Alcohol compound catalyst preparation process parameter optimization method

Abstract

The invention discloses a method for optimizing preparation process parameters of a catalyst for alcohol compounds, which specifically comprises the following steps of: (1) inputting preparation process parameters of a catalyst for preparing alcohol compounds by carbon dioxide hydrogenation into a performance prediction model of the catalyst for preparing alcohol compounds by carbon dioxide hydrogenation, the carbon dioxide hydrogenation alcohol compound catalyst performance prediction model outputs the prediction performance of the carbon dioxide hydrogenation alcohol compound catalyst; and (2) updating the preparation process parameters of the catalyst for preparing the alcohol compounds through carbon dioxide hydrogenation, and executing the step (1) until the preparation process parameters corresponding to the target performance of the catalyst for preparing the alcohol compounds through carbon dioxide hydrogenation are found. According to the method, a small amount of test data is utilized, an artificial neural network model based on data driving is provided, a rapid prediction method for obtaining different preparation process parameters-catalyst performance is obtained, and the method is used for rapidly optimizing a preparation process parameter combination under the target performance of the catalyst, so that technical reference is provided for actual emission reduction and carbon reduction.

Description

Technical Field

[0001] The present invention belongs to the technical field of alcohol compound catalysts, and more specifically, the present invention relates to a method for optimizing process parameters for preparing alcohol compound catalysts. Background Art

[0002] Alcohol compounds are a class of organic compounds with one or more alcohol (-OH) functional groups, common types include methanol, ethanol, propanol, etc. Alcohol compounds have important applications in many fields, can be used as fuels and solvents (such as ethanol), to make medicinal tinctures and drug carriers; used as raw materials in the chemical industry to produce various chemicals and materials, such as plastics, rubber, fibers, etc.; used in medicine, pesticides, coatings and other fields, such as glucose decomposed into ethanol and carbon dioxide under the action of yeast.

[0003] In recent years, with the tightening of the national "dual carbon" strategy and people's increasing attention to environmental issues, the process route of converting carbon dioxide (CO2) as raw material into alcohol compounds by hydrogenation has attracted widespread attention. It can not only effectively reduce CO2 gas emissions, but also obtain important alcohol chemicals, which has extremely high research value and economic benefits. However, this route still has prominent problems such as low CO2 conversion rate and low product selectivity. It is necessary to further improve the catalytic performance of this reaction to meet the needs of industrial production.

[0004] Generally, the factors that affect the performance of CO2 hydrogenation to alcohol-based compounds mainly include the type of catalyst selected (such as copper-based, nickel-based catalysts, etc.), the preparation method, and the catalytic reaction process conditions (such as temperature, pressure, etc.). These factors are intertwined, making the performance fluctuations of the final catalyst highly nonlinear and uncertain. In order to obtain catalysts with higher performance indicators, a large number of experiments are often required, resulting in problems such as long development cycles and high experimental costs. Therefore, it is necessary to use some new technical means to construct a mathematical mapping relationship model between "catalyst performance indicators-different process factors", and then use the model to quickly predict the optimal preparation process under different catalyst target performance.

[0005] In recent years, data-driven research methods have been widely used due to their strong nonlinear fitting ability and wide application range. The invention patent (CN118746924A) announced "a method for optimizing parameters of a cement production decomposition furnace temperature control system based on genetic algorithm"; the invention patent (CN118230854A) announced "a method for predicting tea moisture content based on genetic algorithm optimization of BP neural network"; the invention patent (CN118658551A) announced "a method for outputting a catalyst stable structure under specified reaction conditions based on machine learning"; the invention patent (CN118098404B) announced "a catalyst molecule reaction performance prediction method, equipment and medium", which predicts the reaction performance of catalyst molecules through machine learning modeling. The various algorithm models constructed by the above research work have significantly improved the prediction accuracy compared with ordinary linear statistical methods, which provides a good reference for the research work on catalysts for the hydrogenation of carbon dioxide to alcohol compounds. However, there are only a small amount of experimental data for catalysts for alcohol compounds, and the number of samples is small, which limits the construction of big data prediction models. Summary of the invention

[0006] The present invention provides a method for optimizing process parameters for preparing a catalyst for an alcohol compound, aiming to improve at least one of the above problems.

