Training method and system of prediction model, prediction method and system, equipment and storage medium

Abstract

The invention provides a prediction model training method and system, a prediction method and system, equipment and a storage medium. The training method of the prediction model comprises the following steps that S1, a data set is obtained, and the data set comprises capacity retention rates of a first lithium ion battery under different charge and discharge rates; and S2, training the data set by adopting machine learning to obtain the prediction model. The method can be used for predicting the fast charging performance of the lithium ion battery; moreover, on the basis of a machine learning algorithm and interface impedance parameters, the scheme of the invention can make parameters in the model more stable and objective and less overfitting problems occur, and can rapidly and accurately predict the fast charging performance of the lithium ion battery and make the result visualized.

Description

Technical Field

[0001] The present invention relates to a prediction model training method, a prediction method and system, a device and a storage medium. Background Art

[0002] In the field of new energy, various types of batteries, including lithium batteries, are often used in new energy vehicles and various energy storage systems; however, with the growing market demand, lithium-ion batteries also need to be improved to varying degrees.

[0003] In order to improve the fast-charging performance of lithium-ion batteries, researchers generally use the traditional "experience-trial and error" strategy, which is not only time-consuming and labor-intensive but also highly dependent on the personal experience of researchers, limiting the research and development efficiency of lithium-ion batteries with excellent fast-charging performance.

[0004] Therefore, there is an urgent need to provide a prediction method for the charge and discharge performance of lithium-ion batteries, which can quickly predict the fast charging performance of lithium-ion batteries, provide guidance for the research and development of lithium-ion batteries and improve research and development efficiency. Summary of the invention

[0005] In order to overcome the defect of the prior art that it is difficult to predict the fast charging performance of lithium-ion batteries, the present invention provides a prediction model training method, prediction method and system, device, and storage medium. The present invention can be used to predict the fast charging performance of lithium-ion batteries; and based on the machine learning algorithm and interface impedance parameters, the solution of the present application can make the parameters in the model more stable and objective, with less overfitting problems, and can quickly and accurately predict the fast charging performance of lithium-ion batteries, and visualize the results.

[0006] The present invention solves the above technical problems through the following technical solutions.

[0007] The present invention provides a prediction model training method, which comprises the following steps:

[0008] S1. Acquire a data set, wherein the data set includes a capacity retention rate of a first lithium-ion battery at different charge and discharge rates;

[0009] S2. Using machine learning to train the data set to obtain the prediction model.

[0010] In some embodiments, the bias term of the prediction model is an impedance parameter.

[0011] In a specific embodiment, the types of impedance parameters include contact impedance, SEI impedance, charge transfer impedance and diffusion impedance.

[0012] In a specific implementation, the impedance parameter is obtained by integrating the Gaussian function fitting; it represents the contribution of different polarization processes to the impedance, and is used as a bias term of the deep learning algorithm to further improve the prediction accuracy and reliability of the algorithm.

[0013] In a specific implementation, the method for obtaining the impedance parameter includes the following steps:

[0014] ss1. Testing the EIS test data of the first lithium-ion battery at different charge and discharge rates;

[0015] ss2. Analyze the EIS test data using DRT to obtain different types of impedance parameters as bias terms; adjust the weight of the bias term in the prediction model;

[0016] ss3. Screen impedance parameters and combine them.

[0017] Among them, the DRT analysis (distribution relaxation time analysis) is an analysis method based on electrochemical impedance spectroscopy (EIS) data, which is mainly used to study the complex kinetic processes in batteries.

[0018] In a specific embodiment, the capacity retention rate of the first lithium-ion battery at different charge and discharge rates is merged with the EIS test data analyzed by DRT; the fusion method is to use Python code for fusion.

[0019] Among them, there is a correlation between the capacity retention rate under the charge and discharge rate and the EIS test data of the DRT analysis. In the fusion layer, the two are fused according to the correlation to obtain a "joint probability density". According to the impact on the loss function, multiple iterations are performed to approximate the "solution" of the model. It can also be considered that they influence each other and adjust the weights.

[0020] Among them, the Nyquist diagram obtained from the EIS test data can be converted into a DRT diagram through the python algorithm library.

[0021] Among them, DRT analysis of EIS test data can optimize the features used in the prediction model training process, thereby improving the prediction performance of the prediction model.

[0022] Among them, for the graph obtained from the EIS test data at different charge and discharge rates analyzed by the DRT, since the physical meaning of each peak area is the impedance of the polarization process corresponding to the horizontal axis time constant, the peak area of ​​different peaks is calculated to obtain the impedance characteristic value.

