A method for regulating the doping / coating amount of lithium battery positive electrode materials based on neural network
By predicting the element doping/covering amount of the positive electrode material of lithium-ion battery by a neural network-based method, the problems of short cycle life and poor thermal stability in the prior art are solved, and the performance status of lithium-ion batteries are accurately determined and tested. promote.
Patent Information
- Application Number
- CN202210832194.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-07-15
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2042-07-15
AI Technical Summary
The prior art cannot effectively predict the element doping/covering amount of lithium-ion battery positive electrode material, resulting in short cycle life and poor thermal stability, and manual testing is time-consuming and labor-intensive.
Using a neural network-based method, an initial data set is established by determining parameters related to the performance of lithium-ion batteries, and a neural network model is constructed to predict the doping/covering of the element of the positive electrode material of lithium batteries. The neural network model is used to dopant different elements. The miscellaneous/coverage amount is predicted to obtain the charge and discharge cycle number of the lithium battery.
It realizes accurate acquisition of lithium battery performance status in element doping/covering information, improves testing efficiency, saves resources, and is suitable for lithium battery status measurement application scenarios.
Smart Images

Figure CN115130388B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of battery material technology, and more specifically, to a method for regulating the doping / coating amount of elements in a lithium battery positive electrode material based on a neural network. Background Art
[0002] Lithium-ion batteries are currently the secondary batteries with the best comprehensive performance and are widely used in consumer electronics, electric vehicles, energy storage and other fields. They have the advantages of high operating voltage, high energy density and low self-discharge. Like other electrochemical batteries, lithium-ion batteries are mainly composed of three basic components: cathode (or positive electrode), anode (or negative electrode) and electrolyte (used as a medium for ion movement). In terms of positive electrode materials, high-nickel layered oxides LiNixM1-xO2 (x≥0.6, M=Co, Mn, Al, etc.) have high specific capacity and are considered to be the most important type of positive electrode material for high-energy density lithium-ion batteries such as power batteries, but their shortcomings such as short cycle life and poor thermal stability restrict their widespread application. Element bulk doping and material surface coating are the most effective improvement methods, but the actual effect of different contents on high-nickel positive electrodes is still unknown. If the materials of each element and concentration are tested manually, it will be a very time-consuming and laborious task. With the intersection of computers and other fields, the combination of machine learning and chemistry is becoming more and more popular, and its combination with the field of materials has inspired promising development methods. The potential of machine learning is becoming increasingly popular in the application of material design, performance prediction, structural optimization and classification. Furthermore, the neural network is the collective behavior of a large number of neurons, not the simple addition of the behavior of each unit, and generally exhibits the characteristics of a complex nonlinear dynamic system. It is precisely because of these characteristics of neural networks that they can handle some problems with very complex environmental information, unclear knowledge background and unclear reasoning rules, which corresponds exactly to the information of material chemistry and is conducive to promoting the development of material chemistry.
[0003] The prior art discloses a method for regulating the energy density of negative electrode materials of lithium-ion batteries based on a neural network, which includes: step 1: collecting material data, and obtaining various parameters of the BP neural network model for initializing the total sample after preprocessing the material data; step 4: training the BP neural network model to obtain the best prediction network; step 5: testing the fit of the BP neural network model; step 6: inputting the sample to be predicted into the BP neural network model after the fit test for prediction, and obtaining the energy density of the negative electrode material. The solution of the prior art predicts the energy density of the electrode material, rather than the doping / coating amount of the electrode material. Summary of the invention
[0004] The present invention proposes a method for regulating the element doping / coating amount of lithium battery positive electrode materials based on a neural network, which can simultaneously and accurately obtain the performance status of the lithium battery in the element doping / coating amount information, and can be better suitable for lithium battery status measurement application scenarios.
