Method, system, device, and medium for generating molecular structure formula of polymer electrolyte

Abstract

The application discloses a polymer electrolyte molecular structure formula generation method, system, device and medium, and belongs to the technical field of lithium batteries. The polymer electrolyte molecular structure formula generation method comprises the following steps: obtaining a performance data set and a molecular structure formula data set of a polymer electrolyte; constructing a discriminant model by using the performance data set; constructing a generative model by using the molecular structure formula data set; training the generative model based on the discriminant model; if a molecular structure design task is received, generating a plurality of new molecular structure formulas of polymer electrolytes by using the trained generative model, and screening the new molecular structure formulas of polymer electrolytes by using the discriminant model to obtain a molecular structure formula meeting the molecular structure design task. The application can automatically generate a molecular structure formula of a polymer electrolyte, and improve the generation efficiency and effectiveness of the molecular structure formula.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of lithium batteries, in particular to a molecular structure formula generation method, system, device and medium of a polymer electrolyte. BACKGROUND

[0002] Lithium metal batteries, as a kind of batteries with very high theoretical energy density, have been widely studied. However, due to uncontrollable lithium dendrite growth, the cycle performance and safety of the battery are reduced. Replacing the liquid electrolyte with a polymer solid electrolyte is expected to solve this problem. However, the polymer electrolyte still faces problems such as low electrical conductivity and low lithium ion transference rate, which hinder its application in lithium metal batteries. Therefore, it is urgent to develop new high-performance polymer electrolytes to improve the performance of lithium metal batteries.

[0003] In related technologies, the design method of the polymer electrolyte mainly relies on laboratory tests and experience accumulation. This process is time-consuming and labor-intensive, resulting in a long research and development cycle and high cost. In addition, due to the complex molecular structure of the polymer electrolyte, its performance is affected by many factors, and a single experimental method cannot fully characterize and optimize it.

[0004] Therefore, how to automatically generate the molecular structure formula of the polymer electrolyte and improve the generation efficiency and effectiveness of the molecular structure formula is a technical problem that needs to be solved by those skilled in the art at present. SUMMARY

[0005] The purpose of the present application is to provide a molecular structure formula generation method, system, device and medium of a polymer electrolyte, which can automatically generate the molecular structure formula of the polymer electrolyte and improve the generation efficiency and effectiveness of the molecular structure formula.

[0006] To solve the above technical problems, the present application provides a molecular structure formula generation method of a polymer electrolyte, comprising:

[0007] obtaining a performance data set and a molecular structure formula data set of the polymer electrolyte;

[0008] constructing a discriminant model using the performance data set; wherein the discriminant model is used to calculate a performance index of the molecular structure formula, and the performance index includes any one or a combination of several of the electrical conductivity, lithium ion transference number, glass transition temperature and thermal decomposition temperature;

[0009] constructing a generative model using the molecular structure formula data set; wherein the generative model is used to generate the molecular structure formula of the polymer electrolyte;

[0010] training the generative model based on the discriminant model;

[0011] If a molecular structure design task is received, a plurality of new molecular structure formulas of polymer electrolytes are generated using the trained generative model, and the new molecular structure formulas of polymer electrolytes are screened using the discriminative model to obtain a molecular structure formula that meets the molecular structure design task.

[0012] Optionally, after obtaining the performance dataset and the molecular structure formula dataset of the polymer electrolyte, the method further comprises:

[0013] The performance dataset and the molecular structure formula dataset are subjected to data cleaning and data visualization operations.

[0014] Optionally, the discriminative model is constructed using the performance dataset, comprising:

[0015] Data features are extracted from the performance dataset, and the data features are subjected to standardization and normalization processing;

[0016] The data features are filtered to obtain target data features;

[0017] Machine learning modeling is performed based on the target data features and corresponding labels in the performance dataset to obtain the discriminative model.

[0018] Optionally, the performance dataset comprises any one or a combination of the conductivity dataset, the lithium ion transference number dataset, the glass transition temperature dataset, and the thermal decomposition temperature dataset; the conductivity dataset comprises a polymer structure, a polymer molecular weight, a lithium salt structure, a lithium salt concentration, a temperature, and a conductivity label; the lithium ion transference number dataset comprises a polymer structure, a lithium salt structure, a lithium salt concentration, a temperature, an additive type, an additive content, and a lithium ion transference number label; the glass transition temperature dataset comprises a polymer structure and a glass transition temperature label; and the thermal decomposition temperature dataset comprises a polymer structure and a thermal decomposition temperature label.

[0019] Correspondingly, the machine learning modeling based on the target data features and corresponding labels in the performance dataset to obtain the discriminative model comprises:

[0020] Machine learning modeling is performed based on the target data features and the conductivity label in the conductivity dataset to obtain a conductivity discriminative model;

[0021] and / or, machine learning modeling is performed based on the target data features and the lithium ion transference number label in the lithium ion transference number dataset to obtain a lithium ion transference number discriminative model;

[0022] and / or, machine learning modeling is performed based on the target data features and the glass transition temperature label in the glass transition temperature dataset to obtain a glass transition temperature discriminative model.

[0023] And / or, based on the target data features in the thermal decomposition temperature dataset and the thermal decomposition temperature label, machine learning modeling is performed to obtain a thermal decomposition temperature discriminant model.

[0024] Optionally, a generative model is constructed using the molecular structure formula dataset, including:

[0025] The molecular structure in the molecular structure formula dataset is converted into a graph structure; wherein, the nodes in the graph structure represent atoms, and the edges represent chemical bonds;

[0026] A corresponding graph syntax is generated according to the graph structure, and the generative model is constructed based on the graph syntax.

[0027] Optionally, the generative model is trained based on the discriminant model, including:

[0028] A target function of the generative model is constructed; wherein, the value of the target function is the weighted calculation result of a plurality of evaluation indexes, and the evaluation indexes include the scores corresponding to the performance indexes calculated by the discriminant model;

[0029] The generative model is iteratively trained based on the target function.

[0030] Optionally, the generative model is iteratively trained based on the target function, including:

[0031] The i-th round of output result of the generative model is controlled to be generated;

[0032] The value of the target function corresponding to the i-th round of output result is calculated;

[0033] It is judged whether the value of the target function calculated last time is within a preset interval;

[0034] If yes, it is determined that the training of the generative model is completed;

[0035] If no, the parameters of the generative model are adjusted, the value of i is increased by 1, and the operation of controlling the generative model to generate the i-th round of output result is entered;

[0036] The evaluation indicators further include an effectiveness indicator, a uniqueness indicator, a novelty indicator, a diversity indicator, an inverse synthesis score indicator, and a membership degree indicator. The effectiveness indicator is used to describe a proportion of effective molecular structures in the output results of the generative model. The uniqueness indicator is used to describe a proportion of non-repeated molecular structures in the output results of the generative model. The novelty indicator is used to describe a proportion of molecular structures not in the molecular structure formula data set in the output results of the generative model. The diversity indicator is used to describe structural similarity between the output results of the generative model. The inverse synthesis score indicator is used to describe a synthesis success rate of the output results of the generative model. The membership degree indicator is used to describe a proportion of molecular structures belonging to the molecular structure formula data set in the output results of the generative model.

[0037] The application further provides a molecular structure formula generation system of a polymer electrolyte, which comprises:

[0038] a data set acquisition module, configured to acquire a performance data set and a molecular structure formula data set of the polymer electrolyte;

[0039] a first model construction module, configured to construct a discriminative model by using the performance data set; wherein the discriminative model is used to calculate a performance indicator of a molecular structure formula, and the performance indicator comprises any one or a combination of any several of the following: electrical conductivity, lithium ion transference number, glass transition temperature, and thermal decomposition temperature;

[0040] a second model construction module, configured to construct a generative model by using the molecular structure formula data set; wherein the generative model is used to generate a molecular structure formula of the polymer electrolyte;

[0041] a model training module, configured to train the generative model based on the discriminative model;

[0042] a molecule generation module, configured to, if a molecular structure design task is received, generate a plurality of new molecular structure formulas of the polymer electrolyte by using the trained generative model, and screen the new molecular structure formulas of the polymer electrolyte by using the discriminative model to obtain a molecular structure formula meeting the molecular structure design task.

[0043] The application further provides a storage medium having a computer program stored thereon, wherein the computer program is configured to implement the steps of the above-mentioned molecular structure formula generation method of the polymer electrolyte when executed.

[0044] The application further provides an electronic device comprising a memory and a processor, wherein the memory has a computer program stored therein, and the processor is configured to implement the steps of the above-mentioned molecular structure formula generation method of the polymer electrolyte when the computer program in the memory is invoked.

[0045] The application builds a discriminant model through a performance data set of the polymer electrolyte, the discriminant model can learn the relationship between different molecular structures and performances, and the discriminant model can be used to evaluate whether a newly generated molecular structure meets specific performance requirements. The application builds a generative model through a molecular structural formula data set of the polymer electrolyte, the generative model can automatically generate a new molecular structural formula of the polymer electrolyte according to a known molecular structure. The application trains the generative model based on the discriminant model, so that the generative model can generate a molecular structural formula with a performance index meeting requirements. After receiving a molecular structure design task, the application can generate a plurality of new molecular structural formulas of the polymer electrolyte by using the generative model, and screen the molecular structural formulas by using the discriminant model to obtain a molecular structural formula meeting the molecular structure design task. The above process uses the discriminant model to guide the generative model to generate the molecular structural formula, without human intervention, so that the application can automatically generate the molecular structural formula of the polymer electrolyte, and improve the generation efficiency and effectiveness of the molecular structural formula. The application also provides a molecular structural formula generation system of a polymer electrolyte, a storage medium and an electronic device, which have the above beneficial effects, and details are not described herein. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the application, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0047] Figure 1 A flowchart of a molecular structural formula generation method of a polymer electrolyte provided by an embodiment of the application;

[0048] Figure 2 A schematic diagram of a representation method of a molecular graph provided by an embodiment of the application;

[0049] Figure 3 A schematic diagram of a graph syntax generation rule and a molecular generation process applying the syntax provided by an embodiment of the application;

[0050] Figure 4 A probability density distribution diagram of a logarithmic conductivity provided by an embodiment of the application;

[0051] Figure 5 A schematic diagram of a feature correlation matrix of screened conductivity data features provided by an embodiment of the application;

[0052] Figure 6A schematic diagram of a high-conductivity molecule generated by a trained graph grammar generation model provided in an embodiment of this application;

[0053] Figure 7 This is a probability density distribution diagram of glass transition temperature data provided in an embodiment of this application;

[0054] Figure 8 This is a probability density distribution diagram of thermal decomposition temperature data provided in an embodiment of this application;

[0055] Figure 9 A schematic diagram of the correlation matrix of the characteristics of glass transition temperature data provided in an embodiment of this application;

[0056] Figure 10 A schematic diagram of the correlation matrix of the characteristics of thermal decomposition temperature data provided in an embodiment of this application;

[0057] Figure 11 A schematic diagram of a molecule with excellent thermal properties provided in the embodiments of this application;

[0058] Figure 12 This is a probability density distribution diagram of lithium-ion transport number data provided in an embodiment of this application;

[0059] Figure 13 A schematic diagram of the correlation matrix of the characteristics of lithium-ion transport number data provided in an embodiment of this application;

[0060] Figure 14 A schematic diagram of a molecule with excellent overall performance provided in an embodiment of this application;

[0061] Figure 15 This is a flowchart illustrating a method for designing lithium-ion battery polymer electrolytes using coupled discriminative and generative models, as provided in an embodiment of this application. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0063] Please see below. Figure 1 , Figure 1 This is a flowchart illustrating a method for generating the molecular structure of a polymer electrolyte, as provided in an embodiment of this application.

