Interactive catalysis method and platform based on molecular simulation and artificial intelligence

Abstract

The invention discloses an interactive catalysis method and platform based on molecular simulation and artificial intelligence. According to the method, intelligent design of a catalyst is achieved through data classification, condition optimization, structure modeling and mechanism analysis. Catalytic data uploaded by a user is subjected to standardization processing and then is divided into two types: a data set A is used for recommending an optimal reaction condition by a multi-objective optimization algorithm; a three-dimensional model is constructed after a catalyst structure of the data set B is optimized through quantum chemistry software, a plurality of intermediates and transition state structures are optimized, and geometric / electronic descriptors are calculated. And constructing a dual-model collaborative prediction system of the model E and the model F through a feature screening and dimension reduction technology, and finally screening the high-performance catalyst by utilizing a model F driven molecule generation algorithm. According to the method, flexible adaptation to specific application scenes can be achieved, the urgent requirements of the industry for efficient and accurate catalyst performance prediction and intelligent design are met, cost reduction and efficiency improvement of the industry are effectively assisted, and the research and development process of the catalyst is accelerated.

Description

Technical Field

[0001] This invention relates to the field of molecular catalysis technology, and more specifically to an interactive catalysis method and platform based on molecular simulation and artificial intelligence. Background Technology

[0002] The improvement and design of catalytic systems play a crucial role in chemical engineering research and development. However, the systematic study of catalytic systems faces numerous challenges, including the complexity and variability of influencing factors and the vast spatial range of catalyst structures. With the continuous development of artificial intelligence technology, combining molecular simulation methods with AI algorithms has become a new trend in academia and industry for exploring catalytic systems. To lower the technical barriers and improve research and development efficiency, developing simple, easy-to-use, platform-based integrated tools is particularly important.

[0003] Several catalyst databases and platforms are currently available, such as Catalysis-Hub and Open CatalystProject (see review Nat. Catal. 2024, 7, 624–635). However, these platforms still lack sufficient customization support, especially in terms of flexible adaptation to specific application scenarios. Furthermore, existing tools often fail to effectively integrate molecular simulation with artificial intelligence methods, making it difficult to meet the industry's urgent need for efficient and accurate catalyst performance prediction and intelligent design.

[0004] Therefore, how to invent an interactive catalysis method based on molecular simulation and artificial intelligence that can meet the industry's urgent need for efficient and accurate prediction of catalyst performance and intelligent design, and effectively help the industry reduce costs and increase efficiency and accelerate the catalyst research and development process, has become an urgent problem to be solved. Summary of the Invention

[0005] To this end, the present invention provides an interactive catalysis method and platform based on molecular simulation and artificial intelligence, which can be flexibly adapted to specific application scenarios, meet the urgent needs of industry for efficient and accurate catalyst performance prediction and intelligent design, and effectively help industry reduce costs and increase efficiency, and accelerate the catalyst research and development process.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an interactive catalysis method based on molecular simulation and artificial intelligence, comprising:

[0007] The catalytic research and development data uploaded by users are standardized to obtain standardized data; data with the same catalyst parameters but different experimental conditions are classified into dataset A; data with the same experimental conditions but different catalyst parameters are classified into dataset B.

[0008] Based on the dataset A, the optimal catalytic conditions are determined using a multi-objective optimization algorithm;

[0009] Based on the catalyst structure information in dataset B, a three-dimensional structure of the catalyst and a three-dimensional structure of the intermediate are constructed. The three-dimensional structure of the catalyst and the three-dimensional structure of the intermediate are then geometrically optimized using quantum chemistry software to obtain a set of catalyst structure molecules and a set of intermediate structure molecules.

[0010] The quantum chemistry software is used to perform transition state search and reaction path calculation on intermediate molecules in the intermediate structure molecule set, generating a transition state structure molecule set and a candidate feature set C;

[0011] For the molecular structures in the catalyst structure molecular set, the intermediate structure molecular set, and the transition state structure molecular set, respectively, calculate the geometric structure descriptor and the electronic structure descriptor to generate candidate feature bindings D;

[0012] By using correlation analysis and dimensionality reduction techniques, the dataset B, the candidate feature set C, and the candidate feature combination D are integrated and analyzed to screen out the feature subset related to performance optimization parameters.

[0013] Based on the feature subset, a performance prediction model E and a performance prediction model F are constructed; the performance prediction model E and the performance prediction model F are trained using a machine learning algorithm to obtain the trained performance prediction model E and the performance prediction model F.

[0014] Candidate catalyst structures are generated using a molecular generation algorithm; a catalyst candidate library is constructed based on the candidate catalyst structures; and the performance of the catalyst structures in the catalyst candidate library is predicted using the trained performance prediction model F to select the optimal catalyst molecular structure.