[0023] Among them, in the graph obtained by the EIS test data at different charge and discharge rates analyzed by the DRT, different positions can be considered as different polarization processes, and the x-axis represents the relaxation time of the polarization process; the peak areas of different peaks are the impedance parameters of different polarization processes.

[0024] In a specific implementation, in step ss1, the EIS test data is cleaned again through the pandas library in python; the cleaning method includes noise reduction and outlier removal.

[0025] The bias term is the initial condition or adjustment term of the model.

[0026] The updating of the bias term and the weight is automatically completed by the training data, without human intervention. During the training process, only the training data plays a role, and the weights and biases are continuously adjusted through an iterative process to minimize the loss function.

[0027] In a specific implementation, in step ss2, by obtaining a bias term and adjusting its weight in the prediction model, the accuracy and robustness of the prediction model for battery performance prediction can be improved.

[0028] In a specific embodiment, in step ss3, the screening method includes using a feature selection method in machine learning to screen and obtain an optimal impedance parameter combination for predicting battery performance; preferably, the feature selection method in machine learning includes principal component analysis (PCA); thereby reducing model complexity and the number of features, avoiding overfitting, and improving computational efficiency.

[0029] Among them, when training large-scale models, the stochastic gradient descent method is adopted to balance the model accuracy in step S2 and the computational efficiency in step S3; the stochastic gradient descent method is a variant of the gradient descent method, which reduces the computational cost of each update by using only a small batch of samples each time to calculate the gradient, and is believed to be able to balance computational efficiency and model accuracy.

[0030] Among them, gradient descent is an iterative optimization algorithm used to find the minimum value of a function. In machine learning, this function is usually a loss function, which represents the difference between the model's predicted value and the true value. The basic steps of gradient descent are as follows:

[0031] 1) Initialization parameters: Randomly initialize model parameters;

[0032] 2) Calculate the gradient: Calculate the gradient of the loss function with respect to all parameters, that is, calculate the partial derivative of J(θ) with respect to the parameter θ in all directions of the vector;

[0033] 3) Update parameters: Update the parameters based on the gradient and the learning rate (a hyperparameter that controls the step size). The update rule is usually:

[0034]

[0035] Among them, θ is the parameter, α is the learning rate, is the loss function, J(θ) is the gradient with respect to the parameter θ;

[0036] 4) Repeat: Repeat steps 2 and 3 until the stopping condition is met, such as the gradient is small enough (reaching a local minimum) or the maximum number of iterations is reached.

[0037] In some embodiments, the method for obtaining the data set includes the following steps:

[0038] a1. Stack the positive electrode sheet, the first diaphragm, the negative electrode sheet, and the second diaphragm to obtain a bare cell, and insert the first reference electrode into the bare cell;

[0039] a2. Carry out packaging, liquid injection, formation, and capacity division in sequence to obtain a three-electrode lithium-ion soft-pack battery;

[0040] a3. Testing the capacity retention rate of the three-electrode lithium-ion soft-pack battery at different charge and discharge rates, and capturing EIS test data in real time through the dynamic link library of the test equipment to form the data set.

[0041] In a specific embodiment, in step a1, the active material of the positive electrode plate includes lithium iron phosphate.

[0042] In a specific embodiment, in step a1, the coating surface density of the positive electrode sheet is 12 mg / cm 2 .

[0043] In a specific embodiment, in step a1, the active material of the negative electrode plate includes graphite.

[0044] In a specific embodiment, in step a1, the coating surface density of the negative electrode sheet is 4 mg / cm 2 .

[0045] In a specific implementation, in step a2, the packaging material includes an aluminum-plastic film.

[0046] In a specific embodiment, in step a2, the formation process is: 0.1C constant current charging to 50% SOC, set aside for 10 to 20 minutes, and then charged to full at 0.5C constant current, with a cut-off voltage of 3.95V and a temperature of 25°C.

[0047] In a specific implementation, in step a2, the capacity division process is: firstly fully charged at 0.33C, 3.95V with constant current and constant voltage, then discharged in stages at 1C, 0.33C, 0.2C, and 0.1C, and finally charged at 0.33C with constant current to 5% SOC.

[0048] Among them, in step a3, a data monitoring python script is used to test the dynamic The link library (DDL) captures the charge and discharge test data in real time and manages the database through the MySQL language.

[0049] In a specific implementation manner, in step a3, each rate charge and discharge is performed and left to stand for 2 hours.

