A model training method and device for simulating water decomposition reaction

Abstract

The application discloses a model training method and device for simulating water decomposition reaction, obtains structure parameters and experimental characterization data of a sample material, constructs a surface model corresponding to the sample material based on the structure parameters and the experimental characterization data, obtains first target continuous change data based on dynamic evolution of the surface model in a water decomposition reaction environment by AIMD simulation, and determines the first target continuous change data as an initial training set; a time step sampling strategy is used to determine second target continuous change data from the initial training set; and model training is performed in a neural network based on the second target continuous change data, so as to obtain an interatomic potential function model; and the interatomic potential function model is used to simulate a target material, so that the calculation speed is greatly improved under the premise of ensuring a certain accuracy, and the model is suitable for long-time scale and large-system simulation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of molecular simulation calculation of material science, and in particular to a model training method and device for simulating water decomposition reaction. BACKGROUND

[0002] In practical application environments where metal materials are in contact with water (such as electrochemical reactions, corrosion protection, energy conversion systems, etc.), the structure and properties of the metal / water interface have a decisive influence on the performance of the system. At present, a variety of theoretical models have been used to explain the behavior of the metal / water interface, and experimental results obtained under ultra-high vacuum conditions can be reasonably explained by existing theoretical frameworks. However, practical application scenarios often involve normal pressure, impurities, and liquid water environments, which are significantly different from the idealized ultra-high vacuum conditions, resulting in limited applicability of the interface structure description obtained based on idealized environments under real working conditions.

[0003] In related technologies, the AIMD (Ab Initio Molecular Dynamics) method is used to simulate the metal / water interface. This method can explicitly consider the presence of liquid water molecules and their complex interactions with the metal surface at the atomic scale, thus more realistically reflecting the actual interface behavior. Although the AIMD method has high accuracy in revealing the microscopic mechanism, it has high computational cost, and the time and spatial scales of the simulation are limited, making it difficult to meet the needs of long-time and large-system interface behavior research.

[0004] Furthermore, to further expand the time and space scales of the simulation, ML (Machine Learning) methods are introduced to construct efficient interatomic potential function models for metal / water interfaces and metal oxide / water interfaces. The ML model trained based on AIMD data not only retains high physical and chemical fidelity, but also significantly improves the simulation efficiency, making it possible to study the dynamic behavior of interfaces in larger systems and longer time scales.

[0005] However, taking the interaction between platinum material and water as an example, the exchange process between bulk water and the interface region requires about 500 picoseconds, which is far beyond the capability of the AIMD method, which can usually only simulate for tens of picoseconds. Since the length of the AIMD simulation is not sufficient to allow the system to reach a state of thermodynamic equilibrium, the results obtained based on this short time scale simulation about the microstructure of liquid water and its related electrochemical properties may not be accurate or representative. That is, the existing AIMD simulation cannot fully capture the complete dynamic exchange process of water molecules between the interface and the bulk phase, resulting in limitations in describing the interface structure and corresponding electrochemical properties, making it difficult to fully reflect the real situation under actual working conditions. SUMMARY

[0006] Therefore, the application provides a model training method and device for simulating water decomposition reaction to solve the problem of low simulation efficiency in the prior art.

[0007] The object of the application can be achieved by the following technical solutions:

[0008] The first aspect of the application provides a model training method for simulating water decomposition reaction, comprising:

[0009] Obtaining the structure parameters and experimental characterization data of the sample material;

[0010] Constructing a surface model corresponding to the sample material based on the structure parameters and experimental characterization data;

[0011] Based on the dynamic evolution process of the surface model in the water decomposition reaction environment simulated by AIMD, the first target continuous change data is obtained, and the first target continuous change data is determined as the initial training set;

[0012] The time step sampling strategy is used to determine the second target continuous change data from the initial training set;

[0013] Based on the second target continuous change data input into the neural network for model training, an interatomic potential function model is obtained;

[0014] Based on the interatomic potential function model, ML-MD simulation is performed on the target material.

[0015] In an optional embodiment, it further comprises:

[0016] Determining the minimum energy path of the reaction system of the target material;

[0017] Based on the minimum energy path, the transition state structure and the reaction energy barrier are determined;

[0018] Based on the transition state structure and the reaction energy barrier, the atomic scale reaction mechanism is revealed.

[0019] In an optional embodiment, it further comprises:

[0020] Coupling the interatomic potential function model with the pre-set coarse-grained model to construct a cross-scale calculation framework;

[0021] Based on the cross-scale calculation framework, the mass transfer of the reactants, the surface adsorption of the reactants and the desorption process of the reactant products corresponding to the target material are simulated to obtain the kinetic parameters of the target material.

[0022] In an optional embodiment, it further comprises:

[0023] The multi-dimensional index based on the interatomic potential function model verifies the calculation accuracy and generalization ability of the interatomic potential function model.