[0064] The specific steps can include:

[0065] S101: Obtain a performance dataset of a polymer electrolyte and a molecular structure formula dataset;

[0066] The embodiment can be applied to an electronic device with data processing and modeling functions, and realizes the automatic design of the lithium battery polymer electrolyte by establishing a discriminant model and a generative model. The performance dataset of the polymer electrolyte includes related data of performance indicators such as conductivity, lithium ion transference number, glass transition temperature, and thermal decomposition temperature; and the molecular structure formula dataset includes molecular structure formulas of a plurality of polymer electrolytes.

[0067] S102: Construct a discriminant model using the performance dataset;

[0068] The discriminant model is used to calculate the performance indicators of the molecular structure formula, and the performance indicators include any one or a combination of any number of the conductivity, lithium ion transference number, glass transition temperature, and thermal decomposition temperature. Each performance indicator can have its corresponding performance dataset, and then a discriminant model corresponding to each performance indicator can be constructed. The discriminant model can be a supervised learning model.

[0069] Specifically, the performance dataset can be split into a training set and a test set, and the data in the training set is used to train the discriminant model, so that the discriminant model learns the relationship between the molecular structure and the performance indicators. The trained discriminant model can be verified by a test machine to ensure the prediction ability of the discriminant model for new data.

[0070] S103: Construct a generative model using the molecular structure formula dataset;

[0071] The generative model can learn the patterns and features in the molecular structure formula dataset, and then generate new molecular structure formulas of polymer electrolytes.

[0072] S104: Train the generative model based on the discriminant model;

[0073] Based on the constructed discriminant model and generative model, the embodiment can use the discriminant model as a guide to evaluate the quality of the molecular structure created by the generative model; through a feedback mechanism, the discriminant model provides performance prediction and optimization direction, so that the generative model can learn and generate new molecular structures that meet the expected performance indicators.

[0074] S105: If a molecular structure design task is received, a plurality of new polymer electrolyte molecular structure formulas are generated using the trained generative model, and the new polymer electrolyte molecular structure formulas are screened using the discriminative model to obtain a molecular structure formula that meets the molecular structure design task.

[0075] Wherein, after receiving a molecular structure design task, a plurality of new polymer electrolyte molecular structure formulas can be quickly generated using the trained generative model, and then the newly generated molecular structure formulas are evaluated and screened by the discriminative model to ensure that the selected molecular structure meets the predetermined performance requirements and design goals, and finally obtain a molecular structure formula that meets the task requirements. The performance requirements, quantity requirements, etc. can be included in the above-mentioned molecular structure design task.

[0076] The discriminative model of the embodiment is constructed by the performance data set of the polymer electrolyte, which can learn the relationship between different molecular structures and performance. The discriminative model can be used to evaluate whether the newly generated molecular structure meets the specific performance requirements. The generative model of the embodiment is constructed by the molecular structure formula data set of the polymer electrolyte, which can automatically generate new polymer electrolyte molecular structure formulas according to known molecular structures. The generative model is trained based on the discriminative model, so that the generative model can generate molecular structure formulas with performance indicators meeting the requirements. After receiving a molecular structure design task, the embodiment can generate a plurality of new polymer electrolyte molecular structure formulas using the generative model, and screen them by the discriminative model to obtain a molecular structure formula that meets the molecular structure design task. The above process uses the discriminative model to guide the generative model to generate molecular structure formulas, without human intervention. Therefore, the embodiment can automatically generate polymer electrolyte molecular structure formulas, and improve the generation efficiency and effectiveness of molecular structure formulas.

[0077] as for Figure 1 For further introduction of the corresponding embodiment, after obtaining the performance data set and the molecular structure formula data set of the polymer electrolyte, the performance data set and the molecular structure formula data set can also be subjected to data cleaning and data visualization operations.

[0078] The above data cleaning can include unit conversion, removal of duplicate data, filling of missing values, correction of abnormal values, etc. The data visualization operation can generate distribution graphs and feature correlation graphs of each performance indicator using data visualization methods.

[0079] as for Figure 1Further to the embodiments, the discriminative model can be constructed by extracting data features from the performance dataset, and normalizing and standardizing the data features; filtering the data features to obtain target data features; and performing machine learning modeling based on the target data features and corresponding labels in the performance dataset to obtain the discriminative model.

[0080] Further, the performance dataset can include any one or a combination of conductivity dataset, lithium ion transference number dataset, glass transition temperature dataset, and thermal decomposition temperature dataset. The conductivity dataset includes polymer structure, polymer molecular weight, lithium salt structure, lithium salt concentration, temperature (i.e., temperature corresponding to the conductivity label), and conductivity label; the lithium ion transference number dataset includes polymer structure, lithium salt structure, lithium salt concentration, temperature (i.e., temperature corresponding to the lithium ion transference number label), and lithium ion transference number label; the glass transition temperature dataset includes polymer structure and glass transition temperature label; and the thermal decomposition temperature dataset includes polymer structure and thermal decomposition temperature label. The conductivity label refers to conductivity corresponding to the group of data including polymer structure, polymer molecular weight, lithium salt structure, lithium salt concentration, and temperature; the lithium ion transference number label refers to lithium ion transference number corresponding to the group of data including polymer structure, lithium salt structure, lithium salt concentration, and temperature; the glass transition temperature label refers to glass transition temperature corresponding to the polymer structure; and the thermal decomposition temperature label refers to thermal decomposition temperature corresponding to the polymer structure.

[0081] If the performance dataset includes the conductivity dataset, the machine learning modeling of the discriminative model can be performed by performing machine learning modeling based on the target data features and conductivity label in the conductivity dataset to obtain a conductivity discriminative model.

[0082] If the performance dataset includes the lithium ion transference number dataset, the machine learning modeling of the discriminative model can be performed by performing machine learning modeling based on the target data features and lithium ion transference number label in the lithium ion transference number dataset to obtain a lithium ion transference number discriminative model.

[0083] If the performance dataset includes the glass transition temperature dataset, the machine learning modeling of the discriminative model can be performed by performing machine learning modeling based on the target data features and glass transition temperature label in the glass transition temperature dataset to obtain a glass transition temperature discriminative model.

[0084] If the performance dataset includes the thermal decomposition temperature dataset, the machine learning modeling of the discriminative model can be performed by performing machine learning modeling based on the target data features and thermal decomposition temperature label in the thermal decomposition temperature dataset to obtain a thermal decomposition temperature discriminative model.

[0085] As for the Figure 1 Further to the corresponding embodiment, the process of constructing the generative model by using the molecular structure formula data set comprises: converting the molecular structure in the molecular structure formula data set into a graph structure; wherein the nodes in the graph structure represent atoms, and the edges represent chemical bonds; generating a corresponding graph grammar according to the graph structure, and constructing the generative model based on the graph grammar.

[0086] The embodiment generates a corresponding graph grammar according to the converted graph structure, and defines how to construct a new molecular structure by rule operations (such as adding, deleting or modifying nodes and edges). Based on the above graph grammar, the generative model is constructed, and the generative model can learn and follow these rules to automatically generate new polymer electrolyte molecular structures that conform to chemical logic.

[0087] Under the premise of constructing the generative model based on the graph grammar, the process of training the generative model by using the discriminative model comprises: constructing an objective function of the generative model; wherein the value of the objective function is the weighted calculation result of a plurality of evaluation indexes, and the evaluation indexes include the scores corresponding to the performance indicators calculated by the discriminative model; iteratively training the generative model based on the objective function.

[0088] The above process of iteratively training the generative model based on the objective function comprises: controlling the generative model to generate the i-th round of output results; calculating the value of the objective function corresponding to the i-th round of output results; determining whether the value of the objective function calculated last time is within a preset interval; if yes, determining that the training of the generative model is complete; if no, adjusting the parameters of the generative model, increasing the value of i by 1, and entering the operation of controlling the generative model to generate the i-th round of output results.

[0089] In addition to the scores corresponding to the performance indicators calculated by the discriminative model, the above evaluation indexes can also include: effectiveness indicators, uniqueness indicators, novelty indicators, diversity indicators, inverse synthesis score indicators and membership indicators.

[0090] The effectiveness indicator is used to describe the proportion of effective molecular structures in the output results of the generative model; the above effective molecular structure is a molecular structure that conforms to chemical rules. The output results of the generative model can be any number of newly generated structural formulas.

[0091] The uniqueness indicator is used to describe the proportion of non-repeating molecular structures in the output results of the generative model, i.e. the proportion of unique molecules in the output results.

[0092] The novelty indicator is used to describe the proportion of molecular structures in the output results of the generative model that are not in the molecular structure formula data set, i.e., the proportion of molecular structure formulas in the output results that are not in the molecular structure formula data set.

[0093] The diversity indicator is used to describe the structural similarity between the output results of the generative model, i.e., the average pairwise molecular distance between molecular structure formulas.

[0094] The inverse synthesis score indicator is used to describe the synthesis success rate of the output results of the generative model.

[0095] The membership indicator is used to describe the proportion of molecular structures in the output results of the generative model that belong to the molecular structure formula data set, i.e., the proportion of molecular structure formulas in the output results that belong to the molecular structure formula data set.

[0096] The above-described process is illustrated by the following examples in practical applications.

[0097] This embodiment provides a scheme for combining a discriminative model with a generative model for reverse design of lithium metal battery polymer electrolyte design. In the case of limited data, this scheme can predict the performance of the polymer electrolyte at an early stage of material design and generate a polymer with excellent performance. This scheme not only accelerates the development process of the polymer electrolyte, but also provides effective guidance for the design of new polymers, reduces experimental costs, and improves research and development efficiency. In addition, based on the guidance of the discriminative model, the polymer synthesis route can be optimized, the overall performance of the electrolyte can be improved, and the development and application of high-performance lithium metal batteries can be promoted.

[0098] The above-mentioned scheme for coupling a discriminative model with a generative model for reverse design of polymer solid-state electrolyte includes the following steps A1-A4:

[0099] Step A1, collect and organize the performance data set and molecular structure formula data set of the polymer electrolyte.

[0100] The data set used in this embodiment is divided into two categories: one is the performance data set of the polymer electrolyte, which is used to train and optimize the discriminative model; the other is the molecular structure formula data set containing the monomers of the polymer for lithium metal batteries, which is used to train the generative model.

[0101] The performance data set of the polymer electrolyte is used to construct a discriminative model to predict the key performance of the polymer electrolyte in lithium metal batteries. The key performance of the polymer electrolyte includes electrical conductivity (σ, S cm -1 ), lithium ion transference number (t Li+ ), glass transition temperature (T g , ℃), and thermal decomposition temperature (Td , ℃), which have important influences on the overall performance of lithium metal batteries. Each of the performances is introduced as follows:

[0102] Electrical conductivity is a parameter used to describe the ease of charge flow in a substance, usually measured in units of Siemens per centimeter (S cm -1 ). In batteries, electrical conductivity is the most core performance indicator of polymer electrolytes, as it directly affects the ion transport efficiency within the battery, thereby affecting the overall performance of the battery. The formula for calculating electrical conductivity is usually based on the extended form of Ohm's law, which takes into account the resistance and geometry of the electrolyte, as follows:

[0103] ;

[0104] where d is the thickness of the polymer electrolyte (cm), R is the impedance of the polymer electrolyte (Ω), and S is the area of the polymer electrolyte (cm -2 ).

