Method and device for predicting properties of electrolyte for lithium batteries based on deep learning

Abstract

A method and device for predicting the properties of an electrolyte for a lithium battery based on deep learning, relating to the technical field of molecular property prediction, wherein the method collects a number of electrolyte molecules to form an electrolyte molecule database; performs high-throughput calculations through a first-principle calculation strategy to obtain set property data of electrolyte molecules; encodes and classifies the set property data of electrolyte molecules; divides the set property data of the coded and classified electrolyte molecules into Atom point level features and Edge key level features; constructs an electrolyte molecule property prediction model based on a GCN model and a feedforward neural network FFN, and trains it; inputs the Atom point level features and Edge key level features of the electrolyte molecules into the trained electrolyte molecule property prediction model to predict the properties of the electrolyte molecules. The present invention can efficiently and accurately predict the key physical properties of the electrolyte, and provide guidance for the optimal design of the electrolyte.

Description

Technical Field

[0001] The present invention relates to the technical field of molecular property prediction, and in particular to a method and device for predicting the properties of an electrolyte for a lithium battery based on deep learning. Background Art

[0002] Building an energy system dominated by renewable energy is a key path to achieving sustainable development. Among them, energy storage devices represented by secondary batteries play a vital role. Among the various components of secondary batteries, the electrolyte directly determines the safety, stability, fast charging performance and cycle life of the battery, so it has long been widely concerned and studied in depth.

[0003] First-principles calculation methods can reveal microscopic mechanisms and explain macroscopic phenomena at the electronic, atomic and molecular levels. Deep learning technology has the ability to deeply mine information from experimental and computational simulation results. These two technologies are gradually becoming important tools in electrolyte research. By applying first-principles calculations to study the effects of changes in electrolyte composition and electrode properties on the interactions between components in the electrolyte, researchers can correlate these microscopic interactions with macroscopic physical and chemical properties such as ionic conductivity and chemical stability, thereby guiding the design of favorable electrolyte structures.

[0004] First-principles calculations can not only provide information for analyzing the 3D structure of the electrolyte, but also include information such as atomic charge, system temperature, pressure and energy. By statistically analyzing this information, the theoretical model can derive important physicochemical properties of the electrolyte, such as melting and boiling points, dielectric constants and kinetic properties. On the one hand, the first-principles calculation method helps to deeply understand these macroscopic properties at the atomic and molecular levels, and on the other hand, it supports high-throughput calculations.

[0005] However, pure first-principles calculation methods require high computing resources and lack a theoretical design foundation. If combined with emerging machine learning methods, efficient prediction of electrolyte properties can be performed, promoting a deeper understanding of chemical mechanisms and accelerating the development of high-performance electrolyte systems.

[0006] Therefore, how to invent a method for predicting electrolyte properties based on deep learning has become an urgent problem to be solved. Summary of the invention

[0007] To this end, the present invention provides a method and device for predicting the properties of lithium battery electrolytes based on deep learning, which can efficiently and accurately predict the key physical properties of the electrolyte and provide guidance for the optimal design of the electrolyte. This method not only improves the efficiency and accuracy of the research, but also provides new opportunities and broad prospects for the development of future energy storage technologies.

[0008] In order to achieve the above object, the present invention provides the following technical solution: a method for predicting the properties of lithium battery electrolyte based on deep learning, comprising:

[0009] Collecting a number of electrolyte molecules to form an electrolyte molecule database; performing high-throughput calculations on the electrolyte molecules in the electrolyte molecule database using a first-principle calculation strategy to obtain set property data of the electrolyte molecules;

[0010] According to the set property classification, the set property data of the electrolyte molecules are coded and classified to obtain the set property data of the electrolyte molecules after coding and classification;

[0011] A graph neural network is used to characterize the electrolyte molecules, and the set property data of the electrolyte molecules after encoding and classification are divided into atomic point level features and edge key level features of the electrolyte molecules;

[0012] According to the GCN model and the feedforward neural network FFN, a prediction model for electrolyte molecular properties is constructed; the electrolyte molecular property prediction model is trained through the electrolyte molecular database to obtain the trained prediction model for electrolyte molecular properties;

[0013] The Atom point level features and Edge key level features of the electrolyte molecules are input into the trained electrolyte molecule property prediction model, and the properties of the electrolyte molecules are obtained through the prediction of the trained electrolyte molecule property prediction model.

