pack can be detected:

[0155]

[0156] Among them, the first time M(t)≥M r The time is recorded as t d .

[0157] Wherein, step S32 is specifically as follows:

[0158] Based on the local resultant voltage correlation coefficient and the temperature square prediction error, the fault contribution function is designed:

[0159]

[0160] Among them, C j (t) is the internal short-circuit fault contribution function of battery cell j, and β is the weight parameter. d At this moment, the battery cell with the largest fault contribution is considered to be the faulty cell:

[0161] z f = arg maxC z (t d )

[0162] Among them, z f It is the serial number of the faulty cell.

[0163] More specifically, the solution was implemented by taking the internal short circuit fault detection and location of a battery pack consisting of 24 battery cells as an example. The structure of the battery pack used in the experiment is as follows: Figure 2 As shown, every 4 battery cells are connected in parallel first, and then the 6 parallel bodies are connected in series. The main parameters of the model are listed in Table 1.

[0164] Table 1 Main model parameters

[0165]

[0166] The following four performance indicators are defined for performance evaluation and comparison of internal short-circuit fault detection and location methods:

[0167] Recall rate:

[0168] False Positive Rate:

[0169] Fault detection delay: FDD = t d -t f

[0170] Among them, the recall rate represents the number of fault samples detected N d The total number of fault samples N af The false alarm rate represents the number of false alarm samples N. f The total number of normal samples N an The ratio of d and t f They represent the moment when the fault is first detected and the time when the fault actually occurs, respectively.

[0171] For this battery pack, a total of 10 faults are designed. The detailed fault parameters are shown in Table 2. Among them, 1C represents the constant current corresponding to the discharge from 100% to 0 in one hour. Under the 2C condition, the current doubles, and theoretically it only takes half an hour to complete the full discharge. The waveform of UDDS is as follows Figure 3 The temperature sensor and voltage sensor measure the temperature and voltage data of 24 battery cells from 0 to 2000 seconds respectively. The curves of the voltage and temperature of 24 battery cells in fault 1 changing with time are shown in Figure 4 As shown, Figure 4 (a) is the voltage waveform. Figure 4 (b) in the figure is the temperature waveform.

[0172] Table 2 Fault Setting Table

[0173]

[0174] Using the proposed multi-scale variable analysis method, short-circuit fault detection and location experiments are carried out on faults 1 to 10 in Table 2 in the lithium-ion battery pack.

[0175] Among them, the detection results of fault 6, fault 8, fault 9, and fault 10 are as follows: Figure 5 Fault 6 is an internal short circuit fault with an internal short circuit resistance of 10Ω at 2C rate, which occurs on battery No. 23 at 1000 seconds. Figure 5 As can be seen in (a) of Figure 1, the proposed multi-scale variable analysis method captures the internal short circuit fault in a timely and clear manner. The parameters of fault 8 and fault 6 differ only in the internal short circuit resistance. The internal short circuit resistance of the former is smaller, which means that the internal short circuit phenomenon of fault 8 is more obvious. Figure 5 (a) and Figure 5 In (b), we can see that under the condition of fault 8, the value of the multi-scale statistics changes more significantly after the fault occurs. Compared with fault 6, the fault occurrence time of fault 9 is delayed by 500 seconds. Figure 5 (a) and Figure 5 In (c), we can see that the proposed method detects that fault 9 occurs at about 1500 seconds. Fault 10 is an internal short circuit fault that occurs under dynamic conditions. The detection effect of the proposed method on fault 10 is shown in Figure 5 As shown in (d) in the figure, it can be seen that the internal short circuit fault occurring at the 1000th second is detected quickly.

[0176] The location effects of the proposed method on fault 1, fault 2, fault 3, and fault 4 are shown as follows: Figure 6 As shown, Figure 6 Figures (a), (b), (c) and (d) are the internal short circuit fault location results under the conditions of fault 1, fault 2, fault 3 and fault 4, respectively. It can be seen from the figures that these internal short circuit faults occurred on battery cells No. 4, No. 5, No. 11 and No. 16, respectively. Figure 6 As can be seen from the bar chart in , there is a significant difference in contribution between the faulty battery cell and the normal battery cell, which means that this method can accurately locate the faulty battery cell.

[0177] The method was implemented on 10 groups of faults, and the recall rate, false alarm rate and fault detection delay of different faults are shown in Table 3. It can be seen that for all faults, the proposed method has a higher recall rate, lower false alarm rate and shorter fault detection delay, which shows that the method can quickly detect the internal short circuit fault in the lithium-ion battery pack; at the same time, the positioning result of the method is the same as the parameters in the fault setting, indicating that the method can accurately locate the faulty monomer.

[0178] Table 3 Effects of the proposed method on 10 groups of faults

[0179]

[0180] In order to show the superiority of the proposed method, its effect on fault 1 is compared with the other two methods (correlation coefficient method based only on voltage signal and kernel principal component analysis based only on temperature signal), and the comparison results are shown in Table 4. From the data in the analysis table, it can be seen that although the correlation coefficient method based only on voltage data has a lower false alarm rate, its reduction in recall rate and extension of fault detection delay make it lose its superiority. The kernel principal component analysis method based only on temperature signal will give a false alarm long before the fault occurs, and the effect on fault 1 is very poor.

