Lithium battery electrolyte formula prediction platform based on large language model

Abstract

The invention discloses a lithium battery electrolyte formula prediction platform based on a large language model, and belongs to the technical field of automatic prediction of battery materials, the lithium battery electrolyte formula prediction platform comprises a data acquisition module used for acquiring property parameters of a target electrolyte; the formula prediction module is used for outputting an electrolyte formula and a theoretical viscosity value based on the first large language model, and storing the electrolyte formula into an electrolyte database; the track simulation module is used for pre-constructing a battery model and simulating an electrochemical reaction kinetic track of an electrolyte and a battery material; the formula property calculation module is used for calculating the system property of the electrolyte formula; and the formula output module is used for outputting the optimized electrolyte formula based on the second large language model according to the electrochemical reaction kinetic track of the electrolyte and the battery material and the system property of the electrolyte formula. According to the invention, the problems of difficult deployment, difficult use and inaccurate prediction result in electrolyte material research and development in the prior art are solved.

Description

Technical Field

[0001] The invention relates to a lithium battery electrolyte formula prediction platform based on a large language model, belonging to the technical field of battery material automatic prediction. Background Art

[0002] With the growing global demand for high-performance energy storage devices, especially the rapid development of new energy vehicles, the development of lithium batteries with higher energy density has become a research hotspot. However, the development of high-energy-density lithium batteries faces many technical challenges, among which the performance optimization of electrolyte materials is a key link. Traditional electrolyte material research and development mainly relies on trial and error, screening and optimizing the performance of electrolytes through a large number of experiments. This method is not only time-consuming and costly, but also difficult to fully and deeply understand the interaction mechanism between electrolytes and electrode materials. In recent years, the application of artificial intelligence technology in scientific research has become increasingly extensive, providing new ideas for the development of electrolyte materials. Artificial intelligence technology can quickly predict and optimize the performance of electrolyte materials by analyzing and learning a large amount of data, thereby greatly shortening the R&D cycle and reducing costs. At present, some artificial intelligence-based battery material property prediction models and active generation models have been developed, but these models are still difficult to deploy and use in practical applications, and require professional technicians to operate and maintain.

[0003] Although artificial intelligence technology has shown great potential in the research and development of battery materials, there are still some obvious problems with existing technologies. First, many advanced artificial intelligence models and tools are difficult to be adopted by a wide range of scientific research teams and industries due to their complexity and high technical barriers. This has led to the inability to fully utilize the advantages of artificial intelligence technology and limited its application scope in the research and development of electrolyte materials. Secondly, when predicting the performance of electrolyte materials, existing artificial intelligence models often ignore the interaction mechanism between electrolytes and electrode materials, resulting in inaccurate prediction results. Summary of the invention

[0004] The purpose of the present invention is to provide a lithium battery electrolyte formula prediction platform based on a large language model. Through the trained large language model, rapid screening and optimization of lithium battery electrolyte materials can be achieved, and the problems of difficult deployment, difficult use and inaccurate prediction results faced by the prior art in the research and development of electrolyte materials can be solved, thereby achieving rapid screening and optimization of electrolyte materials, reducing research and development costs and improving research and development efficiency.

[0005] In order to solve the above technical problems, the present invention is implemented by adopting the following technical solutions.

[0006] The present invention provides a lithium battery electrolyte formula prediction platform based on a large language model, comprising: A data acquisition module, used to obtain the property parameters of the target lithium battery electrolyte; A formula prediction module is used to predict the properties of the target lithium battery electrolyte based on the trained first language model, and output the electrolyte formula and the theoretical viscosity value of the electrolyte formula; when the difference between the theoretical viscosity value of the electrolyte formula and the actual viscosity value of the electrolyte formula is within a preset threshold range, the electrolyte formula is stored in the electrolyte database; The trajectory simulation module is used to pre-build a battery model based on the electrolyte formula that meets the battery model viscosity threshold in the electrolyte database, and simulate the electrochemical reaction kinetic trajectory of the electrolyte and the battery material; A formula property calculation module, used to calculate the system properties of the electrolyte formula that meets the battery model viscosity threshold; The formula output module is used to predict and output an optimized electrolyte formula based on the trained second language model according to the electrochemical reaction kinetic trajectory of the electrolyte and the battery material and the system properties of the electrolyte formula that meets the battery model viscosity threshold.