[0015] As a preferred embodiment of an interactive catalysis method based on molecular simulation and artificial intelligence, the catalysis R&D data includes: experimental input parameters, catalyst and auxiliary agent parameters, experimental output parameters, and performance optimization parameters.

[0016] As a preferred embodiment of the interactive catalysis method based on molecular simulation and artificial intelligence, the multi-objective optimization algorithm used to determine the optimal catalytic conditions includes a genetic algorithm and a Bayesian optimization algorithm.

[0017] As a preferred option for interactive catalysis methods based on molecular simulation and artificial intelligence, the quantum chemistry software includes Gaussian and ORCA.

[0018] As a preferred embodiment of the interactive catalysis method based on molecular simulation and artificial intelligence, the machine learning algorithm used to train the performance prediction model E and the performance prediction model F includes: random forest, XGBoost and neural network algorithm.

[0019] This invention also provides an interactive catalysis platform based on molecular simulation and artificial intelligence, and based on the above-mentioned interactive catalysis method based on molecular simulation and artificial intelligence, it includes:

[0020] The catalyst data acquisition and standardization module is used to standardize the catalytic R&D data uploaded by users to obtain standardized data; data with the same catalyst parameters but different experimental conditions are classified into dataset A; data with the same experimental conditions but different catalyst parameters are classified into dataset B.

[0021] The optimal catalytic conditions determination module is used to determine the optimal catalytic conditions based on the dataset A using a multi-objective optimization algorithm.

[0022] The molecular structure construction and optimization module is used to construct the three-dimensional structure of the catalyst and the three-dimensional structure of the intermediate based on the catalyst structure information in the dataset B; and to perform geometric structure optimization on the three-dimensional structure of the catalyst and the three-dimensional structure of the intermediate using quantum chemistry software to obtain the set of catalyst structure molecules and the set of intermediate structure molecules.

[0023] The reaction mechanism calculation module is used to perform transition state search and reaction path calculation on intermediate molecules in the intermediate structure molecule set using the quantum chemistry software, and generate a transition state structure molecule set and a candidate feature set C.

[0024] The molecular descriptor extraction module is used to calculate geometric structure descriptors and electronic structure descriptors for the molecular structures in the catalyst structure molecular set, the intermediate structure molecular set, and the transition state structure molecular set, respectively, and generate candidate feature combinations D.

[0025] The data analysis and feature selection module is used to integrate and analyze the dataset B, the candidate feature set C, and the candidate feature combination D through correlation analysis and dimensionality reduction techniques, and to select a feature subset related to performance optimization parameters.

[0026] The performance prediction model construction and training module is used to construct performance prediction model E and performance prediction model F based on the feature subset; and to train the performance prediction model E and the performance prediction model F using a machine learning algorithm to obtain the trained performance prediction model E and the performance prediction model F.

[0027] The catalyst structure generation and optimization module is used to generate candidate catalyst structures through a molecular generation algorithm; construct a catalyst candidate library based on the candidate catalyst structures; and perform performance prediction on the catalyst structures in the catalyst candidate library using the trained performance prediction model F to select the optimal catalyst molecular structure.

[0028] As a preferred option for an interactive catalysis platform based on molecular simulation and artificial intelligence, the catalyst data acquisition and standardization processing module includes the following catalytic R&D data: experimental input parameters, catalyst and auxiliary agent parameters, experimental output parameters, and performance optimization parameters.

[0029] As a preferred embodiment of an interactive catalysis platform based on molecular simulation and artificial intelligence, the optimal catalysis condition determination module includes a genetic algorithm and a Bayesian optimization algorithm in the process of determining the optimal catalysis conditions through the multi-objective optimization algorithm.

[0030] As a preferred embodiment of an interactive catalysis platform based on molecular simulation and artificial intelligence, the quantum chemistry software in the molecular structure construction and optimization module includes Gaussian and ORCA.

[0031] As a preferred embodiment of an interactive catalysis platform based on molecular simulation and artificial intelligence, the performance prediction model construction and training module includes, during the training of the performance prediction model E and the performance prediction model F using the machine learning algorithm, random forest, XGBoost, and neural network algorithms.

