Ionic liquid electrolyte intelligent design method for extreme high temperature requirement
Through deep learning and intelligent optimization algorithms, intelligent design of the entire process from molecular structure to formula optimization is achieved, solving the problems of low R&D efficiency and insufficient accuracy in traditional methods, and is suitable for the application of ionic liquid electrolytes in extreme high temperature environments.
Patent Information
- Application Number
- CN202510871013.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-26
- Publication Date
- 2025-10-21
AI Technical Summary
Existing technologies rely on traditional trial-and-error methods in the development of extreme high-temperature ionic liquid electrolytes, which are inefficient and costly, and lack intelligent molecular structure coding, performance prediction, and combination optimization, making it difficult to meet performance requirements in high-temperature environments.
By adopting deep learning and intelligent optimization algorithms, through the construction of structured data sets, machine learning model training and multiple filtering mechanisms, we can achieve intelligent design of the entire process from molecular structure to formula optimization, and quickly screen out ionic liquid formulas with both thermal stability and good electrochemical properties.
It significantly improves R&D efficiency, shortens the cycle by 75%, and improves design reliability and accuracy, making it suitable for electrolyte applications in extreme high-temperature environments.
Smart Images
Figure CN120823908A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the interdisciplinary technical field of electrolyte material design and artificial intelligence, and relates to an intelligent design method for ionic liquid electrolytes for extreme high temperature requirements. Background Art
[0002] As high-temperature energy conversion and storage systems become increasingly complex and extreme, traditional organic electrolytes have been unable to meet the operational requirements of demanding operating conditions such as high temperatures due to their poor thermal stability, high volatility, and serious safety hazards. Ionic liquids, room-temperature molten salts composed of anions and cations, have become one of the core electrolyte materials for building a new generation of high-temperature energy storage systems (such as high-temperature lithium / sodium batteries, solid oxide fuel cells, spacecraft power supplies, geothermal / deep well energy devices, etc.) due to their unique low volatility, high thermal decomposition temperature, wide electrochemical window, and excellent electrochemical and interfacial stability. At the same time, their intrinsic non-flammability and chemical inertness are expected to completely break through the safety bottleneck of traditional liquid electrolytes at high temperatures, extend the cycle life of devices, and improve system operational reliability. Therefore, the development of high-temperature ionic liquid electrolytes is not only a key breakthrough in the development of high-temperature, high-safety, and high-efficiency energy devices, but also an important basic support for the construction of green, safe, and intelligent energy systems.
[0003] However, despite the significant performance potential of ionic liquids, rapid screening for ionic liquids that combine strong thermal stability with excellent electrochemical performance remains a major challenge in current research and development. Specifically, current research and development faces the following bottlenecks: 1) Reliance on iterative trial-and-error methods, resulting in low R&D efficiency and high costs: Currently, the development of ionic liquid electrolytes resistant to extreme temperatures relies primarily on the traditional "Edison-style" trial-and-error approach. Researchers must invest considerable time and resources to individually formulate ionic liquids of varying compositions and conduct tedious electrochemical performance tests under various extreme high-temperature conditions. This process is not only time-consuming and resource-intensive, but also faces extremely limited experimental coverage with tens of thousands of potential combinations, making comprehensive and efficient screening difficult. 2) The complex structure-performance relationship of ionic liquids limits the effectiveness of traditional prediction methods: Due to the highly tunable molecular structure of ionic liquids, their performance is significantly influenced by their microstructure and exhibits a high degree of nonlinearity. Current performance prediction methods based on rule-based inference and molecular simulations struggle to fully and accurately describe their electrochemical performance at extreme temperatures, lacking an efficient and systematic performance prediction and screening mechanism. 3) Lack of a systematic intelligent design and screening platform: Existing research mostly focuses on local performance prediction or experimental verification, and lacks an intelligent generation path that integrates molecular structure encoding, performance prediction and combinatorial optimization. This makes it difficult to support targeted matching and precise recommendations of performance requirements under complex working conditions such as extremely high temperatures.
