DFT (Discrete Fourier Transform) and machine learning combined catalytic adsorption material development method and system

Abstract

The invention discloses a DFT (Discrete Fourier Transform) and machine learning combined catalytic adsorption material development method and system. The method comprises the following steps: acquiring material characteristics; establishing a plurality of catalytic adsorption material structures; calculating the adsorption energy of different structures on oxygen molecules; performing prediction by utilizing the trained model, and taking a decision coefficient R2 as an evaluation standard to qualitatively describe the prediction capability of different machine learning models on adsorption energy; important factors with different characteristics are extracted by utilizing machine learning, and through a linear regression method, a coincidence is described so as to predict the adsorption energy of different structures to oxygen molecules. The experimental time can be saved, the cost is reduced, and the computing resources can be saved. And a suitable catalyst can be found and developed to replace an expensive noble metal catalyst. By improving important influencing factors, the method can serve the design of the selective catalytic reaction process to inhibit undesired overreactions and provide insight for controlling chemical and physical principles of the catalytic process.

Description

Technical Field

[0001] The present invention relates to the technical field of material development, and in particular to a method and system for developing catalytic adsorption materials by combining DFT with machine learning. Background Art

[0002] With the continuous advancement of industrial production technology, clean rooms are playing an increasingly important role in high-precision production, and low-concentration gaseous pollutants in the air will also have a significant impact on production. Gases such as SO2 and NO have a wide range of influence and are highly harmful, but their concentration in the air is low. Traditional adsorption materials have a small adsorption capacity for low-concentration gaseous pollutants in clean rooms and are easy to desorb. Therefore, it is necessary to develop new catalytic adsorption materials to convert pollutants in the air into compounds to purify the air.

[0003] Catalytic adsorption materials purify gaseous pollutants such as SO2 and NO in the air through redox reactions. The first step is to combine with oxygen molecules in the air. However, the current development of catalytic adsorption materials is still mainly through experimental preparation of materials and testing of adsorption performance, relying on expert experience and knowledge. There are disadvantages such as long development cycle, high cost and unclear factors affecting material performance. In recent years, with the development of computer simulation technology, the use of first principles to develop new materials through calculation has received more and more attention. However, the diverse structures of catalytic adsorption materials, the controversial identification of active sites, and the unclear factors affecting the adsorption performance of materials have led to excessive consumption of computing resources and blind development process.

[0004] Therefore, in the context of first principles and machine learning, it is of great significance to propose a simple method for the development of catalytic adsorption materials by calculating the system characteristics and adsorption energy of different structures, extracting key features using machine learning. Summary of the invention

[0005] In view of the above-mentioned problems, the present invention is proposed.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for developing catalytic adsorption materials by combining DFT with machine learning, comprising:

[0007] Obtain material properties; establish various catalytic adsorption material structures;

[0008] Calculate the adsorption energy of different structures on oxygen molecules;

[0009] The trained model was used for prediction, and the coefficient of determination R2 was used as the evaluation criterion to qualitatively describe the prediction ability of different machine learning models for adsorption energy;

[0010] Machine learning is used to extract important factors of different characteristics, and the linear regression method is used to describe the conformity to predict the adsorption energy of different structures on oxygen molecules.

[0011] As a preferred scheme of the method for developing catalytic adsorption materials by combining DFT with machine learning described in the present invention, the material properties include the properties of the constituent elements, the intrinsic characteristics of the surrounding coordination environment, and the intrinsic characteristics of the system environment.