[0032] This invention has the following advantages: It standardizes user-uploaded catalytic research data to obtain standardized data; data with the same catalyst parameters but different experimental conditions are categorized into dataset A; data with the same experimental conditions but different catalyst parameters are categorized into dataset B; based on dataset A, the optimal catalytic conditions are determined using a multi-objective optimization algorithm; based on the catalyst structure information in dataset B, a three-dimensional structure of the catalyst and a three-dimensional structure of the intermediate are constructed; the three-dimensional structures of the catalyst and the intermediate are geometrically optimized using quantum chemistry software to obtain a set of catalyst structure molecules and a set of intermediate structure molecules; the intermediate molecules in the set of intermediate structure molecules are searched for transition states and their reaction paths are calculated using the quantum chemistry software to generate a set of transition state structure molecules and a set of candidate features C; and the catalyst structure is optimized accordingly. The molecular structures in the molecular set, the intermediate structure molecular set, and the transition state structure molecular set are used to calculate geometric structure descriptors and electronic structure descriptors, respectively, generating candidate feature combinations D. Through correlation analysis and dimensionality reduction techniques, the dataset B, the candidate feature set C, and the candidate feature combinations D are integrated and analyzed to screen out feature subsets related to performance optimization parameters. Based on these feature subsets, performance prediction models E and F are constructed. Machine learning algorithms are used to train performance prediction models E and F, resulting in well-trained models. A molecular generation algorithm is used to generate candidate catalyst structures. Based on these candidate catalyst structures, a catalyst candidate library is constructed. The performance of catalyst structures in the catalyst candidate library is predicted using the trained performance prediction model F, and the optimal catalyst molecular structure is selected. This invention innovatively achieves a highly customized and intelligent integrated design, effectively compensating for the shortcomings of existing catalytic platforms. It enables flexible adaptation to specific application scenarios, meeting the urgent industrial demand for efficient and accurate catalyst performance prediction and intelligent design. By seamlessly integrating molecular simulation tools (such as quantum chemical calculations) with advanced artificial intelligence algorithms, the platform allows users to upload and flexibly adapt their own R&D data, automatically completing the entire process from data standardization to catalyst structure optimization, significantly reducing the technical threshold and the need for manual intervention. This design not only improves the accuracy of catalyst performance prediction but also accelerates the intelligent recommendation of optimal reaction conditions and novel structures, thereby significantly shortening the R&D cycle, reducing experimental trial-and-error costs, and providing strong support for the rapid iteration and cost reduction of catalysts in industry. Attached Figure Description

[0033] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0034] The structures, proportions, sizes, etc. illustrated in this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed herein, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0035] Figure 1 This is a schematic diagram of the interactive catalysis method based on molecular simulation and artificial intelligence provided in Embodiment 1 of the present invention;

[0036] Figure 2 This is a schematic diagram of the catalyst structure in one possible embodiment provided in Example 1 of the present invention;

[0037] Figure 3 This is a schematic diagram of mechanism calculation in one possible embodiment provided in Embodiment 1 of the present invention;

[0038] Figure 4 This is a schematic diagram illustrating the relationship between an electronic descriptor and a transition state reaction enthalpy barrier in one possible embodiment provided in Embodiment 1 of the present invention;

[0039] Figure 5 This is a schematic diagram of the novel catalyst structure in one possible embodiment provided in Example 1 of the present invention;

[0040] Figure 6 This is a schematic diagram of the architecture of the interactive catalytic platform based on molecular simulation and artificial intelligence provided in Embodiment 2 of the present invention. Detailed Implementation

[0041] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0042] Example 1

[0043] See Figure 1 Embodiment 1 of the present invention provides an interactive catalysis method based on molecular simulation and artificial intelligence, comprising the following steps:

[0044] S1. Standardize the catalytic R&D data uploaded by users to obtain standardized data; classify the standardized data with the same catalyst parameters but different experimental conditions into dataset A; classify the standardized data with the same experimental conditions but different catalyst parameters into dataset B.

[0045] S2. Based on the dataset A, determine the optimal catalytic conditions using a multi-objective optimization algorithm;

[0046] S3. Based on the catalyst structure information in the dataset B, construct the three-dimensional structure of the catalyst and the three-dimensional structure of the intermediate; optimize the geometry of the three-dimensional structure of the catalyst and the three-dimensional structure of the intermediate using quantum chemistry software to obtain the set of catalyst structure molecules and the set of intermediate structure molecules.

[0047] S4. Using the quantum chemistry software, perform transition state search and reaction path calculation on the intermediate molecules in the intermediate structure molecule set to generate a transition state structure molecule set and a candidate feature set C;

[0048] S5. For the molecular structures in the catalyst structure molecular set, the intermediate structure molecular set, and the transition state structure molecular set, calculate the geometric structure descriptor and electronic structure descriptor respectively, and generate candidate feature bindings D;

[0049] S6. Through correlation analysis and dimensionality reduction techniques, the dataset B, the candidate feature set C, and the candidate feature combination D are integrated and analyzed to select a feature subset related to performance optimization parameters.