[0014] As a preferred solution of the method for predicting the properties of lithium battery electrolyte based on deep learning, the set property data of the electrolyte molecules include: ground state structure, atomic charge, Fukui function, vibration entropy, rotation entropy and molecular point group.

[0015] As a preferred scheme of the method for predicting the properties of lithium battery electrolytes based on deep learning, in the process of encoding and classifying the set property data of the electrolyte molecules according to the set property classification, the atomic coordinates and atomic charge structure properties in the ground state structure of the electrolyte molecules are encoded as point properties; the bond length and bond order properties in the ground state structure of the electrolyte molecules are encoded as bond properties; and the Fukui function descriptor of the electrolyte molecules is encoded as molecular properties.

[0016] As a preferred scheme of the method for predicting the properties of electrolytes for lithium batteries based on deep learning, in the process of dividing the set property data of the electrolyte molecules after encoding and classification into Atom point level features and Edge bond level features of the electrolyte molecules, the point properties and the molecular properties of the electrolyte molecules are used as Atom point level features of the electrolyte molecules; and the bond properties of the electrolyte molecules are used as Edge bond level features of the electrolyte molecules.

[0017] As a preferred solution of the method for predicting the properties of electrolytes for lithium batteries based on deep learning, in the process of training the electrolyte molecular properties prediction model through the electrolyte molecular database, the optimizer used by the electrolyte molecular properties prediction model is the adaptively adjusted Adam; the loss function is the mse loss function; the batch_size is 64; the dropout is 0.1, and the activation function is relu.

[0018] The present invention also provides a device for predicting the properties of a lithium battery electrolyte based on deep learning, based on the above method for predicting the properties of a lithium battery electrolyte based on deep learning, comprising:

[0019] The electrolyte molecule property data acquisition module is used to collect a number of electrolyte molecules to form an electrolyte molecule database; perform high-throughput calculations on the electrolyte molecules in the electrolyte molecule database through a first-principle calculation strategy to obtain set property data of the electrolyte molecules;

[0020] A molecular property coding classification module is used to code and classify the set property data of the electrolyte molecules according to the set property classification, and obtain the set property data of the electrolyte molecules after coding and classification;

[0021] An electrolyte molecule characterization processing module, used for characterizing the electrolyte molecules using a graph neural network, and dividing the set property data of the electrolyte molecules after encoding and classification into Atom point level features and Edge key level features of the electrolyte molecules;

[0022] The electrolyte molecular property prediction model construction and training module is used to construct the electrolyte molecular property prediction model according to the GCN model and the feedforward neural network FFN; the electrolyte molecular property prediction model is trained through the electrolyte molecular database to obtain the trained electrolyte molecular property prediction model;

[0023] The electrolyte molecule property prediction module is used to input the Atom point level features and Edge key level features of the electrolyte molecule into the trained electrolyte molecule property prediction model, and obtain the properties of the electrolyte molecule through the prediction of the trained electrolyte molecule property prediction model.

[0024] As a preferred solution of the device for predicting the properties of lithium battery electrolytes based on deep learning, in the electrolyte molecule property data acquisition module, the set property data of the electrolyte molecules include: ground state structure, atomic charge, Fukui function, vibration entropy, rotational entropy and molecular point group.

[0025] As a preferred solution of the device for predicting the properties of lithium battery electrolytes based on deep learning, in the molecular property coding and classification module, in the process of encoding and classifying the set property data of the electrolyte molecules according to the set property classification, the atomic coordinates and atomic charge structure properties in the ground state structure of the electrolyte molecules are encoded as point properties; the bond length and bond order properties in the ground state structure of the electrolyte molecules are encoded as bond properties; and the Fukui function descriptor of the electrolyte molecules is encoded as molecular properties.