[0007] The present invention is achieved by providing a method for optimizing process parameters for preparing a catalyst for an alcohol compound, the method being specifically as follows:

[0008] (1) inputting the preparation process parameters of the carbon dioxide hydrogenation to alcohol compound catalyst into the carbon dioxide hydrogenation to alcohol compound catalyst performance prediction model, and the carbon dioxide hydrogenation to alcohol compound catalyst performance prediction model outputs the predicted performance of the carbon dioxide hydrogenation to alcohol compound catalyst;

[0009] (2) Updating the preparation process parameters of the catalyst for preparing alcohol compounds by hydrogenating carbon dioxide, and executing step (1) until the preparation process parameters corresponding to the target performance of the catalyst for preparing alcohol compounds by hydrogenating carbon dioxide are found.

[0010] Furthermore, a performance prediction model for the catalyst for the hydrogenation of carbon dioxide to produce alcohol compounds is constructed using a BP network model. The preparation process parameters of the catalyst for the hydrogenation of carbon dioxide to produce alcohol compounds are used as characteristics, and the performance of the catalyst for the hydrogenation of carbon dioxide to produce alcohol compounds is used as a label to construct samples. The BP network model is trained based on the training samples, and the BP network model is tested during the training process based on the test samples. When the test accuracy of the BP network model reaches a preset accuracy threshold, the training is completed and the BP network model is used as a performance prediction model for the catalyst for the hydrogenation of carbon dioxide to produce alcohol compounds.

[0011] Furthermore, the preparation process parameters for the alcohol compound include: reaction temperature, reaction pressure, air velocity, and CuNi ratio.

[0012] Furthermore, the performance of the catalyst for hydrogenating carbon dioxide to produce alcohol compounds is conversion rate or selectivity.

[0013] Furthermore, the BP network topology consists of three layers, namely, input layer, hidden layer and output layer. The number of input layer nodes is 4, the number of output layer nodes is 1, and the number of hidden layer nodes ranges from [5-12].

[0014] Furthermore, the genetic algorithm GA is used to optimize the initial weights and thresholds of the BP network.

[0015] Furthermore, before constructing samples based on the collected actual test data, the test data is cleaned, and data cleaning includes: missing data interpolation, erroneous data repair, duplicate data deletion, and data format unification.

[0016] Furthermore, the training function of the BP network model is TRAINLM, the learning function is LEARGDM, the transfer function is Tansig and Purelin, the learning rate is set to 0.01, the maximum number of iterations is 1000, and the target error E is 1x10 -4 .

[0017] Furthermore, the samples are divided into training samples and test samples, and the ratio of training samples to test samples is 8:2.

[0018] The present invention uses a small amount of test data to provide a data-driven artificial neural network model to obtain a fast prediction method for different preparation process parameters-catalyst performance, which is used to quickly optimize the preparation process parameter combination under the catalyst target performance, thereby providing a technical reference for actual emission reduction and carbon reduction. In order to overcome the shortcomings of the existing ordinary BP (Back Propagation, BP) neural network technology, which has poor prediction accuracy for small sample data and is prone to local minimization, the present invention uses a genetic algorithm GA (Genetic Algorithm, GA) to optimize the initial weights and thresholds of the BP network. GA has the characteristics of fast learning speed, strong approximation ability and good generalization ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0020] Figure 1 Iterative curve diagram of maximum conversion rate of catalyst for preparing alcohol compounds by hydrogenation of carbon dioxide provided in an embodiment of the present invention;

[0021] Figure 2 A comparison diagram of the error in optimizing the maximum conversion rate of the catalyst for producing alcohol compounds by hydrogenating carbon dioxide provided in an embodiment of the present invention;

[0022] Figure 3 Iterative curve diagram when the conversion rate of the catalyst for preparing alcohol compounds by hydrogenation of carbon dioxide provided in an embodiment of the present invention is 17%;

[0023] Figure 4 Iterative curve diagram of the maximum selectivity of the catalyst for producing alcohol compounds by hydrogenation of carbon dioxide provided in an embodiment of the present invention;

[0024] Figure 5 A comparison diagram of the error in optimizing the selectivity maximum value of the catalyst for producing alcohol compounds by hydrogenation of carbon dioxide provided in an embodiment of the present invention;

[0025] Figure 6 This is an iterative curve diagram when the selectivity target value of the catalyst for producing alcohol compounds by hydrogenating carbon dioxide provided by an embodiment of the present invention is 90%. DETAILED DESCRIPTION

[0026] The specific implementation modes of the present invention are further explained in detail below by describing the embodiments with reference to the accompanying drawings, so as to help those skilled in the art to have a more complete, accurate and in-depth understanding of the inventive concept and technical solution of the present invention.