[0050] S7. Based on the feature subset, construct performance prediction model E and performance prediction model F; train performance prediction model E and performance prediction model F using machine learning algorithms to obtain trained performance prediction model E and performance prediction model F.

[0051] S8. Generate candidate catalyst structures using a molecular generation algorithm; construct a catalyst candidate library based on the candidate catalyst structures; predict the performance of the catalyst structures in the catalyst candidate library using the trained performance prediction model F, and select the optimal catalyst molecular structure.

[0052] In this embodiment, in step S1, the catalytic R&D data uploaded by the user is standardized to obtain standardized data; data with the same catalyst parameters but different experimental conditions in the standardized data are classified into dataset A; data with the same experimental conditions but different catalyst parameters in the standardized data are classified into dataset B.

[0053] Specifically, the system receives catalytic research and development data uploaded by users and standardizes the format of heterogeneous data (e.g., SMILES standardization and unit conversion). Data with the same (or similar) catalyst and auxiliary parameters but different experimental input parameters are grouped into dataset A; data with different (different or significantly different) catalyst and auxiliary parameters but the same (or similar) experimental input parameters are grouped into dataset B.

[0054] The catalytic research and development data includes:

[0055] Experimental input parameters: such as reaction temperature, reaction time, reaction pressure, etc.;

[0056] Catalyst and auxiliary parameters: such as molecular structure (SMILES), XRD crystal structure, etc.;

[0057] Experimental output parameters: such as catalytic activity, product conversion rate, selectivity, etc.;

[0058] Performance optimization parameters: such as target performance indicators like catalytic efficiency.

[0059] In this embodiment, in step S2, the optimal catalytic conditions are determined based on the dataset A using a multi-objective optimization algorithm;

[0060] Specifically, based on dataset A, by constructing a functional relationship between experimental input parameters and performance optimization indicators, and combining multi-objective optimization algorithms (such as genetic algorithms, Bayesian optimization, etc.), the optimal experimental conditions (such as the best combination of temperature, time, and pressure) are automatically searched and recommended to obtain the optimal catalytic conditions.

[0061] In this embodiment, in step S3, based on the catalyst structure information of the dataset B, a three-dimensional structure of the catalyst and a three-dimensional structure of the intermediate are constructed; the three-dimensional structure of the catalyst and the three-dimensional structure of the intermediate are geometrically optimized using quantum chemistry software to obtain a set of catalyst structure molecules and a set of intermediate structure molecules;

[0062] Specifically, the structural information of different catalysts and promoters in dataset B is encoded and represented, and important reaction intermediate molecules are derived. The three-dimensional structures of the catalysts and intermediates are constructed using crystal structures (e.g., XRD) or molecular structure representations (e.g., SMILES). Quantum chemistry software (e.g., Gaussian, ORCA) is used to optimize the geometric structure of the catalyst and intermediate three-dimensional structures. Successfully optimized catalyst structures are grouped into a catalyst structure molecule set; successfully optimized intermediate structures are grouped into an intermediate structure molecule set.

[0063] In this embodiment, in step S4, the quantum chemistry software is used to perform transition state search and reaction path calculation on the intermediate molecules in the intermediate structure molecule set, generating a transition state structure molecule set and a candidate feature set C;

[0064] Specifically, key intermediate molecules are screened from the set of intermediate structure molecules. Combined with known reaction sites, quantum chemistry software is used to search for transition states and calculate reaction pathways, obtaining thermodynamic parameters such as activation energy and heat of reaction for each elementary reaction, thus forming a candidate feature set C. This process simultaneously generates transition state structures, constituting a set of transition state structure molecules.

[0065] In this embodiment, in step S5, geometric structure descriptors and electronic structure descriptors are calculated for the molecular structures in the catalyst structure molecular set, the intermediate structure molecular set, and the transition state structure molecular set, respectively, to generate candidate feature bindings D;

[0066] Specifically, for each molecular structure in the catalyst structure molecular set, the intermediate structure molecular set, and the transition state structure molecular set, geometric structure descriptors (such as bond length, bond angle, dihedral angle, molecular buried volume, etc.) and electronic structure descriptors (such as HOMO / LUMO energy levels, charge distribution, dipole moment, etc.) are calculated to generate candidate feature bindings D.

[0067] In this embodiment, in step S6, the dataset B, the candidate feature set C, and the candidate feature combination D are integrated and analyzed through correlation analysis and dimensionality reduction techniques to screen out the feature subset related to the performance optimization parameters.

