Method, device, electronic equipment and computer program product for predicting catalyst components

Abstract

The application discloses a kind of prediction method, device, electronic equipment and computer program product of catalyst component.Therein, the method includes: obtaining the activation energy demand that target reactant needs catalyst to provide in the case of generating target product by catalytic reaction, wherein, catalyst includes: multiple catalytic elements, and different catalytic elements provide different activation energy for target reactant;Activation energy prediction model is trained in advance, and the catalyst component that catalytic activation energy provided by catalyst meets the activation energy demand of catalytic activation energy is determined, wherein, activation energy prediction model is used to express the association between the catalytic activation energy provided by catalyst and the catalytic catalyst component corresponding to catalyst, and catalyst component indicates the proportion of multiple catalytic elements in catalyst.The present application solves the technical problem that the design of catalyst component in the prior art is complex.

Description

Technical Field

[0001] This invention relates to the field of machine learning, and more specifically, to a method, apparatus, electronic device, and computer program product for predicting catalyst components. Background Technology

[0002] Catalysts play a vital and irreplaceable role in energy conversion, chemical production, and environmental remediation, as their performance directly impacts reaction efficiency, product selectivity, and process economics. Traditionally, catalyst development has relied heavily on trial and error and experimental screening, resulting in long development cycles, high costs, and low efficiency. While density functional theory (DFT) calculations and microkinetic models can provide some theoretical guidance by calculating adsorption energies and ultimately deriving activation and reaction energies, their high computational complexity makes them difficult to apply to high-throughput optimization of industrial-grade catalysts.

[0003] In recent years, machine learning methods have shown great potential in catalyst design, but many key problems remain. Existing models are mostly built for specific catalyst systems, lacking generalization ability and failing to adapt to complex variations in component combinations and reaction conditions, resulting in large cross-system prediction errors. Literature shows that machine learning without considering error correction mechanisms directly predicts activation energy with an error (RMSE > 0.35 eV, ≈ 33.78 kJ / mol), an accuracy insufficient for catalyst design. Even with neural network methods to improve prediction accuracy, they are typically only applicable to single catalyst systems, involve complex feature engineering, and cannot achieve multi-component synergistic optimization.

[0004] There is currently no effective solution to the problem of complex catalyst component design in the aforementioned existing technologies. Summary of the Invention

[0005] This invention provides a method, apparatus, electronic device, and computer program product for predicting catalyst components, to at least solve the technical problem of complex catalyst component design in the prior art.

[0006] According to one aspect of the present invention, a method for predicting catalyst components is provided, comprising: obtaining the activation energy requirement required by a catalyst for a target reactant to generate a target product through a catalytic reaction, wherein the catalyst comprises multiple catalytic elements, and different catalytic elements provide different activation energies for the target reactant; and using a pre-trained activation energy prediction model to determine the catalyst components whose catalytic activation energy provided by the catalyst satisfies the activation energy requirement, wherein the activation energy prediction model is used to represent the correlation between the catalyst components corresponding to the catalyst and the catalytic activation energy provided by the catalyst, and the catalyst components represent the proportions of multiple catalytic elements in the catalyst.

[0007] Optionally, using a pre-trained activation energy prediction model to determine the catalyst component whose catalytic activation energy meets the activation energy requirement includes: establishing an initial population, wherein each particle in the initial population is determined based on a random catalyst component; using a particle swarm optimization algorithm to iteratively update the particles in the initial population until a preset particle swarm stopping condition is reached; determining the particles from the last iteration of the particle swarm optimization algorithm as the initial differential population for a differential evolution algorithm; using the differential evolution algorithm to iterate the initial differential population until a preset differential stopping condition is reached to obtain the catalyst component that meets the activation energy requirement, wherein the differential stopping condition is that the negative correlation between selectivity and catalytic activation energy reaches a preset target value, the selectivity representing the ability of the target reactant to generate the target product in the catalytic reaction, and the catalytic activation energy corresponding to the unique catalyst component of the catalyst.

[0008] Optionally, obtaining the activation energy requirement of the catalyst required for the target reactant to generate the target product through a catalytic reaction includes: obtaining a pre-set target selectivity for the target reactant to generate the target product through a catalytic reaction; obtaining a selectivity-activation energy fitting curve of the catalyst, wherein the selectivity-activation energy fitting curve represents the mapping relationship between selectivity and activation energy; and determining the activation energy requirement mapped to the target selectivity based on the selectivity-activation energy fitting curve.

[0009] Optionally, before analyzing the activation energy requirement using a pre-trained activation energy prediction model to obtain the catalyst composition, the method further includes: obtaining a selectivity-activation energy fitting curve of the catalyst, wherein the selectivity-activation energy fitting curve represents the mapping relationship between selectivity and activation energy; and training the activation energy prediction model based on the selectivity-activation energy fitting curve.

[0010] Optionally, obtaining the selectivity-activation energy fitting curve of the catalyst includes: obtaining multiple catalyst samples of the catalyst, wherein different catalyst samples represent catalysts with different catalyst components; determining the sample activation energy provided by each catalyst sample; determining the sample selectivity corresponding to each sample activation energy based on the ability of the target reactant to generate the target product under the action of each sample activation energy; and fitting the sample activation energy and sample selectivity of multiple catalyst samples to obtain the selectivity-activation energy fitting curve.

[0011] Optionally, training the activation energy prediction model based on the selectivity-activation energy fitting curve includes: acquiring multiple sets of training data for training the activation energy prediction model, wherein each set of training data includes at least: a sample component for which the catalyst component is known; using the activation energy prediction model to predict the predicted selectivity and predicted activation energy of the target reactant producing the target product under the catalysis of the sample component; determining the expected activation energy mapped by the predicted selectivity in the selectivity-activation energy fitting curve; and optimizing the activation energy prediction model based on the difference between the predicted activation energy and the expected activation energy.

[0012] Optionally, determining the sample activation energy provided by each of the catalyst samples includes: obtaining descriptors for the target reactant and the target product, wherein the descriptor is an adsorbent in the target reactant or the target product with a pre-recorded standard adsorption energy, the adsorbent reacting on the surface of the catalyst, and the standard adsorption energy is the energy that the adsorbent will generate on the surface of a reference catalyst calculated based on density functional theory; calculating a first adsorption energy of each descriptor relative to the catalyst sample and a second adsorption energy of each non-descriptor relative to the catalyst sample based on the catalyst composition of the catalyst sample, wherein the non-descriptor is an adsorbent in the target reactant or the target product for which the standard adsorption energy has not been pre-recorded, and a scaling relationship exists between the adsorption energy of the descriptor and the adsorption energy of the non-descriptor; and determining the activation energy of the target reactant at the surface of the catalyst. The catalyst sample generates multiple elementary reactions that produce the target product under catalysis, and the elementary activation energy of each elementary reaction is defined. Each elementary reaction represents a chemical transformation relationship between species, including the descriptor, the non-descriptor, and the non-adsorbate. The non-adsorbate does not react on the surface of the catalyst. A BEP relationship exists between the elementary activation energy and the elementary reaction energy. The elementary reaction energy is the difference between the product energy and the reactant energy. The product energy and the reactant energy include the first adsorption energy, the second adsorption energy, and the non-adsorption energy. The non-adsorption energy is the internal energy of the non-adsorbate. Based on the elementary activation energy of the target elementary reaction, the sample activation energy is determined. The elementary activation energy of the target elementary reaction meets a preset activation energy condition, and the sample activation energy is a weighted result of the elementary activation energy of the target elementary reaction.

[0013] According to another aspect of the present invention, a catalyst component prediction device is also provided, comprising: an acquisition module, configured to acquire the activation energy requirement required by a catalyst when a target reactant generates a target product through a catalytic reaction, wherein the catalyst comprises: multiple catalytic elements, and different catalytic elements provide different activation energies for the target reactant; and a determination module, configured to use a pre-trained activation energy prediction model to determine the catalyst component whose catalytic activation energy provided by the catalyst satisfies the activation energy requirement, wherein the activation energy prediction model represents the correlation between the catalyst component corresponding to the catalyst and the catalytic activation energy provided by the catalyst, and the catalyst component represents the proportion of multiple catalytic elements in the catalyst.