[0050] In a specific embodiment, in step a3, the electrochemical work includes a working electrode (WE), a counter electrode (CE) and a second reference electrode (RE), the working electrode is connected to the negative electrode plate, the counter electrode is connected to the positive electrode plate, and the second reference electrode is connected to the first reference electrode.

[0051] Among them, the three-electrode lithium-ion soft-pack battery can decouple the negative electrode interface impedance, which is conducive to studying the interface impedance of the negative electrode; while the full battery (the working electrode of the electrochemical workstation is connected to the negative electrode of the battery, and the auxiliary electrode and the reference electrode are short-circuited and then connected to the positive electrode of the battery) is not conducive to studying the interface impedance of the negative electrode.

[0052] In some embodiments, the type of prediction model includes a TensorFlow model.

[0053] Among them, the TensorFlow model is an open source machine learning framework developed by the Google Brain team; the usage scenarios of the TensorFlow model include but are not limited to: (1) Deep learning: building and training deep neural networks, including convolutional neural networks (CNNs), recurrent neural networks (RNNs), long short-term memory networks (LSTMs), etc.; (2) Reinforcement learning: used to develop and train reinforcement learning models to solve complex decision-making problems; (3) Computer vision: used for tasks such as image recognition, image classification, and target detection; (4) Natural language processing: used for tasks such as text classification, sentiment analysis, and machine translation; (5) Speech recognition: used for speech-to-text conversion and speech recognition systems.

[0054] In some embodiments, the learning rate of the prediction model is 0.01-0.1, for example, 0.01.

[0055] In some embodiments, the batch size of the prediction model is 32-128, for example, 32, 64, 128, 256 or 512.

[0056] In some embodiments, the number of iterations of the prediction model is 300-500, for example, 400.

[0057] In the present invention, those skilled in the art should know that the charge and discharge rate refers to the conditions of a certain charge rate and a certain discharge rate.

[0058] In a certain embodiment, the conditions of different charge and discharge rates may be at least one of 0.2C charge rate and 0.2C discharge rate, 1C charge rate and 0.2C discharge rate, 2C charge rate and 0.2C discharge rate, 3C charge rate and 0.2C discharge rate, 4C charge rate and 0.2C discharge rate, and 5C charge rate and 0.2C discharge rate.

[0059] In a certain embodiment, the conditions of different charge and discharge rates may be at least one of 0.2C charge rate and 0.2C discharge rate, 0.33C charge rate and 0.2C discharge rate, 0.5C charge rate and 0.2C discharge rate, 1C charge rate and 0.2C discharge rate, 2C charge rate and 0.2C discharge rate, 3C charge rate and 0.2C discharge rate, 4C charge rate and 0.2C discharge rate, and 5C charge rate and 0.2C discharge rate.

[0060] In some embodiments, in the data set, the number of the data is at least 100.

[0061] In some embodiments, the machine learning includes training using a neural network; the neural network is trained on the training samples The mean square error MSE on is:

[0062] Where m is the number of samples; j=1 means the summation function starts from 1; is the predicted value; k is the actual value.

[0063] In the present invention, EIS test data and experimental data of capacity retention rate under new charge and discharge rates can be continuously collected, and the above prediction model can be iteratively updated.

[0064] In the present invention, the fast charging performance, health status, SOH, life span, etc. of the lithium-ion battery are predicted by the above-mentioned modeling method.

[0065] The present invention also provides a method for predicting the charging and discharging performance of a lithium-ion battery, which comprises the following steps:

[0066] The charge and discharge rate of the lithium-ion battery to be tested is input into the prediction model to obtain the capacity retention rate under the charge and discharge rate; the prediction model is obtained by the prediction model training method as described above.

[0067] In some embodiments, the first lithium-ion battery and the lithium-ion battery to be tested are both lithium-ion soft-pack batteries.

[0068] In some embodiments, the main material of the positive electrode of the first lithium-ion battery is lithium iron phosphate.

[0069] In some embodiments, the main material of the negative electrode of the first lithium-ion battery is graphite.

[0070] In some embodiments, the first lithium-ion battery is a three-electrode soft-pack battery containing a reference electrode, and the reference electrode is disposed in a bare cell of the three-electrode soft-pack battery.

[0071] In some embodiments, the ratio of the capacity of the lithium-ion battery to be tested to the capacity of the first lithium-ion battery is (0.8-1.2):1; preferably, the capacity of the first lithium-ion battery is 1Ah, and the capacity of the lithium-ion battery to be tested is 0.8-1.2Ah.