[0068] Specifically, all catalytic samples in dataset B are integrated and analyzed. These samples include experimental input parameters, experimental output parameters, candidate feature set C, and candidate feature combination D. Through correlation analysis and dimensionality reduction techniques (such as principal component analysis and mutual information evaluation), a subset of features highly correlated with the performance optimization parameters is selected for subsequent model training.

[0069] In this embodiment, in step S7, a performance prediction model E and a performance prediction model F are constructed based on the feature subset; the performance prediction model E and the performance prediction model F are trained by a machine learning algorithm to obtain the trained performance prediction model E and the performance prediction model F.

[0070] Specifically, based on the screening results, two types of performance prediction models are constructed:

[0071] The first candidate feature group includes experimental input parameters and candidate feature sets C and D, which are used to train the performance prediction model E. This model considers the combined effects of experimental conditions and catalyst structure on performance.

[0072] The second candidate feature group contains only candidate feature sets C and D, which are used to train the performance prediction model F, which focuses on the influence of structure on catalytic performance.

[0073] Machine learning algorithms (such as random forest, XGBoost, neural networks, etc.) are used to train the above feature groups to obtain the trained performance prediction model E and performance prediction model F.

[0074] In this embodiment, in step S8, candidate catalyst structures are generated using a molecular generation algorithm; a catalyst candidate library is constructed based on the candidate catalyst structures; and the performance of the catalyst structures in the catalyst candidate library is predicted using the trained performance prediction model F to select the optimal catalyst molecular structure.

[0075] Specifically, based on the predictive power of the performance prediction model F, combined with expert knowledge or molecular generation algorithms (such as molecular generation models based on graph neural networks), a batch of candidate catalyst structures is generated to construct a catalyst candidate library. Subsequently, the performance of these structures is predicted and screened using the performance prediction model F in high throughput, and the catalyst molecular structure with the best performance is selected as the final recommended design scheme.

[0076] In one possible embodiment, an example of catalytic polymerization of ethylene into polyethylene using a catalyst with a defined geometry is provided as follows:

[0077] T1, the collected partial catalyst structure is as follows: Figure 2 As shown in Tables 1 and 2, R&D personnel upload R&D data to the system.

[0078]

[0079] Table 1 Experimental Input Parameters

[0080] Catalyst number Yield active weight average molecular weight Cat_1 30.2g <![CDATA[0.4kg PE ·g Cat -1 ·h -1 ]]> <![CDATA[107x10 3 ]]>

[0081] Table 2 Experimental Output Parameters

[0082] Tables 1 and 2 show the data distribution of the experimental input and output parameters of the catalyst. The performance optimization parameter is the yield. The uploaded data are grouped as follows: the data with the same or similar catalyst and auxiliary parameters but different experimental input parameters are denoted as CGC catalyst dataset A; the data with different catalyst and auxiliary parameters but the same or similar experimental input parameters are denoted as CGC catalyst dataset B.

[0083] T2. Analyze the CGC catalyst dataset A and call the optimization algorithm module to obtain a set of optimal reaction conditions: reaction time: 1h; reaction temperature: 60 degrees Celsius; system pressure: 5 bar; catalyst amount: 10mg; post-treatment temperature: 20 degrees Celsius; additive / catalyst ratio: 100.

[0084] T3. Using the CGC catalyst dataset B, construct the 3D structures of the catalyst molecules and the intermediate IM1 molecule (the CGC catalyst insertion into ethylene). ORCA quantum chemistry software is used for molecular optimization. The computational approach employs B3LYP density functional theory, def2svp basis sets, and an SMD (Solvation Model based on Density) solvation model with heptane as the solvent, using D3(BJ) dispersion correction. A batch of optimized CGC catalysts is denoted as the CGC catalyst molecule set, and the intermediate IM1 molecules are denoted as the intermediate IM1 molecule set.

[0085] T4. Perform transition state calculations for IM1 molecules under different catalytic states, with the bonding site being the bond between carbon atom 1 and carbon atom 2, such as... Figure 3 As shown, this type of transition state is calculated under different catalyst conditions. These transition state structures are denoted as the catalyst CGC transition state structure molecular set. The electronic energy barrier, free energy barrier, and enthalpy barrier of these transition states are obtained and denoted as the CGC catalyst candidate feature set C.

[0086] T5. Based on the CGC catalyst molecule set, intermediate IM1 molecule set, and catalyst CGC transition state structure molecule set obtained in steps T3 and T4, batch calculate the descriptors of these catalysts. The descriptors include geometric descriptors, electronic descriptors, etc., and obtain a batch of candidate features denoted as CGC catalyst candidate feature set D.

[0087] T6. Organize the above data and select catalytic samples from CGC catalyst dataset B. All samples include experimental input parameters, experimental output parameters, CGC catalyst candidate feature set C, and CGC catalyst candidate feature set D. Analyze the correlations among these data and retain the highly correlated portions. An example correlation plot is shown below. Figure 4 As shown.