[0014] According to another aspect of the present invention, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to execute the above-described method for predicting catalyst components through the computer program.

[0015] According to another aspect of the present invention, a computer program product is also provided, including computer instructions that, when executed by a processor, implement the steps of the above-described method for predicting catalyst components.

[0016] In the embodiments described above, the definition of catalyst components encompasses the proportions of multiple catalytic elements in the catalyst, reflecting the influence of different element combinations on catalytic activity. By obtaining the activation energy required for the conversion of the target reactant into the target product, and using a pre-trained activation energy prediction model to analyze the correlation between different element combinations of catalyst components and activation energy, the catalyst components that meet the activation energy requirements are accurately determined. This achieves intelligent optimization of the catalyst composition, improves the accuracy and efficiency of catalyst design, avoids extensive experimental trial and error, significantly reduces R&D costs and time consumption, and provides a powerful tool for the efficient design and rapid screening of industrial catalysts. It achieves the effect of reducing the complexity of catalyst component design and solves the technical problem of complex catalyst component design in the prior art. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0018] Figure 1 This is a flowchart of a method for predicting catalyst components according to an embodiment of the present invention;

[0019] Figure 2 This is a schematic diagram of a selectivity-activation energy prediction curve according to an embodiment of the present invention;

[0020] Figure 3 This is a schematic diagram illustrating the establishment of a multi-parameter correlation function between the content of catalyst components and the activation energy of the reaction according to an embodiment of the present invention;

[0021] Figure 4 This is a schematic diagram of the training results of an activation energy error prediction model according to an embodiment of the present invention;

[0022] Figure 5 A schematic diagram of a catalyst component prediction device according to an embodiment of the present invention;

[0023] Figure 6 This is a structural block diagram of a computer terminal according to an embodiment of the present invention. Detailed Implementation

[0024] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0025] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0026] First, some nouns or terms that appear in the description of the embodiments of this application shall be interpreted as follows:

[0027] Elementary reaction: A chemical reaction in which reactant molecules are directly converted into product molecules in a single reaction step. Its characteristic is that the reaction proceeds through only one transition state (TS) and does not involve any separable reaction intermediates.

[0028] Adsorption energy (Eads) refers to the change in energy of the entire system when one or more gaseous or liquid molecules (adsorbates) are adsorbed onto a solid surface (substrate). The formula for calculating adsorption energy Eads is: Eads = EAB - EA - EB, where EAB is the total energy of system AB after substance A is adsorbed onto substrate B, EA is the energy of substance A, and EB is the energy of substrate B.

[0029] Reaction energy (E): refers to the total energy difference between the products (final state) and reactants (initial state) in a single elementary reaction. The specific formula is E = EFS – EIS, where E is the reaction energy of this elementary reaction step, EIS is the energy of the initial reaction state, and EFS is the energy of the final reaction state.

[0030] Activation energy (Ea): also known as the activation barrier, refers to the energy difference between the transition state (TS) and the reactants (initial state, IS) in a one-step elementary reaction. It represents the minimum energy threshold that must be overcome for the reaction to occur.

[0031] Scaling relationship: refers to the approximately linear relationship between the adsorption energies of different reaction intermediates (especially intermediates with the same central atom and similar structures, representing active sites on the catalyst surface) on the surface of a series of different catalysts (e.g., transition metals of the same group or alloys of different compositions).

[0032] The BEP (Brønsted–Evans–Polanyi) relationship refers to the approximately linear relationship between the activation energy (Ea) and the reaction energy (ΔEr) for a class of elementary reactions with similar chemical properties or structures. It is usually expressed as: Ea = α ΔEr + β, where α and β are characteristic constants of this type of reaction.

[0033] In electrocatalysis research, a descriptor is a concise parameter used to correlate the intrinsic properties of a material with its catalytic performance, significantly accelerating catalyst screening and design. In this application, the descriptor refers to two selected adsorbates, whose adsorption energies are used to describe the adsorption energies of other adsorbates.

[0034] According to an embodiment of the present invention, a method for predicting catalyst components is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0035] Figure 1 This is a flowchart of a method for predicting catalyst components according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0036] Step S102: Obtain the activation energy required by the catalyst for the target reactant to generate the target product through a catalytic reaction. The catalyst includes multiple catalytic elements, and different catalytic elements provide different activation energies for the target reactant.

[0037] Step S104: Using a pre-trained activation energy prediction model, determine the catalyst components whose catalytic activation energy meets the activation energy requirements. The activation energy prediction model is used to represent the correlation between the catalyst components corresponding to the catalyst and the catalytic activation energy provided by the catalyst. The catalyst components represent the proportion of multiple catalytic elements in the catalyst.

[0038] In the embodiments described above, the definition of catalyst components encompasses the proportions of multiple catalytic elements in the catalyst, reflecting the influence of different element combinations on catalytic activity. By obtaining the activation energy required for the conversion of the target reactant into the target product, and using a pre-trained activation energy prediction model to analyze the correlation between different element combinations of catalyst components and activation energy, the catalyst components that meet the activation energy requirements are accurately determined. This achieves intelligent optimization of the catalyst composition, improves the accuracy and efficiency of catalyst design, avoids extensive experimental trial and error, significantly reduces R&D costs and time consumption, and provides a powerful tool for the efficient design and rapid screening of industrial catalysts. It achieves the effect of reducing the complexity of catalyst component design and solves the technical problem of complex catalyst component design in the prior art.

[0039] In step S102 above, the catalyst includes multiple catalytic elements, but the ratio between the multiple catalysts needs to be predicted and designed using a chemical energy prediction model.

[0040] In step S102 above, the target product can be methylchlorosilane (MCS), and the catalyst can be a Cu-based catalyst.

[0041] It should be noted that the catalytic activation energy is the catalytic energy that the catalyst can provide for the target reactant. Among them, the activation energy that the catalyst can provide when catalyzing the target reactant is different for catalysts with different catalyst components.

[0042] It should be noted that the catalytic reaction of the target reactant under the catalytic action of the catalyst involves multiple elementary reactions, which constitute the elementary reaction network in which the target reactant generates the target product.

[0043] It should be noted that different elementary reactions require different activation energies. That is, different activation energies can catalyze different elementary reactions and affect the reaction rate of various elementary reactions. Therefore, by adjusting the catalyst composition, different activation energies can be provided for the catalytic reaction, guiding various elementary reactions in the catalytic reaction, and thus adjusting the formation ratio or rate of the target product.

[0044] Alternatively, the process by which activation energy guides elementary reactions can be selectively represented.

[0045] In step S102 above, the activation energy requirement can be preset. If the catalyst components can meet the activation energy requirement, the target product can reach the preset ideal state.

[0046] In step S102 above, the activation energy prediction model can analyze the catalyst components that provide catalytic activation energy that can meet the activation energy requirements under ideal conditions, based on the guidance of the catalyst activation energy provided by the catalyst components on each elementary reaction in the elementary reaction network.

[0047] The method for predicting catalyst components described above in this application can be used to predict the reaction selectivity of methylchlorosilane (MCS) with Cu-based catalysts.

[0048] It should be noted that methylchlorosilanes, especially dimethyldichlorosilane (CH3)2SiCl2, are the "cornerstone" of organosilicon chemistry and the core monomers for the production of many organosilicon products such as silicone oils, silicone rubbers, and silicone resins. In the presence of a catalyst, silicon (Si) reacts with chloromethane (C... Cl) can be directly reacted in a gas-solid phase to yield dimethyldichlorosilane and byproducts such as monomethyltrichlorosilane (MeSiCl3) and trimethylchlorosilane (CH3)3SiCl. The catalyst used in this process is based on Cu, with the addition of elements such as Ca, Al, Sn, and Zn.

