Method for selecting preparation parameters of iron-carbon-based catalyst

Abstract

The invention discloses an iron-carbon-based catalyst preparation parameter selection method. The method comprises the following steps: acquiring a plurality of groups of iron-carbon-based catalyst preparation parameters, and performing pretreatment to obtain a plurality of groups of pretreated iron-carbon-based catalyst preparation parameters; constructing a multi-task learning model; training the multi-task learning model by using the plurality of groups of preprocessed iron-carbon-based catalyst preparation parameters to obtain a mapping relationship between each group of iron-carbon-based catalyst preparation parameters and a catalyst performance value; and iteratively optimizing each group of iron-carbon-based catalyst preparation parameters by utilizing a genetic algorithm according to the mapping relationship between each group of iron-carbon-based catalyst preparation parameters and the catalyst performance value to obtain an optimal iron-carbon-based catalyst preparation parameter scheme. The method is high in parameter selection efficiency, and the prepared catalyst is good in performance.

Description

Technical Field

[0001] The present application relates to the technical field of catalyst preparation, and more specifically, to a method for selecting parameters for preparing an iron-carbon-based catalyst. Background Art

[0002] In today's chemical engineering and materials science fields, iron-carbon-based catalysts, with their unique properties, play a crucial role in many key chemical reactions, such as carbon dioxide hydrogenation and organic pollutant degradation, and have become a research hotspot. However, traditional methods for preparing iron-carbon-based catalysts, which rely heavily on extensive trial-and-error experiments and empirical research, have significant drawbacks.

[0003] On the one hand, the traditional method of selecting parameters for catalyst preparation has a long experimental cycle and is costly. Researchers need to spend a lot of time and resources to repeatedly change the preparation process parameters, such as the type and ratio of iron salts and carbon sources, calcination temperature, time, reaction atmosphere and other conditions, and repeatedly test the catalyst performance, which is extremely inefficient. On the other hand, it is difficult to accurately control the performance of catalysts using traditional methods. Due to the lack of a deep understanding of the complex relationship between preparation conditions and catalytic performance, the prepared catalysts vary in activity, selectivity and stability, making it difficult to meet actual industrial needs. Even if some catalysts show certain activity in the laboratory, during the industrial scale-up production process, the performance is greatly reduced due to the difficulty in accurately reproducing the preparation conditions, making it impossible to achieve industrial application.

[0004] The prior art discloses a high-dimensional data-driven, artificial intelligence-based next-generation catalyst design method, including: a multi-dimensional data fusion module: using new data processing technology to integrate various experimental and computational data. An adaptive neural network architecture: dynamically adjusting the network structure according to the performance requirements of the catalyst to optimize the selectivity and activity of the catalyst. A knowledge-driven feature engineering module: combining chemical knowledge graphs to automatically extract the features that have the greatest impact on catalytic performance. An iterative feedback optimization module: using online learning mechanisms to adjust parameters, digital twin technology to simulate behavior, and machine learning prediction models and sensor data fusion to monitor status. An interactive visualization and decision support module: providing advanced visualization tools and AI decision support to help R&D personnel understand data and make scientific decisions. It focuses on the structural optimization of general catalysts and does not design special models for the characteristics of iron-carbon systems. Therefore, its performance on iron-carbon-based catalysts is poor. Summary of the Invention

[0005] The present invention addresses the defects of the prior art in low parameter selection efficiency and poor performance of the prepared catalyst, and provides a method for selecting parameters for preparing an iron-carbon-based catalyst. The method has high parameter selection efficiency and produces a catalyst with good performance.