[0004] Therefore, there is an urgent need for an intelligent design method that integrates molecular structure expression, deep learning modeling, performance prediction and formula generation optimization to quickly screen and predict ionic liquid electrolyte formulas with excellent performance under extreme high temperature conditions, and realize the intelligent design and optimization of ionic liquid electrolytes from "data-structure-performance". Summary of the Invention
[0005] In response to the problems in the existing technology of relying on traditional experimental trial and error methods for the development of extreme high-temperature ionic liquid electrolytes, which lead to low R&D efficiency and high costs, and the performance prediction methods based on rule deduction or simple simulation have inaccurate structure-performance relationship modeling and insufficient multi-objective optimization capabilities. The present invention provides an intelligent design method for ionic liquid electrolytes for extreme high-temperature needs. This method is based on deep learning and intelligent optimization algorithms to achieve full-process intelligence from molecular structure and performance prediction to formula optimization, and efficiently screen ionic liquid formulas with both strong thermal stability and good electrochemical properties, significantly improving the R&D efficiency of electrolytes in extreme high-temperature environments, and promoting the application of ionic liquid electrolytes under extreme high-temperature conditions in key fields such as aerospace and new energy.
[0006] The technical solutions of the present invention are as follows:
[0007] The intelligent design method of ionic liquid electrolytes for extreme high temperature requirements includes the following steps:
[0008] (1) Construction of the initial data set: The data platform was retrieved through the structured query language SQL and the electrochemical performance data of ionic liquids under extreme high temperatures were screened. The data were systematically integrated according to the four-element definition model of ion combination-environmental parameters-surface characteristics-electrochemical performance to establish the initial data set. The screening conditions were: the decomposition temperature T when the mass loss was 5% in the thermogravimetric analysis (TGA) was met. d ≥250℃, the viscosity range is 20-40cP when measured by a rotational rheometer at 1450±50rpm in a nitrogen atmosphere at a test temperature of 240℃, the conductivity range is 10-15mS / cm when measured in a N2 atmosphere, and the electrochemical window range is 2.6V-4.8V when measured in an argon atmosphere;
[0009] (2) Supplementation of missing entries in the dataset and cleaning of abnormal data: Retrieve literature related to the electrochemical performance of ionic liquids and input it into the semantic model BERT to identify and extract the electrochemical performance data of ionic liquids in the literature and supplement the blank data; Based on the calculation results of molecular dynamics simulation, set the theoretical threshold of data confidence and automatically clean the data points that deviate from the confidence interval to obtain the optimized dataset;
[0010] (3) Model training: The optimized dataset is pre-processed with ion feature engineering to stack the electrochemical performance data of ionic liquids into 20 dimensions for easy machine learning; the dataset is stratified and sampled according to the temperature range and divided into training and test sets in proportion; a variety of machine learning algorithms are used to deeply train the generative model, and the algorithm with the best training results is selected to preliminarily train the generative model and define the model's hyperparameter space. The above training set is input into the machine learning model for training, and the optimal value of the hyperparameter vector is further converged based on grid search to improve the model accuracy;
[0011] (4) Model accuracy evaluation and optimization: The test set is input into the machine learning model obtained in step (3) and the R 2 The accuracy of ionic liquid generation by the RMAE evaluation model was evaluated; the bird's flock algorithm was adaptively improved by introducing classical fuzzy reasoning and inertial particles to enhance the global optimization capability; the bird's flock update formula was adaptively improved to reduce the impact factor as the number of algorithm iterations increased; chaotic perturbations were used to increase the diversity of the algorithm's later iterations. Combined with the improved bird's flock algorithm IBSA, a global search of the data points in the training set was performed, each data point was marked one by one and shared, a few abnormal data points were excluded, and the learning process converged to data points with higher accuracy. The model's description of the structure-activity relationship of the electrochemical properties of ionic liquids was optimized to further improve the model accuracy. The adaptively improved bird's flock update formula is as follows:
[0012]
[0013] Where i is the ordinal number of the individual in the flock, j is the ordinal number of the dimension, t is the number of iterations, and FL is the adaptive influencing factor. When the number of iterations increases, FL decreases linearly.
[0014] (5) Formula generation: Input the set extreme high temperature environmental parameters and electrochemical surface characteristics into the ionic liquid electrolyte generation model for extreme high temperature requirements, and screen out the ionic liquid electrolyte that can meet the set extreme high temperature working conditions.