[0007] Furthermore, the training method of the first language model includes: Pre-training the first language model to generate a molecular structure and a molecular ratio as an electrolyte formula according to the element composition category and additive type of a given electrolyte; Introducing chemical rule constraints, and performing domain adaptation training on the pre-trained first language model; According to the electrolyte formula and the viscosity range of the electrolyte, an electrolyte formula-viscosity association database is established as a supervisory signal to guide the training process; During the training process, the reinforcement learning strategy cross entropy loss function is used to calculate the sequence generation loss and property prediction loss, and the viscosity value reward feedback is given to the electrolyte formula according to the supervision signal to guide the first language model to consider the viscosity range of the electrolyte when generating molecular structure and molecular ratio, and output the electrolyte formula and the theoretical viscosity value of the electrolyte formula; The sequence generation loss is used to measure the rationality of the generated molecular structure and molecular ratio; The property prediction loss is used to measure the difference between the theoretical viscosity value of the electrolyte formulation and the actual viscosity value of the electrolyte formulation.

[0008] Furthermore, the first language model is pre-trained to generate a molecular structure and a molecular ratio as an electrolyte formula according to the element composition category and additive type of a given electrolyte, including: Obtain the SMILES string corresponding to the electrolyte and the molecular structure and molecular ratio corresponding to the SMILES string from the professional database in the electrolyte field; The SMILES strings corresponding to the electrolyte are annotated according to the molecular structure and molecular ratio corresponding to the SMILES strings, and the Prompt dataset is constructed; The first language model is pre-trained using the Prompt dataset to generate a molecular structure and a molecular ratio as an electrolyte formula according to the element composition category and additive type of a given electrolyte.

[0009] Furthermore, the actual viscosity value of the electrolyte formula is calculated by molecular dynamics simulation, including: According to the electrolyte formula and the theoretical viscosity value of the electrolyte formula, an initial simulation box is constructed, and solvents, cations and anions are added according to the molecular structure and molecular ratio of the electrolyte formula and the theoretical viscosity value of the electrolyte formula; Energy minimization methods were used to optimize the molecular structures and interactions in the initial simulation box; Set the NVT (constant number, volume and temperature) simulation parameters to pre-balance the initial simulation box and generate the initial conformation; According to the energy, temperature, pressure and density of the initial simulation box, determine whether the initial simulation box has reached a stable state; If the initial simulation box does not reach a stable state, adjust the NVT simulation parameters or extend the equilibrium time until the initial simulation box is stable, and record and output the conformation in the stable state as the simulated electrolyte formula; The actual viscosity of the electrolyte formula is calculated using the liquid shear viscosity formula according to the acceleration amplitude and fitting speed of the conformation in the steady state, the density of the initial simulation box, and the length in the Z-axis direction.

[0010] Further, a battery model is pre-built according to an electrolyte formula in an electrolyte database that meets a viscosity threshold of the battery model, including: constructing an electrolyte according to an electrolyte formula that meets a viscosity threshold of a battery model; Define the initial parameters of the battery model, the electrode material lattice structure, the electrolyte wetting layer, and the multi-scale model of the SEI (Solid Electrolyte Interface) membrane. The initial parameters include the molecular structure and molecular ratio of the electrolyte, the type of electrode, and the size of the initial simulation box. For known molecular structures in the electrolyte, the stable conformations of the known molecular structures are obtained from the electrolyte molecule professional database; For unknown molecular structures in the electrolyte, the stable conformation of the unknown molecular structure is calculated using density functional theory software; According to the stable conformations of known molecular structures and the stable conformations of unknown molecular structures, the filling amount of each molecular structure in the initial simulation box is calculated; According to the filling amount of each molecular structure in the initial simulation box, it is distributed in the battery model space according to the molecular structure and proportion; Add electrode layers within the reserved electrode space to obtain a pre-built battery model.

[0011] Furthermore, the electrochemical reaction kinetics of the electrolyte and battery materials are simulated, including: mapping the molecular distribution of the electrolyte within the space of a pre-built battery model; The charge transfer process at the interface between the electrolyte and the electrode is simulated by reaction molecular dynamics, the ion migration energy barrier and interface impedance parameters are calculated, and the voltage-time curve and interface structure evolution trajectory are output as the electrochemical reaction kinetics trajectory between the electrolyte and the battery material.

[0012] Further, calculating the system properties of the electrolyte formula that meets the battery model viscosity threshold includes: The electrolyte formula that meets the viscosity threshold of the battery model is converted into the SMILES representation of the Lewis base target molecule, and the DN value of the electrolyte formula is predicted based on the trained DN value prediction model to obtain the DN value.

[0013] Furthermore, the electrolyte formula that meets the viscosity threshold of the battery model is converted into the SMILES representation of the Lewis base target molecule, including: Extract molecular structures in electrolyte formulations that meet viscosity thresholds for battery models; Using ASE (Atomic Simulation Environment) to read the SMILES string corresponding to the molecular structure; Performing a dehydrogenation treatment on the SMILES string to obtain a dehydrogenated SMILES string; ASE was used to identify Lewis base sites in the dehydrogenated SMILES string, and the Lewis base sites were spliced ​​with standard Lewis acids to generate a SMILES representation of the Lewis base target molecule.