[0072] In some embodiments, the positive electrode main material and the negative electrode main material of the lithium-ion battery to be tested are of the same type as the positive electrode main material and the negative electrode main material of the first lithium-ion battery, respectively.

[0073] In the present invention, the results of the prediction method can be visualized through a python script.

[0074] The present invention also provides a prediction system for the charge and discharge performance of a lithium-ion battery, which includes a prediction module, wherein the prediction module is used to input the charge and discharge rate of the lithium-ion battery to be tested into a prediction model to obtain the capacity retention rate under the charge and discharge rate; the prediction model is obtained using the prediction model training method as described above.

[0075] The present invention also provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the training method of the prediction model as described above or the method for predicting the charging and discharging performance of a lithium-ion battery as described above is implemented.

[0076] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the training method of the prediction model as described above, or the prediction method of the charging and discharging performance of a lithium-ion battery as described above is implemented.

[0077] The present invention also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the training method of the prediction model as described above, or the prediction method of the charging and discharging performance of a lithium-ion battery as described above.

[0078] On the basis of being in accordance with the common sense in the art, the above-mentioned preferred conditions can be arbitrarily combined to obtain the preferred embodiments of the present invention.

[0079] The reagents and raw materials used in the present invention are commercially available.

[0080] The positive and progressive effects of the present invention are:

[0081] (1) The prediction method of the present invention includes real-time data capture, data processing, model training, and visual result output; through charge and discharge tests and EIS / DRT analysis, the interface impedance changes of the battery during the charge / discharge process at different rates are obtained, data is collected in real time, the prediction model is trained, trend prediction is performed, the performance of the battery is predicted quickly and more accurately, and the analysis results are visualized.

[0082] (2) The present invention predicts the high-rate charge and discharge performance of lithium-ion soft-pack batteries through EIS / DRT analysis and fast-charge test data, thereby screening experimental schemes and evaluating materials to reduce costs;

[0083] (3) Compared with the traditional ECM (equivalent circuit model) method based on impedance calculation to extract impedance parameters and train the model, the DRT analysis can obtain more stable and objective parameters and reduce the overfitting problem of the prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Figure 1 This is a schematic flow chart of a method for predicting charging and discharging performance of a lithium-ion battery according to Example 1 of the present application;

[0085] Figure 2 This is a comparison chart of the predicted value and the actual value of the capacity retention rate at different charge and discharge rates by the prediction model of Example 1 of the present application;

[0086] Figure 3 This is a comparison chart of the predicted value and actual value of the capacity retention rate at different charge and discharge rates by the prediction model of Example 2 of the present application. DETAILED DESCRIPTION

[0087] The present invention is further described below by way of examples, but the present invention is not limited to the scope of the examples. The experimental methods in the following examples without specifying specific conditions are carried out according to conventional methods and conditions, or selected according to the product specifications.

[0088] Example 1

[0089] This embodiment discloses a method for predicting the charging and discharging performance of a lithium-ion battery. Figure 1 : is a flow chart of a method for predicting the charge and discharge performance of a lithium-ion battery of this embodiment; the prediction method comprises the following steps:

[0090] S1. Obtaining the capacity retention rate of the first lithium-ion battery at different charge and discharge rates to form a data set;

[0091] Wherein, the first lithium-ion battery is a three-electrode lithium-ion soft-pack battery, and its preparation method is as follows:

[0092] (1) Pole coating: The active material LFP (lithium iron phosphate) is coated on aluminum foil to prepare the positive electrode, and the surface density of the positive electrode coating is controlled to be 12 mg / cm 2 ; The active material graphite A is coated on the copper foil to make Prepare negative electrode sheets and control the surface density of negative electrode coating The density is 4mg / cm 2 ; Among them, the thickness of aluminum foil is 15μm, copper The foil thickness is 8 μm;

[0093] (2) Liquid injection lamination: After injecting the electrolyte, a bare cell is obtained by laminating the positive and negative electrode sheets prepared in step (1) with a separator having a thickness of 8 μm, inserting a reference electrode into the bare cell, and then conventionally encapsulating (the encapsulating material is an aluminum plastic film), injecting (i.e. injecting the above-prepared electrolyte), forming, and dividing the capacity to obtain a three-electrode lithium-ion soft-pack battery; the formation process is: charging at a constant current of 0.1C to 50% SOC, leaving it for 10 to 20 minutes, and then charging at a constant current of 0.5C to full, with a cut-off voltage of 3.95V and a temperature of 25°C; the dividing capacity process is: first charging at a constant current and constant voltage of 0.33C and 3.95V to full, then discharging at 1C, 0.33C, 0.2C, and 0.1C in stages, and finally charging at a constant current of 0.33C to 5% SOC;