[0015] Furthermore, in step (1), the data platform includes NIST ILThermo, v3.2 and Web of Science data platform.
[0016] Furthermore, in step (2), the literature includes SCI-indexed literature and ASTM standard experimental data.
[0017] Furthermore, in step (1), the ion combination is a combination of anion and cation species, the environmental parameters include temperature and humidity, the surface characteristics include roughness and charge density, and the electrochemical properties include viscosity and conductivity.
[0018] Furthermore, step (2) is specifically as follows: for the missing entries of the initial data set, search from the SCI-indexed literature, identify and extract the semantics of ionic liquids in the literature through the BERT model, extract electrochemical performance parameters from the literature to supplement the blank data, and use the graph neural network GNN to map the ionic structural features of image data such as ionic structural formulas and 3D models in the literature, and convert them into SMILES format for encoding and storage; based on the calculation results of molecular dynamics simulation, set the theoretical threshold of data confidence, automatically clean the data points that deviate from the confidence interval, and obtain the optimized data set.
[0019] Furthermore, step (3) is as follows: preprocess the data with ion feature engineering, using RDKit The 2022.09 toolkit converts the SMILES-encoded ionic structural features into 208 feature descriptors, and further stacks the 208 feature descriptors into 20 dimensions that can accurately reflect the differences in ionic structural features of the dataset based on the one-hot encoding and Pearson correlation coefficient heat map shared or not shared by the anion and cation feature descriptors; the dataset is divided into training and test sets in a 7:3 ratio based on stratified sampling of each temperature interval of 10K to ensure that the data can be evenly divided into training and test sets according to temperature, reducing model underfitting or overfitting caused by division differences; a variety of machine learning algorithms are used to deeply train the generative model, and the algorithm with the best training results is selected to preliminarily train the generative model and define the model's hyperparameter space. The dimensions of the hyperparameter space include depth, number of iterations, and learning rate. The above training set is trained using a conditional generative machine learning model that integrates the CatBoost framework and the VAE encoder. Grid search and 10CV cross-validation methods are used to determine the value range of the better hyperparameter vector, narrowing the value range of the hyperparameter vector to improve the accuracy of the model.
[0020] Furthermore, in step (3), the machine learning algorithm includes but is not limited to support vector machine (SVM), random forest (RF), etc.
[0021] Furthermore, in step (4), by R 2 The accuracy of the model in generating ionic liquids was evaluated by ≥0.95 and RMSE≤0.02.
[0022] Furthermore, step (5) is specifically as follows: the set extreme high temperature environmental parameters and electrochemical surface characteristics are input into the ionic liquid electrolyte generation model for extreme high temperature requirements. The model will determine the corresponding hyperparameter vector based on the set temperature range. The generated ion combination will pass through RDKit syntax verification, ΔG<0 thermodynamic criterion, and CatBoost performance prediction triple filtering in sequence, and finally output the ionic liquid candidate formula; the output result is encoded in SMILES and can be directly converted into the structural formula and 3D model of the ion in ChemDraw and Chem3D, which is convenient for structural feature analysis.
[0023] Furthermore, the method further includes step (6): testing the compatibility of candidate ionic liquids with the requirements of extreme high temperature working conditions through artificial synthesis, including high temperature viscosity testing using an Anton Paar MCR302 rheometer, measuring conductivity using a DDS-11A digital conductivity meter, measuring the electrochemical window using an HSV-100 electrochemical workstation and a three-electrode system, and performing thermal stability testing using a thermogravimetric analysis method with a heating rate of 10K / min, to further screen suitable ionic liquids.