[0006] The primary purpose of the present invention is to solve the above technical problems, and the technical solutions of the present invention are as follows: A method for selecting parameters for preparing an iron-carbon-based catalyst, comprising: S1: Acquire multiple sets of iron-carbon-based catalyst preparation parameters and perform pretreatment to obtain multiple sets of pretreated iron-carbon-based catalyst preparation parameters; S2: constructing a multi-task learning model; using the multiple groups of pre-processed iron-carbon-based catalyst preparation parameters to train the multi-task learning model, to obtain a mapping relationship between each group of iron-carbon-based catalyst preparation parameters and catalyst performance values; S3: Based on the mapping relationship between each set of iron-carbon-based catalyst preparation parameters and the catalyst performance value, a genetic algorithm is used to iteratively optimize each set of iron-carbon-based catalyst preparation parameters to obtain an optimal iron-carbon-based catalyst preparation parameter solution.

[0007] Furthermore, the preparation parameters of the iron-carbon based catalyst include: process parameters, material structure images, reaction kinetics curves; Catalyst performance values ​​include: conversion rate prediction value and active site identification value.

[0008] Furthermore, in step S1, the preprocessing includes: S101: Eliminate abnormal values ​​in multiple sets of iron-carbon-based catalyst preparation parameters; supplement missing values ​​in multiple sets of iron-carbon-based catalyst preparation parameters; S102: Mapping each set of iron-carbon-based catalyst preparation parameters to a standard normal distribution interval to obtain multiple sets of pre-processed iron-carbon-based catalyst preparation parameters.

[0009] Furthermore, in step S2, the multi-task learning model includes: a multimodal input layer, a feature fusion layer, and a multi-task output layer connected in sequence.

[0010] Furthermore, the formula of the multimodal input layer is as follows:

[0011] represents the first weight matrix, represents the process parameter vector, represents the first bias term, represents the first output of the multimodal input layer;

[0012] Represents a material image, represents the second output of the multimodal input layer;

[0013] represents the reaction kinetics curve at time t, Represents the third output of the multimodal input layer at time t-1.

[0014] Furthermore, the formula of the feature fusion layer is as follows:

[0015] represents the output of the feature fusion layer, represents the first weight, represents the second weight, represents the third weight; represents the first output of the multimodal input layer; represents the second output of the multimodal input layer; represents the third output of the multimodal input layer;

[0016] represents the joint eigenvector, represents the second weight matrix, represents the second bias term; Represents the weight of the feature fusion layer;

[0017] represents the weight of the feature fusion layer, represents the first weight, represents the second weight, represents the third weight;

[0018] represents the joint eigenvector, represents the first output of the multimodal input layer; represents the second output of the multimodal input layer; Represents the third output of the multimodal input layer.

[0019] Furthermore, the formula of the multi-task output layer is as follows:

[0020] represents the third weight matrix, represents the third bias term, represents the output of the feature fusion layer, Represents the predicted value of conversion rate;

[0021] represents the fourth weight matrix, represents the output of the feature fusion layer, represents the fourth bias term, Indicates the active site identification value.

[0022] Furthermore, in step S3, a genetic algorithm is used to iteratively optimize each set of iron-carbon-based catalyst preparation parameters to obtain an optimal iron-carbon-based catalyst preparation parameter scheme, including: S301: Generate N chromosomes as a first chromosome group according to the preparation parameters of the iron-carbon based catalyst; S302: Calculate the fitness of each chromosome in the first chromosome group to obtain a first fitness of each chromosome; S303: Dividing the first chromosome group into a second chromosome group and a third chromosome group according to the first fitness; S304: performing crossover mutation on the chromosomes in the third chromosome set to obtain a fourth chromosome set; S305: merging the fourth chromosome set with the second chromosome set to form a new first chromosome set; S306: Determine whether the first termination condition is met based on the first chromosome set; if not, execute step S302; if so, execute step S307; S307: Selecting an optimal iron-carbon based catalyst preparation parameter scheme in the first chromosome group according to the Pareto optimal algorithm.

[0023] Furthermore, in step S302, the calculation formula of the fitness is as follows:

[0024] 、 、 represents the Pareto frontier optimization weight coefficient, Represents the predicted value of conversion rate, represents the active point identification value, and E represents energy consumption.