[0012] As a preferred embodiment of the method for developing catalytic adsorption materials by combining DFT with machine learning described in the present invention, the properties of the constituent elements include atomic number, relative atomic mass, number of d orbital valence electrons, number of s orbital valence electrons, Pauling electronegativity, first ionization energy, electron affinity, covalent radius, atomic radius, element periodic number, family number, average effective charge, Hirshfeld charge, and Hirshfeld-I charge;

[0013] The intrinsic characteristics of the surrounding coordination environment include, for example, the sum of the Pauling electronegativity of the surrounding atoms, the sum of the covalent radii of the surrounding atoms, the sum of the atomic radii of the surrounding atoms, the sum of the number of valence electrons of the surrounding atoms, the sum of the average effective charges of the surrounding atoms, the sum of the Hirshfeld charges of the surrounding atoms, and the sum of the Hirshfeld-I charges of the surrounding atoms;

[0014] The intrinsic characteristics of the system environment include D-band center and energy band gap.

[0015] As a preferred embodiment of the method for developing catalytic adsorption materials by combining DFT with machine learning described in the present invention, the various catalytic adsorption material structures include using an infinitely extended graphene model as an unoptimized structure, and constructing an initial single-atom catalytic material model by deleting and replacing the carbon atoms therein;

[0016] The initial single-atom catalytic material model is structurally optimized to obtain an optimal structure.

[0017] As a preferred solution of the method for developing catalytic adsorption materials by combining DFT with machine learning described in the present invention, the structural optimization is specifically: using VASP software for structural optimization, and the setting of various parameters includes:

[0018] Use projected augmented wave pseudopotentials to calculate the interaction between electrons and nuclei;

[0019] Under GGA conditions, the PBE functional is used to calculate the electronic structure of the material;

[0020] The cutoff energy was set to 500 eV, the convergence threshold of energy was set to 10-5 eV, and the convergence threshold of force was set to

[0021] During simulation calculations, a small number of units are taken from the graphene model for calculations and recorded as repeating units;

[0022] The repeating unit is k-point sampled for the Brillouin zone through a 2 × 3 × 1 grid;

[0023] The DFT-D3 method was used to treat the van der Waals interactions between molecules;

[0024] Set the intermolecular spacing to To avoid interactions between layers.

[0025] As a preferred embodiment of the method for developing catalytic adsorption materials by combining DFT with machine learning described in the present invention, in which: when calculating the adsorption energy of oxygen molecules by different structures, spin and van der Waals force are considered, the adsorption free energy ΔG of O2 molecules is:

[0026] ΔG=G M@C / O2 -G O2 -G M@C

[0027] in, is the Gibbs free energy of the system after O2 adsorption, G M@C are the Gibbs free energy of the free O2 molecule and the original material model, respectively.

[0028] As a preferred solution of the DFT and machine learning combined catalytic adsorption material development method of the present invention, the training of the machine learning model includes randomly extracting 20% ​​of the data obtained by density functional theory calculation as a test set, and setting 80% as a training set; inputting the training set into the model, and obtaining the optimal hyperparameters of the model by grid search;

[0029] Machine learning is used to extract important factors of different characteristics, and characteristics with a high proportion are selected. By performing linear regression on each feature, the adsorption energy of different structures on oxygen molecules is described.

[0030] A catalytic adsorption material development system using DFT combined with machine learning using the method of the present invention, characterized in that:

[0031] Acquisition module, to obtain material properties; to establish various catalytic adsorption material structures;

[0032] Calculation module, which calculates the adsorption energy of different structures on oxygen molecules;

[0033] The prediction module uses the trained model to make predictions and takes the coefficient of determination R2 as the evaluation criterion to qualitatively describe the prediction ability of different machine learning models for adsorption energy. It uses machine learning to extract important factors of different characteristics and uses linear regression methods to describe the conformity to predict the adsorption energy of different structures on oxygen molecules.

[0034] A computer device comprises: a memory and a processor; the memory stores a computer program, wherein: the processor implements the steps of any one of the methods of the present invention when executing the computer program.

[0035] A computer-readable storage medium stores a computer program, wherein: when the computer program is executed by a processor, the steps of any one of the methods of the present invention are implemented.