[0024] In an alternative embodiment, further comprising:

[0025] Determine the element doping ratio, surface active site distribution and stability optimization parameters in the ML-MD simulation process.

[0026] Based on the element doping ratio, surface active site distribution and stability optimization parameters, the catalyst is prepared.

[0027] In an alternative embodiment, further comprising:

[0028] The catalyst is subjected to microstructure characterization processing.

[0029] In an alternative embodiment, further comprising:

[0030] The catalyst is subjected to electrochemical performance testing.

[0031] The second aspect of the present application provides a model training device for simulating water decomposition reaction, comprising:

[0032] The acquisition module is configured to acquire the structure parameters and experimental characterization data of the sample material.

[0033] The construction module is configured to construct a surface model corresponding to the sample material based on the structure parameters and experimental characterization data.

[0034] The first simulation module is configured to obtain first target continuous change data based on the dynamic evolution process of the AIMD simulated surface model in the water decomposition reaction environment, and determine the first target continuous change data as an initial training set.

[0035] The determination module is configured to determine second target continuous change data from the initial training set using a time step sampling strategy.

[0036] The training module is configured to input the second target continuous change data into a neural network for model training to obtain an interatomic potential function model.

[0037] The second simulation module is configured to perform ML-MD simulation on the target material based on the interatomic potential function model.

[0038] The third aspect of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method of the first aspect.

[0039] The fourth aspect of the present application provides a computer-readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method of the first aspect.

[0040] Compared with the prior art, the model training method for simulating water decomposition reaction provided by the present application obtains the structural parameters and experimental characterization data of the sample material; constructs a surface model corresponding to the sample material based on the structural parameters and experimental characterization data; obtains first target continuous change data based on the dynamic evolution process of the surface model in the water decomposition reaction environment by AIMD simulation, and determines the first target continuous change data as an initial training set; determines second target continuous change data from the initial training set by using a time step sampling strategy; inputs the second target continuous change data into a neural network for model training to obtain an interatomic interaction potential function model; and simulates the target material using the interatomic interaction potential function model, which can greatly improve the calculation speed under the premise of ensuring a certain accuracy and is suitable for long-time scale and large system simulation. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0042] Figure 1 A flowchart of the model training method for simulating water decomposition reaction provided by the embodiments of the present application;

[0043] Figure 2 A structural block diagram of the model training device for simulating water decomposition reaction provided by the embodiments of the present application;

[0044] Figure 3 A structural block diagram of the electronic device for implementing the model training method for simulating water decomposition reaction provided by the embodiments of the present application. DETAILED DESCRIPTION

[0045] In order to make the person skilled in the art better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application.

[0046] It should be noted that the terms "first", "second" and the like in the description and in the claims of the present application and the above-mentioned accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances so that the embodiments of the present application described herein can be implemented in other than the order illustrated or described herein. In addition, the terms "comprise" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a list of steps or units as an element does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to such processes, methods, products or devices.

[0047] It should be understood that in the embodiments of the present application, "at least one" means one or more, and "multiple" means two or more. "And / or" is only a description of the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B can mean that A exists alone, A and B exist together, and B exists alone. The character " / " generally represents that the associated objects before and after it are in an "or" relationship. "Including A, B and / or C" means including any one or any two or three of A, B and C.

[0048] It should be understood that in the embodiments of the present application, "B corresponding to A", "B corresponding to A", "A corresponding to B" or "B corresponding to A" means that B is associated with A, and B can be determined according to A. Determining B according to A does not mean that B is determined only according to A, but also can be determined according to A and / or other information.

[0049] To solve the technical problems existing in the related art, the embodiments of the present application provide a model training method and device for simulating water decomposition reaction.

[0050] The model training method for simulating water decomposition reaction provided by the embodiments of the present application can be executed by an electronic device, which can be a terminal or a server and the like. The terminal can be a terminal device such as a smart phone, a tablet computer, a notebook computer and the like. The server can be a stand-alone physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing cloud services, cloud databases, cloud computing, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and basic cloud computing services such as big data and artificial intelligence platforms. It can be understood that the present application does not specifically limit the execution subject of the model training method for simulating water decomposition reaction.

[0051] Before the model training method for simulating water decomposition reaction provided by the embodiments of the present application is described in detail, the terms involved in the present application are first explained.

[0052] High-entropy oxide (HEO): refers to an oxide with a configurational entropy greater than or equal to 1.5R, and the specific calculation formula is as follows:

[0053]

[0054] wherein S is the mixed configurational entropy of the high-entropy oxide, N is the number of element types, x i is the mole fraction of the i-th component, and R is the universal gas constant.