[0025] A system for selecting parameters for preparing an iron-carbon-based catalyst, comprising: Preparation parameter acquisition module: obtains multiple sets of iron-carbon based catalyst preparation parameters and performs pretreatment to obtain multiple sets of pretreated iron-carbon based catalyst preparation parameters; Mapping relationship inference module: constructing a multi-task learning model; using the multiple sets of pre-processed iron-carbon-based catalyst preparation parameters to train the multi-task learning model, and obtaining a mapping relationship between each set of iron-carbon-based catalyst preparation parameters and catalyst performance values; Scheme generation module: Based on the mapping relationship between each set of iron-carbon-based catalyst preparation parameters and catalyst performance values, each set of iron-carbon-based catalyst preparation parameters is iteratively optimized using a genetic algorithm to obtain an optimal iron-carbon-based catalyst preparation parameter scheme.

[0026] Compared with the prior art, the present invention has the following beneficial effects: The present invention utilizes the multiple groups of pretreated iron-carbon-based catalyst preparation parameters to train the multi-task learning model, and obtains a mapping relationship between each group of iron-carbon-based catalyst preparation parameters and catalyst performance values, thereby deeply exploring the intrinsic connection between preparation process parameters, material properties and catalytic performance, and realizing the precise design and efficient preparation of iron-carbon-based catalysts.

[0027] At the same time, according to the mapping relationship between each set of iron-carbon-based catalyst preparation parameters and catalyst performance values, a genetic algorithm is used to iteratively optimize each set of iron-carbon-based catalyst preparation parameters to obtain the optimal iron-carbon-based catalyst preparation parameter scheme, thereby improving the efficiency of preparation parameter optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a flow chart of a method for selecting parameters for preparing an iron-carbon-based catalyst provided in Example 1.

[0029] Figure 2 This is a flowchart of the preprocessing provided in Example 1.

[0030] Figure 3 This is a flow chart for iteratively optimizing the preparation parameters of each group of iron-carbon-based catalysts using a genetic algorithm provided in Example 1.

[0031] Figure 4 This is a test result diagram provided for Example 1.

[0032] Figure 5 This is a test result diagram provided for Example 1.

[0033] Figure 6 This is a test result diagram provided for Example 1.

[0034] Figure 7 This is a test result diagram provided for Example 1. DETAILED DESCRIPTION

[0035] The accompanying drawings are for illustrative purposes only and are not to be construed as limiting this patent; In order to better illustrate this embodiment, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product size; It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0036] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0037] Example 1 like Figure 1 As shown, a method for selecting parameters for preparing an iron-carbon based catalyst comprises: S1: Acquire multiple sets of iron-carbon-based catalyst preparation parameters and perform pretreatment to obtain multiple sets of pretreated iron-carbon-based catalyst preparation parameters; S2: constructing a multi-task learning model; using the multiple groups of pre-processed iron-carbon-based catalyst preparation parameters to train the multi-task learning model, to obtain a mapping relationship between each group of iron-carbon-based catalyst preparation parameters and catalyst performance values; S3: Based on the mapping relationship between each set of iron-carbon-based catalyst preparation parameters and the catalyst performance value, a genetic algorithm is used to iteratively optimize each set of iron-carbon-based catalyst preparation parameters to obtain an optimal iron-carbon-based catalyst preparation parameter solution.

[0038] In a specific embodiment, obtaining multiple sets of iron-carbon-based catalyst preparation parameters includes: Step 11: Through the integrated temperature sensor (K-type thermocouple), gas flow controller (MFC) and reactor pressure sensor, process parameters such as iron salt type (FeCl3, Fe(NO3)3), carbon-iron molar ratio (0.5-5.0), calcination temperature (300-900℃) are collected in real time. The 3σ principle is used to eliminate abnormal data. The specific formula is: , where μ is the data mean and σ is the standard deviation. For the missing carbon-iron ratio data, the K nearest neighbor algorithm (K=5) is used for interpolation, and the distance metric is the Euclidean distance. , ensuring data integrity and reliability.