[0005] The primary purpose of the present invention is to solve the above technical problems. The technical solution of the present invention is as follows:
[0006] A method for regulating the doping / coating amount of lithium battery positive electrode materials based on a neural network comprises the following steps:
[0007] S1. Determine parameters directly related to the performance of lithium-ion batteries, and use the relevant parameters to establish an initial data set, wherein the relevant parameters include: element doping concentration, element coating amount, electrode material formula, battery charge and discharge cut-off voltage, electrode material surface loading, and charge and discharge current density;
[0008] S2. Establish a neural network model for predicting the doping / coating amount of lithium battery positive electrode materials, and use the constructed data set to evaluate the model training set to obtain a neural network model that meets the preset accuracy;
[0009] S3. Use the obtained neural network model to predict materials with different element doping / coating amounts to obtain the number of charge and discharge cycles of the lithium battery.
[0010] Furthermore, the structure of the neural network model includes: an input layer, a hidden layer and an output layer, wherein the input layer of the neural network model contains 4 neurons, i.e., four input parameters, corresponding to the element doping concentration, the formula ratio of the positive electrode material, the upper and lower limits of the battery's operating voltage, and the charging and discharging current density, the number of neurons in the hidden layer of the neural network model is 9, and the output layer of the neural network model contains 1 neuron, which is the number of charging and discharging cycles of the lithium battery.
[0011] Furthermore, the input layer of the neural network model includes a first input terminal and a second input terminal, wherein the input data of the first input terminal are: the upper and lower limits of the battery's operating voltage, the current density of charging and discharging, and the first input terminal is connected to the cyclic network layer of the neural network model; the input data of the second input terminal are the element doping concentration and the composition of the positive electrode material, and the second input terminal is connected to the fully connected layer of the neural network model, and the output of the neural network model is the number of charge and discharge cycles of the lithium battery.
[0012] Furthermore, the number of neurons in the hidden layer of the neural network model is estimated using the Kolmogorov theorem, and the equation for the number of hidden layer nodes is:
[0013] m=2n+1
[0014] Among them, m and n are the number of neurons contained in the input layer and output layer respectively.
[0015] Furthermore, the specific training process of the neural network model is:
[0016] The training is performed according to the set network structure, and a trained neural network model is obtained when the preset accuracy is reached. The acquired data is subjected to data sampling processing, data standardization processing, and data dimension change processing. The initial information in the neural network model is subjected to error back propagation until the weights of each layer are obtained. The weights and biases of the optimal neurons in each layer of the neural network are used to judge the current state in real time according to the continuous change of the error of the lithium-ion battery energy storage system, and to provide the current optimal control parameters.
[0017] Furthermore, a ten-fold cross validation method was used to validate the model during the training of the neural network model. The learning rate during model training was less than 1.0 and greater than 10. -6 The default number of training times for the neural network model is 1000, and the error accuracy is set to 1e -5 , the activation function of the hidden layer adopts the ReLU activation function, and the activation function of the output layer adopts the Sigmoid activation function.
[0018] Furthermore, the mean square error (MSE) and the coefficient of determination R 2 As an indicator for evaluating the performance of neural network models.
[0019] Furthermore, the calculation formula of mean square error MSE is:
[0020]
[0021] The calculation formula of the coefficient of determination R2 is:
[0022]
[0023] Where n is the number of samples, y pred is the predicted value of the neural network model, and y is the measured value.
[0024] Furthermore, before the neural network model is trained, it also includes: data initialization and normalization, which limits the data to the interval [0,1] after processing, and data normalization, which removes the unit restriction of the data and converts it into a dimensionless pure value to facilitate the comparison and weighting of indicators of different units or magnitudes.
[0025] Furthermore, the acquired data is subjected to data sampling processing, data standardization processing and data dimension change processing, which specifically includes the following steps:
[0026] Sample the element doping concentration, the formula ratio of the positive electrode material, the upper and lower limits of the battery's operating voltage, and the current density data of charge and discharge, and set the data interval to 1s to generate a data point;
[0027] The sampled data is standardized so that the element doping concentration, the formula ratio of the positive electrode material, the upper and lower limits of the battery's operating voltage, and the charge and discharge current density are all distributed in the [0,1] interval. The formula for the standardized processing is:
[0028]
[0029] Where minA and maxA are the minimum and maximum values of attribute A respectively, and an original value x of A is mapped to a value in the interval x1[0,1] through maximum-minimum normalization;
[0030] The standardized data is subjected to dimension change processing, the parameters of each time point in the data are connected into a vector form, and the parameters of the nth time point are connected as an input number of the neural network.
