A lithium battery thermal power extraction method and device combining a POD and a GRNN, and a medium

Abstract

The application discloses a lithium battery thermal power extraction method and device combining POD and GRNN, and a medium, relates to the technical field of lithium batteries, and mainly comprises the following steps: acquiring a temperature-time change test curve of a lithium battery during discharging from a rated maximum voltage to a rated minimum voltage at each discharge rate, constructing a corresponding database, learning and training the relationship between the heat power and the modal coefficient, acquiring the corresponding modal coefficient, and acquiring a temperature-time change target curve based on the modal coefficient; according to the Pearson correlation coefficient between each target curve and the test curve, matching the heat power with the highest correlation for each time period corresponding to the discharge rate; learning and training the relationship between the modal and the system parameters; and simulating the surface temperature change of the lithium battery at the target discharge rate based on the trained network and the thermal model. The application does not require complex model construction or long-time testing, and improves the efficiency of heat power estimation.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of lithium batteries, in particular to a lithium battery thermal power extraction method combining POD and GRNN, a lithium battery thermal power extraction device and a medium. BACKGROUND

[0002] With the large-scale grid connection of renewable energy and the popularity of interactive devices such as electric vehicles and distributed power sources, lithium ion batteries are widely used in the fields of transportation power, power storage, mobile communication, new energy storage power, aerospace and military industry. Lithium batteries have become the mainstream energy storage method in the market due to their good cycle characteristics, fast response speed and high system comprehensive efficiency.

[0003] The performance of a single lithium battery directly affects the overall performance of an integrated battery module, and temperature is one of the key factors affecting the safety, life and performance of lithium batteries. When evaluating the temperature condition of a battery, the heat generation power is an important parameter commonly used to analyze the temperature distribution of the battery in combination with numerical simulation.

[0004] The existing method is to construct an equivalent circuit model and obtain model parameters by using HPPC (Hybrid Pulse Power Characterization) testing, and then combine a heat generation model to calculate the change of heat generation power with time and working conditions. Although this method is common, it has problems such as complex model, dependence on professional equipment, low estimation accuracy, long construction period, etc. Using an electrochemical model combined with numerical simulation to calculate the heat generation power and temperature change, due to the extremely complex internal electrochemical reaction of lithium batteries, there is no electrochemical model that can fully cover the internal reaction of the battery under all working conditions. SUMMARY

[0005] In order to more effectively and quickly estimate the thermal power of lithium batteries, especially to accurately and quickly evaluate the thermal power under multiple discharge rates, the application provides a lithium battery thermal power extraction method combining POD and GRNN, which comprises the following steps:

[0006] S1: obtaining a temperature-time change test curve of a lithium battery under each discharge rate during discharging from a rated maximum voltage to a rated minimum voltage under a multi-temperature measurement point mean collection;

[0007] S2: constructing a database of battery surface temperature change with time under different heat generation powers by numerical simulation according to the test curve and in combination with lithium battery physical parameters;

[0008] S3: obtaining an orthogonal basis and a modal coefficient by performing intrinsic orthogonal decomposition on the database in combination with system parameters, and learning and training the relationship between the heat generation power and the modal coefficient by a generalized regression neural network;

[0009] S4: The generalized regression neural network trained in the S3 step acquires the corresponding modal coefficients according to the random heating power, and acquires the temperature-time change target curve based on the modal coefficients;

[0010] S5: According to the Pearson correlation coefficient between each target curve and test curve, the highest correlation heating power is matched for each time period corresponding to the discharge rate;

[0011] S6: According to the matrix composed of the heating power corresponding to each time period of each discharge rate, after eigenvalue orthogonal decomposition and modal extraction, the relationship between the modal and the system parameters is learned and trained by the generalized regression neural network;

[0012] S7: Based on the transient three-dimensional heat conduction differential equation of the uniform distribution of internal heat source, the lithium battery thermal model is constructed, and the generalized regression neural network trained in the S6 step is substituted to simulate the surface temperature change of the lithium battery at the target discharge rate.

[0013] Further, in the S1 step, the multiple temperature measuring points are distributed in the vicinity of the lithium battery tab and the middle region of the lithium battery.