[0049] Table 1 is a schematic table of a dimethyl-based reaction network according to an embodiment of the present invention. As shown in Table 1, it represents multiple elementary reactions that generate dimethyldichlorosilane.

[0050] Table 1

[0051]

[0052] Based on the analysis of Table 1, the elementary reaction of dimethylchloromethane can be divided into three stages: CH3Cl dissociation, Si migration, and product formation.

[0053] Following the steps shown in Table 1, the adsorption energies of all adsorbates on a certain reference catalyst surface were calculated using DFT, and different descriptor-non-descriptor combinations were fitted to select the adsorbate with the best fit – CH3. - As a descriptor. The adsorption energies of all other species are compared with the descriptor C. Linear regression was performed on the adsorption energy to establish a scaling relationship. Once established, this scaling relationship can be applied to any other catalyst surface, by calculating C... The adsorption energy can be used to obtain the adsorption energy of all adsorbed species.

[0054] DFT calculations revealed a correlation between the energy of the descriptor and factors such as the composition of the catalyst promoter and the content of doped metals. Therefore, C... The adsorption energy is expressed as a function of the catalyst composition. In this case, 19 catalyst composition parameters were selected as variables.

[0055] C was obtained based on the catalyst composition. After determining the adsorption energy, the adsorption energies of other adsorbates can be calculated using the previously established scaling relationship. Combined with the gaseous molecule energies calculated by DFT (for non-adsorbates, whose energies are independent of the catalyst surface), the reaction energy of each elementary reaction can be calculated. Based on the energy barrier (i.e., the activation energy of the elementary reaction) on the surface of the benchmark catalyst calculated by DFT, the energy barrier is linearly fitted to the reaction energy to establish the BEP relationship. With the BEP relationship, it can be used to calculate the energy barrier of the elementary reaction on the surface of any other catalyst.

[0056] It should be noted that the elementary reaction in dimethylchloromethane that meets the preset activation energy conditions includes three stages, with energy barriers (i.e., the elementary activation energies of the target elementary reaction) of R13, R14, and R11, respectively. The apparent activation energy of the macroscopic reaction that produces dimethylchloromethane can be considered as the weighted sum of these three reaction energy barriers. Based on mechanistic analysis, the weighting coefficients of these three energy barriers are initially set to 0.2, 0.2, and 0.6. These weighting coefficients can be optimized in subsequent parameter tuning if necessary.

[0057] According to the above embodiments of this application, measured data of a methylchlorosilane production line in a factory were collected based on the steps shown in Table 1. The factory laboratory conducted chemical analyses on the catalyst catalyst itself and other elements in the silicon powder, and determined the selectivity of the main product, dimethylchlorosilane, obtained under different catalyst / silicon powder doping element compositions. A total of 234 sets of experimental data were collected. A hybrid optimization strategy combining an improved particle swarm optimization (PSO) algorithm and a differential evolution (DE) algorithm was adopted, aiming at the negative correlation between reaction selectivity and activation energy, to optimize the parameters A1~Ai in the catalyst composition-reaction activation energy correlation function model (i.e., the activation energy prediction model). The optimized parameters yielded kendall tau = -0.86, where A1~Ai represent the importance of each catalytic element in the catalyst in the catalytic reaction.

[0058] As an optional embodiment, the catalyst component that meets the activation energy requirement by using a pre-trained activation energy prediction model includes: establishing an initial population, wherein each particle in the initial population is determined based on a random catalyst component; iteratively updating the particles in the initial population using a particle swarm optimization algorithm until a preset particle swarm stopping condition is reached; determining the particles from the last iteration of the particle swarm optimization algorithm as the initial differential population for the differential evolution algorithm; iterating the initial differential population using the differential evolution algorithm until a preset differential stopping condition is reached to obtain the catalyst component that meets the activation energy requirement, wherein the differential stopping condition is that the negative correlation between selectivity and catalytic activation energy reaches a preset target value, selectivity represents the ability of the target reactant to generate the target product in the catalytic reaction, and the catalytic activation energy corresponds to a unique catalyst component of the catalyst.

[0059] In the embodiments described above, the activation energy prediction model employs a hybrid optimization strategy combining particle swarm optimization (PSO) and differential evolution (DE) algorithms. This strategy optimizes various catalyst components capable of generating the target product from the target reactants to identify those that increase the proportion of the target product. By combining the global search capability of the PSO algorithm with the local refinement capability of the DE algorithm, the optimization process for the catalyst components is both comprehensive and precise. This results in a catalyst that exhibits excellent catalytic activity and selectivity in practical applications. It not only effectively improves the efficiency of finding catalyst compositions that meet performance requirements but also ensures a high degree of matching between the obtained catalyst components and the activation energy requirements, thus providing a more precise guide for catalyst design.

[0060] It's important to note that reaction selectivity is a key concept in chemical reaction engineering. It refers to the ability of a chemical reaction to form a specific target product relative to other possible products. In complex chemical reaction systems, reactants may generate multiple products through different reaction pathways. Higher reaction selectivity means that, under given reaction conditions, more reactants are converted into the target product, while the proportion of byproducts or undesirable products is smaller. Improving reaction selectivity is crucial for reducing costs, increasing production efficiency, and minimizing waste in the chemical industry.

[0061] It's important to note that selectivity and activation energy exhibit a negative correlation. This means that in chemical reactions, pathways with higher activation energies tend to produce fewer target products, while pathways with lower activation energies tend to produce more. In other words, when the activation energy of a reaction pathway decreases, the reaction rate of that pathway increases, thereby increasing the proportion of the specific product formed, and the reaction selectivity increases accordingly. Conversely, if the activation energy of a pathway is high, the reaction rate of that pathway will be slower, the selectivity will decrease, and more reactants may be converted into byproducts or undesirable products.

[0062] Optionally, the negative correlation between selectivity and activation energy can be represented by the Kendall tau coefficient. Ideally, the stopping condition for the difference is to make the Kendall tau coefficient as close as possible to -1.

[0063] Optionally, the particle swarm optimization stopping condition is that the number of particle swarm iterations reaches the required number of iterations, or the convergence of the particle swarm optimization meets the convergence threshold.

[0064] Optionally, the process of iterating the initial differential population using the differential evolution algorithm includes mutation, crossover, and selection operations.

[0065] Optionally, the target value can be a set value that makes the negative correlation between selectivity and catalytic activation energy close to -1.

[0066] It should be noted that the activation energy requirement for generating the target product can be set directly, or it can be indirectly determined by setting an ideal target selectivity based on the requirements for the target product.

[0067] As an optional embodiment, obtaining the activation energy requirement of the catalyst required for the target reactant to generate the target product through a catalytic reaction includes: obtaining a pre-set target selectivity for the target reactant to generate the target product through a catalytic reaction; obtaining a selectivity-activation energy fitting curve of the catalyst, wherein the selectivity-activation energy fitting curve represents the mapping relationship between selectivity and activation energy; and determining the activation energy requirement mapped to the target selectivity based on the selectivity-activation energy fitting curve.

[0068] In the embodiments described above, the selectivity-activation energy fitting curve of the catalyst can clearly define the mapping relationship between selectivity and activation energy. Based on this selectivity-activation energy fitting curve, the specific activation energy requirement corresponding to the target selectivity can be accurately determined. The activation energy requirement determined by the selectivity-activation energy fitting curve can effectively guide the optimization of catalyst composition to meet the requirements of specific reaction activity and selectivity. This not only avoids the blindness and inefficiency of traditional trial and error methods, but also enables rapid positioning within the ideal activation energy range in a data-driven manner, thereby accelerating the design and development process of high-performance catalysts.