[0088] T7. This embodiment is more concerned with the relationship between catalytic structure and catalytic performance. Therefore, the candidate feature set C of CGC catalyst and the candidate feature set D of CGC catalyst are merged and a prediction model is established with the reaction yield. The prediction model F can be obtained through multiple rounds of feature engineering and machine learning model training. The R2 of leave-one-out cross-validation can reach 0.92.

[0089] T8. Based on the prediction model F from step T7, a novel catalyst can be derived using a molecular generation algorithm, with the structure as follows: Figure 5 As shown.

[0090] In summary, this invention standardizes user-uploaded catalytic R&D data to obtain standardized data; data with the same catalyst parameters but different experimental conditions are categorized into dataset A; data with the same experimental conditions but different catalyst parameters are categorized into dataset B; based on dataset A, optimal catalytic conditions are determined using a multi-objective optimization algorithm; based on the catalyst structure information in dataset B, a three-dimensional structure of the catalyst and a three-dimensional structure of the intermediate are constructed; the three-dimensional structures of the catalyst and the intermediate are geometrically optimized using quantum chemistry software to obtain a set of catalyst structure molecules and a set of intermediate structure molecules; the intermediate molecules in the set of intermediate structure molecules are searched for transition states and their reaction paths are calculated using the quantum chemistry software to generate a set of transition state structure molecules and a set of candidate features C; the catalyst structure molecule set... The invention involves calculating geometric and electronic structure descriptors for the molecular structures in the set of intermediate and transition state structures, respectively, to generate candidate feature combinations D. Through correlation analysis and dimensionality reduction techniques, the dataset B, the candidate feature set C, and the candidate feature combinations D are integrated and analyzed to select feature subsets related to performance optimization parameters. Based on these feature subsets, performance prediction models E and F are constructed. Machine learning algorithms are used to train performance prediction models E and F, resulting in well-trained models. A molecular generation algorithm is used to generate candidate catalyst structures. Based on these candidate catalyst structures, a catalyst candidate library is constructed. The performance of catalyst structures in the candidate library is predicted using the trained performance prediction model F, and the optimal catalyst molecular structure is selected. This invention innovatively achieves a highly customized and intelligent integrated design, effectively compensating for the shortcomings of existing catalytic platforms. It enables flexible adaptation to specific application scenarios, meeting the urgent industrial demand for efficient and accurate catalyst performance prediction and intelligent design. By seamlessly integrating molecular simulation tools (such as quantum chemical calculations) with advanced artificial intelligence algorithms, the platform allows users to upload and flexibly adapt their own R&D data, automatically completing the entire process from data standardization to catalyst structure optimization, significantly reducing the technical threshold and the need for manual intervention. This design not only improves the accuracy of catalyst performance prediction but also accelerates the intelligent recommendation of optimal reaction conditions and novel structures, thereby significantly shortening the R&D cycle, reducing experimental trial-and-error costs, and providing strong support for the rapid iteration and cost reduction of catalysts in industry.

[0091] It should be noted that the method of this disclosure embodiment can be executed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method of this disclosure embodiment, and the multiple devices will interact with each other to complete the method described.

[0092] It should be noted that the above description describes some embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0093] Example 2

[0094] See Figure 6 Embodiment 2 of the present invention also provides an interactive catalytic platform based on molecular simulation and artificial intelligence, including:

[0095] The catalyst data acquisition and standardization processing module 001 is used to standardize the catalytic R&D data uploaded by users to obtain standardized data; the standardized data with the same catalyst parameters but different experimental conditions are classified into dataset A; the standardized data with the same experimental conditions but different catalyst parameters are classified into dataset B.

[0096] The optimal catalytic conditions determination module 002 is used to determine the optimal catalytic conditions based on the dataset A using a multi-objective optimization algorithm.

[0097] The molecular structure construction and optimization module 003 is used to construct the three-dimensional structure of the catalyst and the three-dimensional structure of the intermediate based on the catalyst structure information of the dataset B; and to perform geometric structure optimization on the three-dimensional structure of the catalyst and the three-dimensional structure of the intermediate using quantum chemistry software to obtain the set of catalyst structure molecules and the set of intermediate structure molecules.

[0098] The reaction mechanism calculation module 004 is used to perform transition state search and reaction path calculation on intermediate molecules in the intermediate structure molecule set using the quantum chemistry software, and generate a transition state structure molecule set and a candidate feature set C.