[0181] Table 4 Comparison of the effects of three methods on fault 1

[0182]

[0183] According to a second aspect of the present invention, an electronic device is proposed, comprising: a processor and a memory, the memory storing programs or instructions that can be run on the processor, and the programs or instructions, when executed by the processor, implement the steps of the method for detecting and locating short-circuit faults in lithium-ion battery packs based on multi-scale variable analysis as provided in the first aspect.

[0184] According to a third aspect of the present invention, a readable storage medium is provided, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the method for detecting and locating short-circuit faults in a lithium-ion battery pack based on multi-scale variable analysis as provided in the first aspect are implemented.

[0185] In summary, the present invention designs the battery cell voltage anomaly index based on the local resultant correlation analysis of the lithium-ion battery pack cell voltage signal, and obtains the square prediction error of the battery cell temperature according to the battery pack temperature feature extraction based on the kernel principal component analysis, and finally constructs a multi-scale fault detection statistic and a fault contribution function to achieve fault detection and location. This method is a purely data-driven method with low equipment cost and simple operation, so it is suitable for internal short circuit fault detection and location in actual battery packs. At the same time, this method is more sensitive to the temperature and voltage signal changes of the battery cell in the early stage of internal short circuit, and can capture the anomaly in the early stage of internal short circuit. The proposed method has been verified in a battery pack consisting of 24 battery cells. The method is used to diagnose 10 groups of internal short circuit faults set in the battery pack, all of which show a high recall rate, a low false alarm rate and a short fault detection delay, and accurately locate the faulty cell.

[0186] It should be noted that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0187] It should be pointed out that, according to the needs of implementation, the various steps / components described in this application can be split into more steps / components, and two or more steps / components or partial operations of steps / components can be combined into new steps / components to achieve the purpose of the present invention.

[0188] It will be easily understood by those skilled in the art that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for detecting and locating short-circuit faults in lithium-ion battery packs based on multi-scale variable analysis, characterized in that: The method includes: S1. Design the battery cell voltage abnormality index based on the local resultant force correlation analysis of the lithium-ion battery pack cell voltage signal; S2, battery pack temperature feature extraction based on kernel principal component analysis to obtain the square prediction error of battery cell temperature; S3. A multi-scale fault detection statistic is constructed based on the voltage anomaly index and the temperature square prediction error to achieve fault detection, and a fault contribution function is constructed based on the temperature square prediction error to achieve fault location.

2. The method for detecting and locating short-circuit faults in lithium-ion battery packs based on multi-scale variable analysis according to claim 1, characterized in that: The step S1 comprises: S11, calculating the local resultant voltage of the battery cell; S12, performing a local resultant force correlation analysis on the local resultant force of the voltage of the battery cell according to the correlation coefficient to obtain a local resultant force correlation coefficient of the voltage; S13. Design voltage anomaly index based on local resultant force, including: Calculate the average value of the correlation coefficient between each monomer and the local resultant voltage of all monomers; The absolute value of the maximum value of the average value of the local resultant voltage correlation coefficients of all monomers at each moment is taken as the voltage anomaly index.

3. The method for detecting and locating short-circuit faults in lithium-ion battery packs based on multi-scale variable analysis according to claim 2, characterized in that: The step S11 comprises: Standardize the voltage signal of the battery cell to obtain the standard score of the voltage: Among them, V i ′(t) is the voltage standard fraction of battery cell i at the tth second, V i (t) is the voltage signal of battery cell i at the tth second, and σ(t) represent the mean and standard deviation of the voltage signal of battery cell i in a given time window with t as the end point, respectively; Based on this, the local voltage force of the battery cell is calculated: Among them, F i (t) is the local resultant force generated by the battery cell i due to the influence of n′ other cells, V j ′(t) and V i ′(t) represents the voltage standard fraction of monomer j and monomer i at time t, respectively, and n′ is the number of monomers other than monomer i; The step S12 comprises: According to the correlation coefficient, the local resultant force correlation analysis of the voltage local resultant force of the battery cell is performed: Among them, F i and F j are the local resultant voltage forces on monomer i and monomer j, ρ i,j (t) represents the correlation coefficient of the local resultant voltage of monomer i and monomer j in the time window from t0 to t, cov i,j (t) represents the F in the time window i and F j The covariance of i (t) and σ j (t) represents the F i and F j The variance of i (t) and μ j (t) represents the F i and F j The mean of The step S13 comprises: Calculate the average value of the correlation coefficient between each monomer and the local resultant voltage of all monomers: in, is the average value of the correlation coefficient between each monomer and the local resultant voltage of all monomers, n is the number of monomers, ρ i,j (t) is the correlation coefficient of the local resultant voltage between monomer i and monomer j; Take n monomers at each moment The absolute value of the maximum value is taken as the voltage anomaly index: Wherein, h(t) represents the voltage anomaly index.