[0026] As a preferred solution of the electrolyte property prediction device for lithium batteries based on deep learning, in the electrolyte molecule characterization processing module, in the process of dividing the set property data of the electrolyte molecules after encoding and classification into Atom point level features and Edge bond level features of the electrolyte molecules, the point properties and the molecular properties of the electrolyte molecules are used as Atom point level features of the electrolyte molecules; and the bond properties of the electrolyte molecules are used as Edge bond level features of the electrolyte molecules.

[0027] As a preferred solution of the device for predicting the properties of electrolytes for lithium batteries based on deep learning, in the electrolyte molecular property prediction model construction and training module, in the process of training the electrolyte molecular property prediction model through the electrolyte molecule database, the optimizer used by the electrolyte molecular property prediction model is the adaptively adjusted Adam; the loss function is the mse loss function; the batch_size is 64; the dropout is 0.1, and the activation function is relu.

[0028] The present invention has the following advantages: a plurality of electrolyte molecules are collected to form an electrolyte molecule database; high-throughput calculations are performed on the electrolyte molecules in the electrolyte molecule database by a first-principle calculation strategy to obtain set property data of the electrolyte molecules; the set property data of the electrolyte molecules are encoded and classified according to the set property classification to obtain the set property data of the electrolyte molecules after encoding and classification; the electrolyte molecules are characterized by using a graph neural network to divide the set property data of the electrolyte molecules after encoding and classification into Atom point level features and Edge key level features of the electrolyte molecules; an electrolyte molecule property prediction model is constructed according to a GCN model and a feedforward neural network FFN; the electrolyte molecule property prediction model is trained by the electrolyte molecule database to obtain a trained electrolyte molecule property prediction model; the Atom point level features and Edge key level features of the electrolyte molecules are input into the trained electrolyte molecule property prediction model, and the properties of the electrolyte molecules are obtained through prediction by the trained electrolyte molecule property prediction model. The present invention can efficiently and accurately predict the key physical properties of the electrolyte and provide guidance for the optimal design of the electrolyte. At the same time, it not only improves the efficiency and accuracy of the research, but also provides new opportunities and broad prospects for the development of future energy storage technology. By combining advanced computing technology with deep learning, the present invention has opened up a new path for improving battery performance and material innovation. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0030] The structures, proportions, sizes, etc. illustrated in this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with the technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantial technical significance. Any structural modification, change in proportion or adjustment of size shall still fall within the scope of the technical contents disclosed in the present invention without affecting the effects and purposes that can be achieved by the present invention.

[0031] Figure 1 This is a schematic flow chart of a method for predicting properties of an electrolyte for a lithium battery based on deep learning provided in Example 1 of the present invention;

[0032] Figure 2This is a schematic diagram of the electrolyte molecular property prediction model architecture in the method for predicting electrolyte properties for lithium batteries based on deep learning provided in Example 1 of the present invention;

[0033] Figure 3 A schematic diagram of comparison and fitting of the actual value and predicted value of the vibration entropy y1 of the electrolyte property in a possible embodiment provided in Embodiment 1 of the present invention;

[0034] Figure 4 A schematic diagram showing a comparison and fitting of the actual value and the predicted value of the rotational entropy y2 of the electrolyte property in a possible embodiment provided in Example 1 of the present invention;

[0035] Figure 5 Schematic diagram of the architecture of the device for predicting the properties of the electrolyte for lithium batteries based on deep learning provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0036] The following is a description of the implementation of the present invention by specific embodiments. People familiar with the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] Example 1

[0038] See also Figure 1 Embodiment 1 of the present invention provides a method for predicting the properties of a lithium battery electrolyte based on deep learning, comprising the following steps:

[0039] S1. Collect a number of electrolyte molecules to form an electrolyte molecule database; perform high-throughput calculations on the electrolyte molecules in the electrolyte molecule database using a first-principle calculation strategy to obtain set property data of the electrolyte molecules;

[0040] S2. According to the set property classification, the set property data of the electrolyte molecules are coded and classified to obtain the set property data of the electrolyte molecules after coding and classification;

[0041] S3, using a graph neural network to characterize the electrolyte molecules, and dividing the set property data of the electrolyte molecules after encoding and classification into Atom point level features and Edge key level features of the electrolyte molecules;

[0042] S4. Constructing an electrolyte molecular property prediction model according to the GCN model and the feed-forward neural network FFN; training the electrolyte molecular property prediction model through the electrolyte molecular database to obtain the trained electrolyte molecular property prediction model;

[0043] S5. Inputting the Atom-level features and Edge-key-level features of the electrolyte molecules into the trained electrolyte molecule property prediction model, and obtaining the properties of the electrolyte molecules through prediction by the trained electrolyte molecule property prediction model.

