PINN-based battery performance prediction method
Through the PINN-based battery performance prediction method, the intrinsic and physical parameters of electrolytes and materials are calculated, and the neural network model is built, which solves the problem of high testing costs in the development of new lithium-ion battery materials, and achieves fast and accurate performance prediction and reduces development costs.
Patent Information
- Application Number
- CN202510093452.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-21
- Publication Date
- 2025-05-16
AI Technical Summary
When developing new lithium-ion battery materials, the prior art requires a lot of expensive and time-consuming testing to understand the performance and scope of application of the material, resulting in high economic and time-consuming costs.
Using a battery performance prediction method based on PINN (Physics Informed Neural Networks), a neural network model is built by calculating the intrinsic parameters and physical parameters of electrolytes and materials, training is carried out, and a PINN framework model is built to predict battery performance.
This method can quickly predict the performance of new battery materials, reduce the manpower and material costs during the development process, and adjust the model through transfer learning to adapt to the performance prediction of new materials.
Smart Images
Figure CN120015179A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a battery performance prediction method, and in particular to a battery performance prediction method based on PINN. Background Art
[0002] For lithium-ion batteries, the development of new positive and negative electrode materials has always been a top priority. Positive electrode materials include lithium iron phosphate (LFP), ternary lithium (NCM & NCA), lithium cobalt oxide (LCO), lithium iron manganese phosphate (LMFP) and lithium manganese oxide. Graphite is mostly used for the negative electrode, but an appropriate amount of silicon is added. The selection of electrolyte is also crucial. Different electrolyte formulas need to be adapted for different positive and negative electrode materials. At the same time, in order to improve the initial efficiency of the battery, lithium supplementation technology will also be used in the positive and negative electrodes.
[0003] Whenever a new material is developed, a lot of tests need to be done, including short-term and long-term tests, in order to understand its first effect, DCIR, cycle capacity and other performance. Before testing, it is impossible to know whether a material is suitable for energy storage or power, whether it should be doped with silicon or supplemented with lithium, whether it is suitable for high or low temperature environments, and whether it is suitable for high-rate charging and discharging or low-rate charging and discharging. Even for manufacturers with testing capabilities, the testing fee is quite expensive. And for some tests, it is often necessary to outsource. At the same time, for some long-term tests, such as storage and cycle tests, the testing time is often measured in years, which results in a lot of economic and time costs. Summary of the invention
[0004] The present invention provides a battery performance prediction method based on PINN to solve the above-mentioned technical problems, and specifically adopts the following technical solutions:
[0005] A battery performance prediction method based on PINN, comprising:
[0006] calculating first intrinsic parameters of the electrolyte, the first type of positive electrode, and the negative electrode;
[0007] determining first physical parameters of the electrolyte, the first type of positive electrode, and the negative electrode;
[0008] Based on the first intrinsic parameter and the first physical parameter, a battery performance parameter is calculated by a P2D model;
[0009] Building a neural network model, taking the first intrinsic parameter and the first physical parameter as inputs of the neural network model, taking the battery performance parameter as outputs of the neural network model, training the neural network model, and finally constructing a first PINN framework model;
[0010] A first intrinsic parameter and the first physical parameter are input into the first PINN framework model to obtain a first predicted performance parameter.
[0011] Furthermore, the specific method for calculating the first intrinsic parameters of the electrolyte, the first type of positive electrode and the negative electrode is:
[0012] The first intrinsic parameters of the electrolyte, the first type of positive electrode and the negative electrode are calculated based on density functional theory and molecular dynamics.
[0013] Further, the specific method for determining the first physical parameter of the electrolyte, the first type of positive electrode and the negative electrode is:
[0014] The first physical parameters of the electrolyte, the first type of positive electrode and the negative electrode are obtained through experimental testing.
[0015] Furthermore, the specific method for obtaining the battery performance parameter based on the first intrinsic parameter and the first physical parameter is:
[0016] The first intrinsic parameter and the first physical parameter are input into a P2D model to obtain the battery performance parameter.
[0017] Further, the first intrinsic parameter includes electrical conductivity and ion diffusion coefficient;
[0018] The first physical parameters include particle size, electrolyte concentration and thickness.
[0019] Furthermore, the battery performance parameter and the first predicted performance parameter include voltage and power.
[0020] Furthermore, the first type of positive electrode is a lithium iron phosphate positive electrode.
[0021] Furthermore, the PINN-based battery performance prediction method further comprises:
[0022] The neural network model is adjusted based on transfer learning.
[0023] Furthermore, the specific method for adjusting the neural network model based on transfer learning is:
[0024] calculating second intrinsic parameters of the electrolyte, the second type positive electrode, and the negative electrode;
[0025] determining a second physical parameter of the electrolyte, the second type of positive electrode, and the negative electrode;
[0026] Using the second intrinsic parameter and the second physical parameter as inputs of the neural network model to obtain a second performance parameter;
[0027] The neural network model is verified using the test performance parameters corresponding to the electrolyte, the second type of positive electrode material and the negative electrode, and the weight parameters of the hidden layer are adjusted to obtain the adjusted neural network model, and finally a second PINN framework model is constructed;
[0028] The second intrinsic parameters and second physical parameters of the electrolyte, the second type of positive electrode and the negative electrode are input into the second PINN framework model to obtain second predicted performance parameters.