[0069] Optionally, by adjusting the parameters of the fitted curve or employing different machine learning algorithms, the solution can be adapted to a wider range of application scenarios and improve prediction accuracy. This flexibility enables the technical solution to cope with constantly changing reaction conditions and catalyst systems, providing strong support for the intelligent design of catalysts.

[0070] Figure 2 This is a schematic diagram of a selectivity-activation energy prediction curve according to an embodiment of the present invention, such as... Figure 2 As shown, based on the calculated activation energy values ​​after parameter optimization, a second-order polynomial fitting is performed on the selectivity-activation energy to obtain the selectivity-activation energy prediction curve.

[0071] As an optional embodiment, before analyzing the activation energy requirement using a pre-trained activation energy prediction model to obtain the catalyst composition, the method further includes: obtaining a selectivity-activation energy fitting curve of the catalyst, wherein the selectivity-activation energy fitting curve represents the mapping relationship between selectivity and activation energy; and training an activation energy prediction model based on the selectivity-activation energy fitting curve.

[0072] In the embodiments described above, the selectivity-activation energy fitting curve of the catalyst represents the ideal mapping relationship between selectivity and activation energy. Therefore, based on this selectivity-activation energy fitting curve, the ideal mapping relationship between selectivity and reaction energy in the catalytic reaction can be reflected under different catalyst components. Thus, using this selectivity-activation energy fitting curve as training data for the activation energy prediction model enables the trained activation energy prediction model to more accurately estimate the activation energy requirements under different catalyst compositions. Training the activation energy prediction model based on the selectivity-activation energy fitting curve not only enhances the predictive ability of the activation energy prediction model but also enables effective performance evaluation and composition optimization of catalysts across systems, significantly improving the accuracy and efficiency of catalyst design, reducing unnecessary experimental attempts, and possessing significant industrial application prospects.

[0073] In the above embodiments of this application, the selectivity-activation energy fitting curve can be established by using industrial or experimental data to establish the relationship between selectivity and activation energy, and then using this relationship to train an activation energy prediction model, so that the model can intelligently compensate for the gap between theoretical calculations and actual needs, and ultimately achieve the goal of reverse screening of the optimal catalyst combination from the target selectivity.

[0074] It should be noted that the selectivity-activation energy fitting curve may vary depending on the reaction type and conditions, but by adjusting the model training parameters and algorithm, it is still possible to effectively predict and optimize a specific catalyst system.

[0075] As an optional embodiment, obtaining the selectivity-activation energy fitting curve of the catalyst includes: obtaining multiple catalyst samples, wherein different catalyst samples represent catalysts with different catalyst components; determining the sample activation energy provided by each catalyst sample; determining the sample selectivity corresponding to each sample activation energy based on the ability of the target reactant to generate the target product under the action of each sample activation energy; and fitting the sample activation energy and sample selectivity of multiple catalyst samples to obtain the selectivity-activation energy fitting curve.

[0076] In the embodiments described above, a series of catalyst samples are collected, each representing a catalyst with different component proportions, thus covering a wide design space. Next, the activation energy of each catalyst sample during elementary reactions on its surface is determined, and the selectivity is calculated based on the target product generation capacity at each sample's activation energy. This selectivity is a key indicator for evaluating catalyst performance. The activation energies and selectivity values ​​of these samples are fitted and analyzed to construct a selectivity-activation energy relationship curve. The entire process is data-driven, utilizing advanced statistical analysis and machine learning techniques to accurately reflect the comprehensive influence of catalyst components on activation energy and selectivity, thereby providing a solid scientific basis for the reverse design of catalysts.

[0077] It should be noted that the sample activation energy is calculated based on the pre-established scaling and BEP relationships between the descriptor adsorption energy and the non-descriptor adsorption energy.

[0078] As an optional embodiment, training the activation energy prediction model based on the selectivity-activation energy fitting curve includes: acquiring multiple sets of training data for training the activation energy prediction model, wherein each set of training data includes at least: a sample component with known catalyst components; using the activation energy prediction model to predict the predicted selectivity and predicted activation energy of the target product produced by the target reactant under the catalysis of the sample component; determining the expected activation energy mapped by the predicted selectivity in the selectivity-activation energy fitting curve; and optimizing the activation energy prediction model based on the difference between the predicted activation energy and the expected activation energy.

[0079] The embodiments described above in this application, by collecting a large amount of measured data from industrial production lines or high-throughput experimental data from laboratories, fit the calculated activation energy value using a quadratic polynomial function to construct a selectivity-activation energy relationship curve. During the model training phase, the ideal desired activation energy can be determined based on this selectivity-activation energy relationship curve. Then, a deep learning model is used to analyze the sample components of the catalyst used as training data to obtain the predicted selectivity and predicted activation energy of the sample components. By comparing the desired activation energy with the predicted activation energy, a loss function for the activation energy prediction model can be established. Based on this loss function, the model parameters in the activation energy prediction model can be adjusted to optimize the model and improve its prediction accuracy. Thus, a complete catalyst performance evaluation and reverse design framework is established, enabling reverse derivation of the required activation energy based on the target selectivity, screening catalyst compositions that meet performance requirements, and significantly improving the efficiency and accuracy of catalyst design.

[0080] Optionally, by adjusting the parameters in the hybrid optimization strategy and the architecture of the deep learning model, it is possible to cope with changes in different catalyst systems and reaction conditions, thus achieving a wide range of applications.

[0081] As an optional embodiment, determining the sample activation energy provided by each catalyst sample includes: obtaining descriptors for the target reactant and target product, wherein the descriptor is an adsorbent in the target reactant or target product with a pre-recorded standard adsorption energy, the adsorbent reacts on the surface of the catalyst, and the standard adsorption energy is the energy that the adsorbent will generate on the surface of a reference catalyst, calculated based on density functional theory; calculating a first adsorption energy of each descriptor relative to the catalyst sample and a second adsorption energy of each non-descriptor relative to the catalyst sample, based on the catalyst composition of the catalyst sample, wherein the non-descriptors are adsorbents in the target reactant or target product with no pre-recorded standard adsorption energy, and a scaling relationship exists between the adsorption energies of the descriptors and the adsorption energies of the non-descriptors; determining the target The reaction involves multiple elementary reactions that generate the target product under the catalysis of a catalyst sample, and the elementary activation energy of each elementary reaction. Elementary reactions represent chemical transformation relationships between species, including descriptors, non-descriptors, and non-adsorbed substances. Non-adsorbed substances do not react on the catalyst surface. A BEP relationship exists between the elementary activation energy and the elementary reaction energy. The elementary reaction energy is the difference between the product energy and the reactant energy. The product energy and reactant energy include the first adsorption energy, the second adsorption energy, and the non-adsorption energy. The non-adsorption energy is the internal energy of the non-adsorbed substance. Based on the elementary activation energy of the target elementary reaction, the sample activation energy is determined. The elementary activation energy of the target elementary reaction meets the preset activation energy conditions, and the sample activation energy is the weighted result of the elementary activation energies of the target elementary reaction.

[0082] In the embodiments described above, different catalyst components can provide different sample activation energies for the same target reactant. The sample activation energy provided by each catalyst sample can be calculated based on the adsorption energy between the target reactant and the sample catalytic energy, as well as the non-adsorption energy of the target reactant. Thus, based on the characteristics of the catalyst components and reactants, the sample activation energy can be accurately determined. Furthermore, based on the activation energy prediction model trained from multiple catalyst samples and their corresponding sample activation energies, the efficiency and accuracy of catalyst performance optimization and intelligent design can be improved, providing strong support for high-throughput screening and industrial application of catalysts.

[0083] It should be noted that, based on density functional theory calculations, the adsorption energies of various adsorbates relative to the surface of a standard reference catalyst can be pre-calculated. By using descriptors, among the multiple descriptors calculated by density functional theory, descriptors suitable for the target reactant can be selected. Then, based on the scaling relationship, the adsorption energies of non-descriptors can be derived. Finally, the active energy can be calculated based on the adsorption energies of descriptors and non-descriptors. This can reduce the amount of data required to calculate the active energy and improve computational efficiency.

