A lithium battery cross-scale simulation method, device, equipment and program product

Abstract

The present invention belongs to the technical field of secondary batteries, and specifically relates to a cross-scale simulation method, device, equipment and program product for lithium batteries. The method comprises the following steps: The method comprises: S1: calculating the basic physicochemical parameters of the microscopic scale object of the lithium battery; S2: learning and training the law between the atomic arrangement mode of the microscopic scale object and the basic physicochemical parameters in step S1 to obtain a machine learning molecular dynamics proxy model; S3: performing parallel calculations on the machine learning molecular dynamics model of the microscopic scale object to obtain the key physicochemical parameters of the overall boundary; S4: combining the obtained key physicochemical parameters with the macroscopic electrochemical simulation method to simulate the electrochemical performance and safety of the lithium battery under several different working conditions. The method realizes the simulation of the electrochemical performance and safety of the lithium battery under different working conditions, thereby realizing the accurate prediction of the decay characteristics and potential safety accidents of the lithium battery.

Description

Technical Field

[0001] The present invention belongs to the technical field of secondary batteries, and specifically relates to a lithium battery cross-scale simulation method, device, equipment and program product. Background Art

[0002] In recent years, as the application scope of secondary batteries represented by lithium-ion batteries has become more and more extensive, lithium-ion batteries are widely used in energy storage systems such as hydropower, thermal power, wind power and solar power stations, as well as power tools, electric bicycles, electric motorcycles, electric vehicles, military equipment, aerospace and other fields. As secondary ion batteries have made great progress, higher requirements have been put forward for their energy density, cycle performance and safety performance.

[0003] Lithium batteries are high-energy components. Combustion and explosion accidents may occur during use. However, there is currently no ability to predict potential safety accidents during the decay of lithium batteries. The main form of all lithium battery accidents is thermal runaway, and all thermal runaways start with the degradation and failure of the interface film on the negative electrode surface. It is hoped that accidents can be predicted by understanding the decay mechanism of the interface film on the negative electrode surface through simulation methods. However, the current computational simulation methods based on the first principles of quantum mechanics can only simulate systems of up to hundreds of atoms, which is far from the scale that can reflect actual characteristics.

[0004] At present, there is no method that can simulate the atomic system of lithium batteries with high precision above 10,000 levels. Summary of the invention

[0005] The present invention provides a lithium battery cross-scale simulation method, device, equipment, computer storage medium and computer program product, which can perform high-precision simulation of the lithium battery decay process.

[0006] The technical solution adopted by the present invention is as follows:

[0007] In a first aspect, the present invention provides a lithium battery cross-scale simulation method, the method comprising:

[0008] S1: Use the first principle method to simulate and calculate the basic physical and chemical parameters of lithium battery micro-scale objects;

[0009] The basic physicochemical parameters include at least one of the physicochemical parameters of the electrode material and the physicochemical parameters of the electrolyte constituting the lithium battery micro-scale object;

[0010] S2: using a machine learning method to learn and train the law between the atomic arrangement of the microscopic object and the basic physicochemical parameters in step S1, and obtaining a machine learning molecular dynamics proxy model;

[0011] S3: Using high-performance computing methods to assist, parallelize the machine learning molecular dynamics models of no less than 1,000 adjacent microscopic objects to obtain the key physical and chemical parameters of the overall boundaries of these objects;

[0012] The adjacency mentioned above means that all objects can be adjacent to each other by being connected;

[0013] The key physicochemical parameters include at least one of hardness, electronic conductivity, lithium ion mobility, reaction activation energy, and substance concentration;

[0014] S4: Combine the key physicochemical parameters obtained with the macroscopic electrochemical simulation method to simulate the electrochemical performance and safety of lithium batteries under several different operating conditions.

[0015] Preferably, in step S1, a high-throughput computing method is used to calculate the basic physicochemical parameters of the lithium battery microscopic scale object.

[0016] Furthermore, the high-throughput computing method includes at least one of density functional theory calculation, molecular dynamics simulation and finite element calculation.