[0029] Furthermore, the second type of positive electrode is a ternary lithium positive electrode.
[0030] The benefit of the present invention lies in that the provided PINN-based battery performance prediction method can predict the battery performance of the original material system through a trained model, and can also adjust the original model through various performance parameters of the newly developed battery to adapt it to the newly developed battery material, and quickly predict the performance of the newly developed battery material, which can greatly reduce the manpower and material costs in the development process of the new battery system. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0032] Figure 1 is a flow chart of a battery performance prediction method based on PINN of the present invention;
[0033] Figure 2 is another flow chart of a battery performance prediction method based on PINN of the present invention;
[0034] Figure 3 is a schematic diagram of the PINN framework of the present invention;
[0035] Figure 4 is a schematic diagram of transfer learning of the present invention;
[0036] Figure 5 It is a schematic diagram of neural network model verification and parameter adjustment of the present invention. DETAILED DESCRIPTION
[0037] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.
[0038] like Figure 1-3 As shown, the present application discloses a battery system performance prediction method based on PINN (Physics Informed Neural Networks), comprising: S1: Calculating the first intrinsic parameters of the electrolyte, the first type of positive electrode and the negative electrode. S2: Determining the first physical parameters of the electrolyte, the first type of positive electrode and the negative electrode. S3: Obtaining the battery performance parameters based on the first intrinsic parameters and the first physical parameters. S4: Building a neural network model, using the first intrinsic parameters and the first physical parameters as inputs of the neural network model, using the battery performance parameters as outputs of the neural network model, training the neural network model, and finally constructing a first PINN framework model. S5: Inputting the first intrinsic parameters and the first physical parameters into the first PINN framework model to obtain the first predicted performance parameters.
[0039] The battery performance prediction method based on PINN in this application can quickly predict the performance of newly developed battery materials through a trained model, which can greatly reduce the manpower and material costs in the development process of new battery systems. The above steps are described in detail below.
[0040] For step S1: first intrinsic parameters of the electrolyte, the first type of positive electrode and the negative electrode are calculated.
[0041] In an embodiment of the present application, a specific method for calculating the first intrinsic parameters of the electrolyte, the first type of positive electrode and the negative electrode is:
[0042] The first intrinsic parameters of the electrolyte, the first type of positive electrode and the negative electrode are calculated based on density functional theory (DFT) and molecular dynamics (MD).
[0043] In an embodiment of the present application, the first intrinsic parameters of the electrolyte, the first type of positive electrode and the negative electrode include electrical conductivity and ion diffusion coefficient.
[0044] For step S2: first physical parameters of the electrolyte, the first type of positive electrode and the negative electrode are determined.
[0045] In the embodiment of the present application, the first physical parameters of the electrolyte, the first type of positive electrode and the negative electrode include but are not limited to parameters of the physical level of the material such as particle size, electrolyte concentration and thickness.
[0046] For step S3: obtaining a battery performance parameter based on the first intrinsic parameter and the first physical parameter.
[0047] In an embodiment of the present application, the battery performance parameters include voltage and power.
[0048] In an embodiment of the present application, a specific method for determining the first physical parameters of the electrolyte, the first type of positive electrode and the negative electrode is:
[0049] The first intrinsic parameter and the first physical parameter are input into a P2D (Pseudo Two Dimensional) model to obtain the battery performance parameter.
[0050] For step S4: build a neural network model, use the first intrinsic parameter and the first physical parameter as the input of the neural network model, use the battery performance parameter as the output of the neural network model, train the neural network model, and finally construct a first PINN framework model.
[0051] For step S5: input the first intrinsic parameter and the first physical parameter into the first PINN framework model to obtain the first predicted performance parameter.
[0052] For a newly developed battery material, a corresponding first intrinsic parameter and a first physical parameter are input to obtain a first predicted performance parameter.
[0053] In an embodiment of the present application, the first predicted performance parameter includes voltage and power.
[0054] It can be understood that the first PINN framework is built based on the parameters of a large number of first-type positive electrode (such as lithium iron phosphate) batteries. When a new first-type positive electrode material is developed, the parameters of the battery constructed with the new first-type positive electrode material are input into the first PINN framework model to predict the first predicted performance parameters.
[0055] In an embodiment of the present application, the battery performance prediction method based on PINN further includes:
[0056] Adjust the neural network model based on transfer learning. It can be understood that the aforementioned first PINN framework is more accurate in predicting the performance of batteries constructed with the first type of positive electrode material. When we want to predict the parameters of batteries constructed with the second type of positive electrode material through the first PINN framework, we only need to test a small number of batteries constructed with the second type of positive electrode material. Specifically, the specific method for adjusting the neural network model based on transfer learning is:
[0057] Second intrinsic parameters of the electrolyte, the second type positive electrode, and the negative electrode are calculated.