[0084] It should be noted that the scaling relationship represents an approximately linear relationship between the adsorption energies of the adsorbate on the surfaces of different catalysts.

[0085] It should be noted that the BEP relationship represents an approximately linear relationship between the activation energy (Ea) and the reaction energy (ΔEr), usually expressed as: Ea = α ΔEr + β, where α and β are characteristic constants of this type of reaction.

[0086] Optionally, the preset activation energy condition can indicate the elementary reactions for which the elementary activation energy is greater than a preset activation energy threshold, or the elementary reactions selected in descending order of their elementary activation energies (e.g., 3).

[0087] Figure 3 This is a schematic diagram illustrating a multi-parameter correlation function between the catalyst component content and the reaction activation energy according to an embodiment of the present invention, as shown below. Figure 3 As shown, the steps are as follows:

[0088] Step S301: Select suitable descriptors based on the elementary reaction network of the catalytic reaction system.

[0089] Step S302: Calculate the adsorption energy Gad of the descriptor based on the elemental composition of the catalyst.

[0090] Step S303: Calculate the adsorption energy of all non-descriptor adsorbates based on the scaling relationship between other adsorbates and descriptors.

[0091] Step S304: Calculate the energy (i.e., internal energy) of the non-adsorbed matter (gas phase molecules).

[0092] Step S305: Calculate the reaction energy of all elementary reactions based on the energy of all substances.

[0093] Step S306: Calculate the energy barrier (i.e., the activation energy of the elementary reaction) for all elementary reactions based on the BEP relationship between the reaction energy and the energy barrier.

[0094] Step S307: Calculate the activation energy based on the energy barrier of the rate-determining step of the elementary reaction and the energy barrier of the reaction that has a significant impact on the selectivity of the reaction.

[0095] Step S308: By collecting measured data from industrial production lines or high-throughput experimental data from laboratories, a hybrid optimization strategy combining an improved particle swarm optimization (PSO) algorithm and a differential evolution (DE) algorithm is adopted. With the negative correlation between reaction selectivity and activation energy as the objective, the parameters A1~Ai in the above correlation function model (i.e., activation energy prediction model) are accurately fitted.

[0096] Step S309: Fit the calculated selectivity-activation energy value of the reaction to a quadratic polynomial function to obtain the selectivity-activation energy fitting curve, and calculate the expected activation energy value (such as the expected activation energy) for the selectivity of a specific reaction based on the curve.

[0097] Step S310: Intelligently compensate for the error between the calculated and expected activation energy values ​​using a deep learning model, and predict the selectivity of the reaction products based on the corrected activation energy; that is, train the activation energy prediction model.

[0098] Step S311: Based on the corrected activation energy prediction model, the required activation energy is derived in reverse through the target reaction selectivity, and the catalyst composition that meets the performance requirements is screened out in reverse optimization.

[0099] In step S301 above, the number of descriptors is 1 or 2.

[0100] In step S302 above, the adsorption energy of the descriptor is a function of the elemental composition / ratio of the catalyst, and the catalyst composition / ratio is at least one type. The formula for calculating the adsorption energy of the descriptor is as follows: ,in, The elemental composition or proportion of the catalyst. Use its weighting coefficients.

[0101] In step S303 above, the adsorption energy of the adsorbate without descriptors is calculated by weighting the adsorption energies of the two descriptors. The formula for calculating the adsorption energy of the non-descriptor is as follows: ,in, and These are the adsorption energies of the descriptors. and These are the weighting coefficients of the adsorbate for the two descriptors, respectively.

[0102] In step S305 above, the reaction energy of the elementary reaction (i.e., the elementary reaction energy) is the sum of the energies of all products minus the sum of the energies of all reactants; the formula for calculating the reaction energy (i.e., the elementary reaction energy) is: ,in, The energy of the product, The energy of the reactants is given by the energy of the reactants; if the species is an adsorbate, its energy is the adsorption energy; if the species is a molecule (i.e., a non-adsorbate), its energy is the calculated internal energy.

[0103] In step S306 above, the energy barrier (i.e., the activation energy of the elementary reaction) is calculated from its reaction energy using a linear function. The formula for calculating the energy barrier (i.e., the activation energy of the elementary reaction) is as follows: ,in, The reaction energy of the elementary reaction is calculated in step S5.

[0104] In step S307 above, the catalytic activation energy is the apparent activation energy of the macroscopic reaction described by the elementary reaction network, which is calculated by weighting the energy barriers of the three elementary reactions with the highest energy barriers. The formula for calculating the catalytic activation energy is as follows: .

[0105] Steps S308 to S311 above provide machine learning predictions of catalyst reaction selectivity and catalyst design schemes.

[0106] In step S308 above, the correlation between the selectivity and activation energy of the reaction is represented by the Kendall tau coefficient, and the optimization target is -1, that is, the selectivity and catalytic activation energy are negatively correlated.

[0107] In step S310 above, the model training phase includes the following specific training process:

[0108] a. Divide the catalyst composition / selectivity dataset into training and testing sets in an 8:2 ratio;

[0109] b. Compare different machine learning algorithms to train models, and obtain the optimal machine model based on the coefficient of determination and the mean square error.

[0110] Alternatively, the methods for training the activation energy prediction model using machine learning algorithms include: gradient boosting, random forest, support vector regression, and additional tree regression.

[0111] In step S311 above, the catalyst reverse optimization design specifically includes:

[0112] a. Based on the selectivity index of the target product, the target activation energy is inversely calculated using a selectivity-activation energy relationship model (such as a selectivity-activation energy fitting curve);

[0113] b. Based on the corrected activation energy prediction model, an optimization algorithm is used to search for catalyst compositions that make the predicted activation energy close to the target activation energy.

[0114] c. Output catalyst formulations that meet performance requirements and their predicted activation energy values.

[0115] Figure 4 This is a schematic diagram of the training results of an activation energy error prediction model according to an embodiment of the present invention, as shown below. Figure 4 As shown, using the sklearn gradient boosting regression (GBR) algorithm, the training set (80%) and test set (20%) were split, and 5-fold cross-validation was used to optimize the hyperparameters, resulting in the following... Figure 4 The training results shown have R²=0.92 and RMSE=7.26kJ / mol (0.075ev) on the test set.

[0116] Optionally, the trained activation energy prediction model can be used to predict the selectivity of reaction products based on the corrected activation energy, thereby enabling efficient screening of catalyst composition.

[0117] The embodiments described above in this application significantly improve the accuracy of activation energy prediction by introducing an AI error correction mechanism. A deep learning model is used to intelligently compensate for the error between the theoretical calculation and the expected value, reducing the RMSE of activation energy prediction from 0.35 eV to 0.08 eV using traditional methods, greatly enhancing the reliability and practicality of the model. It possesses strong cross-system applicability and generalization ability. By combining data-driven modeling with a hybrid optimization algorithm (PSO+DE), it is applicable not only to various catalyst systems such as metal alloys and oxide supports, but also to multi-component synergistic optimization, with generalization error controlled within 5%, overcoming the limitations of existing models that are restricted to specific systems. It achieves reverse intelligent design from target selectivity to catalyst composition, significantly improving R&D efficiency. Activation energy is inferred from reaction selectivity, and the optimal catalyst formulation is quickly screened using optimization algorithms, shortening the R&D cycle by 80% and reducing the number of experimental verifications by 70%, significantly reducing development costs and time investment. It has strong industrial application value and solves the problems of low accuracy, poor generalization, and reliance on trial and error in existing catalyst design methods.

[0118] According to an embodiment of the present invention, a catalyst component prediction device embodiment is also provided. It should be noted that the catalyst component prediction device can be used to execute the catalyst component prediction method in the embodiments of the present invention, and the catalyst component prediction method in the embodiments of the present invention can be executed in the catalyst component prediction device.