[0044] In this embodiment, in step S1, a number of electrolyte molecules are collected to form an electrolyte molecule database; a high-throughput calculation is performed on the electrolyte molecules in the electrolyte molecule database using a first-principles calculation strategy to obtain set property data of the electrolyte molecules;

[0045] The set property data of the electrolyte molecules include: ground state structure, atomic charge, Fukui function, vibration entropy, rotation entropy and molecular point group.

[0046] Specifically, the ground state structure refers to the atomic arrangement of a molecule or material in the lowest energy (most stable) state, which is usually calculated using density functional theory (DFT).

[0047] The formula for calculating atomic charge is:

[0048]

[0049] In the formula, Q A is the atomic charge; Z A is the atomic number, indicating the number of protons in an atom; μ and ν are different atomic orbitals; A is the atomic substitution representation, referring to a certain atom in the system; P μν is the density matrix element, representing the electron density distribution between μ and ν atomic orbitals; S μν is the overlap matrix element, representing the spatial overlap between μ and ν atomic orbitals.

[0050] The calculation formula of Fukui function is:

[0051]

[0052] Where N is the total number of electrons in the molecule; v(r) is the external potential at position r; δ is a possible minimum value, N+δ and N-δ represent a small positive and negative range near N respectively; ρ(r) is the electron density obtained from the ground state configuration, and the electrophilic and nucleophilic functions are calculated respectively.

[0053] The derived formulas for vibrational and rotational entropy are:

[0054]

[0055] In the formula, S vib is the vibration entropy; h is Planck's constant; ν i The frequency of the i-th vibration mode, k B is the Boltzmann constant, and T is the temperature.

[0056] The molecular point group is obtained by calculating the rotation axis, symmetry plane and inversion center of the 3D structure, and outputting the symmetry center, which is compared with the standard point group table to determine the specific type of the molecular point group.

[0057] In this embodiment, in step S2, the set property data of the electrolyte molecules are coded and classified according to the set property classification to obtain the set property data of the electrolyte molecules after coding and classification;

[0058] Specifically, the atomic coordinates and atomic charge structure properties in the ground state structure of the electrolyte molecule are encoded as point properties; the bond length and bond order properties in the ground state structure of the electrolyte molecule are encoded as bond properties; and the Fukui function descriptor of the electrolyte molecule is encoded as molecular properties.

[0059] In this embodiment, in step S3, a graph neural network is used to characterize the electrolyte molecules, and the set property data of the electrolyte molecules after encoding and classification are divided into Atom point level features and Edge key level features of the electrolyte molecules;

[0060] Specifically, since molecules have a natural graph attribute architecture, using graph neural networks to characterize molecules has a natural advantage. The molecular graph representation is divided into two parts: Atom point level and Edge bond level. Atom level features are divided into two parts, one is the characteristics of the molecule itself, such as the type of element, the number of atoms, etc., and the other is the quantum computing descriptor, such as: Fukui function, etc. The two features need to be spliced ​​before being input into GCN. Edge bond level features, such as bonding types (single and double bonds) and other features.

[0061] In this embodiment, the point properties and the molecular properties of the electrolyte molecules are used as Atom point-level features of the electrolyte molecules; and the bond properties of the electrolyte molecules are used as Edge bond-level features of the electrolyte molecules.

[0062] In this embodiment, in step S4, a molecular property prediction model of the electrolyte is constructed according to the GCN model and the feed-forward neural network FFN; the molecular property prediction model of the electrolyte is trained through the electrolyte molecule database to obtain the trained molecular property prediction model of the electrolyte;

[0063] Among them, the model GCN mainly consists of two graph convolutional layers and two fully connected layers.