[0039] Step 12: The process parameters are mapped to the [0,1] interval through linear normalization, and the formula is Catalytic performance indicators (such as CO2 conversion rate and product selectivity) are standardized using Z-score TEM image data is grayscale normalized to compress pixel values ​​to [0,1], and XRD data is used to extract diffraction peak positions ( ), full width at half maximum (FWHM) and intensity integral characteristics to construct the material property matrix .

[0040] Step 13: Process parameters, material properties, and performance indicators are integrated to generate a high-dimensional dataset with associated timestamps. The standardized data is stored in HDF5 format. The data structure includes process parameters (numeric), material images (binary matrices), and reaction kinetics curves (time series). Label association enables rapid retrieval and access of multimodal data, providing structured input for subsequent modeling.

[0041] Furthermore, if Figure 2 As shown, the preparation parameters of the iron-carbon based catalyst include: process parameters, material structure images, reaction kinetics curves; Catalyst performance values ​​include: conversion rate prediction value and active site identification value.

[0042] Process parameters include: iron salt type ( , etc.), carbon source ratio (0.5-5.0), calcination temperature (range 300-900°C), reaction time (1-24 hours) and atmosphere conditions (inert gas or reducing gas).

[0043] In a specific embodiment, X-ray diffraction (XRD) and transmission electron microscopy (TEM) are used to obtain structural images of the crystal structure and morphological characteristics (peak area ratio, nanoparticle size distribution, micropore volume) of the iron-carbon composite material.

[0044] In a specific embodiment, during the training process, Gaussian noise is added to the input data ( ), generate adversarial samples to improve robustness. Use Bayesian optimization to determine the learning rate (range to ), batch size (16-128), number of LSTM hidden units (32-256). Use the Adam optimizer with early stopping (terminating when validation loss does not decrease for 50 consecutive epochs).

[0045] It should be noted that the catalyst performance values ​​need to be processed by the Z-score standardization method.

[0046] Furthermore, in step S1, the preprocessing includes: S101: Eliminate abnormal values ​​in multiple sets of iron-carbon-based catalyst preparation parameters; supplement missing values ​​in multiple sets of iron-carbon-based catalyst preparation parameters; S102: Mapping each set of iron-carbon-based catalyst preparation parameters to a standard normal distribution interval to obtain multiple sets of pre-processed iron-carbon-based catalyst preparation parameters.

[0047] Furthermore, in step S2, the multi-task learning model includes: a multimodal input layer, a feature fusion layer, and a multi-task output layer connected in sequence.

[0048] It should be noted that this model can accurately predict catalyst performance and provide direction for the selection of preparation parameters.

[0049] Furthermore, the formula of the multimodal input layer is as follows:

[0050] represents the first weight matrix, represents the process parameter vector, represents the first bias term, represents the first output of the multimodal input layer;

[0051] Represents a material image, represents the second output of the multimodal input layer;

[0052] represents the reaction kinetics curve at time t, Represents the third output of the multimodal input layer at time t-1.

[0053] Furthermore, the formula of the feature fusion layer is as follows:

[0054] represents the output of the feature fusion layer, represents the first weight, represents the second weight, represents the third weight; represents the first output of the multimodal input layer; represents the second output of the multimodal input layer; represents the third output of the multimodal input layer; In a specific embodiment, the weight distribution is a first weight of 0.45, a second weight of 0.35, and a third weight of 0.2.

[0055]

[0056] represents the joint eigenvector, represents the second weight matrix, represents the second bias term; Represents the weight of the feature fusion layer;

[0057] represents the weight of the feature fusion layer, represents the first weight, represents the second weight, represents the third weight;

[0058] represents the joint eigenvector, represents the first output of the multimodal input layer; represents the second output of the multimodal input layer; Represents the third output of the multimodal input layer.