[0024] Compared with the prior art, the present invention has the following advantages:
[0025] (1) Achieve efficient screening of extremely high-temperature ionic liquid electrolytes: Using AI's high computing power to learn the electrochemical structure-activity relationship of ionic liquids instead of traditional trial-and-error experiments can shorten the R&D cycle by 75%;
[0026] (2) Multimodal data fusion improves design reliability: The domain-specific BERT model is used to accurately extract literature performance parameters, and GNN is used to analyze ionic structures and 3D models, which are uniformly converted into SMILES encoding and storage to solve the problem of heterogeneous data integration;
[0027] (3) Generative model-driven active design: The CatBoost machine learning model is integrated with the VAE encoder to build a conditional generative model, which automatically rolls back and generates candidate recipes based on the input working conditions. The improved bird flock algorithm (IBSA) is introduced to globally optimize data point selection, allowing the model to converge to a higher accuracy area, breaking through the accuracy bottleneck of traditional models.
[0028] (4) Triple filtering mechanism ensures the feasibility of the formula: through RDKit syntax verification, ΔG<0 thermodynamic criterion, and CatBoost performance prediction, invalid solutions are triple filtered. The output SMILES code can be directly converted into a structural formula and 3D model to accelerate experimental verification;
[0029] (5) Good scalability: The method framework is universal and can be adapted to more complex material screening tasks by expanding descriptors, improving model structure, or integrating other electrochemical performance indicators. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] Figure 1 It is an automatic extraction platform for electrochemical performance data of ionic liquids from literature based on the BERT model.
[0031] Figure 2 It is a visual annotation relationship data set constructed by the method of the present invention.
[0032] Figure 3 It is a heatmap visualization of the Pearson correlation coefficient matrix (part of the output results).
[0033] Figure 4 It is the ionic liquid data obtained by searching and screening on demand in the NIST ionic liquid database.
[0034] Figure 5 It is the quantitative linear fitting graph and verification graph of conductivity.
[0035] Figure 6 The ionic liquid output of the generative model is displayed as a graphic in the chemical software. DETAILED DESCRIPTION
[0036] The technical solution of the present invention is further described below with reference to specific embodiments and accompanying drawings.
[0037] Example 1
[0038] An intelligent design method for ionic liquid electrolytes for extreme high temperature requirements, screening ionic liquid electrolytes for satellite electronic systems suitable for 250°C environments, includes the following steps:
[0039] 1. Initial dataset construction:
[0040] (1) Through the coordinated call of structured query language (SQL) and application program interface (API), a joint search was performed on the following databases: NIST ionic liquid database (ILThermo v2.0), v3.2 and Web of Science Core Collection (indexed time range 2010-2025).
[0041] (2) Establish a three-level screening system, using a logical "AND" relationship to connect the following conditions:
[0042] (a) Thermal stability condition: decomposition temperature T at which 5% mass loss is observed in thermogravimetric analysis (TGA) d ≥250℃;
[0043] (b) Rheological conditions: The viscosity range of the rheometer was 20–40 cP at a test temperature of 240°C.
[0044] (c) Electrochemical conditions: The conductivity range measured under N2 atmosphere was 10-15 mS / cm, and the electrochemical window range measured under argon atmosphere was 2.6 V-4.8 V.
[0045] like Figure 1 As shown, the data platform v3.2 shows 12,340 records. SQL filtering obtains 632 valid records, and the visual annotation of the filtered data set is as follows: Figure 2 shown.
[0046] (3) Data standardization processing - constructing a four-dimensional feature matrix:
[0047] (a) Ionic combination: cation / anion type, using IUPAC nomenclature;
[0048] (b) Environmental parameters: temperature, humidity, etc.;
[0049] (c) Surface characteristics: roughness, charge density, etc.;
[0050] (d) Electrochemical properties: viscosity, conductivity, and electrochemical window.
[0051] 2. Supplement missing entries in the dataset and clean abnormal data:
[0052] (1) Supplement missing data through literature data mining: Use the scientific field pre-trained model (BERT) to automatically parse the electrochemical performance data of ionic liquids and ASTM standard experimental data. For image data such as structural formulas and 3D models appearing in the literature, their structural features are mapped based on the graph neural network (GNN) and converted to the unified IUPAC ion naming. The extracted molecular structure data is uniformly converted to the SMILES format for encoding and storage.
[0053] (2) Abnormal data cleaning: Based on the calculation results of molecular dynamics simulation, the theoretical threshold of data confidence is set. For data points that deviate from the confidence interval, they are output and manually tested and verified to remove abnormal data.