[0055] LAMMPS (Large-scale Atomic / Molecular Massively Parallel Simulator): is a classical molecular dynamics program focusing on material modeling. The software has the potential function of simulating solid materials (metals and semiconductors), soft materials (biological molecules and polymers), and coarse-grained or mesoscopic systems. It can be used for atomic scale modeling, and can also be more generally used as a parallel particle simulation platform for atomic, mesoscopic or continuous medium scale.

[0056] AIMD: is a molecular dynamics simulation method based on the basic principles of quantum mechanics. Its core feature is to directly solve the interatomic interaction force through first-principle calculation (i.e. independent of empirical parameters), thereby simulating the motion trajectory of atoms or molecules.

[0057] NEB (Nudged Elastic Band) method: in the field of computational chemistry and materials science, it specifically refers to a numerical simulation algorithm for finding transition states (such as reaction paths and phase transition paths) of chemical / physical processes. The core idea is to "nudge" the path points on the potential energy surface through the elastic band model to converge to the minimum energy path.

[0058] The technical solutions of the present application will be described in detail through specific embodiments. It should be noted that the following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments described below are used to explain the technical solutions of the present application and do not limit the actual use.

[0059] To solve the technical problems existing in the related art, the embodiments of the present application provide a model training method for simulating water decomposition reaction, as shown in Figure 1 Figure 1 ​An example flowchart of the model training method for simulating water decomposition reaction provided by the embodiments of the present application is shown. It should be noted that the steps shown can be executed in a different logical order than that shown in the method flowchart. The method can include the following steps S101-S106.

[0060] Step S101: Obtain the structural parameters and experimental characterization data of the sample material.

[0061] In an alternative embodiment, the material capable of promoting water decomposition reaction includes at least one of the following materials: metal, metal oxide, carbon-based material, sulfide, phosphide, nitride, and perovskite-type oxide. For example, the sample material is a high-entropy oxide. It should be noted that in addition to the exemplary materials given above, other materials can also be used, which are not limited by the present application.

[0062] In an alternative embodiment, the structural parameters of the material mainly refer to those capable of defining the arrangement of atoms or molecules inside the material and its macroscopic morphology. The structural parameters include at least one of the following parameters: lattice parameters, atomic positions, space groups, defect information, thermal vibration parameters, and microstructure characteristics. In addition, other structural parameters can also be included, which are not limited by the present application.

[0063] Experimental characterization data is data obtained by various experimental means, which is used to verify the structural parameters of the material and further explore the performance of the material. The experimental characterization data includes at least one of the following parameters: X-ray diffraction: used to determine the crystal structure and lattice constant. Transmission electron microscopy: can directly observe the atomic arrangement at the nanoscale, especially useful for the study of surfaces and interfaces. Scanning tunneling microscope: can observe the surface morphology at the atomic level and obtain information on the local electronic density of states. X-ray photoelectron spectroscopy: used to analyze the surface elemental composition and chemical state. Raman spectroscopy: can be used to study molecular vibration modes, indirectly reflecting the bonding characteristics in the crystal structure. Of course, the experimental characterization data can also be other experimental characterization data, which is not limited by the present application.

[0064] Step S102: Construct a surface model corresponding to the sample material based on the structural parameters and experimental characterization data.

[0065] It should be noted that if the sample material refers to a metal oxide, the surface model can be a metal oxide surface model.

[0066] In an alternative embodiment, the surface model corresponding to the sample material is constructed based on the structural parameters and experimental characterization data, which specifically includes the following steps:

[0067] According to the experimental characterization data and the structure parameters, the main exposed crystal surface of the sample material is determined; and the selected unit cell is cut along a preset direction to generate a surface model corresponding to the sample material.

[0068] In this step, a surface model corresponding to the sample material is constructed based on the structure parameters and the experimental characterization data, so as to ensure that the surface model conforms to the actual crystal structure and the surface exposed crystal characteristics.

[0069] Step S103: obtaining first target continuous change data based on the dynamic evolution process of the surface model in the water decomposition reaction environment through AIMD simulation, and determining the first target continuous change data as an initial training set.

[0070] In an optional embodiment, the water decomposition reaction includes an electrocatalytic oxygen evolution reaction and an electrocatalytic hydrogen evolution reaction.

[0071] In an optional embodiment, obtaining the first target continuous change data based on the dynamic evolution process of the surface model in the water decomposition reaction environment through AIMD simulation specifically includes the following steps:

[0072] The constructed surface model is placed in the AIMD model to ensure that the surface model is under correct temperature and pressure conditions; simulation conditions such as simulation time step and total simulation time are set; and simulation is performed, and the program will update the atomic position at each time step and obtain the energy and force field of the system by solving the Schrödinger equation.

[0073] In an optional embodiment, the first target continuous change data includes at least one of the following data: atomic type change data, atomic coordinate change data, system energy change data, and atomic force field change data.

[0074] Step S104: determining second target continuous change data from the initial training set by using a time step sampling strategy.