[0014] Furthermore, the DN value prediction model is constructed based on a graph convolutional neural network, wherein the data processing method of the DN value prediction model includes: Convert the SMILES representation of the Lewis base target molecule into a graph structure representation, extract node features and edge features, the node features include atomic number, chirality, degree, formal charge, number of hydrogen atoms, number of free radical electrons, orbital hybridization, aromaticity and ring determination, and the edge features include bonding type, bond stereo configuration and conjugation; Using the input layer to receive the node features and the edge features, and output them to the graph convolution layer; Using a graph convolutional layer to weight and aggregate neighbor nodes of the node features and update the node features; The output layer is used to map the output of the graph convolutional layer to a single DN value, and the DN value of the electrolyte formula is output.

[0015] Furthermore, the training process of the second largest language model includes: Collecting electrolyte formulas that meet the battery model viscosity threshold and system properties of the electrolyte formulas; constructing an electrolyte according to the electrolyte formula, and simulating the electrochemical reaction kinetics trajectory of the electrolyte and the battery material; The electrochemical reaction kinetics trajectory of the electrolyte and the battery material, the system properties of the electrolyte formula and the electrolyte formula are used as a training set, and the second largest language model is trained to perform predictions and output an optimized electrolyte formula.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention realizes the intelligent prediction of the whole chain from inputting the property parameters of the target electrolyte to outputting the optimized electrolyte formula through the data acquisition module, the formula prediction module, the trajectory simulation module, the formula property calculation module and the formula output module. The present invention significantly improves the research and development efficiency and formula quality of the electrolyte formula, reduces the experimental cost and time cost, accelerates the research and development process of battery materials, solves the problems of difficult deployment, difficult use and inaccurate prediction results faced by the prior art in the research and development of electrolyte materials, realizes the rapid screening and optimization of electrolyte materials, reduces the research and development cost and improves the research and development efficiency.

[0017] 2. The present invention can efficiently generate a reasonable molecular structure and molecular ratio as an electrolyte formula according to the elemental composition category and additive type of the electrolyte by using the pre-trained first language model, combined with chemical rule constraints and reinforcement learning strategies. At the same time, the electrolyte formula-viscosity association database is used as a supervisory signal, and the cross entropy loss function is used to calculate the sequence generation loss and property prediction loss for training, ensuring that the predicted electrolyte formula is not only structurally reasonable, but also has a viscosity value close to the actual value, thereby greatly improving the accuracy and efficiency of the electrolyte formula prediction.

[0018] 3. The present invention constructs an initial simulation box according to the predicted electrolyte formula through molecular dynamics simulation, and finally calculates the actual viscosity value of the electrolyte after energy minimization, pre-balancing and other steps. In addition, the present invention can also pre-construct a battery model according to the electrolyte formula that meets the viscosity threshold of the battery model, simulate the electrochemical reaction kinetics trajectory of the electrolyte and the battery material, calculate the key performance indicators such as the ion migration energy barrier and the interface impedance parameters, and provide reliable data support for the further optimization of the electrolyte formula.

[0019] 4. Based on the training process of the second largest language model, the present invention can collect electrolyte formulas and electrolyte formula system properties that meet the battery model viscosity threshold, predict through the second largest language model, and output the optimized electrolyte formula, which not only realizes the intelligent optimization of the electrolyte formula, but also greatly improves the research and development efficiency and reduces the research and development cost. At the same time, the DN value prediction model constructed by the graph convolutional neural network can accurately predict the DN value and other system properties of the electrolyte formula, providing a more comprehensive evaluation index for the screening and optimization of the electrolyte formula. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic diagram of the module composition of a lithium battery electrolyte formula prediction platform based on a large language model provided by an embodiment of the present invention;

[0021] Figure 2 It is a schematic diagram of the workflow of a lithium battery electrolyte formula prediction platform based on a large language model provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0022] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention. The embodiments of the present invention and the technical features in the embodiments may be combined with each other unless there is a conflict.

[0023] Example 1

[0024] like Figure 1 As shown, this embodiment introduces a lithium battery electrolyte formula prediction platform based on a large language model, including:

[0025] The data acquisition module is used to obtain the property parameters of the target lithium battery electrolyte, including the viscosity range, element composition category and additive type.

[0026] In the present invention, the data acquisition module is responsible for collecting the property parameters of the target lithium battery electrolyte, including the viscosity range, element composition category such as the content of chemical elements and additive types such as thickeners and conductive agents.