[0036] Beneficial effects of the present invention: The catalytic adsorption material development method provided by the present invention in combination with DFT and machine learning can not only save experimental time and reduce costs, but also save computing resources. It can help find and develop suitable catalysts to replace expensive precious metal catalysts. A simple and clear descriptor has universality and promotes material development towards standardization. By improving important influencing factors, it can serve the design of selective catalytic reaction processes to inhibit undesirable overreactions and provide insights into the chemical and physical principles that control catalytic processes. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0038] Figure 1 An overall flow chart of a method for developing catalytic adsorption materials using DFT combined with machine learning provided in the first embodiment of the present invention;

[0039] Figure 2 Comparison of the SVR model DFT calculated value and the machine learning model predicted value in a catalytic adsorption material development method using DFT and machine learning provided in the third embodiment of the present invention;

[0040] Figure 3 A comparison between the DFT calculated value of the MLP model and the predicted value of the machine learning model in a method for developing catalytic adsorption materials using DFT and machine learning provided in the third embodiment of the present invention;

[0041] Figure 4A method for developing catalytic adsorption materials by combining DFT and machine learning provided in the third embodiment of the present invention is a comparison between the DFT calculated value of the RFR model and the predicted value of the machine learning model;

[0042] Figure 5 A comparison between a DFT calculated value of a GBR model and a predicted value of a machine learning model in a method for developing catalytic adsorption materials using a combination of DFT and machine learning provided in a third embodiment of the present invention;

[0043] Figure 6 A comparison between the XGBR model DFT calculated value and the machine learning model predicted value in a catalytic adsorption material development method using DFT and machine learning provided in a third embodiment of the present invention;

[0044] Figure 7 A comparison between the DFT calculated value of the ETR model and the predicted value of the machine learning model in a method for developing catalytic adsorption materials using DFT and machine learning provided in the third embodiment of the present invention;

[0045] Figure 8 The proportion of important factors of various characteristics in a catalytic adsorption material development method using DFT and machine learning provided in the third embodiment of the present invention;

[0046] Fig. 9 A comparison between a DFT calculated value and a formula predicted value in a method for developing a catalytic adsorption material using a combination of DFT and machine learning provided in a third embodiment of the present invention;

[0047] Fig.10 A catalytic adsorption material structure of a catalytic adsorption material development method combining DFT and machine learning is provided in the first embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0049] Example 1, reference Figure 1 , Fig.10 , which is an embodiment of the present invention, provides a method for developing catalytic adsorption materials by combining DFT with machine learning, comprising:

[0050] S1: Obtain material properties; establish various catalytic adsorption material structures.

[0051] The material properties include the properties of the constituent elements, the intrinsic characteristics of the surrounding coordination environment, and the intrinsic characteristics of the system environment. The properties of the constituent elements include atomic number, relative atomic mass, number of d orbital valence electrons, number of s orbital valence electrons, Pauling electronegativity, first ionization energy, electron affinity, covalent radius, atomic radius, element period number, family number, average effective charge, Hirshfeld charge, and hirshfeld-I charge; the intrinsic characteristics of the surrounding coordination environment include the sum of the Pauling electronegativity of the surrounding atoms, the sum of the covalent radii of the surrounding atoms, the sum of the atomic radii of the surrounding atoms, the sum of the number of valence electrons of the surrounding atoms, the sum of the average effective charge of the surrounding atoms, the sum of the Hirshfeld charge of the surrounding atoms, and the sum of the hirshfeld-I charge of the surrounding atoms; the intrinsic characteristics of the system environment include D band center and band gap.

[0052] The "multiple catalytic adsorption material structures" mentioned here are obtained through optimization. The unoptimized structure uses an infinitely extended graphene model, and the carbon atoms in it are deleted and replaced to construct the initial single-atom catalytic material model. Therefore, the energy at this time is not the lowest, the structure is not stable, and structural optimization is required. Structural optimization is performed through density functional theory software to obtain the optimal structure.