[0054] (3) The final dataset outputs a structured CSV file, including the following fields:
[0055] (a) Cation / anion SMILES string; (b) Environmental parameters such as temperature and humidity; (c) Surface properties such as roughness and charge density; (d) Viscosity, conductivity, and electrochemical window.
[0056] 15 sets of ASTM D4172-22 standard experimental data were added. A total of 1,248 data items were added, and 43 abnormal data items were removed.
[0057] 3. Generative model training for ionic liquid electrolytes for extreme high temperature requirements:
[0058] (1) Ion feature engineering and data set division: RDKit 2022.09 toolkit is used to calculate 208 molecular descriptors, among which the Wiener index describes the topological structure, the highest occupied molecular orbital can describe the electronic state, and the moment of inertia is used to describe the geometric characteristics. The environmental parameters (temperature, humidity) are Min-Max normalized, and the ionic liquid structural feature descriptors and surface characteristics are further screened and simplified based on One-Hot encoding and Pearson correlation coefficient heat map, and all ionic liquid data dimensions are stacked to 20 dimensions. According to the ratio of 7:3, based on stratified sampling of each temperature interval of 10K, the data set is divided into training set, validation set and test set. The output results of the Pearson correlation coefficient heat map are as follows: Figure 3 shown.
[0059] (2) Model training for the intelligent design of ionic liquid electrolytes for extreme high temperature requirements: The model hyperparameter space was defined, and a CatBoost regression model with a depth of 8 was used. The number of early stopping rounds was set to 50, the learning rate was 0.03, and the L2 regularization coefficient was 5. Grid search and 10CV cross-validation were used to determine the optimal value range of the hyperparameter vector, narrowing the value range of the hyperparameter vector to improve the accuracy of the model.
[0060] 4. Model accuracy evaluation and optimization:
[0061] (1) Accuracy evaluation index: R 2 The accuracy of the ionic liquid generated by the RMAE evaluation model is plotted, and the actual conductivity value-calculated value fitting line is drawn to compare the learning accuracy of the training set and test set models. The fitting results are as follows Figure 5 As shown, the left figure is the training set fitting relationship of conductivity-corresponding state principle analysis, and the determination coefficient R 2 =0.9987; the right figure shows the test set verification relationship of conductivity-corresponding state principle analysis, the determination coefficient R 2 =0.9914.
[0062] (2) Model optimization: Combined with the improved bird swarm algorithm (IBSA), a global search is performed on the data points of the training set to ensure that the learning process converges to data points with higher accuracy, optimize the model's description of the structure-activity relationship of the electrochemical properties of ionic liquids, and further improve the model accuracy.
[0063] 5. Recipe generation:
[0064] (1) Generate multiple filters of results: After the VAE model receives the input of environmental parameters temperature T, humidity ρ and target performance parameter conductivity σ, it first generates 100 candidate molecular structures. Through three levels of filtering: the first level uses RDKit to verify the validity of SMILES syntax; the second level calls ORCA for DFT calculation to screen thermodynamically stable structures with Gibbs free energy ΔG<0; the third level uses the trained CatBoost model to predict electrochemical performance and retains candidate molecules that meet the viscosity, conductivity and electrochemical window requirements. Each run generates 10-20 optimal candidate solutions and ranks them according to CatBoost prediction performance. Evaluation indicator requirements R 2 ≥0.95 and RMSE ≤0.02. For batches that do not meet the quality requirements, the VAE latent space sampling radius is automatically adjusted and regenerated until the quality requirements are met or the maximum number of iterations is reached. The maximum number of iterations is 5 by default.
[0065] The ionic liquid data obtained by searching and screening on demand in the NIST ionic liquid database are as follows: Figure 4 As shown. According to the evaluation index requirements, that is, R 2 ≥0.95 and RMSE≤0.02, the optimal formula is obtained: [EMIM][BF4], the anion and cation structure diagram is as follows Figure 6 As shown, the viscosity of the obtained ionic liquid [EMIM][BF4] is predicted to be 20-40 cP, the thermal decomposition temperature is predicted to be 250°C, the conductivity is predicted to be 10-15 mS / cm, and the electrochemical window is predicted to be 2.6V-4.8V.