[0027] The formula prediction module is used to predict according to the property parameters of the target lithium battery electrolyte based on the trained first language model, and output the electrolyte formula and the theoretical viscosity value of the electrolyte formula, wherein the electrolyte formula includes a molecular structure and a molecular ratio; when the difference between the theoretical viscosity value of the electrolyte formula and the actual viscosity value of the electrolyte formula is within a preset threshold range, the electrolyte formula is stored in the electrolyte database.

[0028] In the present invention, the formula prediction module predicts the electrolyte formula including the molecular structure and molecular ratio and the theoretical viscosity value of the electrolyte formula based on the property parameters of the target lithium battery electrolyte and the trained first language model through its powerful data processing and pattern recognition capabilities. The output electrolyte formula may be one group or multiple groups.

[0029] When the difference between the theoretical viscosity value of the electrolyte formula and the actual viscosity value of the electrolyte formula is within a preset threshold range, it indicates that the electrolyte formula has good prediction accuracy, so it is stored in the electrolyte database to accumulate reliable electrolyte formula data.

[0030] The trajectory simulation module is used to pre-build a battery model based on the electrolyte formula in the electrolyte database that meets the viscosity threshold of the battery model, and simulate the electrochemical reaction kinetic trajectory of the electrolyte and the battery material.

[0031] In the present invention, the trajectory simulation module uses the electrolyte formula in the electrolyte database that meets the battery model viscosity threshold to pre-construct the battery model, and simulates the electrochemical reaction kinetic trajectory between the electrolyte and the battery material. It can evaluate the actual performance of the electrolyte constructed according to the electrolyte formula in the battery, including its charge and discharge performance, cycle stability, etc.

[0032] The formula property calculation module is used to calculate the system properties of the electrolyte formula that meets the battery model viscosity threshold.

[0033] For electrolyte formulas that meet the viscosity threshold of the battery model, key properties such as DN value, viscosity, diffusion coefficient, etc. of the electrolyte constructed according to the electrolyte formula are calculated to evaluate important indicators of electrolyte performance.

[0034] The formula output module is used to predict and output an optimized electrolyte formula based on the trained second language model according to the electrochemical reaction kinetic trajectory of the electrolyte and the battery material and the system properties of the electrolyte formula that meets the battery model viscosity threshold.

[0035] In the present invention, the formula output module is used to perform a comprehensive analysis based on the trained second language model according to the electrochemical reaction kinetic trajectory provided by the trajectory simulation module and the system properties provided by the formula property calculation module, to predict and output the optimized electrolyte formula, and to provide the final electrolyte formula optimization result, thereby providing a practical guidance program for battery manufacturers and R&D personnel.

[0036] Example 2

[0037] Based on the same inventive concept as in Example 1, Figure 2 As shown, this embodiment introduces a workflow of a lithium battery electrolyte formula prediction platform based on a large language model, including:

[0038] Step 1: Obtain the property parameters of the target lithium battery electrolyte, including viscosity range, element composition category and additive type.

[0039] Step 2: According to the property parameters of the target lithium battery electrolyte, prediction is performed based on the trained first language model, and the electrolyte formula and the theoretical viscosity value of the electrolyte formula are output, wherein the electrolyte formula includes a molecular structure and a molecular ratio; when the difference between the theoretical viscosity value of the electrolyte formula and the actual viscosity value of the electrolyte formula is within a preset threshold range, the electrolyte formula is stored in the electrolyte database.

[0040] In some embodiments, the training method of the first language model includes: Step 2.1: pre-train the first language model to generate a molecular structure and a molecular ratio as an electrolyte formula according to the element composition category and additive type of a given electrolyte; In this embodiment, the first language model is pre-trained to generate a molecular structure and a molecular ratio as an electrolyte formula according to the element composition category and additive type of a given electrolyte, including: Step 2.1.1: Obtain the SMILES string corresponding to the electrolyte and the molecular structure and molecular ratio corresponding to the SMILES string from the electrolyte professional database; Step 2.1.2: Label the SMILES strings corresponding to the electrolyte according to the molecular structure and molecular ratio corresponding to the SMILES strings, and construct the Prompt dataset; Step 2.1.3: Pre-training the first language model using the Prompt dataset, generating a molecular structure and a molecular ratio as an electrolyte formula according to the element composition category and additive type of a given electrolyte; Step 2.2: Introduce chemical rule constraints and perform domain adaptation training on the pre-trained first language model; Step 2.3: Based on the electrolyte formula and the viscosity range of the electrolyte, an electrolyte formula-viscosity association database is established as a supervisory signal to guide the training process.

[0041] During the training process, the cross entropy loss function of the reinforcement learning strategy is adopted to calculate the sequence generation loss and the property prediction loss. The viscosity value reward feedback is given to the electrolyte formula according to the supervision signal to guide the first language model to consider the viscosity range of the electrolyte when generating molecular structure and molecular ratio, and output the electrolyte formula and the theoretical viscosity value of the electrolyte formula.