[0105] In terms of data collection, the electrical conductivity in polymer electrolytes is influenced by both the polymer and the lithium salt. Therefore, the data in the performance dataset corresponding to electrical conductivity (i.e., the electrical conductivity dataset) includes the polymer structure, polymer molecular weight, lithium salt structure, lithium salt concentration, and temperature corresponding to electrical conductivity, ensuring that the influence of different factors on electrical conductivity is fully considered. Specifically, the performance dataset corresponding to electrical conductivity includes multiple data in the following format: {polymer structure, polymer molecular weight, lithium salt structure, lithium salt concentration, test temperature, label (i.e., electrical conductivity)}.

[0106] Lithium ion transference number (t Li+ ) refers to the ability of lithium ions to move in the electrolyte and is an important parameter for measuring the performance of electrolytes. A high lithium ion transference number means that lithium ions can move more effectively during the charging and discharging process of the battery, thereby improving the performance of the battery. For the lithium ion transference number of polymer electrolytes, it is usually measured at 60 ℃, using the Bruce-Vincent-Evans method, which applies a 30 mV voltage to a symmetric Li|Li battery for 90 minutes (chronoamperometric measurement), and collects the impedance spectrum of the battery before and after polarization. The amplitude of the alternating voltage bias measured by electrochemical impedance spectroscopy (EIS) is 10 mV, and the research frequency range is 500 kHz to 100 mHz. The specific calculation formula is as follows:

[0107] ;

[0108] where ΔV is the chronoamperometric voltage (i.e., 30 mV), i0 is the initial current during polarization, and i​ss is the final current at the time of polarization, R0 and R ss are the interfacial resistances of the electrode and electrolyte before and after polarization, respectively, measured by EIS.

[0109] In terms of data collection, the data in the performance dataset corresponding to the number of lithium ion transference (i.e., the number of lithium ion transference dataset) includes the structure of the polymer, the structure and concentration of the lithium salt, the test temperature corresponding to the number of lithium ion transference, and the type and content of the additive in the polymer electrolyte (such as fillers and plasticizers) to ensure that the effect of the additive on it can be reflected. Specifically, the performance dataset corresponding to the number of lithium ion transference includes multiple data in the following format: {polymer structure, lithium salt structure, lithium salt concentration, test temperature, additive type, additive content, label (i.e., number of lithium ion transference)}.

[0110] The thermal properties of the polymer electrolyte also have a certain impact on the performance of the battery, among which the glass transition temperature T g and the thermal decomposition temperature T d are relatively important.

[0111] The glass transition temperature refers to the temperature corresponding to the transition of the polymer from the glass state to the high-elasticity state, which is generally measured by a differential scanning calorimeter (DSC). Its importance lies in the fact that the migration of lithium ions in the polymer electrolyte mainly depends on the coordination between lithium ions and the polymer and the movement of the polymer chain segment. A lower glass transition temperature can enhance the flexibility of the polymer chain and promote the migration of lithium ions, thereby improving the conductivity of the electrolyte. On the other hand, however, a lower glass transition temperature can also reduce the mechanical strength and stability of the polymer. Therefore, a suitable glass transition temperature is very important for the polymer electrolyte.

[0112] Thermal decomposition refers to the process of decomposition of a compound when heated. The polymer will also undergo thermal decomposition under the action of heat. The thermal decomposition temperature of the polymer electrolyte generally refers to the temperature corresponding to a 5% mass loss when the material is thermally decomposed, also known as T d5 . The thermal decomposition temperature is usually measured by thermogravimetric analysis (TGA) and is a key parameter for measuring the thermal stability of the polymer electrolyte.

[0113] In terms of data collection, since the thermal properties of the polymer electrolyte are mainly related to the structure of the polymer, the effects of other components can be approximately ignored, so the features in the collected thermal performance dataset only include the features of the polymer. The molecular structure of one polymer in the dataset corresponds to one T g or T d .

[0114] The performance data set corresponding to the glass transition temperature (i.e., the glass transition temperature data set) includes the polymer structure and the corresponding label (i.e., the glass transition temperature), and the performance data set corresponding to the thermal decomposition temperature (i.e., the thermal decomposition temperature data set) includes the polymer structure and the corresponding label (i.e., the thermal decomposition temperature).

[0115] The molecular structure formula of the polymer monomer used for the lithium metal battery polymer electrolyte is collected as the training data (i.e., the molecular structure formula data set) of the generation model to construct the molecular generation model for generating new polymers.

[0116] To ensure the quality and consistency of the data set, the performance data set and the molecular structure formula data set collected in this embodiment are systematically cleaned and visualized to facilitate subsequent discriminant model performance prediction and generation model training and optimization.

[0117] The performance data set and the molecular structure formula data set collected in this embodiment are systematically cleaned to ensure the uniformity and accuracy of the data. The cleaning steps include unit conversion, removal of duplicate data, filling of missing values, and correction of abnormal values.

[0118] The unit conversion is explained as follows: Since there are various units of lithium salt concentration used in the original data, such as mass fraction (denoted by ω, unit: wt%), amount-of-substance fraction (denoted by x, unit: mol%, where the amount of substance of the polymer is the amount of substance of the repeating unit), mass molar concentration (denoted by m, defined as the amount of substance of lithium salt contained in each unit mass of polymer, unit: mol kg -1 ), and amount-of-substance concentration (denoted by c, unit: mol L -1 ), the data units need to be unified before machine learning modeling. In this embodiment, the lithium salt concentration unit is unified as mass molar concentration.

[0119] The formula for obtaining the mass molar concentration m from the lithium salt mass fraction ω s is as follows:

[0120] ;

[0121] where M s is the molar mass of lithium salt, unit: g mol -1 .

[0122] The formula for obtaining the mass molar concentration m from the lithium salt amount-of-substance fraction x is as follows:

[0123] ;

[0124] where M p is the molar mass of the repeating unit of the polymer, unit: g mol -1.

[0125] For in-situ polymerized polymer electrolyte, the polymer is in the form of monomer as solvent of lithium salt before preparation, using the amount concentration c p As the unit of lithium salt concentration, it represents the amount of lithium salt contained in per unit volume, unit: mol L -1 . The formula of mass molar concentration m from the amount concentration c p The formula of mass molar concentration m from the amount concentration c

[0126] ;

[0127] Where ρ p The density of polymer monomer, unit: g mL -1 .

[0128] The above formula takes into account the case of polymer electrolyte with two components of polymer and lithium salt, and other additives can also be present in the polymer electrolyte.

[0129] One of the additives, plasticizer, can be expressed in different units. The preparation process of gel polymer electrolyte is to soak the dry electrolyte in the plasticizer containing lithium salt, forming a plasticizer fixed in the polymer matrix. Uptake% (uptake rate) is an important parameter to measure the performance of gel polymer electrolyte, which represents the ability of polymer matrix to absorb liquid electrolyte, and also reflects the content of lithium salt and plasticizer in polymer electrolyte. The formula of Uptake% is as follows:

[0130] ;

[0131] Where M1 and M0 represent the mass of electrolyte after soaking and dry electrolyte before soaking, respectively.

[0132] The formula of mass molar concentration m of lithium salt from Uptake% in gel polymer solid-state electrolyte is as follows:

[0133] ;

[0134] Where ρ l The density of plasticizer, unit: g mL -1 ; c l The molar concentration of lithium salt in plasticizer, unit: mol L -1 .

[0135] The unit of plasticizer content also needs to be unified, and the formula of plasticizer mass percentage ω l is as follows:

[0136] ;

[0137] The missing value filling is described as follows: During the data collection process, data missing may inevitably occur. If the missing data is not particularly important, the missing values can be filled. For the missing temperature of lithium ion transference number measurement, since the lithium ion transference number of most polymer electrolytes is measured at 60℃, the missing value is filled with 60℃; for a small number of samples with missing weight average molecular weight, the number average molecular weight or viscosity average molecular weight is used for approximate filling; if three kinds of molecular weight data are not given, only the source of the reagent is given, and the molecular weight of the reagent from the source is given in other data, then the data is filled with other data.

[0138] The abnormal value processing is described as follows: the data points inconsistent with the actual situation are removed, and the long tail data points seriously deviating from other data are deleted to eliminate possible extreme errors.

[0139] The above cleaning steps improve the consistency and reliability of the data set, and provide a reliable data basis for the accurate training and prediction of the subsequent model.

[0140] To further analyze the data distribution, the embodiment generates a distribution graph and a feature correlation graph of each performance indicator by using a data visualization method. For example, the distribution of various performances is displayed by using a probability density distribution graph, and the correlation between different features is displayed by using a heat map, which provides a basis for the next feature engineering and model construction.

[0141] Step A2, constructing an efficient discriminant model based on the performance data set of the polymer electrolyte.

[0142] The discriminant model in this embodiment belongs to a supervised learning model, which aims to predict the key performance (such as conductivity, lithium ion transference number, glass transition temperature and thermal decomposition temperature) of the polymer electrolyte by inputting the features of the polymer electrolyte. Each performance has a corresponding discriminant model, but the construction process of each model is similar. The discriminant model can effectively support the prediction of the performance of the polymer electrolyte, and can select the performance excellent molecules for the generation model. Steps A21-A27 are the specific steps for constructing the discriminant model:

[0143] Step A21: feature extraction;

[0144] In the data set, the molecular structure of the polymer and the lithium salt is represented by SMILES (Simplified Molecular Input Line Entry System). In order to adapt to the modeling needs of machine learning, the structure features of SMILES are extracted by using Mordred descriptor. This step can extract multiple features including different atomic numbers, molecular weight, polarization performance, acid-base property, hydrogen bond related properties, etc., which can be used to describe the molecular structure.

[0145] Step A22: Data Splitting;

[0146] To ensure the training effect and evaluation accuracy of the model, the data is divided into input (X) and output (Y), where X is the characteristics of the polymer electrolyte and Y is the performance to be predicted. At the same time, the data set is randomly divided into training set and test set, the training set is used to train the model; the test set is used to evaluate the performance and generalization ability of the model to ensure the prediction ability of the model for new data.

[0147] Step A23: Data Standardization and Normalization;

[0148] To improve the convergence speed, stability and prediction effect of the model, the input feature data is standardized and normalized to maintain the consistency of different feature scales. The standardized features are achieved by subtracting the mean and scaling to unit variance. The calculation formula is as follows:

[0149] ;

[0150] Where μ is the mean of the feature x, σ is the standard deviation of the feature x, and z represents the feature vector of the sample.

[0151] Normalization can scale the feature vector of each sample to a unit norm (usually a unit length), and the calculation formula is as follows:

[0152] ;

[0153] ;

[0154] Where ||x|| is the norm of the sample (usually L2 norm, i.e. Euclidean norm); n is the dimension of the feature vector, i.e. the number of features.

[0155] Through standardization and normalization, the dimensional differences of each feature can be eliminated to ensure that the model can more efficiently capture the relationship between the features and the output.