[0058] A second physical parameter of the electrolyte, the second type of positive electrode, and the negative electrode is determined.
[0059] The second intrinsic parameter and the second physical parameter are used as the input of the neural network model to obtain the second performance parameter. Figure 4 As shown, the hidden layer of the original model and the adjusted weight parameters are used to calculate the corresponding performance.
[0060] In the embodiment of the present application, the second intrinsic parameters of the electrolyte, the second type of positive electrode and the negative electrode include conductivity and ion diffusion coefficient. The second physical parameters of the electrolyte, the second type of positive electrode and the negative electrode include but are not limited to the physical parameters of the material such as particle size, electrolyte concentration and thickness. Figure 5 As shown, the actual test performance parameters corresponding to the electrolyte, the second type of positive electrode material and the negative electrode are used to verify the neural network model, and the weight parameters of the hidden layer are fine-tuned to obtain the adjusted neural network model, and finally the second PINN framework model is constructed. In this way, the second intrinsic parameters and the second physical parameters of the electrolyte, the second type of positive electrode and the negative electrode are input into the second PINN framework model to obtain the second predicted performance parameters.
[0061] In an embodiment of the present application, the first type of positive electrode is a lithium iron phosphate positive electrode, and the second type of positive electrode is a ternary lithium positive electrode.
[0062] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the above embodiments do not limit the present invention in any form, and any technical solution obtained by equivalent replacement or equivalent transformation falls within the protection scope of the present invention.
Claims
1. A battery performance prediction method based on PINN, characterized in that: Include: calculating first intrinsic parameters of the electrolyte, the first type of positive electrode, and the negative electrode; determining first physical parameters of the electrolyte, the first type of positive electrode, and the negative electrode; Based on the first intrinsic parameter and the first physical parameter, a battery performance parameter is calculated by a P2D model; Building a neural network model, taking the first intrinsic parameter and the first physical parameter as inputs of the neural network model, taking the battery performance parameter as outputs of the neural network model, training the neural network model, and finally constructing a first PINN framework model; A first intrinsic parameter and the first physical parameter are input into the first PINN framework model to obtain a first predicted performance parameter.
2. The battery performance prediction method based on PINN according to claim 1, characterized in that: The specific method for calculating the first intrinsic parameters of the electrolyte, the first type of positive electrode and the negative electrode is: The first intrinsic parameters of the electrolyte, the first type of positive electrode and the negative electrode are calculated based on density functional theory and molecular dynamics.
3. The battery performance prediction method based on PINN according to claim 1, characterized in that: The specific method for determining the first physical parameter of the electrolyte, the first type positive electrode and the negative electrode is: The first physical parameters of the electrolyte, the first type of positive electrode and the negative electrode are obtained through experimental testing.
4. The battery performance prediction method based on PINN according to claim 1, characterized in that: The specific method for obtaining the battery performance parameter based on the first intrinsic parameter and the first physical parameter is: The first intrinsic parameter and the first physical parameter are input into a P2D model to obtain the battery performance parameter.
5. The battery performance prediction method based on PINN according to claim 1, characterized in that: The first intrinsic parameters include electrical conductivity and ion diffusion coefficient; The first physical parameters include particle size, electrolyte concentration and thickness.
6. The battery performance prediction method based on PINN according to claim 1, characterized in that: The battery performance parameter and the first predicted performance parameter include voltage and power.
7. The battery performance prediction method based on PINN according to claim 1, characterized in that: The first type of positive electrode is a lithium iron phosphate positive electrode.
8. The battery performance prediction method based on PINN according to claim 1, characterized in that: The PINN-based battery performance prediction method further comprises: The neural network model is adjusted based on transfer learning.
9. The battery performance prediction method based on PINN according to claim 8, characterized in that: The specific method for adjusting the neural network model based on transfer learning is: calculating second intrinsic parameters of the electrolyte, the second type positive electrode, and the negative electrode; determining a second physical parameter of the electrolyte, the second type of positive electrode, and the negative electrode; Using the second intrinsic parameter and the second physical parameter as inputs of the neural network model to obtain a second performance parameter; The neural network model is verified using the test performance parameters corresponding to the electrolyte, the second type of positive electrode material and the negative electrode, and the weight parameters of the hidden layer are adjusted to obtain the adjusted neural network model, and finally a second PINN framework model is constructed; The second intrinsic parameters and second physical parameters of the electrolyte, the second type of positive electrode and the negative electrode are input into the second PINN framework model to obtain second predicted performance parameters.
10. The battery performance prediction method based on PINN according to claim 9, characterized in that: The second type of positive electrode is a ternary lithium positive electrode.