[0059] Furthermore, the formula of the multi-task output layer is as follows:

[0060] represents the third weight matrix, represents the third bias term, represents the output of the feature fusion layer, Represents the predicted value of conversion rate;

[0061] represents the fourth weight matrix, represents the output of the feature fusion layer, represents the fourth bias term, Indicates the active site identification value.

[0062] In a specific embodiment, during the training process, the loss function is:

[0063] is a hyperparameter (default 0.7).

[0064] The mean square error loss is used for conversion rate prediction:

[0065] Cross entropy loss is used for the active point identification value:

[0066] Furthermore, if Figure 3 As shown, in step S3, a genetic algorithm is used to iteratively optimize each set of iron-carbon-based catalyst preparation parameters to obtain an optimal iron-carbon-based catalyst preparation parameter scheme, including: S301: Generate N chromosomes as a first chromosome group according to the preparation parameters of the iron-carbon based catalyst; S302: Calculate the fitness of each chromosome in the first chromosome group to obtain a first fitness of each chromosome; S303: Dividing the first chromosome group into a second chromosome group and a third chromosome group according to the first fitness; S304: performing crossover mutation on the chromosomes in the third chromosome set to obtain a fourth chromosome set; S305: merging the fourth chromosome set with the second chromosome set to form a new first chromosome set; S306: Determine whether the first termination condition is met based on the first chromosome set; if not, execute step S302; if so, execute step S307; S307: Selecting an optimal iron-carbon based catalyst preparation parameter scheme in the first chromosome group according to the Pareto optimal algorithm.

[0067] In a specific embodiment, the first termination condition is to terminate the optimization when the fitness improvement is less than 1% for 10 consecutive generations or when the preset maximum number of iterations (200 times) is reached.

[0068] In a specific embodiment, N is 100 chromosomes. The parameters for preparing the iron-carbon-based catalyst with T>800°C need to be excluded. The number of the first chromosome group is always N.

[0069] In a specific embodiment, in step S303, the second chromosome group and the third chromosome group are divided using the tournament algorithm (Tournament Size=5); the chromosomes with the highest fitness of 20% are divided into the second chromosome group.

[0070] In a specific embodiment, in step S304, the crossover operation is to perform a single-point crossover on the third chromosome group (with a probability of ), randomly select crossover points for exchange. The mutation operation is to randomly flip binary bits according to the adaptive mutation rate pm (initial 0.1, maximum 0.4): Among them diversity = , HD is the Hamming distance.

[0071] In a specific embodiment, when generating chromosomes, the encoding relationship between specific parameters and chromosomes is as follows: calcination temperature (300-900°C, accuracy ±5°C), carbon-iron molar ratio (0.5-5.0, accuracy ±0.01), and reaction pressure (0.1-5.0MPa, accuracy ±0.1MPa) are encoded as binary chromosomes. Using binary encoding, each parameter is assigned a binary string with a fixed number of bits: chromosome = [ ]; Calcination temperature (10 bits): Map 300-900°C to integers from 0 to 1023 (step size 0.59°C); Carbon-iron ratio (8 bits): Map 0.5-5.0 to integers from 0 to 255 (step size 0.0176); Pressure (8 bits): Map 0.1-5.0 MPa to integers from 0 to 255 (step size 0.0192 MPa). The corresponding correspondence is as follows: Calcination temperature:

[0072] Carbon to iron molar ratio:

[0073] Reaction pressure:

[0074] In a specific embodiment, the above parameters are encoded as follows: Calcination temperature: 10-bit binary (step size 0.59°C), example: 680°C → 1011011100 Carbon-to-iron ratio: 8-bit binary (resolution 0.0176), example: 3.2 → 11001010 Reaction pressure: 8-bit binary (step size 0.0192 MPa), example: 2.1 MPa→10010111 Furthermore, in step S302, the calculation formula of the fitness is as follows:

[0075] 、 、 represents the Pareto frontier optimization weight coefficient, Represents the predicted value of conversion rate, represents the active point identification value, and E represents energy consumption.

