Large-size lithium battery thermal anomaly detection method and storage medium

Abstract

The application discloses a large-size lithium battery thermal abnormality detection method and a storage medium, and comprises the following steps: acquiring temperature data of the large-size lithium battery and converting the temperature data into a temperature data matrix; calculating a first detection index and a second detection index according to the FastICA algorithm by using the temperature data matrix; in the training stage, calculating corresponding first control limits and second control limits by using the first detection index and the second detection index and the KDE algorithm; in the detection stage, detecting the first detection index value and the second detection index value at time t online; comparing the first detection index value and the second detection index value with the corresponding first control limits and second control limits respectively, and performing fault judgment. According to the distributed temperature characteristics of the large-size lithium battery as a whole, the FastICA algorithm and the KDE algorithm are used to calculate the control limits, and the limited temperature data is used to realize the thermal abnormality fault detection of the large-size lithium ion battery without a model structure.

Description

Technical Field

[0001] This invention relates to the field of new energy technology, and in particular to a method for detecting thermal anomalies in large-size lithium batteries and a storage medium thereon. Background Technology

[0002] Temperature is a critical factor affecting the availability, safety, and lifespan of lithium-ion batteries used in vehicles. As the size of lithium-ion batteries gradually increases, the surface area to volume ratio decreases, making thermal safety issues increasingly prominent. Localized overheating inside the lithium battery accelerates side reactions during battery failures, such as internal and external short circuits. This causes heat release, leading to smoke, and in severe cases, even thermal runaway. Therefore, real-time thermal fault detection of large-size lithium batteries is crucial for the safety of electric vehicles. The temperature distribution of a large-size lithium battery is a typical distributed parameter system, with its inputs, outputs, and even parameters varying with time and space. If only radial temperature variations are considered, the thermal process of a cylindrical lithium battery can be simplified to a one-dimensional distributed parameter system.

[0003] However, for prismatic or pouch lithium batteries, their distributed thermal processes are two-dimensional distributed parameter systems, with temperatures varying significantly along their length and width. This variation makes thermal fault detection in large-size lithium-ion batteries challenging: 1) the fault detection model must consider two-dimensional spatiotemporal thermodynamics; 2) due to installation limitations, only a limited number of sensors can be used. Therefore, there is an urgent need to develop a thermal fault detection method for large-size lithium-ion batteries that can accommodate a limited number of sensors and consider the two-dimensional spatiotemporal heat distribution. Summary of the Invention

[0004] Existing large-size square or pouch lithium batteries have fewer sensors installed, but their distributed thermal processes are two-dimensional distributed parameter systems with large temperature variations along the length and width directions, making thermal fault detection very challenging.

[0005] To address the aforementioned issues, a method and storage medium for detecting thermal anomalies in large-size lithium batteries are proposed. This method involves placing a limited number of temperature sensors at appropriate locations on the lithium battery. Based on spatial location and temperature data, a global spatial function matrix is ​​constructed using spectral methods to obtain a time coefficient matrix. Then, based on this time coefficient matrix, a first detection index and a second detection index are calculated using the FastICA algorithm. Finally, a first control limit and a second control limit are obtained using the KDE algorithm. By comparing the real-time values ​​of the first and second detection indexes with the corresponding control limits, the thermal anomaly status of the large-size lithium battery is determined, thus achieving localized thermal fault detection in large-size batteries.

[0006] Firstly, a method for detecting thermal anomalies in large-size lithium batteries includes: Step 100: Obtain temperature data of large-size lithium batteries and convert it into a temperature data matrix; Step 200: Using the temperature data matrix and the FastICA algorithm, calculate the first detection index and the second detection index; Step 300: During the training phase, the first detection index and the second detection index are used to calculate the corresponding first control limit and second control limit according to the KDE algorithm. Step 400: During the detection phase, the first detection index value and the second detection index value are detected online at time t. Step 500: Compare the first detection index value and the second detection index value with the corresponding first control limit and second control limit respectively to determine the fault.

[0007] In conjunction with the large-size lithium battery thermal anomaly detection method described in the first aspect of the present invention, in a first possible embodiment, step 100 includes: Step 110: Convert the temperature data into a preset temperature data matrix, wherein the temperature data matrix is: (1) in, This represents a matrix of temperature data output from N sensors. This represents the spatial location of N sensors.