[0119] Figure 5 A schematic diagram of a catalyst component prediction device according to an embodiment of the present invention, as shown below. Figure 5 As shown, the device may include: an acquisition module 52, used to acquire the activation energy required by the catalyst to generate the target product through a catalytic reaction of the target reactant, wherein the catalyst includes multiple catalytic elements, and different catalytic elements provide different activation energies for the target reactant; and a determination module 54, used to determine the catalyst components that meet the activation energy requirements by using a pre-trained activation energy prediction model, wherein the activation energy prediction model represents the correlation between the catalyst components corresponding to the catalyst and the catalytic activation energy provided by the catalyst, and the catalyst components represent the proportion of multiple catalytic elements in the catalyst.

[0120] It should be noted that the acquisition module 52 in this embodiment can be used to execute step S102 in the embodiments of this application, and the determination module 54 in this embodiment can be used to execute step S104 in the embodiments of this application. The examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the content disclosed in the above embodiments.

[0121] In the embodiments described above, the definition of catalyst components encompasses the proportions of multiple catalytic elements in the catalyst, reflecting the influence of different element combinations on catalytic activity. By obtaining the activation energy required for the conversion of the target reactant into the target product, and using a pre-trained activation energy prediction model to analyze the correlation between different element combinations of catalyst components and activation energy, the catalyst components that meet the activation energy requirements are accurately determined. This achieves intelligent optimization of the catalyst composition, improves the accuracy and efficiency of catalyst design, avoids extensive experimental trial and error, significantly reduces R&D costs and time consumption, and provides a powerful tool for the efficient design and rapid screening of industrial catalysts. It achieves the effect of reducing the complexity of catalyst component design and solves the technical problem of complex catalyst component design in the prior art.

[0122] As an optional embodiment, the determining module includes: a establishing unit for establishing an initial population, wherein each particle in the initial population is determined based on a random catalyst composition; a first iteration unit for iteratively updating the particles in the initial population using a particle swarm optimization algorithm until a preset particle swarm stopping condition is reached; a first determining unit for determining the particles obtained from the last iteration of the particle swarm optimization algorithm as the initial differential population for the differential evolution algorithm; and a second iteration unit for iterating the initial differential population using the differential evolution algorithm until a preset differential stopping condition is reached to obtain a catalyst composition that meets the activation energy requirement, wherein the differential stopping condition is that the negative correlation between selectivity and catalytic activation energy reaches a preset target value, selectivity represents the ability of the target reactant to generate the target product in the catalytic reaction, and catalytic activation energy corresponds to a unique catalyst composition.

[0123] As an optional embodiment, the acquisition module includes: a first acquisition unit, used to acquire a pre-set target selectivity when the target reactant generates the target product through a catalytic reaction; an acquisition submodule, used to acquire a selectivity-activation energy fitting curve of the catalyst, wherein the selectivity-activation energy fitting curve represents the mapping relationship between selectivity and activation energy; and a second determination unit, used to determine the activation energy requirement mapped to the target selectivity based on the selectivity-activation energy fitting curve.

[0124] As an optional embodiment, the apparatus further includes: an acquisition submodule, configured to acquire a selectivity-activation energy fitting curve of the catalyst before analyzing the activation energy requirement using a pre-trained activation energy prediction model to obtain the catalyst composition, wherein the selectivity-activation energy fitting curve represents the mapping relationship between selectivity and activation energy; and a training submodule, configured to train the activation energy prediction model based on the selectivity-activation energy fitting curve.

[0125] As an optional embodiment, the acquisition submodule includes: a first acquisition subunit for acquiring multiple catalyst samples, wherein different catalyst samples represent catalysts with different catalyst components; a first determination subunit for determining the sample activation energy provided by each catalyst sample; a second determination subunit for determining the sample selectivity corresponding to each sample activation energy based on the ability of the target reactant to generate the target product under the action of each sample activation energy; and a fitting subunit for fitting the sample activation energy and sample selectivity of multiple catalyst samples to obtain a selectivity-activation energy fitting curve.

[0126] As an optional embodiment, the training submodule includes: a second acquisition subunit for acquiring multiple sets of training data for training the activation energy prediction model, wherein each set of training data includes at least: a sample component with known catalyst components; a prediction subunit for using the activation energy prediction model to predict the predicted selectivity and predicted activation energy of the target product produced by the target reactant under the catalysis of the sample component; a third determination subunit for determining the expected activation energy of the predicted selectivity mapping in the selectivity-activation energy fitting curve; and an optimization subunit for optimizing the activation energy prediction model based on the difference between the predicted activation energy and the expected activation energy.

[0127] As an optional embodiment, the first determining subunit includes: a third acquiring subunit, used to acquire descriptors of the target reactant and the target product, wherein the descriptor is an adsorbent in the target reactant or target product with a pre-recorded standard adsorption energy, the adsorbent reacts on the surface of a catalyst, and the standard adsorption energy is the energy that the adsorbent will generate on the surface of a reference catalyst, calculated based on density functional theory; a calculation subunit, used to calculate, based on the catalyst composition of the catalyst sample, a first adsorption energy of each descriptor relative to the catalyst sample and a second adsorption energy of each non-descriptor relative to the catalyst sample, wherein the non-descriptors are adsorbents in the target reactant or target product with no pre-recorded standard adsorption energy, and there is a scaling relationship between the adsorption energies of the descriptors and the adsorption energies of the non-descriptors; and a fourth determining subunit, used for The process involves identifying multiple elementary reactions that generate the target product from the target reactants under the catalysis of a catalyst sample, and the elementary activation energy of each elementary reaction. Elementary reactions represent chemical transformation relationships between species, including descriptors, non-descriptors, and non-adsorbed substances. Non-adsorbed substances do not react on the catalyst surface. A BEP relationship exists between the elementary activation energy and the elementary reaction energy. The elementary reaction energy is the difference between the product energy and the reactant energy. The product energy and reactant energy include the first adsorption energy, the second adsorption energy, and the non-adsorption energy. The non-adsorption energy is the internal energy of the non-adsorbed substance. A fifth determining subunit is used to determine the sample activation energy based on the elementary activation energy of the target elementary reaction. The elementary activation energy of the target elementary reaction meets preset activation energy conditions, and the sample activation energy is a weighted result of the elementary activation energies of the target elementary reaction.

[0128] Embodiments of the present invention can provide an electronic device, which can be a computer terminal, and the computer terminal can be any one of a group of computer terminal devices. Optionally, in this embodiment, the computer terminal can also be replaced by a mobile terminal or other terminal device.

[0129] Optionally, in this embodiment, the computer terminal may be located in at least one of a plurality of network devices in a computer network.

[0130] In this embodiment, the computer terminal described above can execute the program code for the following steps in the method for predicting catalyst components: obtaining the activation energy requirement that the catalyst needs to provide when the target reactant generates the target product through a catalytic reaction, wherein the catalyst includes: multiple catalytic elements, and different catalytic elements provide different activation energies for the target reactant; using a pre-trained activation energy prediction model, determining the catalyst components whose catalytic activation energy provided by the catalyst meets the activation energy requirement, wherein the activation energy prediction model is used to represent the correlation between the catalyst components corresponding to the catalyst and the catalytic activation energy provided by the catalyst, and the catalyst components represent the proportion of multiple catalytic elements in the catalyst.

[0131] Figure 6 This is a structural block diagram of a computer terminal according to an embodiment of the present invention, such as... Figure 6 As shown, the computer terminal 60 may include one or more (only one is shown in the figure) processors 62 and memory 64.

[0132] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the catalyst component prediction method and apparatus in this embodiment of the invention. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby realizing the aforementioned catalyst component prediction method. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the terminal 60 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0133] The processor can access information and applications stored in the memory via a transmission device to perform the following steps: obtaining the activation energy required by the catalyst to generate the target product through a catalytic reaction of the target reactants, wherein the catalyst includes multiple catalytic elements, and different catalytic elements provide different activation energies for the target reactants; using a pre-trained activation energy prediction model, determining the catalyst components whose catalytic activation energy meets the activation energy requirement, wherein the activation energy prediction model is used to represent the correlation between the catalyst components corresponding to the catalyst and the catalytic activation energy provided by the catalyst, and the catalyst components represent the proportion of multiple catalytic elements in the catalyst.