[0156] Step A24: Feature Selection;

[0157] To further improve the prediction accuracy of the model, reduce the computational complexity and reduce the risk of overfitting, the extracted features need to be selected. Specifically, variance filtering, correlation filtering and RFECV (Recursive Feature Elimination with Cross-Validation) are used, which is a feature selection method based on recursive feature elimination and cross-validation.

[0158] The explanation of variance filtering is as follows:

[0159] The basic idea of variance filtering is to calculate the variance of each feature, and only the features with a variance greater than a set threshold are retained, and the rest are deleted. The calculation formula of variance is as follows:

[0160] ;

[0161] where X is the feature, N is the number of samples, x i is the i-th sample value in the feature, and μ is the mean value of the feature. Some features have very small variance (for example, all values are similar), and these features may not contribute much to the prediction ability of the model, and these low-variance features need to be removed.

[0162] The description of correlation filtering is as follows:

[0163] Correlation filtering uses Pearson correlation coefficient to calculate the correlation r between features, and the formula is as follows:

[0164] ;

[0165] where X i and Y i are the i-th observation values of two features, and are the mean values of the two features. After obtaining the correlation between the features, a threshold is set, and the features with a correlation higher than the threshold are deleted.

[0166] The description of recursive feature elimination cross-validation (RFECV) is as follows:

[0167] RFECV is a feature selection method that combines recursive feature elimination (RFE) and cross-validation (CV). It automatically determines the optimal number of features by iteratively removing the features that have the least impact on model performance and using cross-validation to evaluate the performance of the feature subset at each iteration.

[0168] Specifically, RFECV starts with all features, uses the current feature set for K-fold cross-validation to evaluate the performance of the model. Then, according to the output of the model, the importance of each feature is evaluated, and the feature with the lowest importance is removed. This process is repeated until a predetermined number of features is reached or other stopping conditions are met. Finally, by comparing the average performance of different numbers of features, the number of features with the best performance is selected.

[0169] The advantage of RFECV is that it can automatically optimize the number of features, reduce the risk of overfitting, and improve the performance and interpretability of the model.

[0170] The above three methods can reduce the dimensionality of the feature space and improve the generalization ability of the model. ​

[0171] Step A25: Algorithm Selection;

[0172] According to the demand of polymer electrolyte performance prediction and the characteristics of data structure, three machine learning regression algorithms are selected, including random forest, gradient boosting tree and XGBoost.

[0173] The description of random forest is as follows:

[0174] Random forest is an ensemble learning method based on decision trees. It improves the prediction accuracy and stability by generating multiple decision trees and integrating the results of these trees. Each tree randomly selects a subset of samples and features from the dataset during training to reduce the variance of the model.

[0175] Specifically, the random forest algorithm first creates multiple decision trees by randomly sampling subsets from the training data, and each tree uses these different subsets for training to form diverse decision trees. When training each tree, random selection of features is used to determine the splitting of nodes, which can avoid the over-influence of certain features on the model.

[0176] Then, the test set is predicted, and the random forest will input the test set samples into all the decision trees. Each tree will output a prediction value, and finally, the random forest will take the average of these prediction values to get the final prediction result, as follows:

[0177] ;

[0178] where, is the final prediction result, B is the number of trees, is the prediction value of the bth tree.

[0179] The description of gradient boosting tree (GBDT) is as follows:

[0180] GBDT is a decision tree algorithm based on gradient boosting. It gradually builds decision trees and adjusts the error of each tree to obtain better results. GBDT uses additive model to combine multiple weak learners (usually decision trees) and uses gradient descent to minimize the loss function.

[0181] Assuming the prediction value of the model is , the update of the th iteration is:

[0182] ;

[0183] where, F m-1 (x) is the prediction result of the first m-1 rounds; h m(x) is the value of the mth decision tree; η is the learning rate, which controls the update step size. The training objective of each step is to minimize the residual. In the regression task, the loss function L(y, F(x)) chooses the mean square error loss function, y represents the true value, denotes the partial derivative, and the negative gradient g m The calculation formula is as follows:

[0184] ;

[0185] When the loss function chooses the mean square loss function, the negative gradient g m of the loss function is the residual, that is, the true value minus the value predicted by the current model, so the training objective is approximately to minimize the negative gradient of the current loss function. If the loss function is not the mean square loss function, the method of using the negative gradient of the loss function instead of the residual is also applicable.

[0186] The description of XGBoost is as follows:

[0187] XGBoost is an improved gradient boosting tree algorithm. Compared with the traditional GBDT, XGBoost introduces many optimizations, such as regularization, parallelization, missing value processing, and split finding optimization, so that the model training speed is faster and the performance is better. Unlike the objective function of GBDT which only contains the loss function, the objective function obj of XGBoost is composed of the loss function and the regularization term:

[0188] ;

[0189] Among them, is the loss function; is the regularization term, denotes the actual output result, denotes the output result of XGBoost. In order to optimize the objective function. The Taylor second-order expansion formula is used to express the approximate objective function obj:

[0190] ;

[0191] Among them, g i and h i are the first-order derivative and the second-order derivative of the loss function, respectively, f t denotes the tree model added in the tth iteration (i.e., the tth tree), f t (x i ) denotes the prediction value of the tth tree for the i th sample x i .

[0192] XGBoost then selects the optimal split point by maximizing the gain of the objective function. For each possible split point, the difference in the objective function value before and after the split is calculated, and the split point that maximizes the gain is selected. Once the loop stopping condition is reached, the iteration of generating new trees is stopped, and an optimal model is obtained.

[0193] The above three algorithms have good performance in processing complex structured data and can effectively capture the nonlinear relationship between features and performance. In the subsequent steps, these three algorithms are used for machine learning modeling. By comparing the performance of different models in the prediction task, the most suitable model is determined to meet the requirements of prediction accuracy and computational efficiency.

[0194] Step A26: Hyperparameter tuning;

[0195] To further improve the prediction accuracy of the model, hyperparameter tuning is needed. The tool used for hyperparameter tuning can be Optuna. The tuned hyperparameters include the number of trees, tree depth, and learning rate, etc. Bayesian optimization method is used for efficient parameter search. Through this optimization process, the generalization ability and prediction effect of the model are improved, ensuring its reliability under different electrolyte structures and performance combinations.

[0196] Step A27: Model evaluation;

[0197] The regression model trained on the training set is applied to the test set to obtain the predicted values, and then a variety of indicators are used to comprehensively evaluate the prediction performance of the model. This step can use mean squared error (MSE), root mean squared error (RMSE), mean absolute error (MAE), or determination coefficient (R 2 ) to evaluate the prediction performance of the model.

[0198] The explanation of mean squared error (MSE) is as follows:

[0199] MSE measures the average of the squared errors between the predicted values and the actual values, and the calculation formula is as follows:

[0200] ;

[0201] where n is the number of samples, y i is the actual value of the i th sample, and y is the predicted value of the i

[0202] th sample.

[0203] The explanation of root mean squared error (RMSE) is as follows:

[0204] .

[0205] The mean absolute error (MAE) is described as follows:

[0206] MAE represents the average of the absolute errors between the predicted values and the actual values, and the formula is as follows:

[0207] .

[0208] The coefficient of determination R 2 is described as follows:

[0209] R 2 represents the proportion of the model to explain the variance of the data, and the value is less than 1. The closer to 1, the stronger the explanation ability of the model. The calculation formula is as follows:

[0210] ;

[0211] wherein, is the mean of the actual values.

[0212] These evaluation indexes can comprehensively measure the prediction accuracy and stability of the model, and verify the generalization ability of the model on the test set. Through the analysis of the evaluation results, the actual application effect of the model in the performance prediction of the polymer electrolyte is further determined.

[0213] Step A3, constructing a molecular generation model based on graph grammar.

[0214] The molecular generation model algorithm used in this embodiment has the significant feature of effectively learning and generating high-quality, synthesizable molecules on a small data set. This feature is very suitable for research fields with limited data, such as the polymer electrolyte research in this embodiment. The construction and optimization process of the generation model includes the following steps A3.1-A3.4:

[0215] Step A3.1: molecular graph representation;

[0216] The molecular structure can be represented as a graph structure, in which nodes represent atoms and edges represent chemical bonds. For a molecule M, its graph H M = (V, E H ) contains a set of nodes V and a set of edges E H , wherein the node V is the set of all atoms, and the edge E H includes adding an edge for each bond connecting only two nodes, and adding an edge for all nodes in each ring (including aromatic rings). Please refer to Figure 2 , Figure 2This is a schematic diagram of a molecular diagram representation method provided in an embodiment of this application. H represents hydrogen, C represents carbon, O represents oxygen, S represents sulfur, F represents fluorine, Carbon Atom Node represents a carbon atom node, Oxygen Atom Node represents an oxygen atom node, Sulphur Atom Node represents a sulfur atom node, Fluorine Atom Node represents a fluorine atom node, Single Bond represents a single bond, Double Bond represents a double bond, Aromatic Bond represents an aromatic bond, and Hyperedge represents a hyperedge. Figure 2 As shown, the four benzene rings in the molecule correspond to the four sides in the diagram.

[0217] Step A3.2: Generate graph syntax rules;

[0218] The formal grammar is introduced below: A grammar G = (N, Σ, P, X) includes an initial symbol X, a finite set of non-terminal symbols N, and a finite set of terminal symbols Σ. The grammar G describes how to use a set of generation rules P = {p i |i = 1, …, k} constructs a string from the alphabet of the language, in the form p i LHS→RHS, where LHS is the left-hand symbol and RHS is the right-hand symbol. Based on this syntax, the rule that matches the non-terminal symbol N in the current string on its left is iteratively selected and replaced with the right-hand symbol of the rule until the string contains no non-terminal symbol N.

[0219] Graph grammar is an extension of formal grammar; see [link to related documentation]. Figure 3 , Figure 3 This diagram illustrates a graph grammar generation rule and a molecular generation process using this grammar, as provided in an embodiment of this application. "Production rules" represents the generation rules, "Generation process" represents the generation process, p1, p2, p3, and p4 represent specific generation rules, X represents a string, and N represents a non-terminal symbol set. For example... Figure 3 As shown, each generation rule is flanked by graphs on both sides. These graphs contain an initial node X, non-terminal nodes N, and terminal nodes representing atoms. The nodes corresponding to the hollow circles (i.e., the white nodes) are anchor nodes, which do not change from the left to the right. The molecular graph generation process based on graph grammar is also... Figure 3 It is displayed in the middle.

[0220] The graph grammar construction is a bottom-up process, by iteratively sampling a set of edges and contracting them to a non-terminal node, for each contraction step, a production rule is constructed and added to the grammar, resulting in a new graph with fewer nodes and edges. Specifically for a molecule, at iteration t, the current graph H M , t = (V, E H ) is sampled m edges, whose set is denoted as where denotes the i-th hyperedge sampled at the t-th round. The set of nodes connected by these edges is denoted as . Then all connected components about these edges are extracted, where and denote the nodes and edges in the i-th connected component sampled at the t-th round, respectively, and each connected component is transformed into a production rule. The nodes connected to the nodes in the connected component but not belonging to the connected component are defined as anchor nodes in the production rule; the edges connecting the anchor nodes and the internal structure are denoted as , whose formulae are:

[0221] ;

[0222] ;

[0223] where s denotes the node belonging to the connected component, and v denotes the node not belonging to the connected component but connected to the node s in the connected component.