[0008] In conjunction with the first possible embodiment of the first aspect of the present invention, in the second possible embodiment, step 200 includes: Step 210: Decompose the battery thermodynamic equation of the lithium battery using the spectral method to obtain the spatial function matrix and the corresponding time coefficient matrix; Step 220: Calculate the first detection index and the second detection index using the spatial function matrix and the time coefficient matrix, and according to the FastICA algorithm.

[0009] In conjunction with the second possible implementation of the first aspect of the present invention, in the third possible implementation, step 210 includes: Step 211: Obtain the battery thermodynamic equation: (2) The boundary conditions for the battery thermodynamic equation are as follows: (3) The initial conditions for the battery thermodynamic equations are: (4) in, , and These represent spatiotemporal temperature, ambient temperature, and initial temperature, respectively. Indicates the spatial location of the sensor; The length and width of the battery are represented by t; t represents the time variable. Indicates density; Indicates specific heat capacity; Indicates the heat transfer coefficient; , in Indicates the thermal convection coefficient; and These are the internal heat generation and lateral heat flow of the battery, respectively; f represents an unknown thermal fault or anomaly. Step 212: Construct auxiliary functions : (5) Step 213: Utilize the auxiliary function The battery thermodynamic equation is transformed into: (6) The boundary conditions for equation (6) are: (7) Where V is , representing the temperature variable after homogenization of the boundary conditions.

[0010] In conjunction with the third possible implementation of the first aspect of the present invention, in the fourth possible implementation, step 210 further includes: Step 214: Decompose the two-dimensional spatiotemporal battery thermodynamic equations using spectral methods as follows: (8) (9) in, Indicates the time coefficient. Let n represent the basis functions in two-dimensional space, and n represent the model order. Let j be the spatial basis function in the x-direction, and j be a natural number: (10); Spatial basis functions in the y-direction: (11); in, The constant coefficient, For eigenvalues; Step 215: Process the temperature data matrix By performing homogenization, its homogenized matrix can be obtained. : (12) Step 216: Homogenize equation (8) to obtain the space function matrix. : (13) Where N is the number of sensors; Step 217: Using the spatial function matrix and the homogenized temperature data matrix, obtain the time coefficient matrix. : (14) in, It is a space function matrix The pseudo-inverse is given by the number of sensors N ≤ model order n.

[0011] In conjunction with the fourth possible implementation of the first aspect of the present invention, in the fifth possible implementation, step 220 includes: Step 221: Calculate independent sources using the FastICA algorithm. : (15) in, , , , Representing the time coefficient matrix respectively Orthogonal mixing matrix, whitening matrix, unmixing matrix, columns; Step 222: Calculate the dominant independent sources using the FastICA algorithm. : (16) in, For aliasing matrix The row with the largest sum of squares; Step 223: Utilize the dominant independent source aliasing matrix The column with the largest sum of squares Whitening matrix , Calculate the first detection index respectively Second detection indicator .

[0012] In conjunction with the fifth possible implementation of the first aspect of the present invention, in the sixth possible implementation, step 223 includes: Step 2231: The first detection index Defined as: (17) The second detection indicator Defined as: (18) in, For space function matrix The column; in, (19).

[0013] In conjunction with the sixth possible implementation of the first aspect of the present invention, in the seventh possible implementation, step 300 includes: Step 310: Obtain the kernel density function of the KDE algorithm: (20) in, For sampled data, and These are the number of samples and the bandwidth, respectively. It is a Gaussian kernel function. ; Step 320: Separately, the first detection index... The second detection indicator The point occupying 99% of the area of ​​the kernel density function is used as the first control limit. Second control limit .

[0014] In conjunction with the seventh possible embodiment of the first aspect of the present invention, in the eighth possible embodiment, step 500 includes: Step 510: Compare the first detection index value and the second detection index value with the corresponding first control limit and second control limit, respectively; Step 520: If the value of the first detection index is greater than the first control limit or the value of the second detection index is greater than the second control limit, then the lithium battery is determined to be faulty.

[0015] In a second aspect, a storage medium stores a computer program that, when loaded and executed by a processor, implements the anomaly detection method as described in the first aspect.