[0014] In the S3 step, the eigenvalue orthogonal decomposition specifically includes the following steps:

[0015] S31: According to the temperature-time change curve, a sample matrix is constructed in combination with system parameters;

[0016] S32: The sample matrix is normalized and the correlation matrix is calculated to obtain the corresponding eigenvalue matrix and eigenvector matrix;

[0017] S33: The eigenvalue matrix and the eigenvector matrix are used to obtain the orthogonal basis of each order and the corresponding modal coefficient.

[0018] Further, in the S31 step, the system parameters include the discharge rate and the environmental variable parameter, and the sample matrix is divided by rows according to different system parameter combinations and divided by columns according to different time points.

[0019] Further, in the S32 step, the eigenvalue matrix and the eigenvector matrix are obtained by the following formula:

[0020]

[0021] RA=λA

[0022] In the formula, D is the sample matrix, N is the total number of time changes, T is the matrix transpose, is the normalization of the sample matrix D, R is the correlation matrix, A is the eigenvector matrix, and λ is the eigenvalue matrix.

[0023] Further, in the S33 step, the orthogonal basis and the modal coefficient are obtained by the following formula:

[0024]

[0025] wherein, is the normalization of the sample matrix D, λ j is the jth column of the eigenvalue matrix, A j is the jth column of the eigenvector matrix, Φ j is the orthogonal basis of the jth order; SP i is the combination of the ith system parameter, t q is the qth time point, T is the matrix transpose, a j (SP i , t q ) is the modal coefficient of the jth order at the qth time point under the combination of the ith system parameter.

[0026] Further, in the S4 step, the temperature-time change target curve is obtained by the following formula:

[0027]

[0028] wherein, is the temperature field data at the qth time point under the combination of the ith system parameter.

[0029] The application also includes a computer readable storage medium having stored thereon a computer program, which, when executed by a processor, implements the steps of the lithium battery thermal power extraction method combining POD and GRNN.

[0030] Also included is a data processing device, comprising:

[0031] a memory having stored thereon a computer program;

[0032] a processor configured to execute the computer program in the memory to implement the steps of the lithium battery thermal power extraction method combining POD and GRNN.

[0033] Compared with the prior art, the application has at least the following beneficial effects:

[0034] (1) The lithium battery thermal power extraction method, device and medium combining POD and GRNN can quickly and accurately obtain the curve of the thermal power of a lithium battery changing with time under different discharge rates, only using a battery simulation load and a temperature recording device, compared with the traditional method, which does not require complex model construction or long-term testing, greatly improving the efficiency of thermal power estimation.

[0035] (2) Through the construction of temperature database and the use of POD method for dimension reduction, combined with the learning of the relationship between working conditions and modal coefficients by GRNN neural network, the thermal power can be more accurately estimated, which ensures that in the new working condition, the thermal power value close to the true value can be quickly output, and the estimation accuracy is improved;

[0036] (3) Through the pre-established POD-GRNN model, the complex calculation from the beginning each time in the new working condition is avoided, the process of thermal power estimation is simplified, the dependence on professional equipment is reduced, and the method is more convenient and practical in industrial application. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 A flow chart of a lithium battery thermal power extraction method combining POD and GRNN. DETAILED DESCRIPTION

[0038] The following is a specific embodiment of the present application and further describes the technical solutions of the present application in combination with the drawings, but the present application is not limited to these embodiments.

[0039] Example 1

[0040] In view of the shortcomings of the prior art, the present application proposes a lithium battery thermal power extraction method combining POD and GRNN, wherein the POD (proper orthogonal decomposition) method used in the present application is as follows:

[0041] (1) A matrix set D is constructed, which is composed of M system parameter combinations (system parameters include discharge rate, environmental variables, etc.) and N time instant temperature field data of the lithium battery energy storage system. Wherein, the system parameter is represented by SP (system parameter), SPi (i = 1, 2, …, M) represents the combination of the i-th system parameter, and the time variable is represented by t. In the i-th system parameter combination mode, the data list containing all temperature measurement point information at the q-th time is represented as D (SPi, tq), and the sample matrix form of all temperature measurement point positions can be represented as formula (1) as follows:

[0042] D = [D (SP1,t) ,D (SP2,t) ,D (SP3,t) ,…,D (SPM,t) ] (1)

[0043] (2) Then, the sample matrix is normalized, the correlation matrix R of the normalized sample matrix is calculated, and the eigenvalues and eigenvectors are calculated, and the formulas (2) and (3) are as follows:

[0044]

[0045] RA = λA (3)

[0046] Where D is the sample matrix, N is the total number of time changes, T is the matrix transpose, is the normalized sample matrix D, R is the correlation matrix, A is the eigenvector matrix, and λ is the eigenvalue matrix.