[0042] In this embodiment, the sequence generation loss is used to measure the rationality of the generated molecular structure and molecular ratio.

[0043] In this embodiment, the property prediction loss is used to measure the difference between the theoretical viscosity value of the electrolyte formula and the actual viscosity value of the electrolyte formula.

[0044] In some embodiments, the actual viscosity value of the electrolyte formulation is calculated by molecular dynamics simulation, including: According to the electrolyte formula and the theoretical viscosity value of the electrolyte formula, an initial simulation box is constructed, and solvents, cations and anions are added according to the molecular structure and molecular ratio of the electrolyte formula and the theoretical viscosity value of the electrolyte formula; Energy minimization methods were used to optimize the molecular structures and interactions in the initial simulation box; Set the NVT simulation parameters to pre-balance the initial simulation box and generate the initial conformation; According to the energy, temperature, pressure and density of the initial simulation box, determine whether the initial simulation box has reached a stable state; If the initial simulation box does not reach a stable state, adjust the NVT simulation parameters or extend the equilibrium time until the initial simulation box is stable, and record and output the conformation in the stable state as the simulated electrolyte formula; The actual viscosity of the electrolyte formula is calculated using the liquid shear viscosity formula according to the acceleration amplitude and fitting speed of the conformation in the steady state, the density of the initial simulation box, and the length in the Z-axis direction.

[0045] In this embodiment, the liquid shear viscosity formula is expressed as: ; In the formula, Indicates the actual viscosity value of the electrolyte formula, represents the acceleration amplitude of the conformation in the steady state, represents the density of the initial simulation box, Indicates the length of the initial simulation box in the Z-axis direction. represents the fitting speed of the conformation in the stable state.

[0046] Step 3: Pre-build the battery model based on the electrolyte formula in the electrolyte database that meets the battery model viscosity threshold, and simulate the electrochemical reaction kinetic trajectory of the electrolyte and battery materials.

[0047] In the study of electrolytes, theoretical calculation methods such as density functional theory (DFT) and molecular dynamics (MD) simulation play an important role. Density functional theory (DFT) calculations can reveal the interactions between electrolyte components from a quantum chemical level. For example, by calculating the binding energy between solvents and lithium ions, the differences in the solvation ability of different solvents for lithium ions can be explained. Molecular dynamics (MD) simulations focus on studying the dynamic behavior of electrolytes at the atomic and molecular levels. They can calculate the macroscopic physical and chemical properties of electrolytes, such as ion diffusion coefficients and viscosity, and reveal the relationship between these properties and microstructures.

[0048] Step 3.1: Pre-build a battery model based on an electrolyte formula in the electrolyte database that meets the battery model viscosity threshold, including: Step 3.1.1: constructing an electrolyte according to an electrolyte formula that meets the viscosity threshold of the battery model; Step 3.1.2: Define the initial parameters of the battery model, the electrode material lattice structure, the electrolyte wetting layer, and the multi-scale model of the SEI film; In this embodiment, the initial parameters include the molecular structure and molecular ratio of the electrolyte, the type of electrode, and the size of the initial simulation box; Step 3.1.3: For known molecular structures in the electrolyte, obtain stable conformations of the known molecular structures from the electrolyte molecule professional database; Step 3.1.4: For the unknown molecular structure in the electrolyte, the stable conformation of the unknown molecular structure is calculated by density functional theory software; Step 3.1.5: Calculate the filling amount of each molecular structure in the initial simulation box based on the stable conformations of the known molecular structures and the stable conformations of the unknown molecular structures; Step 3.1.6: According to the filling amount of each molecular structure in the initial simulation box, distribute them in the battery model space according to the molecular structure and proportion; Step 3.1.7: Add electrode layers within the reserved electrode space to obtain a pre-built battery model.

[0049] Step 3.2: Simulate the electrochemical reaction kinetics of the electrolyte and battery materials, including: Step 3.2.1: Mapping the molecular distribution of the electrolyte in the space of a pre-built battery model; Step 3.2.2: The charge transfer process at the interface between the electrolyte and the electrode is simulated by reaction molecular dynamics, the ion migration energy barrier and the interface impedance parameters are calculated, and the voltage-time curve and the interface structure evolution trajectory are output as the electrochemical reaction kinetic trajectory of the electrolyte and the battery material.

[0050] Step 4: Calculate the system properties of the electrolyte formula that meets the battery model viscosity threshold, including DN value, viscosity and diffusion coefficient.