[0016] The large-size lithium battery thermal anomaly detection method described in this invention involves setting a limited number of temperature sensors at corresponding locations on the lithium battery. A global spatial function matrix is ​​constructed based on spatial location data and temperature data using spectral methods, thereby obtaining a time coefficient matrix. Then, based on the aforementioned time coefficient matrix, a first detection index and a second detection index are calculated using the FastICA algorithm. Finally, a first control limit and a second control limit are obtained using the KDE algorithm. By comparing the online real-time acquired values ​​of the first and second detection indices with the corresponding control limits, the thermal anomaly situation of the large-size lithium battery is determined, thus achieving localized thermal fault detection for large-size batteries. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram showing the distribution of four thermocouples and faults in an embodiment of the large-size lithium battery thermal anomaly detection method of the present invention; Figure 2 The input current under fault-free conditions provided in the embodiment of the large-size lithium battery thermal anomaly detection method of the present invention is... ; Figure 3 The output voltage under fault-free conditions provided in the embodiment of the large-size lithium battery thermal anomaly detection method of the present invention is as follows: ; Figure 4 The first detection index diagram of fault 1 provided in the embodiment of the large-size lithium battery thermal anomaly detection method of the present invention; Figure 5 The second detection index diagram of fault 1 is provided in the embodiment of the large-size lithium battery thermal anomaly detection method of the present invention; Figure 6 The first detection index diagram of fault 3 is provided in the embodiment of the thermal anomaly detection method for large-size lithium batteries of the present invention; Figure 7 The second detection index diagram of fault 3 is provided in the embodiment of the large-size lithium battery thermal anomaly detection method of the present invention; Figure 8 This is a first schematic diagram of the process for detecting thermal anomalies in large-size lithium batteries in this invention; Figure 9 This is a second schematic diagram of the process for detecting thermal anomalies in large-size lithium batteries in this invention; Figure 10This is the third schematic diagram of the large-size lithium battery thermal anomaly detection method in this invention; Figure 11 This is the fourth schematic diagram of the large-size lithium battery thermal anomaly detection method in this invention; Figure 12 This is the fifth schematic diagram of the large-size lithium battery thermal anomaly detection method in this invention; Figure 13 This is the sixth schematic diagram of the large-size lithium battery thermal anomaly detection method in this invention; Figure 14 This is the seventh schematic diagram of the large-size lithium battery thermal anomaly detection method in this invention; Detailed Implementation

[0019] The technical solutions of this invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are all within the scope of protection of this invention.

[0020] Name Explanation Spectral method: A numerical method for solving partial differential equations. Its key point is to approximately expand the solution into a finite series expansion of smooth functions (usually orthogonal polynomials), i.e., the so-called approximate spectral expansion of the solution. Then, based on this expansion and the original equation, a system of equations with coefficients of the expansion is derived. For unsteady problems, the system of equations also depends on time t. The spectral method is essentially a generalization of the standard separation of variables technique.

[0021] FastICA Algorithm: ICA (Independence Component Analysis), also known as the Fixed-Point algorithm, is a fast iterative optimization algorithm. Unlike ordinary neural network algorithms, it uses batch processing, meaning a large number of samples are involved in the computation at each iteration. However, from the perspective of distributed parallel processing, it can still be considered a neural network algorithm. FastICA algorithms exist in forms based on fourth-order cumulants, likelihood maximization, and negative entropy maximization. Furthermore, this algorithm employs a fixed-point iterative optimization algorithm, making convergence faster and more robust.

[0022] KDE Algorithm: KDE (Kernel Density Estimation) is used in probability theory to estimate unknown density functions, which are then used to infer the distribution of overall data based on a finite sample. Therefore, the result of kernel density estimation is the probability density function estimate of the sample. Based on this estimated probability density function, we can obtain some properties of the data distribution, such as the clustering regions of the data.

[0023] Existing large-size square or pouch lithium batteries have fewer sensors installed, but their distributed thermal processes are two-dimensional distributed parameter systems with large temperature variations along the length and width directions, making thermal fault detection very challenging.

[0024] To address the aforementioned issues, a method for detecting thermal anomalies in large-size lithium batteries and a readable storage medium are proposed.

[0025] I. Method Implementation Firstly, a method for detecting thermal anomalies in large-size lithium batteries, preferably, as follows: Figure 8 , Figure 8 This is a first schematic diagram of the large-size lithium battery thermal anomaly detection method of the present invention; it includes: Step 100, acquiring temperature data of the large-size lithium battery and converting it into a temperature data matrix; Step 200, using the temperature data matrix and calculating a first detection index and a second detection index according to the FastICA algorithm; Step 300, in the training phase, using the first detection index and the second detection index and calculating the corresponding first control limit and second control limit respectively according to the KDE algorithm; Step 400, in the detection phase, detecting the first detection index value and the second detection index value at time t online; Step 500, comparing the first detection index value and the second detection index value with the corresponding first control limit and second control limit respectively to determine the fault. By setting a limited number of temperature sensors at corresponding locations on the lithium battery, a global spatial function matrix is ​​constructed based on spatial location data and temperature data using spectral methods, thereby obtaining a time coefficient matrix. Then, based on the aforementioned time coefficient matrix, the first detection index and the second detection index are calculated using the FastICA algorithm. Subsequently, the first control limit and the second control limit are obtained using the KDE algorithm. By comparing the online real-time acquired values ​​of the first and second detection indexes with the corresponding control limits, the thermal anomaly of the large-size lithium battery can be obtained, thus realizing the detection of local thermal faults in large-size batteries.