[0224] Next, the production rule p i : LHS→RHS is constructed, where the left side LHS and the right side RHS are as follows, H denotes the graph composed of edges E and nodes V, V L , E L , V R and E R denote the nodes on the left side, the edges on the left side, the nodes on the right side and the edges on the right side, respectively, and the left side contains a non-terminal node N:

[0225] ;

[0226] ;

[0227] Then each connected component is replaced by a non-terminal node N, and the graph is updated to H M , t+1 , and the iteration continues until the whole graph is contracted to a single non-terminal node, which is replaced by the initial node X.

[0228] Step A3.3: Optimization and learning of graph grammar rules;

[0229] The optimization goal of graph grammar is to maximize the evaluation indicators (e.g. diversity and synthetic possibility) of the input. Since the optimization goal is usually non-differentiable and cannot be directly calculated by gradient descent to optimize the path, Monte Carlo sampling (MC sampling) and reinforcement learning policy gradient algorithm are used to optimize the objective function.

[0230] Step A3.4: evaluation indicators;

[0231] In order to evaluate the performance of the molecule generation model, the evaluation indicators introduced in this embodiment are effectiveness, uniqueness, novelty, diversity, reverse synthesis score (RS) and membership. This embodiment can evaluate the quality and diversity of the generated molecules, or be used to evaluate the similarity and extrapolation ability of the generated molecules to the training data distribution, etc.

[0232] Effectiveness is the proportion of chemically effective molecules in the generated molecules. Uniqueness is the proportion of unique molecules in the generated molecules; Novelty is the proportion of molecules in the generated molecules that are not in the training data set. Diversity is the average pairwise molecular distance between the generated molecules. Reverse synthesis score (RS) is the success rate of synthesizing the current molecule from commercially available raw materials. Membership is the proportion of molecules in the generated molecules that belong to the training data.

[0233] Step A4, using a discriminative model to guide the generation model.

[0234] Although the molecules generated by the generation model generated by creating graph grammar rules from the training data are similar in chemical structure to the training data for polymer electrolytes, it cannot be ensured that these generated molecules have good conductivity and other performances. Therefore, the present application proposes a method of combining a discriminative model with a generative model, so that the discriminative model can effectively guide the construction of the generative model, thereby ensuring that the generated molecules have excellent performance. The specific method includes the following steps A4.1 and A4.2:

[0235] Step A4.1: constructing a prediction function based on the discriminative model;

[0236] The input of the function is the SMILES formula of the molecule generated by the generative model, and the output is one or more performances predicted by the discriminative model according to the SMILES formula of the molecule, such as σ, t Li+ , T g and T d . The function specifically includes feature extraction of the SMILES formula and performance prediction, wherein the feature extraction part is the same as the step of constructing the discriminative model to ensure consistency with the input of the trained discriminative model.

[0237] Step A4.2: updating the objective function of the graph grammar rule;

[0238] The objective function in the optimization of the graph grammar rule is where M i and λ i represent the value and weight of the i-th evaluation index respectively. The evaluation indexes are mentioned above, which can evaluate the quality and diversity of the generated molecules, or the similarity and extrapolation ability of the generated molecules to the training data distribution, etc. In order to combine the generative model with the discriminative model, the performance value predicted by the discriminative model is integrated into the objective function of the graph grammar rule, so that the optimization of the graph grammar rule iterates in the direction of generating molecules with more excellent performance.

[0239] The normalization is explained as follows:

[0240] The value range of the above evaluation index is between 0 and 1. In order to make the performance score and the evaluation index score consistent in the same objective function, the predicted performance value needs to be normalized to obtain a performance score, whose value range is also between [0, 1]. Since the output of the discriminative model does not have a fixed range, the min-max normalization method cannot be directly used. The method of threshold clipping mapping is used in the present application, which defines an effective range based on the existing data set, so as to map the extreme value output by the discriminative model to between 0 and 1.

[0241] First, the maximum and minimum values are defined, and the 5th and 95th percentiles of the existing data set are denoted as P5 and P95, which are used as references for the minimum and maximum values to avoid the influence of extreme abnormal values in the data. Then the output x of the discriminative model is threshold clipped and normalized, and the formula is as follows:

[0242] ;

[0243] Where threshold clipping means that if x is lower than the set minimum value, it is normalized to 0, and if it is higher than the maximum value, it is normalized to 1; and x within the maximum and minimum value range is simply linearly normalized to map them to [0, 1].

[0244] In addition, not all indexes or performances are better the lower. For example, the glass transition temperature, in practical application, we hope to get a polymer electrolyte with a lower glass transition temperature. At this time, the above formula needs to be modified, that is:

[0245] ;

[0246] The weight of the objective function is explained as follows:

[0247] After the performance score is obtained by normalizing the performance value predicted by the discriminant model, the graph grammar rule objective function is updated as where M i represents the evaluation index (such as diversity, RS, etc.) or the normalized performance score (such as conductivity score, lithium ion migration number score, etc.), and λ i is the corresponding weight. Subsequently, the weight λ needs to be set and optimized.

[0248] The initial weight setting can be set according to the importance of different indexes or performance scores. If the actual demand is to focus on generating molecules with better performance, the weight of the performance score should be set higher, so that the performance will have a greater impact on the objective function, and the optimized graph grammar rule will tend to generate molecules with better performance.

[0249] If the actual demand is multi-objective optimization, therefore the weight should balance each index, so that the generated molecules perform best in comprehensive performance. Balanced weight method can be used, that is, a set of weights that make multiple objective function values higher are found through grid search or Bayesian optimization.

[0250] The discriminant model guided generation model in this embodiment can generate high-quality molecules that meet the specific electrolyte requirements, providing support for the development of polymer electrolyte materials.

[0251] The above scheme of coupling discriminant and generative models to design lithium battery polymer electrolytes is illustrated below by Examples B, C and D.

[0252] Example B: High conductivity molecules are generated by combining a conductivity discriminant model with a generative model, which specifically includes the following steps B1-B3.

[0253] Step B1: Construct a conductivity discriminant model.

[0254] Specifically, the process of constructing the conductivity discriminant model includes the following steps B1.1-B1.5:

[0255] Step B1.1: Construct a conductivity dataset of polymer electrolytes;

[0256] The conductivity dataset of polymer electrolytes can include multiple data (such as 9117 pieces), each piece of data includes 5 features, which are polymer monomer SMILES, polymer weight average molecular weight, lithium salt SMILES, lithium salt mass molar concentration and conductivity temperature, and a label: conductivity. The features and labels are subjected to unit conversion, missing value filling and outlier processing.

[0257] See Figure 4 , Figure 4A probability density distribution diagram of the logarithmic conductivity is provided by the embodiments of the present application. Log Conductivity (log S cm -1 ) represents the logarithmic value of conductivity, with a unit of siemens per centimeter, and the ordinate represents probability.

[0258] The data processing and visualization work is implemented by using Python 3.11.7, the data processing uses the pandas library, and the visualization uses the matplotlib library.

[0259] Step B1.2: Feature extraction;

[0260] The Mordred descriptor is used to extract the structural features of the polymer and lithium salt SMILES. Including about 28 specific features related to polarization performance, hydrogen bond, rotatable bond, mass, acidity and basicity, and the number of atoms. In addition to the original features of the polymer weight average molecular weight, lithium salt mass molar concentration and conductivity temperature, the total number of features is 59 in the next step of feature selection.

[0261] All the features extracted by the above electrolyte discriminant model are as follows:

[0262] 'polymer RotRatio' (polymer molecule rotatable bond ratio), 'polymer Vabc' (polymer molecule van der Waals volume), 'polymer MW' (polymer molecule exact molecular weight), 'polymer AMW' (polymer molecule average exact molecular weight), 'polymer nAcid' (polymer molecule number of acidic groups), 'polymer nBase' (polymer molecule number of basic groups), 'polymer nAtom' (polymer molecule number of atoms), 'polymer nHeavyAtom' (polymer molecule number of heavy atoms), 'polymer nSpiro' (polymer molecule number of spiro atoms), 'polymer nBridgehead' (polymer molecule number of bridgehead atoms), 'polymer nHetero' (polymer molecule number of hetero atoms), 'polymer nH' (polymer molecule number of H atoms), 'polymer nB' (polymer molecule number of B atoms), 'polymer nC' (polymer molecule number of C atoms), 'polymer nN' (polymer molecule number of N atoms), 'polymer nO' (polymer molecule number of O atoms), 'polymer nS' (polymer molecule number of S atoms), 'polymer nP' (polymer molecule number of P atoms), 'polymer nF' (polymer molecule number of F atoms), 'polymer nCl' (polymer molecule number of Cl atoms), 'polymer nBr' (polymer molecule number of Br atoms), 'polymer nI' (polymer molecule number of I atoms), 'polymer nX' (polymer molecule number of halogen atoms), 'polymer Kierl' (polymer molecule kappa shape index 1), 'polymer Kier2' (polymer molecule kappa shape index 2),'salt apol' (lithium salt molecule atomic polarizability),'salt bpol' (lithium salt molecule bond polarizability),'salt nHBAcc' (lithium salt molecule number of hydrogen bond acceptors),'salt nHBDon' (lithium salt molecule number of hydrogen bond donors),'salt nRot' (lithium salt molecule number of rotatable bonds),'salt MW' (molecular weight of lithium salt),'salt AMW' (average molecular weight of lithium salt),'salt nAcid' (number of acidic groups of lithium salt),'salt nBase' (number of basic groups of lithium salt),'salt nAtom' (number of atoms of lithium salt),'salt nHeavyAtom' (number of heavy atoms of lithium salt),'salt nSpiro' (number of spiro atoms of lithium salt),'salt nBridgehead' (number of bridgehead atoms of lithium salt),'salt nHetero' (number of hetero atoms of lithium salt),'salt nH' (number of H atoms of lithium salt),'salt nB' (number of B atoms of lithium salt),'salt nC' (number of C atoms of lithium salt),'salt nN' (number of N atoms of lithium salt),'salt nO' (number of O atoms of lithium salt),'salt nS' (number of S atoms of lithium salt),'salt nP' (number of P atoms of lithium salt),'salt nF' (number of F atoms of lithium salt),'salt nCl' (number of Cl atoms of lithium salt),'salt nBr' (number of Br atoms of lithium salt),'salt nI' (number of I atoms of lithium salt),'salt nX' (number of halogen atoms of lithium salt), polymer molecular weight, lithium salt mass molar concentration, and conductivity temperature.

[0263] Step B1.3: data splitting

[0264] The data set is split into a training set and a test set, wherein the training set is 80% of the total data (e.g., 7293 pieces); and the test set is 20% of the total data (e.g., 1824 pieces).

[0265] Step B1.4: feature screening

[0266] After data standardization and normalization, the features are screened. First, variance filtering is performed, and the threshold is set to 0.005. Among them, the variances of 22 features are lower than the threshold and are deleted, and there are still 37 features left. Then, correlation filtering is performed, and the threshold is set to 0.9. Among them, the correlation of 10 features is higher than the threshold and is deleted, and there are still 27 features left. Then, RFECV is used to select the features. Among them, the model is a random forest model, and the K-fold cross-validation is set to 5 folds. After screening, the best feature number is finally determined to be 12, and the correlation matrix of these features is as shown in Figure 5 Figure 5 It is a feature correlation matrix diagram after conductivity data feature screening provided by the embodiment of the application.