[0075] In an optional embodiment, the second target continuous change data includes the first target continuous change data containing at least one of the following points: an energy mutation point, a bond reconstruction point, and a reactant and product conversion point.

[0076] In an optional embodiment, the second target continuous change data is determined from the initial training set by using a time step sampling strategy, specifically including the following steps: extracting frame configurations at a fixed interval; dynamically adjusting the sampling frequency according to the energy or structural change rate; performing energy mutation point detection, bond reconstruction point detection, and / or reactant / product conversion point identification detection to obtain the second target continuous change data.

[0077] In this step, not only the information content of the data set can be improved, but also the burden of redundant data on model training can be reduced.

[0078] Step S105: inputting the second target continuous change data into the neural network for model training to obtain an interatomic interaction potential function model.

[0079] In an optional embodiment, the neural network can be a graph neural network or a neural network potential, or other neural networks according to actual needs, which are not limited in the present application.

[0080] In a specific embodiment, the multi-dimensional index includes at least one of the following indexes: energy prediction error, force field convergence, and reaction energy barrier reproducibility.

[0081] In another optional embodiment, the model training method for simulating water decomposition reactions provided in the embodiments of the present application further includes: verifying the calculation accuracy and generalization ability of the interatomic interaction potential function model based on the multi-dimensional index of the interatomic interaction potential function model.

[0082] Step S106: performing ML-MD (Machine Learning Molecular Dynamics) simulation on the target material based on the interatomic interaction potential function model.

[0083] In the embodiments of the present application, the structure parameters and experimental characterization data of the sample material are obtained, a surface model corresponding to the sample material is constructed based on the structure parameters and experimental characterization data, the dynamic evolution process of the surface model in a water decomposition reaction environment is obtained based on AIMD simulation, the first target continuous change data is determined as the initial training set, the second target continuous change data is determined from the initial training set by using a time step sampling strategy, and the interatomic interaction potential function model is obtained by inputting the second target continuous change data into the neural network for model training. Using the interatomic interaction potential function model to simulate the target material can greatly improve the calculation speed under the premise of ensuring a certain accuracy, and is suitable for long-time scale and large system simulation. Moreover, the dynamic evolution behavior of the surface atoms of the catalytic material in the electrochemical reaction environment can be tracked at the nanoscale time scale.

[0084] In another optional embodiment, the model training method for simulating water decomposition reactions provided in the embodiments of the present application further includes:

[0085] determining the minimum energy path of the reaction system of the target material; determining the transition state structure and the reaction energy barrier based on the minimum energy path; and revealing the atomic scale reaction mechanism based on the transition state structure and the reaction energy barrier.

[0086] It should be noted that the minimum energy path refers to the path with the lowest energy from reactants to products, typically involving one or more transition states. The transition state structure refers to the point of highest energy during the reaction, determining the reaction rate. The reaction energy barrier refers to the energy difference between reactants and the transition state, reflecting the ease with which the reaction occurs. Revealing atomic-scale reaction mechanisms means that by analyzing these paths and structures, we can gain a deeper understanding of the specific reaction steps occurring on the surface of catalytic materials, including microscopic processes such as adsorption, dissociation, and recombination. In a specific embodiment, the minimum energy path of the target material's reaction system is determined using the NEB method. The NEB method is used to find the optimal path from reactants to products, ensuring that the energy changes at each step are accurately captured.

[0087] In one specific embodiment, determining the transition state structure and reaction energy barrier based on the minimum energy path includes the following steps:

[0088] The transition state structure is identified based on the minimum energy path, and the reaction energy barrier is calculated based on the transition state structure.

[0089] Identifying the highest energy point in the reaction pathway, i.e., the transition state, is crucial for understanding reaction mechanisms. Calculating the reaction energy barrier quantifies the energy barrier at which a reaction occurs, helping to assess the likelihood and rate of the reaction.

[0090] In another optional embodiment, the model training method for simulating water splitting reaction provided in this application further includes:

[0091] A cross-scale computational framework is constructed by coupling the interatomic interaction potential function model with a pre-defined coarse-grained model.

[0092] Based on a cross-scale computational framework, the mass transfer, surface adsorption, and product desorption processes of the target material are simulated to obtain the kinetic parameters of the target material.

[0093] In one specific embodiment, the kinetic parameters include at least one of the following parameters: reaction rate constant and conversion frequency.

[0094] In this embodiment, the catalytic activity and stability of the target material are cross-validated by combining kinetic parameters, experimentally measured polarization curves, and electrochemical data such as the Tafel slope.

[0095] In another optional embodiment, the model training method for simulating water splitting reaction provided in this application further includes:

[0096] The elemental doping ratio, surface active site distribution, and stability optimization parameters were determined during the ML-MD simulation process; based on the elemental doping ratio, surface active site distribution, and stability optimization parameters, the catalyst was prepared.