[0047] (3) According to the obtained eigenvector matrix and eigenvalue matrix, the orthogonal basis and modal coefficients of each order can be obtained by the following formulas (4) and (5), respectively:

[0048]

[0049] Where λ j is the jth column of the eigenvalue matrix, A j is the jth column of the eigenvector matrix, Φ j is the jth order orthogonal basis, a j (SP i ,t q ) is the j-th modal coefficient at the q-th time point under the i-th combination of system parameters.

[0050] It should be noted that the energy captured by the first n POD modes accounts for the energy of all modes as shown in formula (6), where k is the highest-order mode:

[0051]

[0052] Where E is the energy occupied by the nth-order mode.

[0053] Then, based on the temperature field data of the lithium battery energy storage system at any time, it can be reconstructed based on the average value of the field data and a set of modal coefficients and bases. The formula is expressed as follows:

[0054]

[0055] Where, is the temperature field data at the qth time point under the i-th combination of system parameters.

[0056] On this basis, a generalized regression neural network (GRNN) is combined. The GRNN neural network is a four-layer feedforward propagation neural network with good nonlinear approximation capabilities and is a variant of the RBF neural network. This neural network also has no hyperparameters to be estimated, thus having strong engineering adaptability. In the present invention, the relationship between the corresponding operating conditions and the nth-order modal coefficients obtained by the POD method is learned through the GRNN neural network, thereby constructing the required GRNN neural network.

[0057] In the constructed GRNN neural network, a new set of modal coefficients can be quickly output by inputting new system parameters, and then the to-be-sought quantity can be quickly reconstructed from the modal coefficients and the original orthogonal basis. Figure 1 as shown, comprising the following steps:

[0058] S1: Obtain the temperature-time variation test curve of the lithium battery during discharging from the rated maximum voltage to the rated minimum voltage at each discharge rate under the condition of collecting the mean value of multiple temperature measuring points;

[0059] S2: According to the test curve and combined with the physical parameters of the lithium battery, a database of the change of the battery surface temperature with time under different heat generation powers is constructed by numerical simulation;

[0060] S3: Obtain the orthogonal basis and modal coefficients by performing intrinsic orthogonal decomposition on the database combined with system parameters, and learn and train the relationship between heat generation power and modal coefficients through generalized regression neural network;

[0061] S4: Obtain the corresponding modal coefficients according to the random heat generation power through the generalized regression neural network trained in step S3, and obtain the temperature-time variation target curve based on the modal coefficients;

[0062] S5: According to the Pearson correlation coefficient between each target curve and the test curve, match the heat generation power with the highest correlation for each time period corresponding to the discharge rate;

[0063] S6: According to the matrix composed of the heat generation power corresponding to each discharge rate and each time period, learn and train the relationship between the modal and the system parameters through the generalized regression neural network after intrinsic orthogonal decomposition and modal extraction;

[0064] S7: Based on the transient three-dimensional heat conduction differential equation of the uniform distribution of internal heat source, construct the lithium battery thermal model, and substitute it into the generalized regression neural network trained in step S6 to simulate the surface temperature variation of the lithium battery under the target discharge rate.

[0065] First, we need to obtain the temperature variation curve of the lithium battery under multiple discharge rates. In order to ensure the accuracy and integrity of the data, the lithium battery is charged to the highest rated voltage by using the constant current and constant voltage charging method. After the battery surface temperature cools down to room temperature or no longer changes, multiple temperature measuring points are arranged at the lithium battery tab and the middle position of the lithium battery. Then, the temperature-time variation test curve of the lithium battery during discharging to the rated minimum voltage under each discharge rate (1C, 2C, 3C) is recorded in real time.

[0066] According to the POD-GRNN method explained above, we can divide the lithium battery discharge heat generation process into N time periods according to the actual measured lithium battery surface temperature-time variation test curve, and the value of N is selected by human, theoretically the larger the value of N, the more characteristics the obtained battery heat generation power-time variation curve contains. Here, we combine the physical parameters of the battery, and through numerical simulation (such as ICEPAK) under different heat generation power setting values, we can simulate the battery temperature rise in each time period, obtain the battery temperature rise curve in each time period, and compare and calculate the correlation coefficient R value with the test curve, thereby constructing a small temperature database H of the battery surface temperature-time variation under different heat generation powers. The row of the database H represents different heat generation powers, and the column represents the temperature value of the battery at different times.