[0094] Among them, the test method of capacity retention rate under different charge and discharge rates is as follows: double charge test: the above-mentioned three-electrode soft-pack battery is sequentially tested at 0.2C, 0.2C, 0.2C, 1C, 2C, 3C, 4C, 5C, 7C, 10C charging rate, and the discharge rate is 0.2C. The condition of fast charging test, and then the ordinary soft-pack battery prepared according to the above materials and methods is sequentially tested at 0.2C, 0.2C, 0.2C, 1C, 2C, 3C, 4C, 5C charging rate, and discharge rate are all 0.2C. The WE (working electrode) of the electrochemical workstation is connected to the graphite negative electrode of the three-electrode soft-pack battery, the CE (counter electrode) is connected to the LFP positive electrode, and the RE is connected to the reference electrode, and an EIS (electrochemical impedance spectroscopy) test is performed;

[0095] The data set is obtained as follows: a special data monitoring Python script is written to capture the capacity retention rate of the three-electrode lithium-ion soft-pack battery at different charge and discharge rates in real time through the DDL (dynamic link library) of the test equipment to obtain a large amount of data; impedance test: each time the rate charge and discharge is performed, it is left to stand for 2 hours, and DRT (relaxation time distribution) analysis is performed to obtain a series of DRT spectra of the battery at different rates. The impedance parameters corresponding to different polarization processes are obtained by Gaussian function fitting and integration. The types of impedance parameters are contact impedance, SEI impedance, charge transfer impedance and diffusion impedance;

[0096] S2. The steps of using machine learning to train the model for the above data set are as follows: conduct a large number of the above tests, establish a data set, and perform denoising and normalization, manage the database through MySQL language, and divide the data set into training set, validation set and test set; the data of the three-electrode soft-pack battery is the training set, and the data of the ordinary battery is the validation set, with a ratio of 8:2; use TensorFlow to define the model, set the hyperparameters (learning rate is 0.01, batch size is 64, number of iterations is 400), and use the above impedance parameters as bias items; monitor indicators during the training process, such as loss value and accuracy, to evaluate the performance of the model and adjust the training strategy. The neural network is used in the training samples. Mean square error MSE on: Where m is the number of samples; j=1 means the summation function starts from 1; is the predicted value; k is the actual value.

[0097] Based on the minimization principle, the model is deduced and iterated, and the hyperparameters of the model are adjusted according to the evaluation results to find the optimal model.

[0098] Figure 2 This is a comparison chart of the predicted value and actual value of the capacity retention rate under different charge and discharge rates of the prediction model of this embodiment. As can be seen from the figure, the capacity retention rate of the three-electrode lithium-ion soft-pack battery at a charge rate of 0.2C, 0.2C, 0.2C, 1C, 2C, 3C, 4C, 5C, 7C, 10C and a discharge rate of 0.2C and the capacity retention rate of the ordinary battery at a charge rate of 0.2C, 0.2C, 0.2C, 1C, 2C, 3C, 4C, 5C and a discharge rate of 0.2C are used as data sets;

[0099] In addition, the above prediction model was used to test the capacity retention rate of the ordinary electrode under the conditions of a charge rate of 7C and a discharge rate of 0.2C, as well as the capacity retention rate under the conditions of a charge rate of 10C and a discharge rate of 0.2C. The capacity retention rate values ​​were very close to the capacity retention rates of the three electrodes under the charging and discharging conditions.

[0100] A total of 50 batches of capacity retention data at different charge and discharge rates were predicted, and the prediction results of the model were: mean square error (MSE) ≤ 1, and the generalization ability of the model was good. This result can be considered a good prediction.

[0101] S3. Result output: Import the new data set, calculate the prediction results based on the model, and display them in the form of a visual analysis report.