[0017] Preferably, the step S2 specifically includes the following steps:

[0018] S2-1: creating training samples based on basic physicochemical parameters of each different lithium battery microscopic object and atomic arrangement information of each microscopic object;

[0019] S2-2: According to the training samples obtained in step S2-1, the original machine learning molecular dynamics proxy model is trained until the training stop condition is met to obtain the machine learning molecular dynamics proxy model.

[0020] Preferably, the specific operation in step S2-2 is to perform the following steps for each training sample:

[0021] S2-2-1: inputting the training sample into the original machine learning molecular dynamics proxy model to obtain machine learning physical and chemical parameters of microscopic objects;

[0022] S2-2-2: Determine the loss function value of the machine learning molecular dynamics proxy model according to the machine learning physicochemical parameters and basic physicochemical parameter information;

[0023] S2-2-3: When the loss function value does not meet the training stop condition, adjust the model parameters of the machine learning molecular dynamics proxy model, and use the training samples to train the machine learning molecular dynamics proxy model with adjusted parameters until the training stop condition is met, thereby obtaining the trained machine learning molecular dynamics proxy model.

[0024] Furthermore, the training stopping condition is that the result of the machine learning molecular dynamics model differs from the result of the computational simulation using first principles molecular dynamics by no more than 10%.

[0025] Preferably, the physicochemical parameters of the electrode material include at least one of the atomic arrangement structure, electronic structure, thermodynamic energy, molecular frontier orbital energy level, electron affinity, redox potential, theoretical specific capacity, electronic conductivity, lithium ion conductivity, magnetic permeability, specific heat capacity, hardness, Raman spectrum, infrared spectrum, and dielectric constant of the electrode material.

[0026] Preferably, the physicochemical parameters of the electrolyte include at least one of the geometric structure, electronic structure, thermodynamic energy, molecular frontier orbital energy level, dipole moment, electron affinity, redox potential, lithium ion conductivity, Raman spectrum, infrared spectrum, dielectric constant, viscosity, density, melting point and boiling point of the electrolyte solvent molecules.

[0027] Preferably, the atomic arrangement of the microscopic-scale object includes the element types, proportions and arrangement order of the atoms in space.

[0028] Preferably, the machine learning method in step S2 includes at least one of a linear regression analysis method, a logistic regression analysis method, a support vector machine, a nearest neighbor algorithm, a K-means clustering algorithm, a decision tree model method, a naive Bayes method, a random forest method, a dimensionality reduction algorithm, and a gradient enhancement algorithm.

[0029] Preferably, the lithium battery includes an aqueous lithium battery and a non-aqueous lithium battery.

[0030] In a second aspect, the present invention provides a lithium battery cross-scale simulation device for implementing the method described in the first aspect, wherein the device comprises an acquisition module, a training module and a simulation module;

[0031] The acquisition module is used to acquire the basic physical and chemical parameters of the lithium battery microscopic object;

[0032] The training module is used to train the multiple groups of basic physicochemical parameters to obtain a machine learning molecular dynamics proxy model, wherein the machine learning molecular dynamics proxy model is obtained by training with training samples obtained based on a high-throughput computing method;

[0033] The simulation module is used to perform computational simulation on a mesoscopic object composed of several lithium battery microscopic objects through a parallel computing method, obtain key physicochemical parameters of the overall boundary of the mesoscopic object, and combine the obtained key physicochemical parameters with a macroscopic electrochemical simulation method to simulate the electrochemical performance and safety of the lithium battery under different operating conditions.

[0034] In a third aspect, the present invention provides a lithium battery cross-scale simulation device, the device comprising: a processor and a memory storing computer program instructions of the method described in the first aspect;

[0035] When the processor executes the computer program instructions, a lithium battery cross-scale simulation method is implemented.

[0036] In a fourth aspect, the present invention provides a computer-readable storage medium, on which computer program instructions of the method described in the first aspect are stored, and when the computer program instructions are executed by a processor, a cross-scale simulation method for lithium batteries is implemented.

[0037] In a fifth aspect, the present invention provides a computer program product. When the instructions in the computer program product are executed by a processor of an electronic device, the electronic device executes the lithium battery cross-scale simulation method described in the first aspect.