[0076] In a specific embodiment, the NSGA-II algorithm is used to solve the Pareto front optimization weight coefficient 、 、 This method can standardize the data and filter outliers, thus enhancing the model's anti-interference ability.

[0077] The energy consumption is calculated as follows:

[0078] t represents the reaction time (hours), =900, =24, k=0.1 is the pressure weight coefficient.

[0079] In a specific embodiment, a penalty term is imposed on individuals whose calcination temperature exceeds the limit: (T-800, 0).

[0080] In a specific embodiment, after the parameters are generated, the specific parameters can be put into production for testing, and the model can be updated based on the test results. The specific steps are as follows: Step 41: Perform the sol-gel method according to the optimized parameters: accurately weigh the iron salt and carbon source (accuracy ±0.01 mol), adjust the hydrochloric acid flow rate by a PID controller to maintain the pH = 3.5 ± 0.1, place the dried product in a tube furnace and heat it at 5°C / min to the target temperature for 3 hours. The atmosphere conditions are precisely controlled by a mass flow controller (MFC).

[0081] Step 42: Mass spectrometer monitors CO2 conversion rate in real time X(t) = (1- ) In-situ XRD data were collected every 30 minutes, and Fe was calculated by Rietveld refinement. 0 Phase ratio Fe 0%= I is the diffraction peak intensity of the corresponding crystal plane Step 43: The fixed bed reactor runs continuously for 100 hours and the activity decay rate k is calculated. deact = - Thermogravimetric analysis of carbon deposits Coke wt% = The experimental data was processed by the isolation forest algorithm Anomaly Score = Filter outliers and update the model using elastic weight consolidation (EWC) = + Among them, F i For parameters The Fisher information matrix of =1000 is the penalty coefficient.

[0082] In a specific embodiment, corresponding charts and reports can be generated according to the actual processing process. The specific steps are as follows: Step 51: Construct a three-dimensional coordinate system using calcination temperature (X-axis), carbon-iron ratio (Y-axis), and pressure (Z-axis). The heat map color mapping formula is Color (T, R, P) = , users can interactively rotate the view and click to query the predicted performance (conversion rate, selectivity, energy consumption) of any parameter point.

[0083] Step 52: The directed graph nodes represent the best individuals in each generation. The attributes include process parameters (T, R, P) and fitness F. The edge weight is the improvement in intergenerational fitness. F = F t+1 - F t The mutation operation marks the flipping site in red (such as the 5th position of the temperature code), and the crossover operation marks the gene segment exchange range in blue.

[0084] Step 53: Values ​​by SHAP = Quantify the parameter contribution and identify the inflection point solution in the Pareto frontier ( A natural language report is generated based on the GPT-3 framework. Example output: "Under the current optimal parameters (T=725°C, R=2.3, P=2.1 MPa), the predicted conversion rate is 89%, but the energy consumption is relatively high (0.35). It is recommended to try parameters around T=700°C, which is expected to reduce energy consumption by 15% and reduce conversion rate by less than 3%.

[0085] A system for selecting parameters for preparing an iron-carbon-based catalyst, comprising: Preparation parameter acquisition module: obtains multiple sets of iron-carbon based catalyst preparation parameters and performs pretreatment to obtain multiple sets of pretreated iron-carbon based catalyst preparation parameters; Mapping relationship inference module: constructing a multi-task learning model; using the multiple sets of pre-processed iron-carbon-based catalyst preparation parameters to train the multi-task learning model, and obtaining a mapping relationship between each set of iron-carbon-based catalyst preparation parameters and catalyst performance values; Scheme generation module: Based on the mapping relationship between each set of iron-carbon-based catalyst preparation parameters and catalyst performance values, each set of iron-carbon-based catalyst preparation parameters is iteratively optimized using a genetic algorithm to obtain an optimal iron-carbon-based catalyst preparation parameter scheme.