[0267] ​Figure 5 Mw represents molecular weight, salt nN represents the number of nitrogen atoms of the salt, salt nC represents the number of carbon atoms of the salt, salt nAcid represents the number of acid groups of the salt, salt AMW represents the average molecular weight of the salt, salt nHBAcc represents the number of hydrogen bond acceptors of the salt, salt bpol represents the polarity of the salt, salt apol represents the non-polarity of the salt, salt nS represents the number of sulfur atoms of the salt, polymer nN represents the number of nitrogen atoms of the polymer, salt nP represents the number of phosphorus atoms of the salt, polymer AMW represents the average molecular weight of the polymer, polymer RotRatio represents the rotation ratio of the polymer, polymer nRot represents the number of rotatable bonds of the polymer, polymer nHBDon represents the number of hydrogen bond donors of the polymer, temperature represents temperature, molality represents molarity, conductivity represents conductivity, Upper Triangle represents upper triangle, Spearman Correlation represents Spearman Correlation, Lower Triangle represents lower triangle, and Pearson Correlation represents Pearson Correlation.

[0268] Step B1.4: Model training and evaluation

[0269] After feature screening, machine learning modeling was performed on the data. Three algorithms, random forest, gradient boosting tree and XGBoost, were used, and the training set was trained through the scikit-learn library. At the same time, Optuna was used to find the best combination of hyperparameters for modeling in the given hyperparameter space. For different algorithms, the optimized hyperparameters mainly include the number of evaluators, the depth of numbers, the learning rate, etc., and the optimization index is the R 2 of the model on the validation set. The number of optimization experiments is 100 times. After the model training of the three algorithms is completed, the prediction performance of the model is evaluated using the test set. The evaluation indexes include RMSE, MAE and R 2 , and the evaluation results are shown in Table 1:

[0270] Table 1 Evaluation index calculation result table

[0271]

[0272] RMSE Train represents the root mean square error of the training set, MAE Train represents the mean absolute error of the training set, R 2 Train represents the determination coefficient of the training set, RMSE Test represents the root mean square error of the test set, MAETest Mean Absolute Error of the test set, R 2 Test R-squared of the test set, XGBoost represents distributed gradient boosting library, RF represents random forest, and GBDT represents gradient boosting tree.

[0273] The comprehensive comparison shows that the prediction effect of the XGBoost model is relatively good, and the algorithm model has the fastest training speed and the shortest time consumption, so the model constructed by the algorithm is selected as the conductivity discriminant model.

[0274] Step B2: constructing a molecular generation model based on graph grammar and combining it with the discriminant model;

[0275] Specifically, the process of implementing this step includes the following steps B2.1-B2.2:

[0276] Step B2.1: constructing a polymer monomer molecular structure formula data set;

[0277] Since the purpose of the generation model in this embodiment is to generate molecules that can be used in lithium metal battery polymer electrolytes, the training data is currently applied molecules of this type, so that the generation model learns the structural characteristics of such molecules when learning the graph grammar generation rules, and generates new molecules of this type. The training set used in this embodiment includes a total of 183 applied polymer monomer molecular structure formulas.

[0278] Step 2.2: setting the objective function for optimizing the graph grammar rules.

[0279] The objective function for optimizing the graph grammar rules of the generation model is , wherein M i and λ i represent the value and weight of the i-th evaluation index, respectively, and the evaluation index and weight to be optimized need to be selected before optimization. The evaluation indexes selected in this embodiment are diversity, RS, and conductivity score, and the corresponding weights are λ1=1, λ2=1, and λ3=2. The weights set are only based on the importance of the evaluation indexes of the molecules expected to be generated in this embodiment and have not been optimized.

[0280] The conductivity score is obtained from the conductivity prediction function. The prediction function includes normalizing the predicted conductivity of the molecules generated by the generation model by the conductivity discriminant model to obtain the conductivity score. The threshold truncation normalization method based on the existing conductivity data is used for normalization, and the P5 and P95 of the logarithmic conductivity of the existing data are-8.648 and-2.928, respectively, with the unit of log S cm -1 . Therefore, the threshold truncation normalization formula is:

[0281] ;

[0282] Step B3: using discriminant model to guide the construction of graph grammar generation model;

[0283] Firstly, the RDKit package is used to extract the graph of the molecules in the training data, including a set of nodes and a set of edges. Then, a feature extraction function f(e) is used to extract features for each edge e. In this example, the feature extraction function is a pre-trained graph neural network implemented by the pytorch library, and the extracted feature dimension is 300. The edge weight function is defined as a composite function F(e; θ) = F θ (f(e)) dependent on the parameter θ, which is a neural network model containing two fully connected layers. The input layer dimension is set to 300, the hidden layer dimension is set to 128, the Dropout rate is set to 0.5, the output layer dimension is set to 2, and the softmax activation function is used to generate the probability distribution of X as 0 and 1. X follows the Bernoulli distribution based on the selection probability of edge e The edge weight function F θ is a parameterized function dependent on the parameter θ, which further maps the output of the edge feature extraction function f(e) to the final result, that is, sets the corresponding weight for each edge.

[0284] Then, the graph grammar rule is optimized, that is, the parameter θ of the edge weight function F(e; θ) is optimized. Each graph grammar rule needs to be evaluated to obtain the value of the objective function. The specific method is to use the graph grammar to generate a certain number of molecules, and then use the selected evaluation index to evaluate these molecules to obtain the value of the evaluation index, and then the objective function is obtained. The number of generated molecules for evaluation is set to 100. After obtaining the objective function, the neural network of the edge weight function is iteratively optimized, the maximum number of iterations is set to 50, the learning rate is set to 0.01, and the optimization is realized by the pytorch library.

[0285] After iterative training, an optimized graph grammar rule generation model is finally obtained.

[0286] Step B4: generating model generates samples and discriminant model predicts conductivity;

[0287] 2K molecules are randomly generated by the best graph grammar rule generation model, and the conductivity is predicted by the discriminant model. The generated high conductivity molecule is shown in Figure 6 . Figure 6 The high conductivity molecule generated by the trained graph grammar generation model provided by the embodiment of the application is shown in the schematic diagram.

[0288] Figure 6 A represents a specific atom or chemical group, log represents logarithm, and Conductivity represents conductivity.-1 Siemens per centimeter, N represents a nitrogen atom, O represents an oxygen atom, and F represents a fluorine atom. The logarithm of the conductivity can be: -3.569 log S cm -1 , -3.587 log S cm -1 , -3.662 log S cm -1 , -4.057 log S cm -1 , -3.623 log S cm -1 , -4.176 log S cm -1 , -4.135 log S cm -1 , -4.119 log S cm -1 , -3.587 log S cm -1 .

[0289] Embodiment C: generating molecules with excellent thermal performance by combining multiple discriminant models with generative models, specifically including steps C1-C4.

[0290] Step C1: constructing a glass transition temperature discriminant model and a thermal decomposition temperature discriminant model;

[0291] Specifically, the process of constructing the glass transition temperature discriminant model and the thermal decomposition temperature discriminant model includes steps C1.1-C1.5:

[0292] Step C1.1: Glass transition temperature and thermal decomposition temperature data set;

[0293] The glass transition temperature data set has 8075 data, and the thermal decomposition temperature data set has 1742 data. The initial features of these two data sets are only the polymer monomer SMILES formula, and there is also a label: glass transition temperature or thermal decomposition temperature. Among them, the thermal decomposition temperature is filtered to only retain the temperature corresponding to a 5% mass reduction when heated, that is, T d5 .

[0294] See Figure 7 , Figure 7 The probability density distribution diagram of the glass transition temperature data provided by the embodiment of the application is shown in Figure 7 , where the abscissa represents the glass transition temperature (°C), and the ordinate represents the probability density.

[0295] See Figure 8 , Figure 8 The probability density distribution diagram of the thermal decomposition temperature data provided by the embodiment of the application is shown in Figure 8The abscissa Thermal decomposition temperature (°C) indicates the thermal decomposition temperature, and the ordinate indicates the probability density.

[0296] Step C1.2: Feature extraction;

[0297] The Mordred descriptor is used to extract the structural features of the polymer SMILES formula, including 31 specific features related to polarization performance, hydrogen bond, rotation bond, mass, acidity and basicity, and the number of atoms.

[0298] Step C1.3: Data splitting;

[0299] The data splitting part is the same as example B, and the glass transition temperature and thermal decomposition temperature data set is split into training set and test set, wherein the training set is 80% of the total data, and the test set is 20% of the total data.

[0300] Step C1.4: Feature screening;

[0301] The feature screening part is roughly the same as example B, and this example does not use RFECV. After screening, the number of features of the glass transition temperature and thermal decomposition temperature data is 18 and 20, and the correlation matrix of the features is shown in Figure 9 and Figure 10 .

[0302] Figure 9An illustration of a correlation matrix featuring glass transition temperature data provided by embodiments of the present application, polymer apol represents polymer nonpolarity, polymer bpol represents polymer polarity, polymer nHBAcc represents polymer number of hydrogen bond acceptors, polymer nHBDon represents polymer number of hydrogen bond donors, polymer nRot represents polymer number of rotatable bonds, polymer AMW represents polymer average molecular weight, polymer nBase represents polymer number of bases, polymer nBridgehead represents polymer number of bridgehead atoms, polymer nHetero represents polymer number of heteroatoms, polymer nH represents polymer number of hydrogen atoms, polymer nN represents polymer number of nitrogen atoms, polymer nO represents polymer number of oxygen atoms, polymer nS represents polymer number of sulfur atoms, polymer nP represents polymer number of phosphorus atoms, polymer nF represents polymer number of fluorine atoms, polymer nCl represents polymer number of chlorine atoms, polymer nBr represents polymer number of bromine atoms, polymer Kier2 represents polymer Kier index 2, Tg represents glass transition temperature, Upper Triangle represents upper triangle, Spearman Correlation represents Spearman correlation coefficient, Lower Triangle represents lower triangle, Pearson Correlation represents Pearson correlation coefficient.

[0303] Figure 10A correlation matrix diagram for the characteristics of thermal decomposition temperature data provided by the embodiments of the present application is shown in the following table. polymer apol represents the non-polarity of the polymer, polymer bpol represents the polarity of the polymer, polymer nHBAcc represents the number of hydrogen bond acceptors of the polymer, polymer nHBDon represents the number of hydrogen bond donors of the polymer, polymer nRot represents the number of rotatable bonds of the polymer, polymer MW represents the molecular weight of the polymer, polymer AMW represents the average molecular weight of the polymer, polymer nBase represents the number of bases of the polymer, polymer nHeavyAtom represents the number of heavy atoms of the polymer, polymer nBridgehead represents the number of bridgehead atoms of the polymer, polymer nHetero represents the number of heteroatoms of the polymer, polymer nC represents the number of carbon atoms of the polymer, polymer nN represents the number of nitrogen atoms of the polymer, polymer nO represents the number of oxygen atoms of the polymer, polymer nS represents the number of sulfur atoms of the polymer, polymer nP represents the number of phosphorus atoms of the polymer, polymer nF represents the number of fluorine atoms of the polymer, polymer nCl represents the number of chlorine atoms of the polymer, polymer Kier1 represents the Kier index 1 of the polymer, polymer Kier2 represents the Kier index 2 of the polymer, T d5 represents the thermal decomposition temperature.