[0067] Then the POD method is used to explain the small temperature database H, and the corresponding modal coefficients and orthogonal bases are obtained. According to the explanation of the POD method above, the specific steps are as follows:

[0068] S31: According to the temperature-time variation curve, combine the system parameters to construct the sample matrix;

[0069] S32: Normalize the sample matrix and calculate the correlation matrix to obtain the corresponding eigenvalue matrix and eigenvector matrix;

[0070] S33: Obtain the orthogonal basis of each order and the corresponding modal coefficient according to the eigenvalue matrix and the eigenvector matrix.

[0071] After obtaining the modal coefficient, the non-linear relationship between the heat generation power and the modal coefficient is learned through the GRNN neural network, and then the trained GRNN neural network is used to input new power values to quickly output new modal coefficients. At the same time, the orthogonal basis is multiplied by the modal coefficient to obtain the temperature-time variation curve under the condition of new heat generation power, and the correlation coefficient R value is calculated. Based on the correlation coefficient R, the heat generation power P1~P N with the highest matching degree in each time period is selected, and the heat generation power-time interval dt variation relationship {P1, P2, …, P N} of the lithium battery under different discharge rates is obtained, that is, the heat power relationship of the lithium battery at different times under different discharge rates.

[0072] We will obtain the thermal power value of N time stage under different discharge rate (for example, 1C, 2C, 3C three discharge rates) combined into an N row 3 column matrix A. In this, the POD method is used to reduce the order of matrix A and extract the mode, and then the corresponding relationship between the system parameters and the mode is obtained through the GRNN neural network, so as to build a lithium battery discharge heat power change proxy model under multi-rate conditions. Through the proxy model, the N thermal power values under the new input discharge rate can be quickly obtained.

[0073] At this time, we regard the lithium battery as a three-dimensional heat conduction problem with internal heat source and anisotropic thermal conductivity, and construct the lithium battery thermal model through the transient three-dimensional heat conduction differential equation, wherein the thermal conductivity, specific heat capacity and other parameters required for the construction of the lithium battery thermal model can be obtained by consulting or calculating the lithium battery factory parameters, and are known quantities, and the change relationship of the lithium battery thermal power with time is simulated by inputting the previously trained proxy model, so as to obtain the change of the lithium battery surface temperature under the corresponding discharge rate.

[0074] The application also includes a computer readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the lithium battery thermal power extraction method combining POD and GRNN.

[0075] It also includes a data processing device, which includes:

[0076] A memory having a computer program stored thereon;

[0077] A processor for executing the computer program in the memory to implement the steps of the lithium battery thermal power extraction method combining POD and GRNN.

[0078] In summary, the lithium battery thermal power extraction method, device and medium combining POD and GRNN can quickly and accurately obtain the thermal power curve of the lithium battery under different discharge rates, compared with the traditional method, which does not need complex model construction or long time test, greatly improving the efficiency of thermal power estimation.

[0079] By constructing the temperature database and using the POD method for dimension reduction, and combining the GRNN neural network to learn the relationship between the working condition and the modal coefficient, the thermal power can be more accurately estimated, which ensures that the thermal power value close to the real value can be quickly output under the new working condition, and the estimation accuracy is improved.

[0080] Through the pre-established POD-GRNN model, the complex calculation from the beginning is avoided every time in the new working condition, the process of the thermal power estimation is simplified, the dependence on professional equipment is reduced, so that the method is more convenient and practical in industrial application.

[0081] It should be noted that all the directionality indications in the embodiments of the present application, such as up, down, left, right, front, back, etc., are only used to explain the relative position relationship, movement condition, etc. between components in a certain specific posture (as shown in the drawings), and if the specific posture changes, the directionality indications will also change accordingly.

[0082] In addition, the descriptions such as "first", "second", "one" and the like in the present application are only for the purpose of description, and cannot be understood as indicating or implying the relative importance of the indicated technical features or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "a plurality of" is at least two, for example, two, three, etc., unless otherwise specifically limited.