[0051] In the field of electrochemistry, researchers often rely on the Gutmann donor number (DN) to interpret the Lewis basicity of chemicals as electrolytes and additives to estimate the degree of their solvation or the degree of interaction with Li+ as a Lewis acid. Therefore, the Gutmann donor number (DN) is often cited as an important reference indicator for the study of advanced lithium-oxygen and lithium-sulfur battery electrolytes. At the same time, the latest research found that the Gutmann donor number (DN) is an effective indicator for screening the main solvent and diluent of locally high-concentration electrolytes, where the diluent DN value is ≤10 and the main solvent DN value is ≥10.

[0052] In some embodiments, calculating the system properties of the electrolyte formulation that meets the battery model viscosity threshold includes: The electrolyte formula that meets the viscosity threshold of the battery model is converted into the SMILES representation of the Lewis base target molecule, and the DN value of the electrolyte formula is predicted based on the trained DN value prediction model to obtain the DN value.

[0053] Step 4.1: Convert the electrolyte formula that meets the battery model viscosity threshold into a SMILES representation of the Lewis base target molecule, including: Step 4.1.1: Extract the molecular structure of the electrolyte formula that meets the viscosity threshold of the battery model; Step 4.1.2: Use ASE to read the SMILES string corresponding to the molecular structure; Step 4.1.3: performing a dehydrogenation treatment on the SMILES string to obtain a dehydrogenated SMILES string; Step 4.1.4: Use ASE to identify the Lewis base sites in the SMILES string after dehydrogenation treatment, and splice the Lewis base sites with standard Lewis acids to generate the SMILES representation of the Lewis base target molecule.

[0054] Step 4.2: Construct the DN value prediction model based on graph convolutional neural network.

[0055] In this embodiment, the data processing method of the DN value prediction model includes: Step 4.2.1: Convert the SMILES representation of the Lewis base target molecule into a graph structure representation, extract node features and edge features, the node features include atomic number, chirality, degree, formal charge, number of hydrogen atoms, number of free radical electrons, orbital hybridization, aromaticity and ring determination, and the edge features include bonding type, bond stereo configuration and conjugation; Step 4.2.2: Using the input layer to receive the node features and the edge features, and output them to the graph convolution layer; Step 4.2.3: Use the graph convolution layer to weight and aggregate the neighbor nodes of the node features and update the node features; Step 4.2.4: Use the output layer to map the output of the graph convolutional layer to a single DN value and output the DN value of the electrolyte formula.

[0056] Step 5: According to the electrochemical reaction kinetics trajectory of the electrolyte and the battery material and the system properties of the electrolyte formula that meets the viscosity threshold of the battery model, prediction is performed based on the trained second largest language model to output the optimized electrolyte formula.

[0057] In this embodiment, the training process of the second largest language model includes: Step 5.1: Collecting electrolyte formulas that meet the viscosity threshold of the battery model and system properties of the electrolyte formulas; Step 5.2: constructing an electrolyte according to the electrolyte formula, and simulating the electrochemical reaction kinetics trajectory of the electrolyte and the battery material; Step 5.3: The electrochemical reaction kinetics trajectory of the electrolyte and the battery material, the system properties of the electrolyte formula and the electrolyte formula are used as a training set, and the second largest language model is trained to make predictions, and an optimized electrolyte formula is output.

[0058] Example 3

[0059] Based on the same inventive concept as other embodiments, this embodiment introduces a computer-readable storage medium on which computer instructions are stored. When the computer instructions are executed by a processor, the steps of the method in the above-mentioned embodiment 1 or 2 are implemented.

[0060] Example 4

[0061] Based on the same inventive concept as other embodiments, this embodiment introduces a computer program product, including computer instructions, which implement the steps of the method in the above-mentioned embodiment 1 or 2 when executed by a processor.

[0062] In summary, the present invention realizes the intelligent prediction of the whole chain from inputting the property parameters of the target electrolyte to outputting the optimized electrolyte formula through the data acquisition module, the formula prediction module, the trajectory simulation module, the formula property calculation module and the formula output module. The present invention significantly improves the research and development efficiency and formula quality of the electrolyte formula, reduces the experimental cost and time cost, accelerates the research and development process of battery materials, solves the problems of difficult deployment, difficult use and inaccurate prediction results faced by the prior art in the research and development of electrolyte materials, realizes the rapid screening and optimization of electrolyte materials, reduces the research and development cost and improves the research and development efficiency.

[0063] The present invention can efficiently generate a reasonable molecular structure and molecular ratio as an electrolyte formula according to the element composition category and additive type of the electrolyte by using the pre-trained first language model, combined with chemical rule constraints and reinforcement learning strategies. At the same time, the electrolyte formula-viscosity association database is used as a supervisory signal, and the cross entropy loss function is used to calculate the sequence generation loss and property prediction loss for training, ensuring that the predicted electrolyte formula is not only structurally reasonable, but also has a viscosity value close to the actual value, thereby greatly improving the accuracy and efficiency of the electrolyte formula prediction.