[0031] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:
[0032] The present invention obtains the optimal number of cycles by predicting the optimal doping / coating amount of elements in the high-nickel positive electrode material of the lithium-ion battery, thereby achieving simultaneous and accurate acquisition of the performance status of the lithium battery in the element doping / coating amount information, which can be better applied to the application scenario of lithium battery status determination, thereby saving test resources and improving test efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] Figure 1 The figure is a flow chart of the method of the present invention.
[0034] Figure 2 Schematic diagram of the neural network model structure in an embodiment of the present invention.
[0035] Figure 3 It is the performance index of the neural network model in the embodiment of the present invention and the performance index after ten-fold cross validation.
[0036] Figure 4 This is a diagram showing the effect of the neural network model of an embodiment of the present invention after ten-fold cross validation.
[0037] Figure 5 This is a cycle performance diagram of lithium ions with an Al doping concentration of 4% according to an embodiment of the present invention. DETAILED DESCRIPTION
[0038] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.
[0039] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.
[0040] Example 1
[0041] like Figure 1 As shown, a method for regulating the amount of element doping / coating of a lithium battery positive electrode material based on a neural network comprises the following steps:
[0042] S1. Determine parameters directly related to the performance of lithium-ion batteries, and use the relevant parameters to establish an initial data set, wherein the relevant parameters include: element doping concentration, element coating amount, electrode material formula, battery charge and discharge cut-off voltage, electrode material surface loading, and charge and discharge current density;
[0043] It should be noted that by analyzing the parameters related to the performance of lithium-ion batteries, each time the input parameters are different, each time the model is trained, each time the weights and deviations obtained will be different, to calculate the significance level of each input, and finally determine the main operating parameters directly related to the cycle performance of lithium-ion batteries, and use the main operating parameters to establish a data set. The main operating parameters include: element doping concentration, element coating amount, electrode material formula, battery charge and discharge cut-off voltage, electrode material surface loading, and charge and discharge current density. Table 1 shows a list of lithium-ion battery data sets.
[0044] Table 1
[0045]
[0046] S2. Establish a neural network model for predicting the doping / coating amount of lithium battery positive electrode materials, and use the constructed data set to evaluate the model training set to obtain a neural network model that meets the preset accuracy;
[0047] It should be noted that the structure of the neural network model includes: an input layer, a hidden layer and an output layer, wherein the input layer of the neural network model contains 4 neurons, i.e., four input parameters, corresponding to the element doping concentration, the formula ratio of the positive electrode material, the upper and lower limits of the battery's operating voltage, and the charging and discharging current density. The number of neurons in the hidden layer of the neural network model is 9, and the output layer of the neural network model contains 1 neuron, which is the number of charge and discharge cycles of the lithium battery.
[0048] More specifically, the input layer of the neural network model includes: a first input terminal and a second input terminal, wherein the input data of the first input terminal are: the upper and lower limits of the battery's operating voltage, the current density of charging and discharging, and the first input terminal is connected to the cyclic network layer of the neural network model; the input data of the second input terminal are the element doping concentration and the formula ratio of the positive electrode material, and the second input terminal is connected to the fully connected layer of the neural network model, and the output of the neural network model is the number of charge and discharge cycles of the lithium battery.
[0049] The number of neurons in the hidden layer of the neural network model is estimated using the Kolmogorov theorem, and the equation for the number of hidden layer nodes is:
[0050] m=2n+1
[0051] Among them, m and n are the number of neurons contained in the input layer and output layer respectively.
[0052] Example 2
[0053] This embodiment describes in detail the training process of the neural network model. The specific training process is as follows:
[0054] First, the data is initialized and normalized. The data is processed and restricted to the interval [0,1]. The data normalization removes the unit restriction of the data and converts it into a dimensionless pure value, which makes it easier for indicators of different units or magnitudes to be compared and weighted.