[0099] The molecular descriptor extraction module 005 is used to calculate the geometric structure descriptor and electronic structure descriptor for the molecular structures in the catalyst structure molecular set, the intermediate structure molecular set and the transition state structure molecular set, respectively, and generate candidate feature bindings D.

[0100] The data analysis and feature selection module 006 is used to integrate and analyze the dataset B, the candidate feature set C, and the candidate feature combination D through correlation analysis and dimensionality reduction techniques, and to select a feature subset related to performance optimization parameters.

[0101] The performance prediction model construction and training module 007 is used to construct performance prediction model E and performance prediction model F based on the feature subset; and to train the performance prediction model E and the performance prediction model F through machine learning algorithms to obtain the trained performance prediction model E and the performance prediction model F.

[0102] The catalyst structure generation and optimization module 008 is used to generate candidate catalyst structures through a molecular generation algorithm; construct a catalyst candidate library based on the candidate catalyst structures; and perform performance prediction on the catalyst structures in the catalyst candidate library using the trained performance prediction model F to select the optimal catalyst molecular structure.

[0103] In this embodiment, the catalyst data acquisition and standardization processing module 001 includes the following catalytic R&D data: experimental input parameters, catalyst and additive parameters, experimental output parameters, and performance optimization parameters.

[0104] In this embodiment, the optimal catalytic condition determination module 002, in the process of determining the optimal catalytic conditions through the multi-objective optimization algorithm, includes a genetic algorithm and a Bayesian optimization algorithm.

[0105] In this embodiment, the quantum chemistry software in the molecular structure construction and optimization module 003 includes Gaussian and ORCA.

[0106] In this embodiment, in the performance prediction model construction and training module 007, during the training of the performance prediction model E and the performance prediction model F using the machine learning algorithm, the machine learning algorithm includes: random forest, XGBoost and neural network algorithm.

[0107] It should be noted that the information interaction and execution process between the modules of the above system are based on the same concept as the method embodiment in Embodiment 1 of this application, and the resulting technical effects are the same as those in the method embodiment of this application. For details, please refer to the description in the method embodiment shown above in this application, and it will not be repeated here.

[0108] Example 3

[0109] Embodiment 3 of the present invention provides a non-transitory computer-readable storage medium storing program code of an interactive catalytic method based on molecular simulation and artificial intelligence. The program code includes instructions for executing the interactive catalytic method based on molecular simulation and artificial intelligence of Embodiment 1 or any possible implementation thereof.

[0110] Computer-readable storage media can be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives, SSDs).

[0111] Example 4

[0112] Embodiment 4 of the present invention provides an electronic device, including: a memory and a processor;

[0113] The processor and the memory communicate with each other via a bus; the memory stores program instructions that can be executed by the processor, and the processor can execute the interactive catalysis method based on molecular simulation and artificial intelligence in Embodiment 1 or any possible implementation thereof by calling the program instructions.

[0114] Specifically, a processor can be implemented in hardware or software. When implemented in hardware, the processor can be a logic circuit, an integrated circuit, etc. When implemented in software, the processor can be a general-purpose processor that reads software code stored in memory. This memory can be integrated into the processor or located outside the processor and exist independently.

[0115] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable system. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0116] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing systems. They can be centralized on a single computing system or distributed across a network of multiple computing systems. Optionally, they can be implemented using program code executable by a computing system, thereby storing them in a storage system for execution by the computing system. In some cases, the steps shown or described can be performed in a different order than those presented herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0117] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. An interactive catalytic method based on molecular simulation and artificial intelligence, characterized in that, include: The catalytic R&D data uploaded by users is standardized to obtain standardized data; Data with the same catalyst parameters but different experimental conditions in the standardized data are grouped into dataset A; data with the same experimental conditions but different catalyst parameters in the standardized data are grouped into dataset B. Based on the dataset A, the optimal catalytic conditions are determined using a multi-objective optimization algorithm; Based on the catalyst structure information in the dataset B, the three-dimensional structure of the catalyst and the three-dimensional structure of the intermediate are constructed. The three-dimensional structures of the catalyst and the intermediate were geometrically optimized using quantum chemistry software to obtain the catalyst structure molecular set and the intermediate structure molecular set. The quantum chemistry software is used to perform transition state search and reaction path calculation on intermediate molecules in the intermediate structure molecule set, generating a transition state structure molecule set and a candidate feature set C; For the molecular structures in the catalyst structure molecular set, the intermediate structure molecular set, and the transition state structure molecular set, respectively, calculate the geometric structure descriptor and the electronic structure descriptor to generate candidate feature bindings D; By using correlation analysis and dimensionality reduction techniques, the dataset B, the candidate feature set C, and the candidate feature combination D are integrated and analyzed to screen out the feature subset related to performance optimization parameters. Based on the feature subset, a performance prediction model E and a performance prediction model F are constructed; the performance prediction model E and the performance prediction model F are trained using a machine learning algorithm to obtain the trained performance prediction model E and the performance prediction model F. Candidate catalyst structures are generated using a molecular generation algorithm; a catalyst candidate library is constructed based on the candidate catalyst structures; and the performance of the catalyst structures in the catalyst candidate library is predicted using the trained performance prediction model F to select the optimal catalyst molecular structure.