[0097] In a specific embodiment, based on the element doping ratio, the surface active site distribution and the stability optimization parameters, the catalyst is prepared by high-temperature solid-phase synthesis method and sol-gel method and the like.

[0098] In another alternative embodiment, the model training method for simulating water decomposition reaction provided by the embodiments of the present application further comprises:

[0099] The catalyst is subjected to microstructure characterization processing.

[0100] In a specific embodiment, the microstructure characterization processing of the catalyst specifically comprises the following steps: the surface morphology and particle distribution of the catalyst are analyzed by a scanning electron microscope, the crystal phase composition and lattice distortion degree are determined by X-ray diffraction, and the atomic scale structure characteristics are analyzed by a transmission electron microscope and selected area electron diffraction.

[0101] In another alternative embodiment, the model training method for simulating water decomposition reaction provided by the embodiments of the present application further comprises:

[0102] The catalyst is subjected to electrochemical performance testing.

[0103] In a specific embodiment, the electrochemical performance testing of the catalyst specifically comprises the following steps: the oxygen evolution reaction polarization curve is obtained by linear sweep voltammetry on an electrochemical workstation by using a three-electrode system, the interface reaction kinetics is analyzed by electrochemical impedance spectroscopy, the material stability is evaluated by constant current cycling method, and the device-level performance parameters are tested in an anion exchange membrane or a proton exchange membrane electrolytic cell.

[0104] The performance parameters include voltage efficiency, gas purity and life attenuation rate and the like.

[0105] The traditional DFT (Density Functional Theory) method needs thousands of core hours for DFT simulation of a single surface catalytic reaction, and the system size is usually limited within 200 atoms. The method provided by the present application can shorten the single-step calculation time from seconds to milliseconds at the DFT level of accuracy. The method provided by the embodiments of the present application supports nanosecond-level molecular dynamics (ML-MD) simulation of a million-atom supercell, and reveals the migration dynamics of interface water molecules.

[0106] In the model training method for simulating water decomposition reaction provided in the embodiments of the present application, in order to solve the configuration space explosion problem caused by the heterogeneity of the local chemical environment of multiple atoms, a quantitative characterization system based on a smooth overlapping atomic position descriptor or an atomic center symmetry function is constructed; by analyzing the bond angle distribution, coordination number variation coefficient and valence charge transfer amount of each element within a local radius of 1-5 Å, the coordination environment difference of the surface active site is accurately encoded, and the discrimination degree of different chemical environments is ≥95%.

[0107] Machine learning molecular dynamics (ML-MD) is used to replace the traditional empirical force field to realize nanosecond-level surface evolution process simulation; by real-time tracking of the dynamics correlation between oxygen vacancy formation energy and surface order parameter, the influence mechanism of non-equilibrium behavior such as phase segregation on catalytic activity is quantified, and the simulation accuracy reaches the level of quantum mechanics (force field error ≤0.1 eV / Å) (electron volts).

[0108] In another specific embodiment, in view of the defects that the traditional micro-elastic band method needs to preset the reaction path and the single calculation time is ≥72 hours (taking a 100-atom system as an example), a self-pathfinding algorithm coupling machine learning potential function and reinforcement learning is developed; a Markov decision process is established, the atomic force gradient is taken as the reward function, and through the Q-learning strategy, multiple candidate paths on the potential energy surface are dynamically explored, and the optimal reaction channel with an energy barrier ≤0.8 eV is automatically selected.

[0109] In another optional embodiment, the problem that classical molecular dynamics cannot accurately describe the double-layer structure of the electrode-electrolyte interface is solved, and a pre-trained neural network potential is used to analyze the orientation distribution function of the interface water molecules and the hydrogen bond lifetime in real time; through a graph convolution network, the dynamic hydrogen bond configuration (bond length 1.5-2.0 Å, bond angle 155°-175°) of the reaction intermediate and water molecules is captured, and the influence of hydrogen bond network stability on adsorption free energy is quantitatively output.

[0110] In another optional embodiment, in view of the defects that the traditional trial-and-error method needs to synthesize ≥200 material samples (research and development period ≥6 months), a D-optimal experimental design is used to construct a sample screening model; based on the confidence interval of the material performance predicted by Gaussian process regression, key component points on the Pareto frontier are preferentially selected, the experimental sample requirement is reduced, and the research and development period is shortened.

[0111] In order to solve the system error that the deviation between the DFT prediction and the measured performance is >0.5 eV, an in-situ characterization data fusion channel is established, the local electronic structure data of the synchrotron X-ray absorption spectrum and the lattice strain field distribution of the four-dimensional scanning transmission electron microscope are accessed in real time, the graph neural network with attention mechanism weighting is used to dynamically correct the machine learning potential function, and the root mean square error between the theoretical adsorption energy and the experimental value is reduced.