[0304] Step C1.5: Model training and evaluation

[0305] The model training and evaluation part is the same as that of Example B. The evaluation results of the glass transition temperature discriminant model are shown in Table 2 below:

[0306] Table 2 Evaluation results of glass transition temperature discriminant model

[0307]

[0308] Table 3 Evaluation results of glass transition temperature discriminant model

[0309]

[0310] It is found through comprehensive comparison that for the glass transition temperature and the thermal decomposition temperature, the prediction effect of the XGBoost model is relatively good, and the algorithm model training speed is the fastest and the time consumption is the shortest, so the model constructed by this algorithm is selected as the glass transition temperature and the thermal decomposition temperature discriminant model.

[0311] Step C2: Constructing a molecular generation model based on graph grammar and combining it with the discriminant model

[0312] Specifically, the process of constructing the molecular generation model and combining it with the discriminant model includes the following steps C2.1-C2.2:

[0313] Step C2.1: Constructing a polymer monomer molecular structure data set;

[0314] This step is the same as Example B.

[0315] Step C2.2: Setting the objective function of graph grammar rule optimization;

[0316] This step is substantially the same as Example B. The evaluation indicators selected in this embodiment include diversity, RS, glass transition temperature score, and thermal decomposition temperature score, and the corresponding weights are λ1=1, λ2=1, λ3=2, and λ4=2.

[0317] The glass transition temperature threshold truncation normalization formula is:

[0318] ;

[0319] The thermal decomposition temperature threshold truncation normalization formula is:

[0320] ;

[0321] Step C3: Using the discriminant model to guide the construction of the graph grammar generation model:

[0322] This part is the same as Example B. Finally, an optimized graph grammar rule generation model can be obtained, which combines the glass transition temperature and thermal decomposition temperature discriminant models and can be used to generate molecules with excellent thermal performance.

[0323] Step C4: Generating samples by the generation model and predicting the performance of the discriminant model:

[0324] Randomly generate 2K molecular samples using the best graph grammar rule generation model, and predict their glass transition temperature and thermal decomposition temperature using the discriminant model. Examples of molecules with excellent thermal performance generated are shown in Figure 11 , Figure 11 is a schematic diagram of a molecule with excellent thermal performance provided by an embodiment of the present application, A represents a specific atom or chemical group, O represents an oxygen atom, N represents a nitrogen atom, P represents a phosphorus atom, F represents a fluorine atom, T d represents the thermal decomposition temperature (such as 429.589°C, 400.651°C, 424.797°C, 397.963°C), T g represents the glass transition temperature (such as -47.960°C, -57.374°C, -48.916°C, -60.827°C).

[0325] Example D: generating molecules with excellent comprehensive performance by combining multiple discriminant models with generative models, specifically including the following steps D1-D4.

[0326] Step D1: constructing discriminant models for each performance:

[0327] Among them, the discriminant models of conductivity, glass transition temperature and thermal decomposition temperature are the same as those in Example B and Example C. The construction of the discriminant model of lithium ion transference number is introduced below in steps D1.1-D1.5:

[0328] Step D1.1: constructing a lithium ion transference number dataset;

[0329] The lithium ion transference number dataset has a total of 3959 data, and the initial features include polymer monomer SMILES, polymer molecular weight and lithium salt. Please refer to Figure 12 , Figure 12 The probability density distribution diagram of the lithium ion transference number data provided by the embodiments of the present application is shown in FIG. 1, where the abscissa Li + transference number represents the lithium ion transference number, and the ordinate represents the probability density.

[0330] Step D1.2: feature extraction;

[0331] This step uses Mordred descriptors to extract structural features for polymer SMILES, including 31 specific features related to polarization performance, hydrogen bonding, rotation bond, mass, acidity and basicity, and number of atoms.

[0332] Step D1.3: data splitting;

[0333] The data splitting part is the same as that in Example 1, and the lithium ion transference number dataset is split into a training set and a test set, wherein the training set is 80% of the total data and the test set is 20% of the total data.

[0334] Step D1.4: feature selection;

[0335] The feature selection part is roughly the same as that in Example 1, and RFECV is not used in this embodiment. After the lithium ion transference number data is filtered, the number of features is 12, and the correlation matrix of the features is shown in Figure 13 Figure 13 ​A correlation matrix diagram provided by the embodiment of the present application for the characteristics of lithium ion transference number data, wherein molality represents molar concentration, polymer_molecular_weight represents polymer molecular weight, nHBAcc represents the number of hydrogen bond acceptors, nHBDon represents the number of hydrogen bond donors, nRot represents the number of rotatable bonds, RotRatio represents the rotation ratio, AMW represents the average molecular weight, nAcid represents the number of acid groups, nBase represents the number of bases, nN represents the number of nitrogen atoms, nS represents the number of sulfur atoms, nF represents the number of fluorine atoms, and transference_number represents transference number.

[0336] Step D1.5: Model training and evaluation

[0337] The model training and evaluation part is the same as that in Embodiment One. The evaluation results of the lithium ion transference number discriminant model are shown in Table 4 below:

[0338] Table 4 Evaluation results of lithium ion transference number discriminant model

[0339]

[0340] It is found through comprehensive comparison that, for the lithium ion transference number, the prediction effect of the XGBoost model is relatively good, and the algorithm model training speed is the fastest and the time consumption is the shortest, so the model constructed by the algorithm is selected as the lithium ion transference number discriminant model.

[0341] Step D2: Constructing a molecule generation model based on graph grammar and combining it with the discriminant model

[0342] Specifically, step D2 can specifically include step D2.1 and step D2.2.

[0343] Step D2.1: Polymer monomer molecular structure data set

[0344] This step is the same as Embodiment B.

[0345] Step D2.2: Setting the objective function for optimizing the graph grammar rule

[0346] This step is substantially the same as Embodiment B. In this embodiment, the evaluation indexes selected include diversity, RS, conductivity score, glass transition temperature score, thermal decomposition temperature score, and lithium ion transference number score, and the corresponding weights are λ1=1, λ2=1, λ3=2, λ4=2, λ5=2, and λ6=2.

[0347] The threshold cut-off normalization formula of the three performance prediction functions of conductivity score, glass transition temperature score, and thermal decomposition temperature score is the same as that in Embodiments B and C. The threshold cut-off normalization formula of lithium ion transference number is:

[0348] ;

[0349] Step D3: Guidance of building graph grammar generation model by discriminant model:

[0350] This step is the same as example B, and finally an optimized graph grammar rule generation model is obtained, which combines the discriminant models of conductivity, glass transition temperature, thermal decomposition temperature and lithium ion transference number, and can be used to generate molecules with excellent comprehensive performance.

[0351] Step D4: Generation of model generated samples and discriminant model prediction performance:

[0352] 2K molecules samples are randomly generated by the best graph grammar rule generation model, and the glass transition temperature and thermal decomposition temperature are predicted by the discriminant model. Please refer to Figure 14 , Figure 14 A schematic diagram of a generated molecule with excellent comprehensive performance provided by the embodiments of the present application. In the figure, A represents a specific atom or chemical group, O represents an oxygen atom, H represents a hydrogen atom, N represents a nitrogen atom, the logarithmic value of conductivity log Conductivity is -7.242 log S cm -1 , -7.110 log S cm -1 , -7.633 log S cm -1 , -7.517 log S cm -1 , the lithium ion transference number Li + transference number is 0.204, 0.195, 0.201 and 0.201, the thermal decomposition temperature T d is 421.673 °C, 403.030 °C, 423.244 °C and 427.649 °C, and the glass transition temperature T g is 197.984 °C, 109.603 °C, 117.496 °C and 190.446 °C.

[0353] Please refer to Figure 15 , Figure 15 A flow chart of a method for designing lithium battery polymer electrolyte by coupling discriminant and generative model provided by the embodiments of the present application, comprising the following steps: (1) collection and arrangement of data set; (2) establishment of discriminant model; (3) establishment of generative model; (4) guidance of discriminant model to generative model.

[0354] The embodiment adopts a machine learning model to directly predict the ionic conductivity of a solid-state electrolyte, constructs a structure-performance relationship of a polymer electrolyte through learning and analysis of a large amount of data sets, and thus predicts the ionic conductivity. The embodiment uses a machine learning method to construct a discriminant model to predict the core performance indicators of a lithium metal battery polymer electrolyte. The embodiment uses a graph grammar generation rule to ensure the performance of the generated molecules.

[0355] The embodiment uses a generative model based on graph grammar that can be trained with very little data to solve the problem of few polymer types of a lithium metal battery polymer solid-state electrolyte. The embodiment constructs a structured data set of a lithium metal battery solid-state polymer electrolyte. The embodiment establishes a discriminant model of the core performance indicators of a polymer for a lithium metal battery. The embodiment couples a multi-objective discriminant model and a generative model of a polymer solid-state electrolyte to inversely design a polymer solid-state electrolyte.

[0356] The system can include:

[0357] A data set acquisition module is configured to acquire a performance data set and a molecular structure formula data set of a polymer electrolyte.

[0358] A first model construction module is configured to construct a discriminant model using the performance data set. The discriminant model is configured to calculate a performance indicator of a molecular structure formula. The performance indicator includes any one or a combination of several of ionic conductivity, lithium ion transference number, glass transition temperature, and thermal decomposition temperature.

[0359] A second model construction module is configured to construct a generative model using the molecular structure formula data set. The generative model is configured to generate a molecular structure formula of a polymer electrolyte.

[0360] A model training module is configured to train the generative model based on the discriminant model.

[0361] A molecule generation module is configured to, if a molecular structure design task is received, generate a plurality of new molecular structure formulas of polymer electrolytes using the trained generative model, and screen the new molecular structure formulas of polymer electrolytes using the discriminant model to obtain a molecular structure formula that meets the molecular structure design task.

[0362] The embodiment builds a discriminant model through a performance data set of the polymer electrolyte, the discriminant model can learn the relationship between different molecular structures and performances, and the discriminant model can be used to evaluate whether a newly generated molecular structure meets specific performance requirements. The embodiment builds a generative model through a molecular structure formula data set of the polymer electrolyte, the generative model can automatically generate a new molecular structure formula of the polymer electrolyte according to a known molecular structure. The embodiment trains the generative model based on the discriminant model, so that the generative model can generate a molecular structure formula with a performance index meeting requirements. After receiving a molecular structure design task, the embodiment can generate a plurality of new molecular structure formulas of the polymer electrolyte by using the generative model, and screen the molecular structure formulas by using the discriminant model to obtain a molecular structure formula meeting the molecular structure design task. The above process uses the discriminant model to guide the generative model to generate the molecular structure formula, without human intervention, so that the embodiment can automatically generate the molecular structure formula of the polymer electrolyte, and improve the generation efficiency and effectiveness of the molecular structure formula.

[0363] Further, the method further comprises:

[0364] The data set processing module is configured to perform data cleaning and data visualization on the performance data set and the molecular structure formula data set after obtaining the performance data set and the molecular structure formula data set of the polymer electrolyte.

[0365] Further, the process in which the first model construction module builds the discriminant model by using the performance data set comprises: extracting data features from the performance data set, and performing standardization and normalization processing on the data features; filtering the data features to obtain target data features; and performing machine learning modeling based on the target data features and corresponding labels in the performance data set to obtain the discriminant model.