[0026] A limited number of temperature sensors, such as thermocouples, are placed at certain locations in large-size lithium batteries to detect and acquire limited temperature data, which is then converted into a temperature data matrix. Specifically, step 100 includes: Step 110, converting the temperature data into a preset temperature data matrix, wherein the temperature data matrix is: (1) in, This represents a matrix of temperature data output from N sensors. This represents the spatial location of N sensors.

[0027] Preferably, such as Figure 9 , Figure 9 This is a second schematic diagram of the large-size lithium battery thermal anomaly detection method in this invention; step 200 includes: step 210, decomposing the battery thermodynamic equation of the lithium battery using a spectral method to obtain a spatial function matrix and a corresponding time coefficient matrix; step 220, using the spatial function matrix and the time coefficient matrix, and calculating a first detection index and a second detection index according to the FastICA algorithm.

[0028] To obtain the spatial function matrix and the corresponding time coefficient matrix, the battery thermodynamic equations need to be decomposed, specifically, as follows: Figure 10 , Figure 10 This is the third schematic diagram of the large-size lithium battery thermal anomaly detection method in this invention; step 210 includes step 211, obtaining the battery thermodynamic equation: (2) The boundary conditions for the battery thermodynamic equations are as follows: (3) The initial conditions for the battery thermodynamic equations are: (4) in, , and These represent spatiotemporal temperature, ambient temperature, and initial temperature, respectively. Indicates the spatial location of the sensor; The length and width of the battery are represented by t; t represents the time variable. Indicates density; Indicates specific heat capacity; Indicates the heat transfer coefficient; , in Indicates the thermal convection coefficient; and These represent the internal heat generation and lateral heat flow of the battery, respectively; f represents an unknown thermal fault or anomaly.

[0029] Step 212: Construct auxiliary functions : (5) Step 213: Using auxiliary functions The battery thermodynamic equation is transformed into: (6) The boundary conditions for equation (6) are: (7) Where V is , representing the temperature variable after homogenization of the boundary conditions.

[0030] Then, the two-dimensional spatiotemporal battery thermodynamic equations are decomposed using spectral methods, specifically, as follows: Figure 11 , Figure 11 This is the fourth schematic diagram of the large-size lithium battery thermal anomaly detection method in this invention; step 210 also includes: Step 214: Decompose the two-dimensional spatiotemporal battery thermodynamic equations using spectral methods as follows: (8) (9) in, Indicates the time coefficient. Let n represent the basis functions in two-dimensional space, and n represent the model order. Let j be the spatial basis function in the x-direction, and j be a natural number: (10) Spatial basis functions in the y-direction: (11) in, The constant coefficient, For eigenvalues; Step 215: Process the temperature data matrix By performing homogenization, its homogenized matrix can be obtained. : (12) Step 216: Homogenize equation (8) to obtain the space function matrix. : (13) Where N is the number of sensors.

[0031] The time coefficient matrix is ​​obtained using the spatial function matrix and the homogenized temperature data matrix. Specifically, in step 217, the time coefficient matrix is ​​obtained using the spatial function matrix and the homogenized temperature data matrix. : (14) in, It is a space function matrix The pseudo-inverse is given by the number of sensors N ≤ model order n.

[0032] Let the time length be L, and the time coefficient matrix A = We can obtain: .

[0033] in, , where L represents the data used for offline training, and L is the time length.

[0034] , representing the offline training data after homogenization of boundary conditions.