[0053] Further, if Fig.10 As shown in the figure, the structures of catalytic adsorption materials are all single-atom catalytic materials. The difference lies in the change of the coordination environment between the central metal and the surrounding (for example: a is four nitrogen atoms around the central intermetallic, b is three nitrogen atoms and one sulfur atom around the central metal, and c is three nitrogen atoms and one phosphorus atom around the central metal). The structure is based on the structure of single-atom materials that have been synthesized in existing research.

[0054] When optimizing the structure, VASP software was used to optimize the structure, and the Projector Augmented Wave (PAW) pseudopotential was used to calculate the interaction between electrons and nuclei. Under the generalized gradient approximation (GGA) condition, the Perdew–Burke–Ernzerhof (PBE) functional was used to calculate the electronic structure of the material. The cutoff energy was set to 500 eV. The convergence threshold of energy was set to 10-5 eV, and the convergence threshold of force was set to The repeating unit is k-point sampled for the Brillouin zone using a 2×3×1 grid. The DFT-D3 method is used to treat the van der Waals interactions between molecules. The intermolecular interlayer spacing is set to To avoid interactions between layers.

[0055] The repeating unit is a small part of the calculation unit. Carbon materials have many carbon atoms, and a small part of the repeating unit will be taken out for calculation during simulation calculation (such as Fig.10 This small unit is the repeating unit.

[0056] After the settings are completed, submit the VASP program for calculation.

[0057] S2: Calculate the adsorption energy of different structures for oxygen molecules.

[0058] The energy of each structure is also calculated by VASP. When calculating the adsorption energy of different structures to oxygen molecules, spin and van der Waals force are considered, and the adsorption free energy ΔG of O2 molecules is:

[0059]

[0060] in, is the Gibbs free energy of the system after O2 adsorption, G M@C are the Gibbs free energy of the free O2 molecule and the original material model, respectively.

[0061] S3: Use the trained model to make predictions and take the coefficient of determination R2 as the evaluation criterion to qualitatively describe the prediction ability of different machine learning models for adsorption energy.

[0062] The obtained data set is divided into training set and test set and normalized.

[0063] The obtained training set is input into the machine learning model for training, and the trained model is used to predict the test set. The determination coefficient R2 is used as the evaluation standard (R2 determination coefficient is an indicator used to evaluate the goodness of fit of a regression model. It indicates the proportion of data variance that the model can explain, and is usually used to compare the performance of different models.) to qualitatively describe the prediction ability of different machine learning models for adsorption energy. 20% of the data obtained by density functional theory (DFT) calculations are randomly selected as the test set, and 80% are set as the training set. The training set is input into the model, and the optimal hyperparameters of the model are obtained through grid search. The determination coefficient R2 is set as the evaluation indicator of the model.

[0064] There are 6 machine learning models, using SVR, MLP, RFR, GBR, XGBR, and ETR. Only the feature importance value of one model is used for linear regression. The six models are used for comparison, and different models are suitable for different adsorption objects and structures.

[0065] S4: We use machine learning to extract important factors of different characteristics and then describe them using linear regression to predict the adsorption energy of different structures on oxygen molecules.

[0066] Use machine learning to extract important factors of different characteristics and select a few characteristics with a higher proportion. Specifically: After the model is established through the machine learning algorithm, the relationship between the eigenvalue and the target value will be learned, and then the algorithm will automatically and directly derive the importance of each eigenvalue. The eigenvalues ​​here are the 23 characteristics obtained above (such as atomic number, relative atomic mass, number of d-orbital valence electrons..., etc.). Do not rely solely on a single indicator, but consider gain, coverage, and frequency comprehensively. For example, a feature may have low gain but high coverage and frequency, which may mean that it is still important in certain situations. Use a weighted average method to combine gain, coverage, and frequency to form a comprehensive score. This part is built into the machine learning algorithm. Specific selection rules can be pre-set in the system according to the needs of the material.