2. The interactive catalysis method based on molecular simulation and artificial intelligence according to claim 1, characterized in that, The catalytic research and development data includes: experimental input parameters, catalyst and additive parameters, experimental output parameters, and performance optimization parameters.

3. The interactive catalysis method based on molecular simulation and artificial intelligence according to claim 2, characterized in that, In determining the optimal catalytic conditions using the multi-objective optimization algorithm, the multi-objective optimization algorithm includes: a genetic algorithm and a Bayesian optimization algorithm.

4. The interactive catalysis method based on molecular simulation and artificial intelligence according to claim 3, characterized in that, The quantum chemistry software includes Gaussian and ORCA.

5. The interactive catalysis method based on molecular simulation and artificial intelligence according to claim 4, characterized in that, During the training of the performance prediction model E and the performance prediction model F using the machine learning algorithms, the machine learning algorithms include: random forest, XGBoost and neural network algorithms.

6. An interactive catalytic platform based on molecular simulation and artificial intelligence, employing the interactive catalytic method based on molecular simulation and artificial intelligence as described in any one of claims 1-5, characterized in that, include: The catalyst data acquisition and standardization processing module is used to standardize the catalytic R&D data uploaded by users to obtain standardized data; Data with the same catalyst parameters but different experimental conditions in the standardized data are grouped into dataset A; data with the same experimental conditions but different catalyst parameters in the standardized data are grouped into dataset B. The optimal catalytic conditions determination module is used to determine the optimal catalytic conditions based on the dataset A using a multi-objective optimization algorithm. The molecular structure construction and optimization module is used to construct the three-dimensional structure of the catalyst and the three-dimensional structure of the intermediate based on the catalyst structure information in the dataset B. The three-dimensional structures of the catalyst and the intermediate were geometrically optimized using quantum chemistry software to obtain the catalyst structure molecular set and the intermediate structure molecular set. The reaction mechanism calculation module is used to perform transition state search and reaction path calculation on intermediate molecules in the intermediate structure molecule set using the quantum chemistry software, and generate a transition state structure molecule set and a candidate feature set C. The molecular descriptor extraction module is used to calculate geometric structure descriptors and electronic structure descriptors for the molecular structures in the catalyst structure molecular set, the intermediate structure molecular set, and the transition state structure molecular set, respectively, and generate candidate feature combinations D. The data analysis and feature selection module is used to integrate and analyze the dataset B, the candidate feature set C, and the candidate feature combination D through correlation analysis and dimensionality reduction techniques, and to select a feature subset related to performance optimization parameters. The performance prediction model construction and training module is used to construct performance prediction model E and performance prediction model F based on the feature subset; and to train the performance prediction model E and the performance prediction model F using a machine learning algorithm to obtain the trained performance prediction model E and the performance prediction model F. The catalyst structure generation and optimization module is used to generate candidate catalyst structures through a molecular generation algorithm; construct a catalyst candidate library based on the candidate catalyst structures; and perform performance prediction on the catalyst structures in the catalyst candidate library using the trained performance prediction model F to select the optimal catalyst molecular structure.

7. The interactive catalytic platform based on molecular simulation and artificial intelligence according to claim 6, characterized in that, In the catalyst data acquisition and standardization processing module, the catalyst R&D data includes: experimental input parameters, catalyst and additive parameters, experimental output parameters, and performance optimization parameters.

8. The interactive catalytic platform based on molecular simulation and artificial intelligence according to claim 7, characterized in that, In the optimal catalytic condition determination module, during the process of determining the optimal catalytic conditions through the multi-objective optimization algorithm, the multi-objective optimization algorithm includes: a genetic algorithm and a Bayesian optimization algorithm.

9. The interactive catalytic platform based on molecular simulation and artificial intelligence according to claim 8, characterized in that, The quantum chemistry software included in the molecular structure construction and optimization module is Gaussian and ORCA.

10. The interactive catalytic platform based on molecular simulation and artificial intelligence according to claim 9, characterized in that, In the performance prediction model construction and training module, during the training of the performance prediction model E and the performance prediction model F using the machine learning algorithms, the machine learning algorithms include: random forest, XGBoost, and neural network algorithms.

Images