[0064] The present invention constructs an initial simulation box according to the predicted electrolyte formula through molecular dynamics simulation, and finally calculates the actual viscosity value of the electrolyte after energy minimization, pre-balancing and other steps. In addition, the present invention can also pre-construct a battery model according to the electrolyte formula that meets the battery model viscosity threshold, simulate the electrochemical reaction kinetics trajectory of the electrolyte and the battery material, calculate the key performance indicators such as the ion migration energy barrier and the interface impedance parameters, and provide reliable data support for the further optimization of the electrolyte formula.

[0065] Based on the training process of the second largest language model, the present invention can collect electrolyte formulas and electrolyte formula system properties that meet the battery model viscosity threshold, predict through the second largest language model, and output the optimized electrolyte formula, which not only realizes the intelligent optimization of the electrolyte formula, but also greatly improves the research and development efficiency and reduces the research and development cost. At the same time, the DN value prediction model constructed by the graph convolutional neural network can accurately predict the DN value and other system properties of the electrolyte formula, providing a more comprehensive evaluation index for the screening and optimization of the electrolyte formula.

[0066] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0067] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0068] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0069] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0070] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

Claims

1. A lithium battery electrolyte formula prediction platform based on a large language model, characterized in that: include: A data acquisition module, used to obtain the property parameters of the target lithium battery electrolyte; A formula prediction module is used to predict the properties of the target lithium battery electrolyte based on the trained first language model, and output the electrolyte formula and the theoretical viscosity value of the electrolyte formula; when the difference between the theoretical viscosity value of the electrolyte formula and the actual viscosity value of the electrolyte formula is within a preset threshold range, the electrolyte formula is stored in the electrolyte database; The trajectory simulation module is used to pre-build a battery model based on the electrolyte formula that meets the battery model viscosity threshold in the electrolyte database, and simulate the electrochemical reaction kinetic trajectory of the electrolyte and the battery material; A formula property calculation module, used to calculate the system properties of the electrolyte formula that meets the battery model viscosity threshold; The formula output module is used to predict and output an optimized electrolyte formula based on the trained second language model according to the electrochemical reaction kinetic trajectory of the electrolyte and the battery material and the system properties of the electrolyte formula that meets the battery model viscosity threshold.

2. The lithium battery electrolyte formula prediction platform based on a large language model according to claim 1, characterized in that: The training method of the first language model includes: Pre-training the first language model to generate a molecular structure and a molecular ratio as an electrolyte formula according to the element composition category and additive type of a given electrolyte; Introducing chemical rule constraints, and performing domain adaptation training on the pre-trained first language model; According to the electrolyte formula and the viscosity range of the electrolyte, an electrolyte formula-viscosity association database is established as a supervisory signal to guide the training process; During the training process, the reinforcement learning strategy cross entropy loss function is used to calculate the sequence generation loss and property prediction loss, and the viscosity value reward feedback is given to the electrolyte formula according to the supervision signal to guide the first language model to consider the viscosity range of the electrolyte when generating molecular structure and molecular ratio, and output the electrolyte formula and the theoretical viscosity value of the electrolyte formula; The sequence generation loss is used to measure the rationality of the generated molecular structure and molecular ratio; The property prediction loss is used to measure the difference between the theoretical viscosity value of the electrolyte formulation and the actual viscosity value of the electrolyte formulation.

3. The lithium battery electrolyte formula prediction platform based on a large language model according to claim 2, characterized in that: The first language model is pre-trained to generate molecular structures and molecular ratios as electrolyte formulas according to the element composition category and additive type of a given electrolyte, including: Obtain the SMILES string corresponding to the electrolyte and the molecular structure and molecular ratio corresponding to the SMILES string from the professional database in the electrolyte field; The SMILES strings corresponding to the electrolyte are annotated according to the molecular structure and molecular ratio corresponding to the SMILES strings, and the Prompt dataset is constructed; The first language model is pre-trained using the Prompt dataset to generate a molecular structure and a molecular ratio as an electrolyte formula according to the element composition category and additive type of a given electrolyte.