[0066] 6. Manual Verification of Output Results: Experimental testing of the candidate ionic liquids' high-temperature viscosity, electrochemical properties, and thermal stability was performed. Specifically, an Anton Paar MCR302 rheometer was used to test the high-temperature viscosity of the candidate ionic liquids; a DDS-11A digital conductivity meter was used to test the electrochemical properties of the candidate ionic liquids under a nitrogen atmosphere; and an HSV-100 electrochemical workstation with a three-electrode system was used to test the electrochemical properties of the candidate ionic liquids under an argon atmosphere. Thermal stability data were obtained by TGA analysis of the ionic liquids at a heating rate of 10°C / min.
[0067] After testing, the viscosity of the ionic liquid [EMIM][BF4] was measured to be 25.2 cP, the thermal decomposition temperature was measured to be 261.1°C, the conductivity was measured to be 14.1 mS / cm, and the electrochemical window was measured to be 4.7 V.
[0068] By comparing the predicted values and the measured values, it can be seen that the method of the present invention is based on deep learning and intelligent optimization algorithms, realizing the intelligence of the entire process from molecular structure and performance prediction to formula optimization, and efficiently screening ionic liquid formulas with both strong thermal stability and good electrochemical properties.
Claims
1. An intelligent design method for ionic liquid electrolytes for extreme high temperature requirements, characterized by: The following steps are involved: (1) Construction of the initial data set: The data platform was searched through the structured query language SQL and the electrochemical performance data of ionic liquids under extreme high temperatures were screened. The data were systematically integrated according to the four-element definition model of ion combination-environmental parameters-surface characteristics-electrochemical performance to establish the initial data set. The screening conditions were: the decomposition temperature T when the mass loss was 5% in the thermogravimetric analysis (TGA) was met. d ≥250°C. Viscosity range is 20-40 cP at 240°C using a rotational rheometer at 1450±50 rpm in a nitrogen atmosphere. Conductivity range is 10-15 mS / cm in a N2 atmosphere. Electrochemical window range is 2.6 V-4.8 V in an argon atmosphere. (2) Supplementation of missing entries in the dataset and cleaning of abnormal data: Retrieve literature related to the electrochemical performance of ionic liquids and input it into the semantic model BERT to identify and extract the electrochemical performance data of ionic liquids in the literature and supplement the blank data; based on the calculation results of molecular dynamics simulation, set the theoretical threshold of data confidence and automatically clean the data points that deviate from the confidence interval to obtain the optimized dataset; (3) Model training: The optimized dataset is pre-processed with ion feature engineering to stack the electrochemical performance data of ionic liquids into 20 dimensions for easier machine learning. The data set is stratified and sampled according to the temperature range, and divided into training set and test set in proportion; Use multiple machine learning algorithms to deeply train the generative model, select the algorithm with the best training results, preliminarily train the generative model and define the model's hyperparameter space. Input the above training set into the machine learning model for training, and further converge on the optimal value of the hyperparameter vector based on grid search to improve model accuracy. (4) Model accuracy evaluation and optimization: Input the test set into the machine learning model obtained in step (3) and use R 2 The accuracy of ionic liquid generation by the RMAE evaluation model was evaluated; the bird's flock algorithm was adaptively improved by introducing classical fuzzy reasoning and inertial particles to enhance the global optimization capability; the bird's flock update formula was adaptively improved to reduce the impact factor as the number of algorithm iterations increased; chaotic perturbations were used to increase the diversity of the algorithm's later iterations. Combined with the improved bird's flock algorithm IBSA, a global search of the data points in the training set was performed, each data point was marked one by one and shared, a few abnormal data points were excluded, and the learning process converged to data points with higher accuracy. The model's description of the structure-activity relationship of the electrochemical properties of ionic liquids was optimized to further improve the model accuracy. The adaptively improved bird's flock update formula is as follows: ， Among them, i is the ordinal number of the individual in the bird flock, j is the ordinal number of the dimension, and t is the number of iterations. FL is the adaptive impact factor, which decreases linearly as the number of iterations increases; (5) Formula generation: Input the set extreme high temperature environmental parameters and electrochemical surface characteristics into the ionic liquid electrolyte generation model for extreme high temperature requirements, and screen out the ionic liquid electrolyte that can meet the set extreme high temperature working conditions.