[0027] The present invention uses a small amount of test data to provide a data-driven artificial neural network model to obtain a fast prediction method for different preparation process parameters-catalyst performance, which is used to quickly optimize the preparation process parameter combination under the catalyst target performance, thereby providing a technical reference for actual emission reduction and carbon reduction. In order to overcome the shortcomings of the existing ordinary BP (Back Propagation, BP) neural network technology, which has poor prediction accuracy for small sample data and is prone to local minimization, the present invention uses a genetic algorithm GA (Genetic Algorithm, GA) to optimize the initial weights and thresholds of the BP network. GA has the characteristics of fast learning speed, strong approximation ability and good generalization ability.

[0028] The present invention provides a method for optimizing preparation process parameters corresponding to target performance of alcohol compound catalysts, comprising the following steps:

[0029] (1) inputting the preparation process parameters of the carbon dioxide hydrogenation to alcohol compound catalyst into the carbon dioxide hydrogenation to alcohol compound catalyst performance prediction model, and the carbon dioxide hydrogenation to alcohol compound catalyst performance prediction model outputs the predicted performance of the carbon dioxide hydrogenation to alcohol compound catalyst;

[0030] (2) Updating the preparation process parameters of the catalyst for preparing alcohol compounds by hydrogenating carbon dioxide, and executing step (1) until the preparation process parameters corresponding to the target performance of the catalyst for preparing alcohol compounds by hydrogenating carbon dioxide are found.

[0031] In an embodiment of the present invention, a performance prediction model of a catalyst for producing alcohol compounds by hydrogenating carbon dioxide is constructed using a BP network model. The preparation process parameters of the catalyst for producing alcohol compounds by hydrogenating carbon dioxide are used as characteristics, and the performance of the catalyst for producing alcohol compounds by hydrogenating carbon dioxide is used as a label. Samples are constructed, and the BP network model is trained based on the samples until the recognition accuracy of the BP network model reaches a preset accuracy threshold. The training is completed and the BP network model is used as a performance prediction model for the catalyst for producing alcohol compounds by hydrogenating carbon dioxide.

[0032] In an embodiment of the present invention, the performance of the catalyst for producing alcohol compounds by hydrogenating carbon dioxide is taken as a research object, actual test data is collected, and the test data is cleaned. The data cleaning includes: missing data interpolation, erroneous data repair, duplicate data deletion, and data format unification; the data conversion operation is to normalize the original data samples.

[0033] The sample is constructed based on the cleaned experimental data, and the preparation process parameters that are more critical to the performance of the carbon dioxide hydrogenation to alcohol compound catalyst are selected as features. The more critical preparation process parameters in the catalyst preparation process include: reaction temperature (unit: ℃), reaction pressure (unit: MPa), air velocity (unit: mL / gcat / h), CuNi ratio, and the performance of the prepared carbon dioxide hydrogenation to alcohol compound catalyst is used as a label. The catalyst performance includes: conversion rate (unit: %) or selectivity (unit: %) to form a small number of data samples.

[0034] Before training the BP network model, it is necessary to first construct the BP network model, set the training function, learning rate, transfer function, number of training steps and maximum allowable error, and train the BP network model for different preparation process parameters-catalyst performance prediction by using the training samples selected from the data samples; wherein, the BP network topology structure is three layers, namely, input layer, hidden layer and output layer, the number of input layer nodes is 4, the number of output layer nodes is 1, the number of hidden layer nodes ranges from [5-12], the training function is TRAINLM, the learning function is LEARGDM, the transfer function is Tansig, Purelin, the learning rate is set to 0.01, the maximum number of iterations is 1000, and the target error E is 1x10 -4 The training sample data comes from 192 groups (about 80%) randomly selected from the 240 groups of data, and the test sample data is the remaining 48 groups (about 20%).

[0035] In order to overcome the shortcomings of the existing common BP (Back Propagation, BP) neural network technology, which has poor prediction accuracy for small sample data and is prone to local minimization, the present invention uses a genetic algorithm GA (Genetic Algorithm, GA) to optimize the initial weights and thresholds of the BP network. GA has the characteristics of fast learning speed, strong approximation ability and good generalization ability. The key parameters of the genetic algorithm to optimize the BP network model are set: number of iterations, population size, crossover probability, and mutation probability, and the initial weights and thresholds of the BP model are optimized. Specifically, the key parameter settings are: the maximum number of iterations is 100 times, the population size is 20, the crossover probability is 0.8, and the mutation probability is 0.1.