[0086] In a specific embodiment, the algorithm of the present invention is used, and after iteration, the parameter combination obtained is calcination temperature: 680° C., carbon-iron ratio: 3.2:1, and reaction pressure: 1.8 MPa.

[0087] Predicted performance: conversion 85.1% ± 1.2%, selectivity 87.3% ± 0.9%, energy consumption index 0.32.

[0088] Based on the above conditions, The following experiments were conducted to prepare the hydrogenation catalyst, and the experimental results were as follows: Figures 4 to 7 As shown: like Figure 4 As shown, weigh 24.3 g (0.06 mol) and 5.59 g (0.031 mol) of glucose were added, and hydrochloric acid was added at a flow rate to maintain pH = 3.5 ± 0.1, and vacuum dried at 120 ° C for 12 hours. The temperature was raised to 680°C at a rate of 5°C / min under an atmosphere and calcined for 3 hours.

[0089] like Figure 5 As shown, with phenol degradation as the target reaction, 6.49g of ferric chloride ( ) was uniformly combined with 1.33 g of biochar at a carbon-to-iron molar ratio of 2.5:1. The temperature was then programmed to 720°C, the optimal calcination temperature (optimized by the genetic algorithm) under a nitrogen atmosphere at a rate of 5°C / min, and held at that temperature for 4 hours. The optimized calcination temperature range was set between 300-800°C. Pareto front analysis determined that 720°C was the optimal balance between degradation efficiency (weighted 0.6) and catalyst life (weighted 0.4).

[0090] like Figure 6 As shown, Figure 4The neural network model only uses process parameters, removes the ResNet-50 and LSTM modules, and does not include XRD images and reaction kinetics timing data in the data input. The optimization target weight only optimizes the conversion rate. It is a unimodal neural network that does not integrate image features.

[0091] like Figure 7 As shown, the calcination temperature is 600℃ instead of Figure 5 The temperature is optimized by the genetic algorithm, and other conditions are the same.

[0092] The same or similar reference numerals correspond to the same or similar components; The terms used in the drawings to describe positional relationships are for illustrative purposes only and should not be construed as limiting this patent; Obviously, the above embodiments of the present invention are merely examples for the purpose of clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Those skilled in the art will appreciate that other variations or modifications can be made based on the above description. It is not necessary and impossible to enumerate all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the claims of the present invention.

Claims

1. A method for selecting parameters for preparing an iron-carbon based catalyst, characterized in that: include: S1: Acquire multiple sets of iron-carbon-based catalyst preparation parameters and perform pretreatment to obtain multiple sets of pretreated iron-carbon-based catalyst preparation parameters; S2: constructing a multi-task learning model; using the multiple groups of pre-processed iron-carbon-based catalyst preparation parameters to train the multi-task learning model, to obtain a mapping relationship between each group of iron-carbon-based catalyst preparation parameters and catalyst performance values; S3: Based on the mapping relationship between each set of iron-carbon-based catalyst preparation parameters and the catalyst performance value, a genetic algorithm is used to iteratively optimize each set of iron-carbon-based catalyst preparation parameters to obtain an optimal iron-carbon-based catalyst preparation parameter solution.

2. The method for selecting parameters for preparing an iron-carbon based catalyst according to claim 1, wherein: The preparation parameters of the iron-carbon-based catalyst include: process parameters, material structure images, and reaction kinetics curves; Catalyst performance values ​​include: conversion rate prediction value and active site identification value.

3. The method for selecting parameters for preparing an iron-carbon based catalyst according to claim 1, wherein: In step S1, the preprocessing includes: S101: Eliminate abnormal values ​​in multiple sets of iron-carbon-based catalyst preparation parameters; supplement missing values ​​in multiple sets of iron-carbon-based catalyst preparation parameters; S102: Mapping each set of iron-carbon-based catalyst preparation parameters to a standard normal distribution interval to obtain multiple sets of pre-processed iron-carbon-based catalyst preparation parameters.