[0064] In this embodiment, the optimizer used in the electrolyte molecular property prediction model is the adaptively adjusted Adam; the loss function is the mse loss function; the batch_size is 64; the dropout is 0.1, and the activation function is relu.

[0065] Use PyTorch's dataset class Dataset as the parent class to construct a new dataset class DataSetOri, and encapsulate the new dataset class by rewriting the init constructor, getitem function, and len function.

[0066] In this embodiment, in step S5, the Atom point level features and Edge key level features of the electrolyte molecules are input into the trained electrolyte molecule property prediction model, and the properties of the electrolyte molecules are obtained through the prediction of the trained electrolyte molecule property prediction model.

[0067] Specifically, Figure 2 As shown, the Atom-level features and Edge-key-level features of the electrolyte molecule are input into the GCN model. The GCN model aggregates and updates the information in the molecular graph to obtain the graph representation of the molecule, and then the properties of the electrolyte molecule are finally obtained through the feedforward neural network FFN.

[0068] In a possible embodiment, a specific prediction example is provided as follows:

[0069] T1. Collect an electrolyte molecule database consisting of 7744 molecules;

[0070] T2. Calculate the ground state configuration of electrolyte molecules by first principles, and count the bond lengths and bond orders of all electrolyte molecules as edge features, and the atomic coordinates of electrolyte molecules as atomic-level features; calculate the atomic charge, Fukui function, molecular point group, vibration entropy (y1) and rotation entropy (y2) of all electrolyte molecules as molecular features and input them into the electrolyte molecular property prediction model;

[0071] T3. Randomly divide the electrolyte molecule database into a training set and a test set according to a ratio of 8:2; train the electrolyte property prediction model using the training set and the test set;

[0072] T4. Input the Edge features and Atom features of the electrolyte molecules into the trained electrolyte property prediction model to predict the kinetic properties y1 and y2 of the electrolyte molecules.

[0073] See also Figure 3It is a comparison and fitting diagram of the real value and predicted value of the vibration entropy y1 of the electrolyte property; Figure 4 The figure shows the comparison between the real value and predicted value of the rotational entropy y2 of the electrolyte property. Among them, the R^2 of y1 is 0.99; the R^2 of y2 is 0.98, indicating that the electrolyte property prediction model has a very high accuracy in describing y1 and y2. The vibration entropy (y1) and rotational entropy (y2) can fully reflect the compatibility and durability of electrolyte molecules.

[0074] In summary, the present invention collects a number of electrolyte molecules to form an electrolyte molecule database; performs high-throughput calculations on the electrolyte molecules in the electrolyte molecule database through a first-principles calculation strategy to obtain set property data of the electrolyte molecules; encodes and classifies the set property data of the electrolyte molecules according to the set property classification to obtain the set property data of the electrolyte molecules after encoding and classification; uses a graph neural network to characterize the electrolyte molecules, and divides the set property data of the electrolyte molecules after encoding and classification into Atom point level features and Edge key level features of the electrolyte molecules; constructs an electrolyte molecule property prediction model based on a GCN model and a feedforward neural network FFN; trains the electrolyte molecule property prediction model through the electrolyte molecule database to obtain a trained electrolyte molecule property prediction model; inputs the Atom point level features and Edge key level features of the electrolyte molecules into the trained electrolyte molecule property prediction model, and obtains the properties of the electrolyte molecules through the prediction of the trained electrolyte molecule property prediction model. The present invention can efficiently and accurately predict the key physical properties of the electrolyte and provide guidance for the optimal design of the electrolyte. At the same time, it not only improves the efficiency and accuracy of the research, but also provides new opportunities and broad prospects for the development of future energy storage technology. By combining advanced computing technology with deep learning, the present invention has opened up a new path for improving battery performance and material innovation.

[0075] It should be noted that the method of the embodiment of the present disclosure can be performed by a single device, such as a computer or a server. The method of the present embodiment can also be applied in a distributed scenario and completed by multiple devices cooperating with each other. In the case of such a distributed scenario, one of the multiple devices can only perform one or more steps in the method of the embodiment of the present disclosure, and the multiple devices will interact with each other to complete the described method.