[0036] The present invention aims at finding the best process parameter combination for the target performance of the catalyst for the hydrogenation of carbon dioxide to alcohol compounds. Based on a very small amount of test data, the genetic algorithm is used to optimize the BP network model, and finally a model with high prediction accuracy and the ability to quickly predict the process parameter combination corresponding to the target performance of the catalyst is obtained. It not only solves the problem of poor prediction accuracy of the traditional modeling method, but also overcomes the common problems of overfitting and poor generalization ability of the ordinary BP network method, providing a new technical means for the development and exploration of catalysts for the hydrogenation of carbon dioxide to alcohol compounds.

[0037] The precipitation method was used to select the process parameters that were most critical to the catalyst performance to form a model sample, as shown in Table 1. The input variables (features) were: reaction temperature, reaction pressure, air velocity, CuNi ratio, and the output variable (label) was catalyst performance: conversion rate.

[0038] Table 1 Preparation process parameters and value distribution table affecting catalyst performance

[0039]

[0040] 240 sets of original data were obtained from the actual experiment. After preprocessing operations such as cleaning, reduction, and transformation, the data sample set for modeling was finally obtained, as shown in Table 2. Among the 240 sets of data, 192 sets (about 80%) were randomly selected as training samples, and the remaining 48 sets (about 20%) were taken as test samples for the model.

[0041] Table 2 Data sample set (catalyst conversion rate)

[0042]

[0043]

[0044]

[0045]

[0046]

[0047]

[0048]

[0049]

[0050]

[0051]

[0052]

[0053]

[0054] In MATLAB R2021 software, we first build a BP neural network model and set the following settings: the number of hidden layer nodes N is 8, the maximum number of iterations is 1000, the learning rate lr is 0.01, and the target error E is 1x10 -4 ; Then, the genetic algorithm GA is used to perform reverse process optimization for the target performance, with the following settings: the maximum number of iterations is 100, the population size is 20, the crossover probability is 0.8, and the mutation probability is 0.1, to obtain the process prediction model corresponding to the catalyst target performance.

[0055] The goal is to find the maximum catalyst conversion rate. Figure 1 The target iteration curve shown in the figure shows that with the increase of iteration times, the target value continues to increase. After nearly 80 iterations, the value tends to be stable and reaches a maximum value of 15.9%, corresponding to the best process parameter combination (such as Figure 2 ) is predicted to be: 320℃, 3.9MPa, 5184mL / gcat / h, 81.2%; this is very close to the 156th and 157th groups of data corresponding to the maximum conversion rate in the original data of Table 1, with a relative error of only 0.6%-1.2%. This shows that the model can accurately realize the rapid prediction of the process corresponding to the maximum conversion rate of the catalyst.

[0056] The goal is to achieve a catalyst conversion rate of 17%. Figure 3 It can be seen from the target iteration curve that as the number of iterations increases, the target error value continues to decrease. After nearly 70 iterations, the error is close to zero, and the corresponding optimal process parameter combination is predicted to be: 318.5℃, 3.7MPa, 5099mL / gcat / h, 80.9%. This shows that the model can quickly predict and output the corresponding process parameter combination for a given catalyst conversion rate target performance.

[0057] The precipitation method was used to select the process parameters that were most critical to the catalyst performance to form a model sample, as shown in Table 1. The selected preparation process parameters were: reaction temperature, reaction pressure, air velocity, CuNi ratio, and the label was catalyst performance: selectivity.

[0058] 240 sets of original data were obtained from the actual experiment. After preprocessing operations such as cleaning, reduction, and transformation, the data sample set for modeling was finally obtained, as shown in Table 3. Among the 240 sets of data, 192 sets (about 80%) were randomly selected as training samples, and the remaining 48 sets (about 20%) were taken as test samples for the model.

[0059] Table 3 Data sample set (catalyst selectivity)

[0060]

[0061]

[0062]

[0063]

[0064]

[0065]

[0066]

[0067]

[0068]

[0069]

[0070]

[0071] In MATLAB R2021 software, we first build a BP neural network model and set the following settings: the number of hidden layer nodes N is 9, the maximum number of iterations is 1000, the learning rate lr is 0.01, and the target error E is 1x10 -4 ; Then, the genetic algorithm GA is used to perform reverse process optimization for the target performance, with the following settings: the maximum number of iterations is 100, the population size is 40, the crossover probability is 0.8, the mutation probability is 0.3, the crossover operator parameter is 10, and the mutation operator parameter is 10, to obtain the process prediction model corresponding to the catalyst target performance.