[0366] Further, the performance data set comprises any one or a combination of the conductivity data set, the lithium ion transference number data set, the glass transition temperature data set and the thermal decomposition temperature data set; the conductivity data set comprises a polymer structure, a polymer molecular weight, a lithium salt structure, a lithium salt concentration, a temperature and a conductivity label; the lithium ion transference number data set comprises a polymer structure, a lithium salt structure, a lithium salt concentration, a temperature, an additive type, an additive content and a lithium ion transference number label; the glass transition temperature data set comprises a polymer structure and a glass transition temperature label; and the thermal decomposition temperature data set comprises a polymer structure and a thermal decomposition temperature label.

[0367] Correspondingly, the process in which the first model construction module performs machine learning modeling based on the target data features and corresponding labels in the performance data set to obtain the discriminant model comprises:

[0368] perform machine learning modeling based on target data features in the conductivity dataset and conductivity labels to obtain a conductivity discriminant model;

[0369] and / or, perform machine learning modeling based on target data features in the lithium ion transference number dataset and lithium ion transference number labels to obtain a lithium ion transference number discriminant model;

[0370] and / or, perform machine learning modeling based on target data features in the glass transition temperature dataset and glass transition temperature labels to obtain a glass transition temperature discriminant model;

[0371] and / or, perform machine learning modeling based on target data features in the thermal decomposition temperature dataset and thermal decomposition temperature labels to obtain a thermal decomposition temperature discriminant model.

[0372] Further, the process of constructing a generative model by the second model construction module using the molecular structure formula dataset includes: converting the molecular structure in the molecular structure formula dataset into a graph structure; wherein the nodes in the graph structure represent atoms and the edges represent chemical bonds; generating a corresponding graph grammar from the graph structure, and constructing the generative model based on the graph grammar.

[0373] Further, the process of training the generative model by the model training module based on the discriminant model includes: constructing an objective function of the generative model; wherein the value of the objective function is the weighted calculation result of a plurality of evaluation indicators, and the evaluation indicators include scores corresponding to performance indicators calculated by the discriminant model; iteratively training the generative model based on the objective function.

[0374] Further, the process of iteratively training the generative model by the model training module based on the objective function includes: controlling the generative model to generate an i-th round of output results; calculating the value of the objective function corresponding to the i-th round of output results; determining whether the value of the objective function calculated last time is within a preset interval; if yes, determining that the training of the generative model is complete; if no, adjusting the parameters of the generative model, increasing the value of i by 1, and entering the operation of controlling the generative model to generate the i-th round of output results;

[0375] The evaluation index further includes an effectiveness index, a uniqueness index, a novelty index, a diversity index, a reverse synthesis score index, and a membership degree index; the effectiveness index is used to describe a proportion of effective molecular structures in the output result of the generative model; the uniqueness index is used to describe a proportion of non-repeated molecular structures in the output result of the generative model; the novelty index is used to describe a proportion of molecular structures not in the molecular structure formula data set in the output result of the generative model; the diversity index is used to describe structural similarity between the output results of the generative model; the reverse synthesis score index is used to describe a synthesis success rate of the output result of the generative model; and the membership degree index is used to describe a proportion of molecular structures belonging to the molecular structure formula data set in the output result of the generative model.

[0376] Since the embodiments of the system part correspond to the embodiments of the method part, the embodiments of the system part are described in the description of the embodiments of the method part, and are not described here.

[0377] The application further provides a storage medium having a computer program stored thereon, and the computer program can implement the steps provided by the above embodiments when executed. The storage medium can include a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0378] The application further provides an electronic device, which can include a memory and a processor, the memory has a computer program stored therein, and the processor can implement the steps provided by the above embodiments when calling the computer program in the memory. Of course, the electronic device can also include various network interfaces, power supplies and other components.

[0379] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts of each embodiment can be referred to each other. For the system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the related parts can be referred to the method part. It should be pointed out that, for those skilled in the art, without departing from the principles of the application, some improvements and modifications can be made to the application, and these improvements and modifications also fall within the protection scope of the application.

[0380] It is further noted that the terminology "first", "second" and the like used in the specification are merely used for differentiating one entity or action from another, and do not necessarily imply any actual physical or logical relationship or order between such entities or actions. Moreover, the use of the term "including", "containing" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or apparatus that comprises a list of elements is not required to comprise only those elements but can include other elements not expressly listed or inherent to such process, method, article or apparatus. An element preceded by "comprises a..." does not, without further restriction, preclude the existence of additional elements of the same nature as those recited.

Claims

1. A method for generating a molecular structure formula of a polymer electrolyte, characterized by, The method comprises the following steps: obtaining a performance dataset and a molecular structure formula dataset of a polymer electrolyte; constructing a discriminant model using the performance dataset; wherein the discriminant model is used to calculate a performance indicator of a molecular structure formula, and the performance indicator comprises any one or a combination of several of the following: conductivity, lithium ion transference number, glass transition temperature, and thermal decomposition temperature; constructing a generative model using the molecular structure formula dataset; wherein the generative model is used to generate a molecular structure formula of a polymer electrolyte; training the generative model based on the discriminant model; if a molecular structure design task is received, generating a plurality of new molecular structure formulas of polymer electrolytes using the trained generative model, and screening the new molecular structure formulas of polymer electrolytes using the discriminant model to obtain a molecular structure formula that meets the molecular structure design task; wherein constructing a generative model using the molecular structure formula dataset comprises: converting the molecular structure in the molecular structure formula dataset into a graph structure; wherein the nodes in the graph structure represent atoms, and the edges represent chemical bonds; generating a corresponding graph grammar from the graph structure, and constructing the generative model based on the graph grammar; wherein training the generative model based on the discriminant model comprises: constructing an objective function of the generative model; wherein the value of the objective function is the weighted calculation result of a plurality of evaluation indicators, and the evaluation indicators include the scores corresponding to the performance indicators calculated by the discriminant model; and iteratively training the generative model based on the objective function.

2. The method for generating the molecular structure formula of the polymer electrolyte according to claim 1, characterized in that, After obtaining the performance dataset and the molecular structure formula dataset of the polymer electrolyte, the method further comprises the following steps: performing data cleaning and data visualization operations on the performance dataset and the molecular structure formula dataset.

3. The method for generating the molecular structure formula of the polymer electrolyte according to claim 1, characterized in that, Constructing a discriminant model using the performance dataset comprises the following steps: extracting data features from the performance dataset, and performing standardization and normalization processing on the data features; filtering the data features to obtain target data features; performing machine learning modeling based on the target data features and corresponding labels in the performance dataset to obtain the discriminant model.

4. The method for generating the molecular structure formula of the polymer electrolyte according to claim 3, characterized in that, The performance dataset comprises any one or a combination of several of the following: a conductivity dataset, a lithium ion transference number dataset, a glass transition temperature dataset, and a thermal decomposition temperature dataset; the conductivity dataset comprises polymer structure, polymer molecular weight, lithium salt structure, lithium salt concentration, temperature, and conductivity label; the lithium ion transference number dataset comprises polymer structure, lithium salt structure, lithium salt concentration, temperature, additive type, additive content, and lithium ion transference number label; the glass transition temperature dataset comprises polymer structure and glass transition temperature label; and the thermal decomposition temperature dataset comprises polymer structure and thermal decomposition temperature label. Correspondingly, performing machine learning modeling based on the target data features and corresponding labels in the performance dataset to obtain the discriminant model comprises the following steps: performing machine learning modeling based on the target data features and conductivity label in the conductivity dataset to obtain a conductivity discriminant model; And / or, based on the target data features and lithium ion transference number labels in the lithium ion transference number data set, machine learning modeling is performed to obtain a lithium ion transference number discriminant model; And / or, based on the target data features and glass transition temperature labels in the glass transition temperature data set, machine learning modeling is performed to obtain a glass transition temperature discriminant model; And / or, based on the target data features and thermal decomposition temperature labels in the thermal decomposition temperature data set, machine learning modeling is performed to obtain a thermal decomposition temperature discriminant model.

5. The method for generating the molecular structure formula of the polymer electrolyte according to claim 1, characterized in that, Iterative training of the generative model based on the target function includes: Controlling the generative model to generate an i-th round output result; Calculating the value of the target function corresponding to the i-th round output result; Determining whether the value of the target function calculated last time is within a preset interval; If yes, it is determined that the training of the generative model is complete; If not, adjust the parameters of the generative model, increase the value of i by 1, and enter the operation of controlling the generative model to generate an i-th round output result; The evaluation index further includes effectiveness index, uniqueness index, novelty index, diversity index, inverse synthesis score index and membership index; the effectiveness index is used to describe the proportion of effective molecular structures in the output results of the generative model; the uniqueness index is used to describe the proportion of non-repeated molecular structures in the output results of the generative model; the novelty index is used to describe the proportion of molecular structures not in the molecular structure formula data set in the output results of the generative model; the diversity index is used to describe the structural similarity between the output results of the generative model; the inverse synthesis score index is used to describe the synthesis success rate of the output results of the generative model; and the membership index is used to describe the proportion of molecular structures belonging to the molecular structure formula data set in the output results of the generative model.

6. A system for generating a molecular structure formula of a polymer electrolyte, characterized by, It includes: The data set acquisition module is used to acquire the performance data set and the molecular structure formula data set of the polymer electrolyte; The first model construction module is used to construct a discriminant model by using the performance data set; wherein the discriminant model is used to calculate the performance index of the molecular structure formula, and the performance index includes any one or combination of any number of items in the conductivity, lithium ion transference number, glass transition temperature and thermal decomposition temperature; The second model construction module is used to construct a generative model by using the molecular structure formula data set; wherein the generative model is used to generate the molecular structure formula of the polymer electrolyte; The model training module is used to train the generative model based on the discriminant model; The molecule generation module is used to generate a plurality of new molecular structure formulas of the polymer electrolyte by using the trained generative model if a molecular structure design task is received, and to screen the new molecular structure formulas of the polymer electrolyte by using the discriminant model to obtain a molecular structure formula meeting the molecular structure design task; The molecule generation module is used to generate a plurality of new molecular structure formulas of the polymer electrolyte by using the trained generative model if a molecular structure design task is received, and to screen the new molecular structure formulas of the polymer electrolyte by using the discriminant model to obtain a molecular structure formula meeting the molecular structure design task; The process of constructing the second model construction module to generate a generative model using the molecular structure formula data set comprises: converting the molecular structure in the molecular structure formula data set into a graph structure; wherein the nodes in the graph structure represent atoms and the edges represent chemical bonds; generating a corresponding graph grammar from the graph structure, and constructing the generative model based on the graph grammar; The process of training the generative model based on the discriminative model by the model training module comprises: constructing an objective function of the generative model; wherein the value of the objective function is the weighted calculation result of a plurality of evaluation indexes, and the evaluation indexes include the scores corresponding to the performance indexes calculated by the discriminative model; and iteratively training the generative model based on the objective function.

7. An electronic device, comprising: A memory and a processor are included, the memory has a computer program stored therein, and the processor realizes the steps of the polymer electrolyte molecular structure formula generation method according to any one of claims 1 to 5 when calling the computer program in the memory.

8. A storage medium, characterized by The storage medium has computer executable instructions stored therein, and the computer executable instructions are loaded and executed by the processor to realize the steps of the polymer electrolyte molecular structure formula generation method according to any one of claims 1 to 5.

Images