[0076] It should be noted that the above describes some embodiments of the present disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the above embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0077] Example 2

[0078] See also Figure 5 Embodiment 2 of the present invention also provides a device for predicting the properties of a lithium battery electrolyte based on deep learning, comprising:

[0079] The electrolyte molecule property data acquisition module 001 is used to collect a number of electrolyte molecules to form an electrolyte molecule database; perform high-throughput calculations on the electrolyte molecules in the electrolyte molecule database using a first-principle calculation strategy to obtain set property data of the electrolyte molecules;

[0080] The molecular property coding classification module 002 is used to code and classify the set property data of the electrolyte molecules according to the set property classification, and obtain the set property data of the electrolyte molecules after coding and classification;

[0081] The electrolyte molecule characterization processing module 003 is used to characterize the electrolyte molecules using a graph neural network, and divide the set property data of the electrolyte molecules after encoding and classification into Atom point level features and Edge key level features of the electrolyte molecules;

[0082] The electrolyte molecular property prediction model construction and training module 004 is used to construct an electrolyte molecular property prediction model according to the GCN model and the feedforward neural network FFN; the electrolyte molecular property prediction model is trained through the electrolyte molecular database to obtain the trained electrolyte molecular property prediction model;

[0083] The electrolyte molecule property prediction module 005 is used to input the Atom point level features and Edge key level features of the electrolyte molecule into the trained electrolyte molecule property prediction model, and obtain the properties of the electrolyte molecule through the prediction of the trained electrolyte molecule property prediction model.

[0084] In this embodiment, in the electrolyte molecule property data acquisition module 001, the set property data of the electrolyte molecules include: ground state structure, atomic charge, Fukui function, vibration entropy, rotation entropy and molecular point group.

[0085] In this embodiment, in the molecular property coding and classification module 002, in the process of coding and classifying the set property data of the electrolyte molecules according to the set property classification, the atomic coordinates and atomic charge structure properties in the ground state structure of the electrolyte molecules are encoded as point properties; the bond length and bond order properties in the ground state structure of the electrolyte molecules are encoded as bond properties; and the Fukui function descriptor of the electrolyte molecules is encoded as molecular properties.

[0086] In this embodiment, in the electrolyte molecule characterization processing module 003, in the process of dividing the set property data of the electrolyte molecules after encoding and classification into Atom point level features and Edge bond level features of the electrolyte molecules, the point properties and the molecular properties of the electrolyte molecules are used as Atom point level features of the electrolyte molecules; and the bond properties of the electrolyte molecules are used as Edge bond level features of the electrolyte molecules.

[0087] In this embodiment, in the electrolyte molecular property prediction model construction and training module 004, in the process of training the electrolyte molecular property prediction model through the electrolyte molecular database, the optimizer used by the electrolyte molecular property prediction model is the adaptively adjusted Adam; the loss function is the mse loss function; the batch_size is 64; the dropout is 0.1, and the activation function is relu.

[0088] It should be noted that the information interaction, execution process and other contents between the modules of the above-mentioned system are based on the same concept as the method embodiment in Example 1 of the present application, and the technical effects they bring are the same as those of the method embodiment of the present application. For specific contents, please refer to the description in the method embodiment shown above in the present application, and will not be repeated here.

[0089] Example 3

[0090] Embodiment 3 of the present invention provides a non-transitory computer-readable storage medium, in which a program code for a method for predicting the properties of an electrolyte for a lithium battery based on deep learning is stored, and the program code includes instructions for executing the method for predicting the properties of an electrolyte for a lithium battery based on deep learning of Embodiment 1 or any possible implementation thereof.

[0091] The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that includes one or more available media. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0092] Example 4

[0093] Embodiment 4 of the present invention provides an electronic device, including: a memory and a processor;

[0094] The processor and the memory communicate with each other via a bus; the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the method for predicting the properties of electrolyte for lithium batteries based on deep learning of Example 1 or any possible implementation thereof.