[0072] The goal is to find the maximum selectivity of the catalyst. Figure 4It can be seen from the target iteration curve that as the number of iterations increases, the target value continues to increase. After nearly 95 iterations, the value tends to be stable and reaches a maximum value of 89.9%, corresponding to the best process parameter combination (such as Figure 5 ) is predicted to be: 264.5℃, 3.4MPa, 17455.9mL / gcat / h, 74.5%; this is very close to the 222nd group of data corresponding to the maximum conversion rate in the original data of Table 2, with a relative error of only 1.2%. This shows that the model can accurately realize the rapid prediction of the process corresponding to the maximum selectivity of the catalyst.

[0073] The goal is to achieve a catalyst selectivity of 90%. Figure 6 It can be seen from the target iteration curve that as the number of iterations increases, the target error value continues to decrease. After nearly 25 iterations, the error is close to zero, and the corresponding optimal process parameter combination is predicted to be: 260.0℃, 1.4MPa, 10412.1mL / gcat / h, 64.6%. This shows that the model can quickly predict and output the corresponding process parameter combination for a given catalyst selectivity target performance.

[0074] The present invention has been described exemplarily. Obviously, the specific implementation of the present invention is not limited to the above-mentioned method. As long as various non-substantial improvements are made using the method concept and technical solution of the present invention, or the concept and technical solution of the present invention are directly applied to other occasions without improvement, they are all within the protection scope of the present invention.

Claims

1. A method for optimizing process parameters for preparing a catalyst for an alcohol compound, characterized in that: The method is specifically as follows: (1) inputting the preparation process parameters of the carbon dioxide hydrogenation to alcohol compound catalyst into the carbon dioxide hydrogenation to alcohol compound catalyst performance prediction model, and the carbon dioxide hydrogenation to alcohol compound catalyst performance prediction model outputs the predicted performance of the carbon dioxide hydrogenation to alcohol compound catalyst; (2) Updating the preparation process parameters of the catalyst for preparing alcohol compounds by hydrogenating carbon dioxide, and executing step (1) until the preparation process parameters corresponding to the target performance of the catalyst for preparing alcohol compounds by hydrogenating carbon dioxide are found.

2. The method for optimizing process parameters for preparing a catalyst for alcohol compounds according to claim 1, characterized in that: The performance prediction model of the catalyst for the hydrogenation of carbon dioxide to produce alcohol compounds is constructed using a BP network model. The preparation process parameters of the catalyst for the hydrogenation of carbon dioxide to produce alcohol compounds are used as characteristics, and the performance of the catalyst for the hydrogenation of carbon dioxide to produce alcohol compounds is used as a label to construct samples. The BP network model is trained based on the training samples, and the BP network model is tested during the training process based on the test samples. When the test accuracy of the BP network model reaches a preset accuracy threshold, the training is completed and the BP network model is used as a performance prediction model for the catalyst for the hydrogenation of carbon dioxide to produce alcohol compounds.

3. The method for optimizing process parameters for preparing a catalyst for alcohol compounds according to claim 1, characterized in that: The preparation process parameters for alcohol compounds include: reaction temperature, reaction pressure, air velocity, and CuNi ratio.

4. The method for optimizing process parameters for preparing a catalyst for alcohol compounds according to claim 1, characterized in that: The performance of the catalyst for hydrogenating carbon dioxide to produce alcohol compounds is conversion rate or selectivity.

5. The method for optimizing process parameters for preparing a catalyst for alcohol compounds as claimed in claim 2, characterized in that: The BP network topology consists of three layers, namely the input layer, hidden layer and output layer. The number of input layer nodes is 4, the number of output layer nodes is 1, and the number of hidden layer nodes ranges from [5-12].

6. The method for optimizing process parameters for preparing a catalyst for alcohol compounds as claimed in claim 5, characterized in that: Genetic algorithm GA is used to optimize the initial weights and thresholds of BP network.

7. The method for optimizing process parameters for preparing a catalyst for alcohol compounds according to claim 1, characterized in that: Before constructing samples based on the actual test data collected, the test data is cleaned. Data cleaning includes: missing data interpolation, erroneous data repair, duplicate data deletion, and data format unification.

8. The method for optimizing process parameters for preparing a catalyst for alcohol compounds as claimed in claim 2, characterized in that: The training function of the BP network model is TRAINLM, the learning function is LEARGDM, the transfer function is Tansig and Purelin, the learning rate is set to 0.01, the maximum number of iterations is 1000, and the target error E is 1x10 -4 .

9. The method for optimizing process parameters for preparing catalysts for alcohol compounds according to claim 1, characterized in that: The samples are divided into training samples and test samples, and the ratio of training samples to test samples is 8:2.

Images