[0083] In the present application, unless otherwise specifically defined and limited, the terms "connection", "fixation" and the like should be understood in a broad sense, for example, "fixation" can be fixed connection, or detachable connection, or integral; can be mechanical connection, or electrical connection; can be directly connected, or indirectly connected through an intermediate medium; can be the internal communication of two elements or the interaction relationship between two elements, unless otherwise specifically limited. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0084] In addition, the technical solutions of each embodiment of the present application can be combined with each other, but it must be based on the fact that a person skilled in the art can realize it, when the combination of technical solutions appears contradictory or unachievable, it should be considered that the combination of technical solutions does not exist, nor is it within the scope of protection required by the present application.

Claims

1. A method for lithium battery thermal power extraction combining POD and GRNN, characterized in that, The method comprises the steps of: S1: obtaining the temperature-time change test curve of the lithium battery during discharging from the rated maximum voltage to the rated minimum voltage at each discharge rate under the average collection of multiple temperature measuring points; S2: constructing a database of the surface temperature change of the battery over time under different heat generation powers through numerical simulation according to the test curve and in combination with the physical parameters of the lithium battery; S3: obtaining the orthogonal basis and modal coefficients by performing intrinsic orthogonal decomposition on the database in combination with the system parameters, and learning and training the relationship between the heat generation power and the modal coefficients through a generalized regression neural network; S4: obtaining the corresponding modal coefficients according to the random heat generation power through the generalized regression neural network trained in the S3 step, and obtaining the temperature-time change target curve based on the modal coefficients; S5: matching the heat generation power with the highest correlation for each time period of the corresponding discharge rate according to the Pearson correlation coefficient between each target curve and the test curve; S6: learning and training the relationship between the modal and the system parameters through a generalized regression neural network after the intrinsic orthogonal decomposition and modal extraction according to the matrix composed of the heat generation powers corresponding to each time period of each discharge rate; S7: constructing a lithium battery thermal model based on the transient three-dimensional heat conduction differential equation of the uniform distribution of internal heat sources, and substituting it into the generalized regression neural network trained in the S6 step to simulate the surface temperature change of the lithium battery at the target discharge rate.

2. The lithium battery thermal power extraction method combining POD and GRNN of claim 1, wherein, In the S1 step, the multiple temperature measuring points are distributed around the lithium battery tab and in the middle region of the lithium battery.

3. The method of claim 1, wherein the POD and GRNN are combined. In the S3 step, the intrinsic orthogonal decomposition specifically comprises the following steps: S31: constructing a sample matrix according to the temperature-time change curve in combination with the system parameters; S32: obtaining the corresponding eigenvalue matrix and eigenvector matrix by performing normalization operation on the sample matrix and calculating the correlation matrix; S33: obtaining the orthogonal basis of each order and the corresponding modal coefficients according to the eigenvalue matrix and the eigenvector matrix.

4. The lithium battery thermal power extraction method combining POD and GRNN of claim 3, wherein, In the S31 step, the system parameters include the discharge rate and the environmental variable parameters, and the sample matrix is divided by rows according to different combinations of system parameters and by columns according to different time points.

5. The lithium battery thermal power extraction method combining POD and GRNN of claim 3, wherein, In the S32 step, the eigenvalue matrix and the eigenvector matrix are obtained through the following formula: RA=λA where D is a sample matrix, N is the total number of time-varying time points, T is a matrix transpose, is a normalization of the sample matrix D, R is a correlation matrix, A is an eigenvector matrix, and λ is an eigenvalue matrix.

6. The lithium battery thermal power extraction method combining POD and GRNN of claim 4, wherein, In the S33 step, the orthogonal basis and the modal coefficients are obtained through the following formula: Where, is the normalization of the sample matrix D, λ j is the jth column of the eigenvalue matrix, A j is the jth column of the eigenvector matrix, Φ j is the jth order orthogonal basis; SP i is the combination of the i-th system parameters, t q is the qth time point, T is the matrix transpose, a j (SP i ,t q ) is the j-th modal coefficient at the q-th time point under the i-th combination of system parameters.

7. The lithium battery thermal power extraction method combining POD and GRNN of claim 6, wherein, In the S4 step, the temperature-time change target curve is obtained through the following formula: In the formula, is the temperature field data at the qth time point under the ith combination of system parameters.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to realize the steps of the extraction method in any one of claims 1 to 7.

9. An apparatus for processing data, the apparatus comprising: Comprise: a memory having a computer program stored thereon; a processor configured to execute the computer program in the memory to realize the steps of the extraction method in any one of claims 1 to 7.

Images