[0095] Specifically, the processor can be implemented by hardware or by software. When implemented by hardware, the processor can be a logic circuit, an integrated circuit, etc.; when implemented by software, the processor can be a general-purpose processor implemented by reading software codes stored in a memory. The memory can be integrated in the processor or can be located outside the processor and exist independently.

[0096] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable systems. The computer instructions can be stored in a computer-readable storage medium, or transmitted from a computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center.

[0097] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing system, they can be concentrated on a single computing system, or distributed on a network composed of multiple computing systems, and optionally, they can be implemented by a program code executable by a computing system, so that they can be stored in a storage system and executed by the computing system, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0098] Although the present invention has been described in detail above by general description and specific embodiments, it is obvious to those skilled in the art that some modifications or improvements can be made to the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention all belong to the scope of protection claimed by the present invention.

Claims

1. A method for predicting the properties of lithium battery electrolyte based on deep learning, characterized in that: include: Collect a number of electrolyte molecules to form an electrolyte molecule database; Performing high-throughput calculations on the electrolyte molecules in the electrolyte molecule database by using a first-principle calculation strategy to obtain set property data of the electrolyte molecules; According to the set property classification, the set property data of the electrolyte molecules are coded and classified to obtain the set property data of the electrolyte molecules after coding and classification; A graph neural network is used to characterize the electrolyte molecules, and the set property data of the electrolyte molecules after encoding and classification are divided into atomic point level features and edge key level features of the electrolyte molecules; According to the GCN model and the feedforward neural network FFN, a prediction model for electrolyte molecular properties is constructed; the electrolyte molecular property prediction model is trained through the electrolyte molecular database to obtain the trained prediction model for electrolyte molecular properties; Inputting the Atom-level features and Edge-key-level features of the electrolyte molecules into the trained electrolyte molecule property prediction model, and obtaining the properties of the electrolyte molecules through prediction by the trained electrolyte molecule property prediction model; The set property data of the electrolyte molecules include: ground state structure, atomic charge, Fukui function, vibration entropy, rotation entropy and molecular point group; The formula for calculating atomic charge is: In the formula, Q A is the atomic charge; Z A is the atomic number, indicating the number of protons in an atom; μ and ν are different atomic orbitals; A is the atomic substitution representation, referring to a certain atom in the system; P μν is the density matrix element, representing the electron density distribution between μ and ν atomic orbitals; S μν is the overlap matrix element, representing the spatial overlap between μ and ν atomic orbitals; The calculation formula of Fukui function is: Where N is the total number of electrons in the molecule; v(r) is the external potential at position r; δ is a desirable minimum value, N+δ and N-δ represent a small positive and negative range near N respectively; ρ(r) is the electron density obtained from the ground state configuration, and the electrophilic and nucleophilic functions are calculated respectively; The derived formulas for vibrational and rotational entropy are: In the formula, S vib is the vibration entropy; h is Planck's constant; ν i is the frequency of the i-th vibration mode; k B is the Boltzmann constant, T is the temperature; The model GCN includes two graph convolutional layers and two fully connected layers; The optimizer used in the electrolyte molecular property prediction model is the adaptively adjusted Adam; the loss function is the mse loss function; and the activation function is relu; Use PyTorch's dataset class Dataset as the parent class to construct a new dataset class DataSetOri, and encapsulate the new dataset class by rewriting the init constructor, getitem function, and len function.

2. The method for predicting properties of lithium battery electrolyte based on deep learning according to claim 1, characterized in that: In the process of encoding and classifying the set property data of the electrolyte molecules according to the set property classification, the atomic coordinates and atomic charge structure properties in the ground state structure of the electrolyte molecules are encoded as point properties; the bond length and bond order properties in the ground state structure of the electrolyte molecules are encoded as bond properties; and the Fukui function descriptor of the electrolyte molecules is encoded as molecular properties.