[0067] For example, by testing 6 machine learning models, the XGBR method achieved the highest correlation coefficient on the test set. In the XGBR algorithm, the specific importance judgment takes into account three factors: (1) weight indicates the number of times a feature is used in all trees. This parameter reflects the importance of the feature, because if a feature is used in more trees, then its contribution to the final prediction result is greater. (2) gain indicates the average gain of a feature on the prediction result in all trees. This parameter reflects the splitting ability of the feature at each node, because if a feature has a stronger splitting ability at each node, its contribution to the final prediction result is greater. (3) cover indicates the average coverage of a feature on samples in all trees. This parameter reflects the coverage ability of the feature on the model, because if a feature has an impact on more samples, its contribution to the final prediction result is greater.

[0068] Through the previous machine learning algorithm, we obtained several important features (features whose importance ratio is greater than the preset value), and then used linear regression again to obtain the relationship between these important features and the adsorption energy of oxygen molecules. The results were used to describe the adsorption energy of different structures on oxygen molecules.

[0069] Example 2 is an embodiment of the present invention, which provides a catalytic adsorption material development system using DFT and machine learning, comprising:

[0070] Acquisition module, obtain material properties; establish various catalytic adsorption material structures.

[0071] Calculation module, calculates the adsorption energy of different structures to oxygen molecules.

[0072] The prediction module uses the trained model to make predictions and takes the coefficient of determination R2 as the evaluation criterion to qualitatively describe the prediction ability of different machine learning models for adsorption energy. It uses machine learning to extract important factors of different characteristics and uses linear regression methods to describe the conformity to predict the adsorption energy of different structures on oxygen molecules.

[0073] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0074] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0075] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0076] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0077] Example 3, reference Figure 2-9 , which is an embodiment of the present invention, provides a method for developing catalytic adsorption materials by combining DFT with machine learning. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0078] Figure 2-7 These are the six machine learning prediction performance indicators. It can be seen that the XGBR method achieved the highest correlation coefficient of 0.955 on the test set.

[0079] from Figure 8 It can be seen that the sum of the Pauling electronegativity of the surrounding atoms, the D band center, the covalent radius, the sum of the Hirshfeld charges of the surrounding atoms and the average effective charge account for the highest proportion, and the characteristic surfaces that do not appear in the figure are too low in importance.

[0080] Based on the importance of the above characteristics, the five most important elements were selected to establish the relationship between characteristics and adsorption energy:

[0081] E ads =2.433N s -0.152DC-0.144R c -2.469H s +0.525Q aec -14.057

[0082] E ads is the adsorption energy of O2 on the single-atom material, N s is the sum of the Pauling electronegativity of the surrounding atoms, DC is the center of the D band, R c is the covalent radius, H s is the sum of the Hirshfeld charges of the surrounding atoms, Q aec is the average effective charge.

[0083] pass Fig. 9It can be seen that the adsorption energy predicted by the above calculation has an R2 of 0.816 on the data set. Although it is slightly lower than that of the XGBR machine learning method, it can quickly obtain the adsorption energy without a model and has universal applicability.

[0084] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for developing catalytic adsorption materials by combining DFT and machine learning, characterized in that: include: Obtain material properties; Establish various catalytic adsorption material structures; Calculate the adsorption energy of different structures on oxygen molecules; The trained model was used for prediction, and the coefficient of determination R2 was used as the evaluation criterion to qualitatively describe the prediction ability of different machine learning models for adsorption energy; Machine learning is used to extract important factors of different characteristics, and the linear regression method is used to describe the conformity to predict the adsorption energy of different structures on oxygen molecules.

2. The method for developing catalytic adsorption materials by combining DFT and machine learning as claimed in claim 1, characterized in that: The material properties include the properties of the constituent elements, the intrinsic characteristics of the surrounding coordination environment, and the intrinsic characteristics of the system environment.