[0035] After obtaining the spatial matrix and the time coefficient matrix, in this embodiment of the application, the FastICA algorithm is used to calculate the first control limit and the second control limit. Specifically, as follows: Figure 12 , Figure 12 This is the fifth schematic diagram of the large-size lithium battery thermal anomaly detection method in this invention; step 220 includes: Step 221: Calculate independent sources using the FastICA algorithm. : (15) in, , , , Representing the time coefficient matrix respectively Orthogonal mixing matrix, whitening matrix, unmixing matrix, columns; Step 222: Calculate the dominant independent sources using the FastICA algorithm. : (16) in, For aliasing matrix The row with the largest sum of squares; Step 223: Utilize the dominant independent source aliasing matrix The column with the largest sum of squares Whitening matrix , Calculate the first detection index respectively Second detection indicator .

[0036] In step 223, it is necessary to test the first indicator. Second detection indicator To define it, specifically, step 223 includes: Step 2231: The first detection indicator Defined as: (17) The second detection indicator Defined as: (18) in, For space function matrix The column; in, (19).

[0037] During the offline phase, the first detection index is calculated. Second detection indicator Then, the KDE algorithm is used to calculate the first control limit respectively. Second control limit Specifically, such as Figure 13 , Figure 13 This is the sixth schematic diagram of the large-size lithium battery thermal anomaly detection method in this invention; step 300 includes: Step 310: Obtain the kernel density function of the KDE algorithm: (20) in, For sampled data, and These are the number of samples and the bandwidth, respectively. It is a Gaussian kernel function. .

[0038] Step 320: Separately, the first detection index... The second detection indicator The point occupying 99% of the area of ​​the kernel density function is used as the first control limit. Second control limit .

[0039] During the online phase, the first control limit is obtained after training in the offline phase. Second control limit Then, the first detection index value of each sample at time t is detected online. Second indicator value Specifically, such as Figure 14 , Figure 14 This is the seventh schematic diagram of the large-size lithium battery thermal anomaly detection method in this invention; step 500 includes: step 510, respectively setting the first detection index value... Second detection index value Each with the corresponding first control limit Second control limit Comparison; Step 520: If the first detection index value Greater than the first control limit Or the second detection index value Greater than the second control limit If so, the lithium battery is determined to be faulty. The determination criteria are as follows: (twenty one) In one detailed embodiment, this application constructs a platform including a battery testing cabinet, a constant temperature chamber, heating elements, an external power supply, and a battery management system to collect battery temperature data and simulate localized thermal faults on the battery surface. The experimental subject is a lithium iron phosphate / graphite soft-pack lithium-ion battery with a length, width, and thickness of 0.20 meters, 0.15 meters, and 0.007 meters (thickness is negligible), and four thermocouples are located on the surface. Figure 1 As shown, Figure 1 This is a schematic diagram showing the distribution of four thermocouples and faults in an embodiment of the large-size lithium battery thermal anomaly detection method of the present invention. Figure 2 The input current under fault-free conditions provided in the embodiment of the large-size lithium battery thermal anomaly detection method of the present invention is... , Figure 3 The output voltage under fault-free conditions provided in the embodiment of the large-size lithium battery thermal anomaly detection method of the present invention is as follows: .

[0040] The input current I(t) and output voltage V(t) under fault-free conditions are shown in Table 1. Fault settings are shown in Table 2, and fault detection results are shown in Table 2. Figure 4 , Figure 5 , Figure 6 , Figure 7 , Figure 4 This is a first detection index diagram for fault 1 provided in an embodiment of the large-size lithium battery thermal anomaly detection method of the present invention. Figure 5 This is a second detection index diagram for fault 1 provided in an embodiment of the large-size lithium battery thermal anomaly detection method of the present invention. Figure 6 The first detection index diagram of fault 3 is provided in the embodiment of the thermal anomaly detection method for large-size lithium batteries of the present invention; Figure 7 This is a second detection index diagram for fault 3 provided in an embodiment of the thermal anomaly detection method for large-size lithium batteries of the present invention. The detection time t is the critical moment.

[0041] Table 1 Fault Setting Table Table 2 Fault Detection Results Compared with other fault detection methods, based on the global distributed temperature characteristics of large-size lithium batteries, the FastICA and KDE algorithms are used to calculate control limits. With limited temperature data, thermal anomaly fault detection of large-size lithium-ion batteries is achieved without a model structure.

[0042] II. Implementation Methods of Readable Storage Media A storage medium storing a computer program, which, when loaded and executed by a processor, implements the anomaly detection method as described in the first aspect.