[0055] According to the set network structure (such as Figure 2 The neural network model is trained to achieve a preset accuracy, and the acquired data is subjected to data sampling, data standardization, and data dimension change. The initial information in the neural network model is subjected to error back propagation until the weight of each layer is obtained, wherein the weight and bias of the optimal neuron in each layer of the neural network are used to judge the current state in real time according to the continuous change of the error of the lithium-ion battery energy storage system, and to provide the current optimal control parameters.
[0056] The acquired data is subjected to data sampling, data standardization and data dimension change processing, which specifically includes the following steps:
[0057] Sample the element doping concentration, the formula ratio of the positive electrode material, the upper and lower limits of the battery's operating voltage, and the current density data of charge and discharge, and set the data interval to 1s to generate a data point;
[0058] The sampled data is standardized so that the element doping concentration, the formula ratio of the positive electrode material, the upper and lower limits of the battery's operating voltage, and the charge and discharge current density are all distributed in the [0,1] interval. The formula for the standardized processing is:
[0059]
[0060] Where minA and maxA are the minimum and maximum values of attribute A respectively, and an original value x of A is mapped to a value in the interval x1[0,1] through maximum-minimum normalization;
[0061] The standardized data is subjected to dimension change processing, the parameters of each time point in the data are connected into a vector form, and the parameters of the nth time point are connected as an input number of the neural network.
[0062] In the training process of the neural network model, in order to improve the accuracy and fitness of the neural network model, and to select the correct model and model parameters, and to intuitively see the impact of different models or parameters on structural accuracy, the ten-fold cross-validation method is used to verify the model. At the same time, the initial parameters of the neural network model are randomly assigned, and the learning rate is set through repeated experiments to find a good (or good enough) learning rate. The learning rate during model training is less than 1.0 and greater than 10 -6 The default number of training times for the neural network model is 1000, and the error accuracy is set to 1e -5 , the activation function of the hidden layer adopts the ReLU activation function, and the activation function of the output layer adopts the Sigmoid activation function.
[0063] In the present invention, the mean square error (MSE) and the determination coefficient R are used. 2 As an indicator for evaluating the performance of the neural network model. The calculation formula of the mean square error MSE is:
[0064]
[0065] Coefficient of determination R 2 The calculation formula is:
[0066]
[0067] Where n is the number of samples, y pred is the predicted value of the neural network model, and y is the measured value.
[0068] S3. Use the obtained neural network model to predict materials with different element doping / coating amounts to obtain the number of charge and discharge cycles of the lithium battery.
[0069] Example 3
[0070] This embodiment specifically describes and verifies this solution in conjunction with the experimental data diagram. Figure 3If the performance indicators of the above evaluation models are all above 85%, it means that the neural network model has an ideal prediction effect on the predicted values of doping concentrations of different elements in lithium-ion batteries, and can be used to become a powerful tool for predicting the effects of different doping concentrations on the number of lithium-ion cycles in the future. Figure 4 This is the effect diagram of the neural network model in the present invention after ten-fold cross validation. Figure 5 The cycle number of the lithium-ion battery with 4% Al doping concentration is shown as 403 when the discharge specific capacity is 120 mAh g-1, which is close to the predicted value of the neural network model for 4% Al doping concentration in lithium-ion batteries.
[0071] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.