[0112] Based on D-optimal experimental design or uncertainty sampling strategy, an iterative data generation loop is constructed; the configuration importance is evaluated by machine learning model prediction error, and the sample with the largest contribution to the potential function accuracy improvement is preferentially selected for DFT calculation, which realizes the directional deployment of computing resources and reduces redundant calculation by ≥70%;

[0113] Integrate DFT energy / force field data of electronic scale, atomic scale molecular dynamics trajectory data, and experimental characterization data such as X-ray photoelectron spectroscopy and extended X-ray absorption fine structure, etc. to construct a unified coded feature space; through transfer learning and multi-task training mechanism, the generalization ability of machine learning model to complex chemical environment is improved, and the average absolute error between prediction result and experimental value is ≤0.03 eV / atom (atom).

[0114] Through D-optimal (D-optimal) design or Bayesian optimization algorithm, key samples are intelligently selected from high-dimensional combination space, reducing DFT calculation by more than 70%. Use negative data mining strategy to build a sample database and improve model generalization ability.

[0115] Joint density functional theory calculation of electronic structure data, molecular dynamics simulation of atomic trajectory data, and experimental characterization of microstructure information (including but not limited to X-ray photoelectron spectroscopy element valence state distribution, extended X-ray absorption fine structure local coordination environment), using interpolation algorithm based on radial basis function or Gaussian process regression model, ≥1000 feature points are discretely sampled in reaction coordinate space (such as bond length-bond angle two-dimensional plane); through energy gradient optimization and transition state verification, a continuous and differentiable three-dimensional potential energy surface is generated, with potential energy value error ≤0.05 eV, and force field root mean square error ≤0.1 eV / Å, which is used to accurately describe the energy evolution process in the catalytic reaction path.

[0116] Integrate Metadynamics (Metadynamics) and reinforcement learning to accelerate oxygen vacancy migration and interface water molecule dynamics simulation, and expand the time scale to nanosecond level. Use quantum-classical hybrid simulation, use interatomic potential function model for interface reaction zone, and use coarse-grained model for bulk phase, to realize catalytic process simulation of 10,000 atom system.

[0117] The pre-trained neural network potential or the graph neural network potential is used to replace the traditional quantum mechanics calculation, the speed of single-point energy / potential field calculation is improved under the premise of maintaining the accuracy of electronic structure, the linear interpolation path between the reactants and the products is initialized, the conjugate gradient method is used for iterative optimization of the intermediate image position, the force calculation of the atom is performed by the interatomic potential function model, and the path convergence condition is set as an average atomic force threshold ≤0.05 eV / Å; the converged path is verified as a transition state, and ≥5 candidate paths are searched in parallel, finally, the path with the lowest reaction energy barrier is selected as the optimal catalytic reaction channel, and the path energy barrier found by the traditional density functional theory (DFT) -NEB method is reduced.

[0118] Corresponding to the model training method for simulating water decomposition reaction provided by the embodiments of the present application, the embodiments of the present application also provide a model training device for simulating water decomposition reaction, as shown in Figure 2 The model training device for simulating water decomposition reaction comprises:

[0119] The acquisition module 201 is configured to acquire structure parameters and experimental characterization data of a sample material.

[0120] The construction module 202 is configured to construct a surface model corresponding to the sample material based on the structure parameters and the experimental characterization data.

[0121] The first simulation module 203 is configured to obtain first target continuous change data by simulating a dynamic evolution process of the surface model in a water decomposition reaction environment based on the AIMD, and determine the first target continuous change data as an initial training set.

[0122] The determination module 204 is configured to determine second target continuous change data from the initial training set by using a time step sampling strategy.

[0123] The training module 205 is configured to input the second target continuous change data into a neural network to perform model training, and obtain an interatomic potential function model.

[0124] The second simulation module 206 is configured to perform ML-MD simulation on a target material based on the interatomic potential function model.

[0125] In an optional embodiment, the model training device for simulating water decomposition reaction further comprises:

[0126] Determine the minimum energy path of the reaction system of the target material.

[0127] Determine the transition state structure and the reaction energy barrier based on the minimum energy path.

[0128] Reveal the atomic-scale reaction mechanism based on the transition state structure and the reaction energy barrier.

[0129] In an optional embodiment, the model training device for simulating water decomposition reaction further comprises:

[0130] coupling the interatomic interaction potential function model and the preset coarse-grained model to construct a cross-scale calculation framework;

[0131] Based on the cross-scale calculation framework, the mass transfer of the reactants, the surface adsorption of the reactants and the desorption process of the reactant products of the target material are simulated to obtain the kinetic parameters of the target material.

[0132] In an optional embodiment, it further comprises:

[0133] Based on the multi-dimensional index of the interatomic interaction potential function model, the calculation accuracy and generalization ability of the interatomic interaction potential function model are verified.