[0102] Example 2

[0103] This embodiment discloses a method for predicting the charging and discharging performance of a lithium-ion battery; the prediction method comprises the following steps:

[0104] S1. Obtaining the capacity retention rate of the first lithium-ion battery at different charge and discharge rates to form a data set;

[0105] Wherein, the first lithium-ion battery is a three-electrode lithium-ion soft-pack battery, and its preparation method is as follows:

[0106] (1) Pole coating: The active material LFP is coated on aluminum foil to prepare the positive electrode, and the surface density of the positive electrode coating is controlled to be 12 mg / cm 2 ; The active material graphite A was coated on the copper foil to prepare the negative electrode sheet, and the surface density of the negative electrode coating was controlled to be 4 mg / cm 2 ; Among them, the thickness of aluminum foil is 15μm, and the thickness of copper foil is 8μm;

[0107] (2) Liquid injection lamination: After injecting the electrolyte, a bare cell is obtained by laminating the positive and negative electrode sheets prepared in step (1) with a separator having a thickness of 8 μm, inserting a reference electrode into the bare cell, and then conventionally encapsulating (the encapsulating material is an aluminum plastic film), injecting (i.e. injecting the above-prepared electrolyte), forming, and dividing the capacity to obtain a three-electrode lithium-ion soft-pack battery; the formation process is: charging at a constant current of 0.1C to 50% SOC, leaving it for 10 to 20 minutes, and then charging at a constant current of 0.5C to full, with a cut-off voltage of 3.95V and a temperature of 25°C; the dividing capacity process is: first charging at a constant current and constant voltage of 0.33C and 3.95V to full, then discharging at 1C, 0.33C, 0.2C, and 0.1C in stages, and finally charging at a constant current of 0.33C to 5% SOC;

[0108] Among them, the test method of capacity retention rate under different charge and discharge rates is as follows: double charge test: the above-mentioned three-electrode soft-pack battery is sequentially subjected to fast charge test at 0.2C, 0.33C, 0.5C, 1C, 2C, 3C, 4C, 5C, 10C, 12C charging rate and discharge rate are all 0.2C, and then the ordinary soft-pack battery prepared according to the above materials and methods is sequentially subjected to fast charge test at 0.2C, 0.33C, 0.5C, 1C, 2C, 3C, 4C, 5C charging rate and discharge rate are all 0.2C, the WE (working electrode) of the electrochemical workstation is connected to the graphite negative electrode of the three-electrode soft-pack battery, the CE (counter electrode) is connected to the LFP positive electrode, and the RE is connected to the reference electrode, and an EIS (electrochemical impedance spectroscopy) test is performed;

[0109] The data set was obtained as follows: a special data monitoring Python script was written to capture the capacity retention rate of the three-electrode lithium-ion soft-pack battery at different charge and discharge rates in real time through the DDL (dynamic link library) of the test equipment to obtain a large amount of data; impedance test: each time the charge and discharge rate was performed, the battery was left to stand for 2 hours, and DRT analysis was performed to obtain a series of DRT spectra of the battery at different rates. The impedance corresponding to different polarization processes was obtained by fitting and integrating the Gaussian function. The types of impedance parameters are contact impedance, SEI impedance, charge transfer impedance and diffusion impedance;

[0110] S2. The steps of using machine learning to train the model of the above data set are as follows: conduct a large number of the above tests, establish a data set, and perform denoising and normalization, manage the database through MySQL language, and divide the data set into a training set, a validation set, and a test set; the data of the three-electrode soft-pack battery is the training set, and the data of the ordinary battery is the validation set, and the ratio is 8:2;

[0111] Use TensorFlow to define the model, set hyperparameters (learning rate of 0.01, batch size of 64, and number of iterations of 400), and use the above impedance parameter as the bias term; monitor indicators during the training process, such as loss value and accuracy, to evaluate the performance of the model and adjust the training strategy. The mean square error (MSE) and mean absolute error (MAE) are usually used to measure the accuracy of the model. MSE measures the average of the squares of the differences between the predicted values ​​and the actual values, while MAE measures the average of the absolute values ​​of the differences.

[0112] The mean square error MSE is:

[0113] The mean absolute error MAE is:

[0114] Where m is the number of samples; j=1 means the summation function starts from 1; is the predicted value;k is the actual value. Based on the minimization principle, multiple iterations are performed to adjust the model's hyperparameters according to the evaluation results to find the optimal model.

[0115] Figure 3 This is a comparison chart of the predicted value and actual value of the capacity retention rate under different charge and discharge rates of the prediction model of this embodiment. As can be seen from the figure, the capacity retention rate of the three-electrode lithium-ion soft-pack battery at a charge rate of 0.2C, 0.33C, 0.5C, 1C, 2C, 3C, 4C, 5C, 10C, 12C and a discharge rate of 0.2C and the capacity retention rate of the ordinary battery at a charge rate of 0.2C, 0.33C, 0.5C, 1C, 2C, 3C, 4C, 5C and a discharge rate of 0.2C are used as data sets;

[0116] In addition, the above prediction model was used to test the capacity retention rate of the ordinary electrode under the conditions of a charging rate of 10C and a discharging rate of 0.2C, as well as the capacity retention rate under the conditions of a charging rate of 12C and a discharging rate of 0.2C. The capacity retention rate values ​​were very close to the capacity retention rates of the three electrodes under the charging and discharging conditions.