[0043] The large-size lithium battery thermal anomaly detection method of the present invention involves setting a limited number of temperature sensors at corresponding locations on the lithium battery, constructing a global spatial function matrix based on spatial location data and temperature data using spectral methods, thereby obtaining a time coefficient matrix, calculating a first detection index and a second detection index using the FastICA algorithm based on the time coefficient matrix, and obtaining a first control limit and a second control limit using the KDE algorithm. By comparing the online real-time acquired values ​​of the first and second detection indexes with the corresponding control limits, the thermal anomaly situation of the large-size lithium battery is obtained, thus realizing the detection of local thermal faults in large-size batteries.

[0044] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for detecting thermal anomalies in large-size lithium batteries, characterized in that, include: Step 100: Obtain temperature data of large-size lithium batteries and convert it into a temperature data matrix; Step 200: Using the temperature data matrix and the FastICA algorithm, calculate the first detection index and the second detection index; Step 300: During the training phase, the points where the first detection index and the second detection index occupy 99% of the area of ​​the kernel density function of the KDE algorithm are used as the first control limit and the second control limit. Step 400: During the detection phase, the first detection index value and the second detection index value are detected online at time t. Step 500: Compare the first detection index value and the second detection index value with the corresponding first control limit and second control limit respectively to determine the fault. Step 200 includes: Constructing auxiliary functions The battery thermodynamic equation of the lithium battery is transformed into: The boundary conditions are: ; in, Indicates spatial dimension; The length and width of the battery are represented by t; t represents the time variable. Indicates density; Indicates specific heat capacity; Indicates the heat transfer coefficient; , in Indicates the thermal convection coefficient; and These are the internal heat generation of the battery and the lateral heat flux, respectively. This is an unknown thermal fault or anomaly. and Representing the spacetime temperature and ambient temperature respectively; V is , representing the temperature variable after homogenization of the boundary conditions; The two-dimensional spatiotemporal battery thermodynamic equations are further decomposed using spectral methods. Homogeneity processing is performed to obtain the space function matrix. ;in, Indicates the time coefficient. Let n denote the basis functions in the two-dimensional space, n represent the model order, and j be a natural number; By homogenizing the temperature data matrix, its homogenized matrix can be obtained. ; Using the space function matrix and the homogenized temperature data matrix Obtain the time coefficient matrix ;in, It is the space function matrix The false reversal; Calculate independent sources using the FastICA algorithm , ;in, , , , Representing the time coefficient matrix respectively The orthogonal mixing matrix, whitening matrix, unmixing matrix, and column vectors; Calculate the dominant independent source using the FastICA algorithm , ;in, For the unmixing matrix A row vector with the maximum sum of squares; Calculate the first detection index , ; Calculate the second detection index , ;in, The homogenized temperature data matrix column vectors, .

2. The method for detecting thermal anomalies in large-size lithium batteries according to claim 1, characterized in that, Step 100 includes: Step 110: Convert the temperature data into a preset temperature data matrix, wherein the temperature data matrix is: （1） in, This represents a matrix of temperature data output from N sensors. This represents the spatial location of N sensors.

3. The method for detecting thermal anomalies in large-size lithium batteries according to claim 2, characterized in that, The battery thermodynamic equation is as follows: （2） The boundary conditions for equation (2) are: （3） The initial conditions for the battery thermodynamic equations are: （4） in, This indicates the initial temperature.

4. The method for detecting thermal anomalies in large-size lithium batteries according to claim 3, characterized in that, The two-dimensional space basis functions It can be obtained through the following formula: （9） （10） （11） in, Let be the spatial basis functions in the x-direction. Let be the spatial basis functions in the y-direction. The constant coefficient, For eigenvalues; The homogenization matrix of the temperature data matrix It can be obtained through the following formula: （12） Step 216: Homogenize equation (8) to obtain the space function matrix. : （13）， in, For the order of the reduced-order model, This represents the number of sensors.

5. The method for detecting thermal anomalies in large-size lithium batteries according to claim 4, characterized in that, The kernel density function of the KDE algorithm is: （20） in, For sampled data, and These are the number of samples and the bandwidth, respectively. It is a Gaussian kernel function. .

6. The method for detecting thermal anomalies in large-size lithium batteries according to claim 5, characterized in that, Step 500 includes: Step 510: Compare the first detection index value and the second detection index value with the corresponding first control limit and second control limit, respectively; Step 520: If the value of the first detection index is greater than the first control limit or the value of the second detection index is greater than the second control limit, then the lithium battery is determined to have a thermal fault.

7. A storage medium storing a computer program, wherein when the computer program is loaded and executed by a processor, it implements the large-size lithium battery thermal anomaly detection method as described in any one of claims 1-6.

Images