[0134] Optionally, the processor may also execute program code for the following steps: establishing an initial population, wherein each particle in the initial population is determined based on a random catalyst composition; using a particle swarm optimization algorithm to iteratively update the particles in the initial population until a preset particle swarm stopping condition is reached; determining the particles from the last iteration of the particle swarm optimization algorithm as the initial differential population for the differential evolution algorithm; using the differential evolution algorithm to iterate the initial differential population until a preset differential stopping condition is reached to obtain a catalyst composition that meets the activation energy requirement, wherein the differential stopping condition is that the negative correlation between selectivity and catalytic activation energy reaches a preset target value, selectivity represents the ability of the target reactant to generate the target product in the catalytic reaction, and catalytic activation energy corresponds to a unique catalyst composition.

[0135] Optionally, the processor may also execute program code for the following steps: obtaining a pre-set target selectivity when the target reactant generates the target product through a catalytic reaction; obtaining a selectivity-activation energy fitting curve of the catalyst, wherein the selectivity-activation energy fitting curve represents the mapping relationship between selectivity and activation energy; and determining the activation energy requirement for the target selectivity mapping based on the selectivity-activation energy fitting curve.

[0136] Optionally, the processor may also execute program code for the following steps: obtaining the selectivity-activation energy fitting curve of the catalyst, wherein the selectivity-activation energy fitting curve represents the mapping relationship between selectivity and activation energy; and training an activation energy prediction model based on the selectivity-activation energy fitting curve.

[0137] Optionally, the processor may also execute program code for the following steps: acquiring multiple catalyst samples, wherein different catalyst samples represent catalysts with different catalyst components; determining the sample activation energy provided by each catalyst sample; determining the sample selectivity corresponding to each sample activation energy based on the ability of the target reactant to generate the target product under the action of each sample activation energy; and fitting the sample activation energy and sample selectivity of multiple catalyst samples to obtain a selectivity-activation energy fitting curve.

[0138] Optionally, the processor may also execute program code for the following steps: acquiring multiple sets of training data for training the activation energy prediction model, wherein each set of training data includes at least: a sample component with known catalyst composition; using the activation energy prediction model to predict the predicted selectivity and predicted activation energy of the target product produced by the target reactant under the catalysis of the sample component; determining the expected activation energy of the predicted selectivity mapping in the selectivity-activation energy fitting curve; and optimizing the activation energy prediction model based on the difference between the predicted activation energy and the expected activation energy.

[0139] Optionally, the processor may also execute program code for the following steps: acquiring descriptors of the target reactant and target product, wherein the descriptor is an adsorbent in the target reactant or target product with a pre-recorded standard adsorption energy, the adsorbent reacts on the surface of a catalyst, and the standard adsorption energy is the energy that the adsorbent will generate on the surface of a reference catalyst, calculated based on density functional theory; calculating, based on the catalyst composition of the catalyst sample, the first adsorption energy of each descriptor relative to the catalyst sample and the second adsorption energy of each non-descriptor relative to the catalyst sample, wherein the non-descriptors are adsorbents in the target reactant or target product with no pre-recorded standard adsorption energy, and a scaling relationship exists between the adsorption energies of the descriptors and the adsorption energies of the non-descriptors; and determining the target reactant. Multiple elementary reactions that generate the target product under the catalysis of the catalyst sample are described, along with the elementary activation energy of each elementary reaction. Elementary reactions represent chemical transformation relationships between species, including descriptors, non-descriptors, and non-adsorbed substances. Non-adsorbed substances do not react on the catalyst surface. A BEP relationship exists between the elementary activation energy and the elementary reaction energy. The elementary reaction energy is the difference between the product energy and the reactant energy. The product energy and reactant energy include the first adsorption energy, the second adsorption energy, and the non-adsorption energy, where the non-adsorption energy is the internal energy of the non-adsorbed substance. Based on the elementary activation energy of the target elementary reaction, the sample activation energy is determined. The elementary activation energy of the target elementary reaction meets the preset activation energy conditions, and the sample activation energy is the weighted result of the elementary activation energies of the target elementary reaction.

[0140] Those skilled in the art will understand that Figure 6 The structure shown is for illustrative purposes only. The computer terminal can also be a smartphone (such as an Android phone, an iOS phone, etc.), a tablet computer, a mobile internet device (MID), a PAD, and other terminal devices. Figure 6 This does not limit the structure of the aforementioned electronic device. For example, computer terminal 60 may also include components that are more... Figure 6 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 6 The different configurations shown.

[0141] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a computer program instructing the hardware related to the terminal device. The computer program can be stored in a non-volatile medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc.

[0142] Embodiments of the present invention also provide a non-volatile storage medium. Optionally, in this embodiment, the aforementioned non-volatile storage medium can be used to store the program code executed by the catalyst component prediction method provided in the above embodiments.

[0143] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0144] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: obtaining the activation energy requirement required by the catalyst for the target reactant to generate the target product through a catalytic reaction, wherein the catalyst includes: multiple catalytic elements, and different catalytic elements provide different activation energies for the target reactant; using a pre-trained activation energy prediction model, determining the catalyst component whose catalytic activation energy provided by the catalyst meets the activation energy requirement, wherein the activation energy prediction model is used to represent the correlation between the catalyst component corresponding to the catalyst and the catalytic activation energy provided by the catalyst, and the catalyst component represents the proportion of multiple catalytic elements in the catalyst.

[0145] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: establishing an initial population, wherein each particle in the initial population is determined based on a random catalyst composition; using a particle swarm optimization algorithm to iteratively update the particles in the initial population until a preset particle swarm stopping condition is reached; determining the particles obtained from the last iteration of the particle swarm optimization algorithm as the initial differential population for the differential evolution algorithm; using the differential evolution algorithm to iterate the initial differential population until a preset differential stopping condition is reached to obtain a catalyst composition that meets the activation energy requirement, wherein the differential stopping condition is that the negative correlation between selectivity and catalytic activation energy reaches a preset target value, selectivity represents the ability of the target reactant to generate the target product in the catalytic reaction, and catalytic activation energy corresponds to a unique catalyst composition.

[0146] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: obtaining a pre-set target selectivity when the target reactant generates the target product through a catalytic reaction; obtaining a selectivity-activation energy fitting curve of the catalyst, wherein the selectivity-activation energy fitting curve represents the mapping relationship between selectivity and activation energy; and determining the activation energy requirement for the target selectivity mapping based on the selectivity-activation energy fitting curve.

[0147] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: obtaining the selectivity-activation energy fitting curve of the catalyst, wherein the selectivity-activation energy fitting curve represents the mapping relationship between selectivity and activation energy; and training an activation energy prediction model based on the selectivity-activation energy fitting curve.

[0148] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: acquiring multiple catalyst samples, wherein different catalyst samples represent catalysts with different catalyst components; determining the sample activation energy provided by each catalyst sample; determining the sample selectivity corresponding to each sample activation energy based on the ability of the target reactant to generate the target product under the action of each sample activation energy; and fitting the sample activation energy and sample selectivity of multiple catalyst samples to obtain a selectivity-activation energy fitting curve.

[0149] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: acquiring multiple sets of training data for training the activation energy prediction model, wherein each set of training data includes at least: a sample component with known catalyst composition; using the activation energy prediction model to predict the predicted selectivity and predicted activation energy of the target product produced by the target reactant under the catalysis of the sample component; determining the expected activation energy of the predicted selectivity mapping in the selectivity-activation energy fitting curve; and optimizing the activation energy prediction model based on the difference between the predicted activation energy and the expected activation energy.