4. The lithium battery electrolyte formula prediction platform based on a large language model according to claim 1, characterized in that: The actual viscosity of the electrolyte formula is calculated by molecular dynamics simulation, including: According to the electrolyte formula and the theoretical viscosity value of the electrolyte formula, an initial simulation box is constructed, and solvents, cations and anions are added according to the molecular structure and molecular ratio of the electrolyte formula and the theoretical viscosity value of the electrolyte formula; Energy minimization methods were used to optimize the molecular structures and interactions in the initial simulation box; Set the NVT simulation parameters to pre-balance the initial simulation box and generate the initial conformation; According to the energy, temperature, pressure and density of the initial simulation box, determine whether the initial simulation box has reached a stable state; If the initial simulation box does not reach a stable state, adjust the NVT simulation parameters or extend the equilibrium time until the initial simulation box is stable, and record and output the conformation in the stable state as the simulated electrolyte formula; The actual viscosity of the electrolyte formula is calculated using the liquid shear viscosity formula according to the acceleration amplitude and fitting speed of the conformation in the steady state, the density of the initial simulation box, and the length in the Z-axis direction.

5. The lithium battery electrolyte formula prediction platform based on a large language model according to claim 1, characterized in that: Pre-build battery models based on electrolyte recipes in the electrolyte database that meet the battery model viscosity threshold, including: constructing an electrolyte according to an electrolyte formula that meets a viscosity threshold of a battery model; Defining the initial parameters of the battery model, the electrode material lattice structure, the electrolyte wetting layer and the multi-scale model of the SEI film, wherein the initial parameters include the molecular structure and molecular ratio of the electrolyte, the type of electrode and the size of the initial simulation box; For known molecular structures in the electrolyte, the stable conformations of the known molecular structures are obtained from the electrolyte molecule professional database; For unknown molecular structures in the electrolyte, the stable conformation of the unknown molecular structure is calculated using density functional theory software; According to the stable conformations of known molecular structures and the stable conformations of unknown molecular structures, the filling amount of each molecular structure in the initial simulation box is calculated; According to the filling amount of each molecular structure in the initial simulation box, it is distributed in the battery model space according to the molecular structure and proportion; Add electrode layers within the reserved electrode space to obtain a pre-built battery model.

6. The lithium battery electrolyte formula prediction platform based on a large language model according to claim 5, characterized in that: Simulate the electrochemical reaction kinetics of electrolytes and battery materials, including: mapping the molecular distribution of the electrolyte within the space of a pre-built battery model; The charge transfer process at the interface between the electrolyte and the electrode is simulated by reaction molecular dynamics, the ion migration energy barrier and interface impedance parameters are calculated, and the voltage-time curve and interface structure evolution trajectory are output as the electrochemical reaction kinetics trajectory between the electrolyte and the battery material.

7. The lithium battery electrolyte formula prediction platform based on a large language model according to claim 6, characterized in that: Calculating the system properties of the electrolyte formulation that meets the battery model viscosity threshold includes: The electrolyte formula that meets the viscosity threshold of the battery model is converted into the SMILES representation of the Lewis base target molecule, and the DN value of the electrolyte formula is predicted based on the trained DN value prediction model to obtain the DN value.

8. The lithium battery electrolyte formula prediction platform based on a large language model according to claim 7, characterized in that: Convert electrolyte formulations that meet the viscosity threshold of the battery model into SMILES representations of Lewis base target molecules, including: Extract molecular structures in electrolyte formulations that meet viscosity thresholds for battery models; Using ASE to read the SMILES string corresponding to the molecular structure; Performing a dehydrogenation treatment on the SMILES string to obtain a dehydrogenated SMILES string; ASE was used to identify Lewis base sites in the dehydrogenated SMILES string, and the Lewis base sites were spliced ​​with standard Lewis acids to generate a SMILES representation of the Lewis base target molecule.

9. The lithium battery electrolyte formula prediction platform based on a large language model according to claim 7, characterized in that: The DN value prediction model is constructed based on a graph convolutional neural network, wherein the data processing method of the DN value prediction model includes: Convert the SMILES representation of the Lewis base target molecule into a graph structure representation, extract node features and edge features, the node features include atomic number, chirality, degree, formal charge, number of hydrogen atoms, number of free radical electrons, orbital hybridization, aromaticity and ring determination, and the edge features include bonding type, bond stereo configuration and conjugation; Using the input layer to receive the node features and the edge features, and output them to the graph convolution layer; Using a graph convolutional layer to weight and aggregate neighbor nodes of the node features and update the node features; The output layer is used to map the output of the graph convolutional layer to a single DN value, and the DN value of the electrolyte formula is output.

10. The lithium battery electrolyte formula prediction platform based on a large language model according to claim 1, characterized in that: The training process of the second largest language model includes: Collecting electrolyte formulas that meet the battery model viscosity threshold and system properties of the electrolyte formulas; constructing an electrolyte according to the electrolyte formula, and simulating the electrochemical reaction kinetics trajectory of the electrolyte and the battery material; The electrochemical reaction kinetics trajectory of the electrolyte and the battery material, the system properties of the electrolyte formula and the electrolyte formula are used as a training set, and the second largest language model is trained to perform predictions and output an optimized electrolyte formula.

Images