[0117] A total of 50 batches of capacity retention data at different charge and discharge rates were predicted, and the prediction results of the model were: mean square error (MSE) ≤ 1, mean absolute error (MAE) ≤ 1, and the generalization ability of the model was good. This result can be considered a good prediction.

[0118] S3. Result output: Import the new data set, calculate the prediction results based on the model, and display them in the form of a visual analysis report.

[0119] Example 3

[0120] This embodiment discloses an electronic device, which includes a memory, a processor, and a computer program stored in the memory and used to run on the processor. When the processor executes the computer program, the method for predicting the charging and discharging performance of a lithium-ion battery provided in the above-mentioned embodiment 1 or 2 is implemented. The electronic device is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0121] The electronic device may be in the form of a general-purpose computing device, for example, it may be a server device. The components of the electronic device may include, but are not limited to: at least one processor, at least one memory, and a bus connecting different system components (including the memory and the processor).

[0122] The bus includes data bus, address bus and control bus.

[0123] The memory may include volatile memory, such as random access memory (RAM) and / or cache memory, and may further include read-only memory (ROM).

[0124] The memory may also include a program tool (or utility) having a set (at least one) of program modules, such program modules including but not limited to: an operating system, one or more application programs, other program modules and program data, each of which or some combination may include the implementation of a network environment.

[0125] The processor executes various functional applications and data processing by running the computer program stored in the memory, such as the method for predicting the charge and discharge performance of a lithium-ion battery provided in the above-mentioned embodiment 1 or 2.

[0126] Electronic equipment can also communicate with one or more external devices (such as keyboards, pointing devices, etc.). This communication can be carried out through an input / output (I / O) interface. In addition, electronic equipment can also communicate with one or more networks (such as local area networks (LANs), wide area networks (WANs) and / or public networks, such as the Internet) through a network adapter. As shown in the figure, the network adapter communicates with other modules of the electronic equipment through a bus. It should be understood that other hardware and / or software modules can be used in conjunction with electronic equipment, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems, etc.

[0127] It should be noted that although several units / modules or sub-units / modules of the electronic device are mentioned in the above detailed description, this division is merely exemplary and not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided into multiple units / modules to be embodied.

[0128] Example 4

[0129] The embodiments of the present disclosure also provide a computer-readable storage medium having a computer program stored thereon, and when the program is executed by a processor, the method for predicting the charging and discharging performance of a lithium-ion battery provided in the above-mentioned embodiment 1 or 2 is implemented.

[0130] The readable storage medium may include but is not limited to: a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device or any suitable combination of the above.

[0131] Example 5

[0132] The embodiments of the present disclosure also provide a computer program product, including a computer program, which, when executed by a processor, implements the method for predicting the charging and discharging performance of a lithium-ion battery provided in the above-mentioned embodiment 1 or 2.

[0133] Among them, the program code for executing the computer program product of the present disclosure can be written in any combination of one or more programming languages, and the program code can be executed completely on the user device, partially on the user device, as an independent software package, partially on the user device and partially on a remote device, or completely on the remote device.

[0134] Although the specific embodiments of the present disclosure are described above, those skilled in the art should understand that this is only an example, and the protection scope of the present disclosure is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present disclosure, but these changes and modifications all fall within the protection scope of the present disclosure.

Claims

1. A prediction model training method, characterized in that: It includes the following steps: S1. Acquire a data set, wherein the data set includes a capacity retention rate of a first lithium-ion battery at different charge and discharge rates; S2. Using machine learning to train the data set to obtain the prediction model.

2. The prediction model training method according to claim 1, characterized in that: The bias term of the prediction model is an impedance parameter; Preferably, the types of impedance parameters include contact impedance, SEI impedance, charge transfer impedance and diffusion impedance; Preferably, the impedance parameter is obtained by fitting and integrating a Gaussian function; Preferably, the method for obtaining the impedance parameter comprises the following steps: ss1. Testing the EIS test data of the first lithium-ion battery at different charge and discharge rates; ss2. Analyze the EIS test data using DRT to obtain different types of impedance parameters as bias terms; Adjusting the weight of the bias term in the prediction model; ss3. Screen impedance parameters and combine them.