4. The method for detecting and locating short-circuit faults in lithium-ion battery packs based on multi-scale variable analysis according to claim 1, characterized in that: The step S2 comprises: S21, calculating the temperature signal kernel matrix; S22, extracting nonlinear features based on the temperature signal kernel matrix to obtain temperature data after feature extraction; S23. Based on the obtained temperature data after feature extraction, derive the temperature square prediction error.

5. The method for detecting and locating short-circuit faults in lithium-ion battery packs based on multi-scale variable analysis according to claim 4, characterized in that: The step S21 comprises: According to the battery cell temperature data measured by the sensor, the temperature signal kernel matrix is ​​constructed: Where K is the temperature signal kernel matrix, l represents the number of training samples, x(t) = [x1(t),…,x m (t)] T is the temperature data of each battery cell at the tth second, m is the number of cells, φ(x(t)) represents a mapping from m dimensions to higher dimensions, and κ(x, x′) is a radial basis function, which is specifically defined as follows: The radial basis function is used to measure the similarity between two vectors x and x′, where σ is called the bandwidth of the function; The step S22 comprises: Solve the characteristic roots of the temperature signal kernel matrix K: λv=Kv Among them, λ is the characteristic root of the temperature signal kernel matrix K, and v is the characteristic vector of the matrix K; According to the order of the characteristic roots from large to small, the corresponding eigenvectors are organized into a characteristic matrix V by column: V=[v1,v2,…,v l ] Among them, v1, v2, …, v l The corresponding eigenvalues ​​are λ1,λ2,…,λ n , and λ1>λ2>…>λ n ; Based on the temperature signal kernel matrix K, nonlinear feature extraction of temperature data is performed: G d =KV d Among them, G d =[g′(1),…,g′(l)] T is the temperature data after feature extraction, V d is the first d columns of V, g′(t)=[g′1(t),…,g′ d (t)] T ; The step S23 comprises: Derive the temperature squared prediction error: Where SPE(t) is the squared prediction error of temperature at time t, g(t) = [g1(t),…,g l (t)] T It is the tth row of G=KV.

6. The method for detecting and locating short-circuit faults in lithium-ion battery packs based on multi-scale variable analysis according to claim 2, characterized in that: The step S3 comprises: S31, constructing a multi-scale fault detection statistic based on the voltage anomaly index and the temperature square prediction error, if the multi-scale fault detection statistic is greater than or equal to a threshold value of the multi-scale statistic, a short circuit fault occurs in the lithium-ion battery pack; S32. A fault contribution function is constructed based on the voltage local resultant force correlation coefficient and the temperature square prediction error. The battery cell with the largest fault contribution is considered to be a faulty cell.

7. The method for detecting and locating short-circuit faults in lithium-ion battery packs based on multi-scale variable analysis according to claim 6, characterized in that: The step S31 comprises: Based on the voltage anomaly index h(t) and the temperature square prediction error SPE(t), a multi-scale fault detection statistic is designed: Where M(t) is the multi-scale fault detection statistic, t is time, h r and SPE r are the reference signals of voltage anomaly index and temperature square prediction error, respectively, and α is the weight parameter; Based on the weight parameter α, the threshold of the multi-scale statistics is calculated: M r =ah r +(1-α)SPE r Among them, M r is the threshold of the multi-scale statistic; The internal short circuit fault in the lithium-ion battery pack is detected by multi-scale fault detection statistics and multi-scale statistical thresholds: Among them, the first time M(t)≥M r The time is recorded as t d ; The step S32 comprises: Based on the local voltage resultant force correlation coefficient and the temperature square prediction error, the fault contribution function is designed: Among them, C j (t) is the internal short-circuit fault contribution function of battery cell j, β is the weight parameter, is the average value of the correlation coefficient between each monomer and the local resultant voltage of all monomers, sgn() is the sign function, x j (t) is the temperature data of battery cell j at the tth second; d At this moment, the battery cell with the largest fault contribution is considered to be the faulty cell: z f =arg maxC z (t d ) Among them, z f It is the serial number of the faulty cell.

8. The method for detecting and locating short-circuit faults in lithium-ion battery packs based on multi-scale variable analysis according to claim 7, characterized in that: The reference signals of voltage anomaly index and temperature square prediction error are calculated by kernel probability density estimation method: Where p is the confidence level, f(s) is the probability density function of variable s, s∈{h(t),SPE(t)}, β r ∈{h r ,SPE r }.

9. An electronic device, characterized in that: include: A processor and a memory, the memory storing programs or instructions that can be run on the processor, and the programs or instructions, when executed by the processor, implement the steps of the method for detecting and locating short-circuit faults in lithium-ion battery packs based on multi-scale variable analysis as described in any one of claims 1 to 8.

10. A readable storage medium, characterized in that: A program or instruction is stored thereon, and when the program or instruction is executed by a processor, the steps of the method for detecting and locating short-circuit faults in a lithium-ion battery pack based on multi-scale variable analysis as described in any one of claims 1 to 8 are implemented.

Images