[0134] In an optional embodiment, it further comprises:

[0135] Determine the element doping ratio, surface active site distribution and stability optimization parameters in the ML-MD simulation process;

[0136] Based on the element doping ratio, surface active site distribution and stability optimization parameters, the catalyst is prepared.

[0137] In an optional embodiment, it further comprises:

[0138] The microstructure of the catalyst is characterized.

[0139] In an optional embodiment, it further comprises:

[0140] The electrochemical performance of the catalyst is tested.

[0141] Corresponding to the model training method for simulating water decomposition reaction provided in the embodiments of the present application, the embodiments of the present application also provide an electronic device for executing the model training method for simulating water decomposition reaction, as shown in Figure 3 The electronic device includes a processor 301 and a memory 302 for storing the program of the model training method for simulating water decomposition reaction. After the device is powered on and the processor runs the program of the model training method for simulating water decomposition reaction, the following steps are performed:

[0142] Obtain the structure parameters and experimental characterization data of the sample material;

[0143] Based on the structure parameters and experimental characterization data, a surface model corresponding to the sample material is constructed;

[0144] Based on the dynamic evolution process of the AIMD simulation surface model in the water decomposition reaction environment, first target continuous change data is obtained, and the first target continuous change data is determined as the initial training set;

[0145] determine second target continuous change data from the initial training set by using a time step sampling strategy;

[0146] perform model training based on the second target continuous change data input into the neural network, to obtain an interatomic interaction potential function model;

[0147] perform ML-MD simulation on the target material based on the interatomic interaction potential function model.

[0148] Corresponding to the model training method for simulating water decomposition reactions provided in the embodiments of the present application, the embodiments of the present application also provide a computer readable storage medium storing a program for the model training method for simulating water decomposition reactions, the program being run by a processor to perform the following steps:

[0149] obtain structure parameters and experimental characterization data of a sample material;

[0150] construct a surface model corresponding to the sample material based on the structure parameters and the experimental characterization data;

[0151] obtain first target continuous change data based on a dynamic evolution process of the surface model in an environment of a water decomposition reaction by AIMD simulation, and determine the first target continuous change data as an initial training set;

[0152] determine second target continuous change data from the initial training set by using a time step sampling strategy;

[0153] perform model training based on the second target continuous change data input into the neural network, to obtain an interatomic interaction potential function model;

[0154] perform ML-MD simulation on the target material based on the interatomic interaction potential function model.

[0155] Corresponding to the model training method for simulating water decomposition reactions provided in the embodiments of the present application, the embodiments of the present application also provide a computer program containing instructions, when the program is executed by a computer, the instructions cause the computer to perform the following steps:

[0156] obtain structure parameters and experimental characterization data of a sample material;

[0157] construct a surface model corresponding to the sample material based on the structure parameters and the experimental characterization data;

[0158] obtain first target continuous change data based on a dynamic evolution process of the surface model in an environment of a water decomposition reaction by AIMD simulation, and determine the first target continuous change data as an initial training set;

[0159] determine second target continuous change data from the initial training set by using a time step sampling strategy;

[0160] Model training is performed based on the second target continuous change data input into the neural network, to obtain an interatomic potential function model;

[0161] ML-MD simulation is performed on the target material based on the interatomic potential function model.

[0162] It should be noted that the detailed description of the model training device for simulating water decomposition reaction, the electronic device, the computer readable storage medium and the computer program provided by the embodiments of the present application can refer to the related description of the model training method for simulating water decomposition reaction provided by the embodiments of the present application. Here, it will not be repeated.

[0163] Although the present application is disclosed with the preferred embodiments as above, it is not intended to limit the present application, and any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of the present application. Therefore, the protection scope of the present application should be limited by the scope defined by the claims of the present application.

[0164] In a typical configuration, the electronic device includes one or more processors (Central Processing Unit), input / output interfaces, network interfaces, and memory.

[0165] The memory can include non-persistent memory in computer readable media, random access memory (Random Access Memory) and / or non-volatile memory, such as read only memory (Read Only Memory) or flash memory. The memory is an example of computer readable media.

[0166] Computer-readable media includes permanent and non-permanent, movable and non-movable media that can be implemented by any method or technology to store information. The information can be computer-readable operations, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, programmable analog modules, static random access memory, dynamic random access memory, other types of random access memory, read-only memory, electrically erasable programmable read-only memory, flash memory or other memory technologies, compact disc read-only memory, digital video disc or other optical storage, magnetic cassette, magnetic tape disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. According to the definition herein, computer-readable media does not include non-transitory computer-readable media, such as modulated data signals and carriers.

[0167] Those skilled in the art will appreciate that embodiments of the present application can be provided as methods, systems or computer program products. Accordingly, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, compact disc read-only memory, optical storage and the like) containing computer-usable program code.

[0168] Although the present application is disclosed with reference to the preferred embodiments above, it is not intended to limit the present application, and any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application should be defined by the scope of the claims.