3. The prediction model training method according to claim 1, characterized in that: The method for obtaining the data set includes the following steps: a1. Stack the positive electrode sheet, the first diaphragm, the negative electrode sheet, and the second diaphragm to obtain a bare cell, and insert the first reference electrode into the bare cell; Preferably, the active material of the positive electrode plate includes lithium iron phosphate; Preferably, the coating surface density of the positive electrode sheet is 12 mg / cm 2 ; Preferably, the active material of the negative electrode plate includes graphite; Preferably, the coating surface density of the negative electrode plate is 4 mg / cm 2 ; a2. Carry out packaging, liquid injection, formation, and capacity division in sequence to obtain a three-electrode lithium-ion soft-pack battery; Preferably, the packaging material includes an aluminum-plastic film; Preferably, the formation process is: 0.1C constant current charging to 50% SOC, leaving for 10 to 20 minutes, and then charging at 0.5C constant current to full, with a cut-off voltage of 3.95V and a temperature of 25°C; Preferably, the process of capacity division is: firstly, fully charged at constant current and constant voltage at 0.33C and 3.95V, then discharged in stages at 1C, 0.33C, 0.2C, and 0.1C, and finally charged at constant current of 0.33C to 5% SOC; a3. Testing the capacity retention rate of the three-electrode lithium-ion soft-pack battery at different charge and discharge rates, and capturing EIS test data in real time through the dynamic link library of the test equipment to form the data set; Preferably, each time the rate charge and discharge is performed, the battery is left to stand for 2 hours; Preferably, the electrochemical work includes a working electrode, a counter electrode and a second reference electrode, the working electrode is connected to the negative electrode plate, the counter electrode is connected to the positive electrode plate, and the second reference electrode is connected to the first reference electrode.

4. The prediction model training method according to claim 1, characterized in that: The prediction model satisfies one or more of the following conditions: ① The type of the prediction model includes a TensorFlow model; ② The learning rate of the prediction model is 0.01 to 0.1, for example, 0.01; ③ The batch size of the prediction model is 32 to 128, for example, 32, 64, 128, 256 or 512; ④ The number of iterations of the prediction model is 300 to 500, for example, 400; ⑤ In the data set, the number of data is at least 100; ⑥ The machine learning includes training using a neural network; the neural network is trained on the training samples The mean square error MSE on is: Where m is the number of samples; j=1 means the summation function starts from 1; is the predicted value; k is the actual value.

5. A method for predicting the charge and discharge performance of a lithium-ion battery, characterized in that: It includes the following steps: The charge and discharge rate of the lithium-ion battery to be tested is input into the prediction model to obtain the capacity retention rate under the charge and discharge rate; the prediction model is obtained by the training method of the prediction model according to any one of claims 1 to 4.

6. The method for predicting the charge and discharge performance of a lithium-ion battery according to claim 5, characterized in that: The first lithium-ion battery and the lithium-ion battery to be tested meet one or more of the following conditions: ① The first lithium-ion battery and the lithium-ion battery to be tested are both lithium-ion soft-pack batteries; ② The main material of the positive electrode of the first lithium-ion battery is lithium iron phosphate; ③ The main material of the negative electrode of the first lithium-ion battery is graphite; ④ The first lithium-ion battery is a three-electrode soft-pack battery containing a reference electrode, and the reference electrode is arranged in a bare cell of the three-electrode soft-pack battery; ⑤ The ratio of the capacity of the lithium-ion battery to be tested to the capacity of the first lithium-ion battery is (0.8-1.2):1; preferably, the capacity of the first lithium-ion battery is 1Ah, and the capacity of the lithium-ion battery to be tested is 0.8-1.2Ah; ⑥ The positive electrode main material and the negative electrode main material of the lithium-ion battery to be tested are of the same type as the positive electrode main material and the negative electrode main material of the first lithium-ion battery, respectively.

7. A prediction system for charging and discharging performance of a lithium-ion battery, characterized in that: It includes a prediction module, which is used to input the charge and discharge rate of the lithium-ion battery to be tested into a prediction model to obtain the capacity retention rate under the charge and discharge rate; The prediction model is obtained by using the prediction model training method as described in any one of claims 1 to 4.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the training method of the prediction model as described in any one of claims 1 to 4 is implemented, or the method for predicting the charging and discharging performance of a lithium-ion battery as described in claim 5 or 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the training method of the prediction model as described in any one of claims 1 to 4, or the prediction method of the charging and discharging performance of a lithium-ion battery as described in claim 5 or 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the training method of the prediction model as described in any one of claims 1 to 4, or the prediction method of the charging and discharging performance of a lithium-ion battery as described in claim 5 or 6 is implemented.

Images