4. The method for selecting parameters for preparing an iron-carbon based catalyst according to claim 2, wherein: In step S2, the multi-task learning model includes: a multimodal input layer, a feature fusion layer, and a multi-task output layer connected in sequence.

5. The method for selecting parameters for preparing an iron-carbon based catalyst according to claim 4, wherein: The formula of the multimodal input layer is as follows: represents the first weight matrix, represents the process parameter vector, represents the first bias term, represents the first output of the multimodal input layer; Represents a material image, represents the second output of the multimodal input layer; represents the reaction kinetics curve at time t, Represents the third output of the multimodal input layer at time t-1.

6. The method for selecting parameters for preparing an iron-carbon based catalyst according to claim 5, wherein: The formula of the feature fusion layer is as follows: represents the output of the feature fusion layer, represents the first weight, represents the second weight, represents the third weight; represents the first output of the multimodal input layer; represents the second output of the multimodal input layer; represents the third output of the multimodal input layer; represents the joint eigenvector, represents the second weight matrix, represents the second bias term; Represents the weight of the feature fusion layer; represents the weight of the feature fusion layer, represents the first weight, represents the second weight, represents the third weight; represents the joint eigenvector, represents the first output of the multimodal input layer; represents the second output of the multimodal input layer; Represents the third output of the multimodal input layer.

7. The method for selecting parameters for preparing an iron-carbon based catalyst according to claim 6, wherein: The formula of the multi-task output layer is as follows: represents the third weight matrix, represents the third bias term, represents the output of the feature fusion layer, Represents the predicted value of conversion rate; represents the fourth weight matrix, represents the output of the feature fusion layer, represents the fourth bias term, Indicates the active site identification value.

8. The method for selecting parameters for preparing an iron-carbon based catalyst according to claim 7, wherein: In step S3, a genetic algorithm is used to iteratively optimize each set of iron-carbon-based catalyst preparation parameters to obtain an optimal iron-carbon-based catalyst preparation parameter solution, including: S301: Generate N chromosomes as a first chromosome group according to the preparation parameters of the iron-carbon based catalyst; S302: Calculate the fitness of each chromosome in the first chromosome group to obtain a first fitness of each chromosome; S303: Divide the first chromosome group into a second chromosome group and a third chromosome group according to the first fitness; S304: performing crossover mutation on the chromosomes in the third chromosome set to obtain a fourth chromosome set; S305: merging the fourth chromosome set with the second chromosome set to form a new first chromosome set; S306: Determine whether the first termination condition is met based on the first chromosome set; if not, execute step S302; if so, execute step S307; S307: Selecting an optimal iron-carbon-based catalyst preparation parameter scheme in the first chromosome group according to the Pareto optimal algorithm.

9. The method for selecting parameters for preparing an iron-carbon-based catalyst according to claim 8, wherein: In step S302, the fitness calculation formula is as follows: 、 、 represents the Pareto frontier optimization weight coefficient, Represents the predicted value of conversion rate, represents the active point identification value, and E represents energy consumption.

10. A parameter selection system for preparing an iron-carbon-based catalyst, applied to the parameter selection method according to any one of claims 1 to 9, characterized in that: include: Preparation parameter acquisition module: obtains multiple sets of iron-carbon based catalyst preparation parameters and performs pretreatment to obtain multiple sets of pretreated iron-carbon based catalyst preparation parameters; Mapping relationship inference module: constructing a multi-task learning model; using the multiple sets of pre-processed iron-carbon-based catalyst preparation parameters to train the multi-task learning model, and obtaining a mapping relationship between each set of iron-carbon-based catalyst preparation parameters and catalyst performance values; Scheme generation module: Based on the mapping relationship between each set of iron-carbon-based catalyst preparation parameters and catalyst performance values, each set of iron-carbon-based catalyst preparation parameters is iteratively optimized using a genetic algorithm to obtain an optimal iron-carbon-based catalyst preparation parameter scheme.

Images