Claims

1. A model training method for simulating a water decomposition reaction, characterized by, The method comprises the following steps: obtaining structural parameters and experimental characterization data of a sample material, wherein the experimental characterization data refers to data obtained through various experimental methods, and the experimental characterization data comprises at least one of the following data: data obtained through X-ray diffraction, data obtained through a transmission electron microscope, data obtained through a scanning tunneling microscope, data obtained through X-ray photoelectron spectroscopy, and data obtained through Raman spectroscopy; determining a main exposed crystal face of the sample material according to the structural parameters and the experimental characterization data; and cutting a selected unit cell along a preset direction to generate a surface model corresponding to the sample material, wherein the preset direction is the main exposed crystal face of the sample material; simulating a dynamic evolution process of the surface model in a water decomposition reaction environment based on an AIMD to obtain first target continuous change data, and determining the first target continuous change data as an initial training set; extracting frame configurations at fixed intervals; dynamically adjusting a sampling frequency according to energy or structural change rates; performing energy mutation point detection, bond reconstruction point detection, and / or reactant / product conversion point identification detection to obtain second target continuous change data, wherein the second target continuous change data comprises the first target continuous change data containing at least one of an energy mutation point, a bond reconstruction point, and a reactant / product conversion point; inputting the second target continuous change data into a neural network to perform model training, and obtaining an interatomic interaction potential function model; performing ML-MD simulation on a target material based on the interatomic interaction potential function model. 2.The model training method for simulating water decomposition reaction according to claim 1, wherein, The method further comprises the following steps: determining a minimum energy path of a reaction system of the target material; determining a transition state structure and a reaction energy barrier based on the minimum energy path; revealing an atomic-scale reaction mechanism based on the transition state structure and the reaction energy barrier. 3.The model training method for simulating water decomposition reaction according to claim 1, wherein, The method further comprises the following steps: coupling the interatomic interaction potential function model with a preset coarse-grained model to construct a cross-scale calculation framework; simulating reactant mass transfer, reactant surface adsorption, and reactant product desorption processes of the target material based on the cross-scale calculation framework to obtain kinetic parameters of the target material. 4.The model training method for simulating water decomposition reaction according to claim 1, wherein, The method further comprises the following steps: verifying calculation accuracy and generalization ability of the interatomic interaction potential function model based on multi-dimensional indexes of the interatomic interaction potential function model. 5.The model training method for simulating water decomposition reaction according to claim 1, wherein, The method further comprises the following steps: determining element doping ratios, surface active site distributions, and stability optimization parameters in the ML-MD simulation process; preparing a catalyst based on the element doping ratios, the surface active site distributions, and the stability optimization parameters. 6.The model training method of simulating water decomposition reaction according to claim 5, characterized in that, The method further comprises the following steps: performing microstructure characterization processing on the catalyst. 7.The model training method of simulating water decomposition reaction according to claim 5, characterized in that, The method further comprises the following steps: performing electrochemical performance testing on the catalyst.

8. A model training device for simulating a water decomposition reaction, characterized by, The method comprises the following steps: an obtaining module, configured to obtain structural parameters and experimental characterization data of a sample material, wherein the experimental characterization data refers to data obtained through various experimental methods, and the experimental characterization data comprises at least one of the following data: data obtained through X-ray diffraction, data obtained through a transmission electron microscope, data obtained through a scanning tunneling microscope, data obtained through X-ray photoelectron spectroscopy, and data obtained through Raman spectroscopy; A construction module configured to determine a main exposed crystal plane of the sample material according to the structure parameters and the experimental characterization data; cutting the selected unit cell along a preset direction to generate a surface model corresponding to the sample material, the preset direction being the main exposed crystal plane of the sample material; a first simulation module configured to simulate a dynamic evolution process of the surface model in a water decomposition reaction environment based on an AIMD, to obtain first target continuous variation data, and to determine the first target continuous variation data as an initial training set; a determination module configured to extract frame configurations at fixed intervals, to dynamically adjust a sampling frequency according to an energy or structure change rate, to perform energy mutation point detection, bond reconstruction point detection, and / or reactant / product conversion point identification detection, to obtain second target continuous variation data, and to determine the second target continuous variation data to include the first target continuous variation data containing at least one of an energy mutation point, a bond reconstruction point, and a reactant / product conversion point; a training module configured to input the second target continuous variation data into a neural network to perform model training, to obtain an interatomic potential function model, and to determine the interatomic potential function model as a model for simulating water decomposition reactions; a second simulation module configured to perform ML-MD simulation on a target material based on the interatomic potential function model.

9. An electronic device, comprising: A computer readable storage medium stores a computer program, and a processor executes the computer program to implement the model training method for simulating water decomposition reactions according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and a processor executes the computer program to implement the model training method for simulating water decomposition reactions according to any one of claims 1-7.

Images