3. The method for predicting properties of lithium battery electrolyte based on deep learning according to claim 2, characterized in that: In the process of dividing the set property data of the electrolyte molecules after coding and classification into the atomic point level features and the edge key level features of the electrolyte molecules, the point properties and the molecular properties of the electrolyte molecules are used as the atomic point level features of the electrolyte molecules; The bond properties of the electrolyte molecules are used as Edge bond level features of the electrolyte molecules.

4. A device for predicting the properties of lithium battery electrolyte based on deep learning, using the method for predicting the properties of lithium battery electrolyte based on deep learning according to any one of claims 1 to 3, characterized in that: include: The electrolyte molecule property data acquisition module is used to collect a number of electrolyte molecules to form an electrolyte molecule database; Performing high-throughput calculations on the electrolyte molecules in the electrolyte molecule database by using a first-principle calculation strategy to obtain set property data of the electrolyte molecules; A molecular property coding classification module is used to code and classify the set property data of the electrolyte molecules according to the set property classification, and obtain the set property data of the electrolyte molecules after coding and classification; An electrolyte molecule characterization processing module, used for characterizing the electrolyte molecules using a graph neural network, and dividing the set property data of the electrolyte molecules after encoding and classification into Atom point level features and Edge key level features of the electrolyte molecules; The electrolyte molecular property prediction model construction and training module is used to construct the electrolyte molecular property prediction model according to the GCN model and the feedforward neural network FFN; the electrolyte molecular property prediction model is trained through the electrolyte molecular database to obtain the trained electrolyte molecular property prediction model; The electrolyte molecule property prediction module is used to input the Atom point level features and Edge key level features of the electrolyte molecule into the trained electrolyte molecule property prediction model, and obtain the properties of the electrolyte molecule through the prediction of the trained electrolyte molecule property prediction model; In the electrolyte molecule property data acquisition module, the set property data of the electrolyte molecule includes: ground state structure, atomic charge, Fukui function, vibration entropy, rotation entropy and molecular point group; The formula for calculating atomic charge is: In the formula, Q A is the atomic charge; Z A is the atomic number, indicating the number of protons in an atom; μ and ν are different atomic orbitals; A is the atomic substitution representation, referring to a certain atom in the system; P μν is the density matrix element, representing the electron density distribution between μ and ν atomic orbitals; S μν is the overlap matrix element, representing the spatial overlap between μ and ν atomic orbitals; The calculation formula of Fukui function is: In the formula, N is the total number of electrons in the molecule; v(r) is the external potential at position r; δ is a desirable minimum value, N+δ and N-δ represent a small positive and negative range near N respectively; ρ(r) is the electron density obtained from the ground state configuration, and the electrophilic and nucleophilic functions are calculated respectively; The derived formulas for vibrational and rotational entropy are: In the formula, S vib is the vibration entropy; h is Planck's constant; ν i is the frequency of the i-th vibration mode; k B is the Boltzmann constant, T is the temperature; The model GCN includes two graph convolutional layers and two fully connected layers; The optimizer used in the electrolyte molecular property prediction model is the adaptively adjusted Adam; the loss function is the mse loss function; and the activation function is relu; Use PyTorch's dataset class Dataset as the parent class to construct a new dataset class DataSetOri, and encapsulate the new dataset class by rewriting the init constructor, getitem function, and len function.

5. The device for predicting properties of lithium battery electrolyte based on deep learning according to claim 4, characterized in that: In the molecular property coding and classification module, in the process of coding and classifying the set property data of the electrolyte molecules according to the set property classification, the atomic coordinates and atomic charge structure properties in the ground state structure of the electrolyte molecules are encoded as point properties; the bond length and bond order properties in the ground state structure of the electrolyte molecules are encoded as bond properties; and the Fukui function descriptor of the electrolyte molecules is encoded as molecular properties.

6. The device for predicting properties of lithium battery electrolyte based on deep learning according to claim 5, characterized in that: In the electrolyte molecule characterization processing module, in the process of dividing the set property data of the electrolyte molecule after coding and classification into the atomic point level features and the edge key level features of the electrolyte molecule, the point properties and the molecular properties of the electrolyte molecule are used as the atomic point level features of the electrolyte molecule; The bond properties of the electrolyte molecules are used as Edge bond level features of the electrolyte molecules.

Images