[0038] Beneficial effects achieved by the present invention:

[0039] The cross-scale simulation method, device, equipment, computer storage medium, and computer program product of lithium batteries of the present invention can obtain a machine learning molecular dynamics proxy model by learning and training the laws between the atomic arrangement mode and basic physicochemical parameters of lithium battery microscopic objects, complete the simulation of cross-scale electrochemical performance through parallel computing, and simulate the electrochemical performance and safety of lithium batteries under different working conditions by combining with macroscopic electrochemical simulation methods, thereby realizing accurate prediction of lithium battery decay characteristics and potential safety accidents. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention is further described in detail in conjunction with the following examples. It should be noted that the present invention is not limited to the following examples. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0041] Example 1

[0042] In lithium batteries, the interface film on the surface of the negative electrode (also called SEI film) has a significant impact on battery capacity decay and safety, but due to its complex structure, it is difficult to simulate and study it using first-principles methods.

[0043] The applicant introduced and designed a machine learning force field training workflow for SEI simulation between the negative electrode and the electrolyte.

[0044] Firstly, the first principles are used to simulate and calculate the different parts of SEI to obtain the basic physicochemical parameters. Then, the machine learning method is applied to learn the simulation results of the basic physicochemical parameters, complete the force field training, and generate the initial machine learning molecular dynamics model.

[0045] Based on the initial machine learning molecular dynamics model, the SEI of other parts is simulated to obtain the basic physicochemical parameters of the system; and at the same time, first-principles molecular dynamics are used for computational simulation.

[0046] The basic physicochemical parameters obtained by the above two methods are compared. If the differences between them are all less than 10%, the next stage of parallel computing research will be carried out. Otherwise, the SEI of other parts will be used to continue the first-principles calculation, generate data, and train the model until the differences between them are all less than 10%.

[0047] Specifically, this embodiment performs first-principles simulations of 500 thousand-atom-level parts on the surface SEI of lithium metal negative electrodes at 300K, uses these 500 sets of data for training, and obtains a machine learning molecular dynamics model to calculate the reaction activation energy of the other 10 random parts. By comparison, the maximum result obtained by first-principles calculation is only 8.7%.

[0048] Using high-performance computing methods, 8,000 SEI parts calculated by the machine learning molecular dynamics model were calculated in parallel. Each thousand-atom-level part is a cube with a side length of 2 nanometers. 8,000 such cube parts are combined into a large cube with a side length of 20 cubes, and the side length of the large cube is 40 nanometers.

[0049] Therefore, a high-performance computing method was used to simulate the SEI part with a side length of 40 nanometers, and the electronic conductivity, lithium ion mobility, reaction activation energy, and substance concentration of the entire part were calculated.

[0050] In addition, for this 40 nanometer cubic SEI part, the chemical reactions and material distribution in it can also be obtained through the above simulation method:

[0051] This example describes the process of decomposing the electrolyte on the surface of lithium metal to generate Li2CO3 and C2H4. The simulation results show that in the initial structure, the oxygen atoms in the electrolyte molecules will preferentially approach the lithium metal surface, and the distance between the two is about As the reaction proceeds, the distance between oxygen and lithium gradually shortens. Finally, the molecular chains between oxygen and the main electrolyte are broken, and the decomposition product CO3 2- The distance between the oxygen in the electrolyte and the surface lithium is reduced to The bond length of Li-O in Li2CO3 crystal Very close, indicating that CO3 2- It forms a Li2CO3 structure with the lithium on the surface. Similarly, we found that the reaction products in SEI are mainly LiF, Li2CO3, Li2O, CO2 and C2H4.

[0052] It can be seen that the method of the present invention can realize cross-scale high-precision simulation of complex systems in lithium batteries, and can calculate and predict chemical reactions, changes in substance concentrations, and key physicochemical properties in cross-scale systems, thereby assisting researchers in understanding the decay and failure mechanisms of lithium batteries, and thereby making the formal decay state of lithium batteries knowable and controllable, and providing early warning of potential safety risks.