Claims
1. A method for controlling the doping / coating amount of lithium battery positive electrode materials based on a neural network, characterized in that: The following steps are involved: S1. Determine parameters directly related to the performance of lithium-ion batteries, and use the relevant parameters to establish an initial data set, wherein the relevant parameters include: element doping concentration, element coating amount, electrode material formula, battery charge and discharge cut-off voltage, electrode material surface loading, and charge and discharge current density; S2. Establish a neural network model for predicting the doping / coating amount of lithium battery positive electrode materials, and use the constructed data set to evaluate the model training set to obtain a neural network model that meets the preset accuracy; The structure of the neural network model includes: an input layer, a hidden layer and an output layer, wherein the input layer of the neural network model contains 4 neurons, i.e., four input parameters, which respectively correspond to the element doping concentration, the formula ratio of the positive electrode material, the upper and lower limits of the working voltage of the battery, and the current density of charge and discharge. The number of neurons in the hidden layer of the neural network model is 9, and the output layer of the neural network model contains 1 neuron, which is the number of charge and discharge cycles of the lithium battery; The input layer of the neural network model includes a first input terminal and a second input terminal, wherein the input data of the first input terminal are: the upper and lower limits of the working voltage of the battery, the current density of charge and discharge, and the first input terminal is connected to the cyclic network layer of the neural network model; the input data of the second input terminal are the element doping concentration and the formula ratio of the positive electrode material, and the second input terminal is connected to the fully connected layer of the neural network model, and the output of the neural network model is the number of charge and discharge cycles of the lithium battery; The specific training process of the neural network model is: The training is performed according to the set network structure, and the trained neural network model is obtained when the preset accuracy is reached. The acquired data is subjected to data sampling processing, data standardization processing, and data dimension change processing. The initial information is subjected to error back propagation in the neural network model until the weight of each layer is obtained, wherein the weight and bias of the optimal neuron of each layer in the neural network are used to judge the current state in real time according to the continuous change of the error of the lithium-ion battery energy storage system, and give the current optimal control parameters; S3. Use the obtained neural network model to predict materials with different element doping / coating amounts to obtain the number of lithium battery charge and discharge cycles.
2. The method for controlling the amount of element doping / coating of lithium battery positive electrode materials based on a neural network according to claim 1, characterized in that: The number of neurons in the hidden layer of the neural network model is estimated using the Kolmogorov theorem, and the equation for the number of hidden layer nodes is: m=2n+1 Among them, m and n are the number of neurons contained in the input layer and output layer respectively.
3. The method for controlling the amount of element doping / coating of lithium battery positive electrode materials based on a neural network according to claim 1, characterized in that: During the training of the neural network model, the ten-fold cross-validation method is used to verify the model. The learning rate during model training is less than 1.0 and greater than 10. -6 The default number of training times for the neural network model is 1000, and the error accuracy is set to 1e -5 , the activation function of the hidden layer adopts the ReLU activation function, and the activation function of the output layer adopts the Sigmoid activation function.
4. The method for controlling the amount of element doping / coating of lithium battery positive electrode materials based on neural network according to claim 1, characterized in that: The mean square error (MSE) and the coefficient of determination (R) were used. 2 As an indicator for evaluating the performance of neural network models.
5. The method for controlling the amount of element doping / coating of lithium battery positive electrode materials based on a neural network according to claim 4, characterized in that: The calculation formula for mean square error (MSE) is: The calculation formula of the coefficient of determination R2 is: Where n is the number of samples, y pred is the predicted value of the neural network model, and y is the measured value.
6. The method for controlling the amount of element doping / coating of lithium battery positive electrode materials based on a neural network according to claim 1, characterized in that: Before training the neural network model, it also includes: data initialization and normalization, which limits the data to the interval [0,1] after processing, and data normalization, which removes the unit restrictions of the data and converts it into a dimensionless pure value to facilitate the comparison and weighting of indicators of different units or magnitudes.
7. The method for controlling the amount of element doping / coating of lithium battery positive electrode materials based on a neural network according to claim 1, characterized in that: The acquired data is subjected to data sampling, data standardization and data dimension change processing, which specifically includes the following steps: Sample the element doping concentration, the formula ratio of the positive electrode material, the upper and lower limits of the battery's operating voltage, and the current density data of charge and discharge, and set the data interval to 1s to generate a data point; The sampled data is standardized so that the element doping concentration, the formula ratio of the positive electrode material, the upper and lower limits of the battery's operating voltage, and the charge and discharge current density are all distributed in the [0,1] interval. The formula for the standardized processing is: Where minA and maxA are the minimum and maximum values of attribute A respectively, and an original value x of A is mapped to a value in the interval x1[0,1] through maximum-minimum normalization; The standardized data is subjected to dimension change processing, the parameters of each time point in the data are connected into a vector form, and the parameters of the nth time point are connected as an input number of the neural network.