3. The method for developing catalytic adsorption materials by combining DFT and machine learning as claimed in claim 2, characterized in that: The characteristics of the constituent elements include atomic number, relative atomic mass, number of d-orbital valence electrons, number of s-orbital valence electrons, Pauling electronegativity, first ionization energy, electron affinity, covalent radius, atomic radius, element period number, family number, average effective charge, Hirshfeld charge, and Hirshfeld-I charge; The intrinsic characteristics of the surrounding coordination environment include, for example, the sum of the Pauling electronegativity of the surrounding atoms, the sum of the covalent radii of the surrounding atoms, the sum of the atomic radii of the surrounding atoms, the sum of the number of valence electrons of the surrounding atoms, the sum of the average effective charges of the surrounding atoms, the sum of the Hirshfeld charges of the surrounding atoms, and the sum of the Hirshfeld-I charges of the surrounding atoms; The intrinsic characteristics of the system environment include D-band center and energy band gap.

4. The method for developing catalytic adsorption materials by combining DFT and machine learning as claimed in claim 3, characterized in that: The various catalytic adsorption material structures include using an infinitely extended graphene model as an unoptimized structure, and constructing an initial single-atom catalytic material model by deleting and replacing the carbon atoms therein; The initial single-atom catalytic material model is structurally optimized to obtain an optimal structure.

5. The method for developing catalytic adsorption materials by combining DFT and machine learning as claimed in claim 4, characterized in that: The structural optimization specifically includes: using VASP software to perform structural optimization, and the settings of various parameters include: Use projected augmented wave pseudopotentials to calculate the interaction between electrons and nuclei; Under GGA conditions, the PBE functional is used to calculate the electronic structure of the material; The cutoff energy was set to 500 eV, the convergence threshold of energy was set to 10-5 eV, and the convergence threshold of force was set to During simulation calculations, a small number of units are taken from the graphene model for calculations and recorded as repeating units; The repeating unit is k-point sampled for the Brillouin zone through a 2 × 3 × 1 grid; The DFT-D3 method was used to treat the van der Waals interactions between molecules; Set the intermolecular spacing to To avoid interactions between layers.

6. The method for developing catalytic adsorption materials by combining DFT and machine learning as claimed in claim 5, characterized in that: When calculating the adsorption energy of oxygen molecules by different structures, considering the spin and van der Waals force, the adsorption free energy ΔG of O2 molecules is: in, is the Gibbs free energy of the system after O2 adsorption, G M@C are the Gibbs free energy of the free O2 molecule and the original material model, respectively.

7. The method for developing catalytic adsorption materials by combining DFT and machine learning as claimed in claim 6, characterized in that: The training of the machine learning model includes randomly extracting 20% ​​of the data obtained through density functional theory calculations as a test set, and setting 80% as a training set; inputting the training set into the model, and relying on grid search to obtain the best hyperparameters of the model; Machine learning is used to extract important factors of different characteristics, and characteristics with a high proportion are selected. By performing linear regression on each feature, the adsorption energy of different structures on oxygen molecules is described.

8. A catalytic adsorption material development system using DFT combined with machine learning according to any one of the methods of claims 1 to 7, characterized in that: Acquisition module, to obtain material properties; to establish various catalytic adsorption material structures; Calculation module, which calculates the adsorption energy of different structures on oxygen molecules; The prediction module uses the trained model to make predictions and takes the coefficient of determination R2 as the evaluation criterion to qualitatively describe the prediction ability of different machine learning models for adsorption energy. It uses machine learning to extract important factors of different characteristics and uses linear regression methods to describe the conformity to predict the adsorption energy of different structures on oxygen molecules.

9. A computer device comprising: Memory and processor; The memory stores a computer program, characterized in that when the processor executes the computer program, the steps of the method for developing catalytic adsorption materials by combining DFT with machine learning are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a method for developing catalytic adsorption materials by combining DFT with machine learning are implemented.

Images