Claims

1. A lithium battery cross-scale simulation method, characterized in that: The method comprises the following steps: S1: Use the first principle method to simulate and calculate the basic physical and chemical parameters of lithium battery micro-scale objects; The basic physicochemical parameters include at least one of the physicochemical parameters of the electrode material and the physicochemical parameters of the electrolyte constituting the lithium battery micro-scale object; The physicochemical parameters of the electrode material include at least one of the atomic arrangement structure, electronic structure, thermodynamic energy, molecular frontier orbital energy level, electron affinity, redox potential, theoretical specific capacity, electronic conductivity, lithium ion conductivity, magnetic permeability, specific heat capacity, hardness, Raman spectrum, infrared spectrum, and dielectric constant of the electrode material; The physicochemical parameters of the electrolyte include at least one of the geometric structure, electronic structure, thermodynamic energy, molecular frontier orbital energy level, dipole moment, electron affinity, redox potential, lithium ion conductivity, Raman spectrum, infrared spectrum, dielectric constant, viscosity, density, melting point, and boiling point of the electrolyte solvent molecules; The atomic arrangement of the microscopic object includes the element types, proportions and arrangement order of the atoms in space; S2: using a machine learning method to learn and train the law between the atomic arrangement of the microscopic object and the basic physicochemical parameters in step S1, and obtaining a machine learning molecular dynamics proxy model; S2-1: creating training samples based on basic physicochemical parameters of each different lithium battery microscopic object and atomic arrangement information of each microscopic object; S2-2: training the original machine learning molecular dynamics proxy model according to the training samples obtained in step S2-1 until the training stop condition is met, thereby obtaining the machine learning molecular dynamics proxy model; S2-2-1: inputting the training sample into the original machine learning molecular dynamics proxy model to obtain machine learning physical and chemical parameters of microscopic objects; S2-2-2: Determine the loss function value of the machine learning molecular dynamics proxy model according to the machine learning physicochemical parameters and basic physicochemical parameter information; S2-2-3: When the loss function value does not meet the training stop condition, adjust the model parameters of the machine learning molecular dynamics proxy model, and use the training samples to train the machine learning molecular dynamics proxy model with the adjusted parameters until the training stop condition is met, thereby obtaining a trained machine learning molecular dynamics proxy model; The training stop condition is: the result of the machine learning molecular dynamics proxy model differs from the result of the computational simulation using first-principles molecular dynamics by no more than 10%; S3: Using high-performance computing methods to assist, machine learning molecular dynamics proxy models of no less than 1,000 adjacent microscopic objects are parallelized to obtain key physical and chemical parameters of the overall boundaries of these objects; The adjacency mentioned above means that all objects can be adjacent to each other by being connected; The key physicochemical parameters include at least one of hardness, electronic conductivity, lithium ion mobility, reaction activation energy, and substance concentration; S4: Combine the key physicochemical parameters obtained with the macroscopic electrochemical simulation method to simulate the electrochemical performance and safety of lithium batteries under several different operating conditions.

2. The lithium battery cross-scale simulation method according to claim 1, characterized in that: In the step S1, a high-throughput calculation method is used to calculate the basic physicochemical parameters of the lithium battery microscopic object.

3. The lithium battery cross-scale simulation method according to claim 2, characterized in that: The high-throughput computational method includes at least one of density functional theory calculations, molecular dynamics simulations, and finite element calculations.

4. A lithium battery cross-scale simulation device, used to implement the method described in any one of claims 1 to 3, characterized in that: The device comprises an acquisition module, a training module and a simulation module; The acquisition module is used to acquire the basic physical and chemical parameters of the lithium battery microscopic object; The training module is used to train the multiple groups of basic physicochemical parameters to obtain a machine learning molecular dynamics proxy model, wherein the machine learning molecular dynamics proxy model is obtained by training with training samples obtained based on a high-throughput computing method; The simulation module is used to perform computational simulation on a mesoscopic object composed of several lithium battery microscopic objects through a parallel computing method, obtain key physicochemical parameters of the overall boundary of the mesoscopic object, and combine the obtained key physicochemical parameters with a macroscopic electrochemical simulation method to simulate the electrochemical performance and safety of the lithium battery under different operating conditions.

5. A lithium battery cross-scale simulation device, characterized in that: The device comprises: a processor and a memory storing computer program instructions of any one of claims 1 to 3; When the processor executes the computer program instructions, a cross-scale simulation method for a lithium battery is implemented.