2. The intelligent design method for ionic liquid electrolytes according to claim 1, wherein: In step (1), the data platform includes NIST ILThermo and Web of Science data platforms.
3. The intelligent design method of ionic liquid electrolyte according to claim 1, characterized in that: In step (2), the literature includes SCI-indexed literature and ASTM standard experimental data.
4. The intelligent design method of ionic liquid electrolyte according to claim 1, characterized in that: In step (1), the ion combination is a combination of anion and cation species, the environmental parameters include temperature and humidity, the surface characteristics include roughness and charge density, and the electrochemical properties include viscosity and conductivity.
5. The intelligent design method of ionic liquid electrolyte according to claim 1, characterized in that: Step (2) is as follows: for the missing entries in the initial data set, search from SCI-indexed literature, identify and extract the semantics of ionic liquids in the literature through the BERT model, extract electrochemical performance parameters from the literature to supplement the blank data, and use the graph neural network (GNN) to map the ionic structural features of the ion structure and 3D model image data in the literature, and convert them into SMILES format for encoding and storage; based on the calculation results of molecular dynamics simulation, set the theoretical threshold of data confidence, automatically clean the data points that deviate from the confidence interval, and obtain the optimized data set.
6. The intelligent design method for ionic liquid electrolytes according to claim 1, wherein: Step (3) is as follows: preprocess the data with ion feature engineering, using RDKit The 2022.09 toolkit converts the SMILES-encoded ionic structural features into 208 feature descriptors, and further stacks the 208 feature descriptors into 20 dimensions that can accurately reflect the differences in ionic structural features of the dataset based on the one-hot encoding and Pearson correlation coefficient heat map shared or not shared by the anion and cation feature descriptors; the dataset is divided into training and test sets in a 7:3 ratio based on stratified sampling of each temperature interval of 10K to ensure that the data can be evenly divided into training and test sets according to temperature, reducing model underfitting or overfitting caused by division differences; a variety of machine learning algorithms are used to deeply train the generative model, and the algorithm with the best training results is selected to preliminarily train the generative model and define the model's hyperparameter space. The dimensions of the hyperparameter space include depth, number of iterations, and learning rate. The above training set is trained using a conditional generative machine learning model that integrates the CatBoost framework and the VAE encoder. Grid search and 10CV cross-validation methods are used to determine the value range of the better hyperparameter vector, narrowing the value range of the hyperparameter vector to improve the accuracy of the model.
7. The intelligent design method of ionic liquid electrolyte according to claim 1, characterized in that: In step (3), the machine learning algorithm is support vector machine or random forest.
8. The intelligent design method of ionic liquid electrolyte according to claim 1, characterized in that: In step (4), the accuracy of the model in generating ionic liquids was evaluated by R² ≥ 0.95 and RMSE ≤ 0.
02.
9. The intelligent design method of ionic liquid electrolyte according to claim 1, characterized in that: Step (5) is specifically as follows: the set extreme high temperature environmental parameters and electrochemical surface characteristics are input into the ionic liquid electrolyte generation model for extreme high temperature requirements. The model will determine the corresponding hyperparameter vector based on the set temperature range. The generated ion combination will pass through RDKit syntax verification, ΔG<0 thermodynamic criterion, and CatBoost performance prediction triple filtering in sequence, and finally output the ionic liquid candidate formula; the output result is encoded in SMILES and can be directly converted into the structural formula and 3D model of the ion in ChemDraw and Chem3D, which is convenient for structural feature analysis.
10. The intelligent design method for ionic liquid electrolytes according to claim 1, wherein: The method also includes step (6): testing the compatibility of candidate ionic liquids with the requirements of extreme high temperature working conditions through artificial synthesis, including high temperature viscosity testing using an Anton Paar MCR302 rheometer, conductivity measurement using a DDS-11A digital conductivity meter, electrochemical window measurement using an HSV-100 electrochemical workstation and a three-electrode system, and thermal stability testing using thermogravimetric analysis at a heating rate of 10 K / min, to further screen suitable ionic liquids.