[0150] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: obtaining descriptors of the target reactant and the target product, wherein the descriptor is an adsorbent in the target reactant or the target product with a pre-recorded standard adsorption energy, the adsorbent reacts on the surface of the catalyst, and the standard adsorption energy is the energy that the adsorbent will have on the surface of the reference catalyst calculated based on density functional theory; calculating a first adsorption energy of each descriptor relative to the catalyst sample and a second adsorption energy of each non-descriptor relative to the catalyst sample based on the catalyst composition of the catalyst sample, wherein the non-descriptor is an adsorbent in the target reactant or the target product with an unrecorded standard adsorption energy, and there is a scaling relationship between the adsorption energy of the descriptor and the adsorption energy of the non-descriptor; Multiple elementary reactions that generate the target product from the target reactants under the catalysis of a catalyst sample are identified, along with the elementary activation energy of each elementary reaction. Elementary reactions represent chemical transformation relationships between species, including descriptors, non-descriptors, and non-adsorbed substances. Non-adsorbed substances do not react on the catalyst surface. A BEP relationship exists between the elementary activation energy and the elementary reaction energy. The elementary reaction energy is the difference between the product energy and the reactant energy. The product energy and reactant energy include the first adsorption energy, the second adsorption energy, and the non-adsorption energy, where the non-adsorption energy is the internal energy of the non-adsorbed substance. Based on the elementary activation energies of the target elementary reactions, the sample activation energy is determined. The elementary activation energy of the target elementary reaction meets the preset activation energy conditions, and the sample activation energy is the weighted result of the elementary activation energies of the target elementary reactions.

[0151] Embodiments of the present invention also provide a computer program product, including a computer program. Optionally, in this embodiment, when the computer program is executed by a processor, it implements the steps of the catalyst component prediction method provided in the above embodiments.

[0152] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0153] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0154] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0155] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0156] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0157] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a non-volatile storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a non-volatile storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned non-volatile storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0158] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for predicting catalyst components, characterized in that, include: When a target reactant is used to generate a target product through a catalytic reaction, the activation energy required by the catalyst is specified. The catalyst includes multiple catalytic elements, and different catalytic elements provide different activation energies for the target reactant. A pre-trained activation energy prediction model is used to determine the catalyst components that provide catalytic activation energy that meets the activation energy requirement. The activation energy prediction model is used to represent the correlation between the catalyst components corresponding to the catalyst and the catalytic activation energy provided by the catalyst. The catalyst components represent the proportion of multiple catalytic elements in the catalyst.

2. The method according to claim 1, characterized in that, Using a pre-trained activation energy prediction model, the catalyst components that provide catalytic activation energy that satisfies the activation energy requirement are determined to include: An initial population is established, wherein each particle in the initial population is determined based on a random catalyst component; The particle swarm optimization algorithm is used to iteratively update the particles in the initial population until a preset particle swarm stopping condition is reached. The particles generated in the last iteration of the particle swarm optimization algorithm are determined as the initial differential population for the differential evolution algorithm. The differential evolution algorithm is used to iterate the initial differential population until a preset differential stopping condition is reached, thereby obtaining the catalyst component that meets the activation energy requirement. The differential stopping condition is that the negative correlation between selectivity and catalytic activation energy reaches a preset target value. Selectivity represents the ability of the target reactant to generate the target product in the catalytic reaction. The catalytic activation energy corresponds to the unique catalyst component of the catalyst.

3. The method according to claim 1, characterized in that, The activation energy required by the catalyst to generate the target product from the target reactants through a catalytic reaction includes: To obtain the pre-defined target selectivity when the target reactant generates the target product through a catalytic reaction; Obtain the selectivity-activation energy fitting curve of the catalyst, wherein the selectivity-activation energy fitting curve represents the mapping relationship between selectivity and activation energy; Based on the selectivity-activation energy fitting curve, the activation energy requirement of the target selectivity mapping is determined.

4. The method according to claim 1, characterized in that, Before analyzing the activation energy requirement using a pre-trained activation energy prediction model to obtain the catalyst composition, the method further includes: Obtain the selectivity-activation energy fitting curve of the catalyst, wherein the selectivity-activation energy fitting curve represents the mapping relationship between selectivity and activation energy; The activation energy prediction model is trained based on the selectivity-activation energy fitting curve.

5. The method according to claim 3 or 4, characterized in that, Obtaining the selectivity-activation energy fitting curve of the catalyst includes: Multiple catalyst samples of the catalyst are obtained, wherein different catalyst samples represent catalysts with different catalyst compositions; Determine the sample activation energy provided by each of the catalyst samples; Based on the ability of the target reactant to generate the target product under the action of each sample activation energy, the sample selectivity corresponding to each sample activation energy is determined; The activation energy and selectivity of multiple catalyst samples are fitted to obtain the selectivity-activation energy fitting curve.

6. The method according to claim 5, characterized in that, Training the activation energy prediction model based on the selectivity-activation energy fitting curve includes: Acquire multiple sets of training data to train the activation energy prediction model, wherein each set of training data includes at least: sample components for which the catalyst components are known; Using the activation energy prediction model, the predicted selectivity and predicted activation energy of the target reactant producing the target product under the catalysis of the sample component are predicted. In the selectivity-activation energy fitting curve, determine the expected activation energy of the predicted selectivity mapping; The activation energy prediction model is optimized based on the difference between the predicted activation energy and the expected activation energy.

7. The method according to claim 5, characterized in that, Determining the sample activation energy provided by each of the catalyst samples includes: Obtain descriptors for the target reactant and the target product, wherein the descriptor is an adsorbent with a pre-recorded standard adsorption energy in the target reactant or the target product, the adsorbent reacts on the surface of the catalyst, and the standard adsorption energy is the energy that the adsorbent will generate on the surface of a reference catalyst, calculated based on density functional theory. Based on the catalyst components of the catalyst sample, a first adsorption energy of each descriptor relative to the catalyst sample and a second adsorption energy of each non-descriptor relative to the catalyst sample are calculated, wherein the non-descriptor is the adsorbate in the target reactant or the target product whose standard adsorption energy has not been recorded in advance, and there is a scaling relationship between the adsorption energy of the descriptor and the adsorption energy of the non-descriptor. Multiple elementary reactions in which the target reactant generates the target product under the catalysis of the catalyst sample are determined, and the elementary activation energy of each elementary reaction is defined. The elementary reaction represents a chemical transformation relationship between species, which includes: the descriptor, the non-descriptor, and the non-adsorbent. The non-adsorbent does not react on the surface of the catalyst. A BEP relationship exists between the elementary activation energy and the elementary reaction energy. The elementary reaction energy is the difference between the product energy and the reactant energy. The product energy and the reactant energy include: the first adsorption energy, the second adsorption energy, and the non-adsorption energy. The non-adsorption energy is the internal energy of the non-adsorbent. The sample activation energy is determined based on the elementary activation energy of the target elementary reaction, wherein the elementary activation energy of the target elementary reaction meets a preset activation energy condition, and the sample activation energy is a weighted result of the elementary activation energy of the target elementary reaction.

8. A device for predicting catalyst components, characterized in that, include: The acquisition module is used to acquire the activation energy required by the catalyst for the target reactant to generate the target product through a catalytic reaction. The catalyst includes multiple catalytic elements, and different catalytic elements provide different activation energies for the target reactant. The determination module is used to determine the catalyst components that provide catalytic activation energy that meets the activation energy requirement by using a pre-trained activation energy prediction model. The activation energy prediction model is used to represent the correlation between the catalyst components corresponding to the catalyst and the catalytic activation energy provided by the catalyst. The catalyst components represent the proportion of multiple catalytic elements in the catalyst.

9. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to execute the method for predicting the catalyst components according to any one of claims 1 to 7 through the computer program.

10. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the steps of the method for predicting catalyst components according to any one of claims 1 to 7.

Images