Heterogeneous catalysis peracetic acid task decomposition method based on machine learning

Abstract

The invention relates to a heterogeneous catalysis peracetic acid task decomposition method based on machine learning, and the method comprises the following steps: S1, collecting experimental data, carrying out the normalization of the data through a data preprocessing method, and removing high-correlation features; s2, decomposing the total task into a performance optimization sub-task and a mechanism identification sub-task, and respectively constructing corresponding data sets for the two tasks; s3, training a machine learning model for processing a regression problem by using the two preprocessed data sets; and S4, optimizing experimental conditions and identifying a mechanism according to machine learning results of the two subtasks. Compared with the prior art, the method is based on machine learning, and can quickly and effectively optimize the performance and the identification mechanism at low cost.

Description

Technical Field

[0001] The invention relates to the technical field of machine learning and water treatment, and in particular to a method for decomposing heterogeneous catalytic peracetic acid tasks based on machine learning. Background Art

[0002] With the development of human society, a large number of pollutants that are difficult to degrade by traditional water treatment technologies are discharged into water bodies, which has caused potential impacts on aquatic ecosystems and humans. The heterogeneous catalytic activation of peracetic acid process is considered to be a promising water treatment technology to deal with this problem, but the performance optimization and mechanism identification are cumbersome, complex, time-consuming and labor-intensive, which hinders the rapid development of this technology. Machine learning methods based on big data have the potential to solve these problems in water treatment. Summary of the invention

[0003] The purpose of the present invention is to overcome the defects of the above-mentioned prior art such as cumbersome, complicated, time-consuming and labor-intensive manual trial and error experiments, and to provide a method for decomposing heterogeneous catalytic peracetic acid tasks based on machine learning, which can effectively optimize performance and identify mechanisms quickly and at low cost.

[0004] The present invention provides a method for decomposing heterogeneous catalytic peracetic acid tasks based on machine learning, comprising the following steps:

[0005] S1: Collect experimental data and use data preprocessing methods to normalize the data and remove high correlation features;

[0006] S2: Decompose the overall task into two subtasks: performance optimization and mechanism identification, and construct corresponding datasets for the two tasks respectively;

[0007] S3: Use the two preprocessed datasets to train a machine learning model for the regression problem;

[0008] S4: Optimize experimental conditions and identify mechanisms based on the machine learning results of the two subtasks.

[0009] Furthermore, in S1, the experimental data collected include: type of pollutant, concentration of pollutant, type of catalyst, concentration of catalyst, water matrix parameters, quenching experimental data, stirring speed, pH value, and temperature.

[0010] Furthermore, in S1, the Sperman correlation coefficient method is used to eliminate features with high correlation.

[0011] Furthermore, in S2, the data set of the performance optimization subtask includes: reaction rate constant, type of pollutant, concentration of pollutant, type of catalyst, concentration of catalyst, water matrix parameters, stirring speed, pH value, and temperature.

[0012] Furthermore, in S2, the data set of the mechanism identification subtask includes: reaction rate constant, type of pollutant, concentration of pollutant, type of catalyst, concentration of catalyst, quenching experimental data, stirring speed, pH value, and temperature.

[0013] Furthermore, in S3, the machine learning models for processing regression problems include: XGBoost, CatBoost, Bagging, LGBM, ANN, SVM, GP, and KNN machine learning models.

[0014] Furthermore, in S3, the two preprocessed data sets are imported into the machine learning model of the regression problem, and R 2 The corresponding machine learning model with the best performance is determined by RMSE, thereby obtaining the optimal models trained for the two problems respectively.

[0015] Furthermore, in S4, the prediction results of the performance optimization subtask optimize the reaction rate constant through an optimization algorithm; the mechanism identification subtask determines the possible reaction mechanism based on the prediction value of the quenching experiment.

[0016] Furthermore, the optimization algorithm is a particle swarm optimization algorithm (PSO) or a differential evolution algorithm (DE) or other related algorithms.

[0017] Furthermore, in S4, for the performance optimization subtask, an optimization algorithm is used to optimize the reaction rate constant, thereby obtaining the optimal reaction rate constant, peracetic acid concentration, pollutant type, pollutant concentration, catalyst type, catalyst concentration, water matrix parameters, stirring speed, pH value, temperature and other parameters; for the mechanism identification subtask, the predicted k t / k0 value, to determine the possible reaction mechanism; where k t is the pseudo-first-order reaction rate constant after adding a certain amount of quencher, and k0 is the pseudo-first-order reaction rate constant without adding any quencher.

[0018] Compared with the prior art, the present invention has the following advantages:

[0019] The present invention is based on machine learning, uses two preprocessed data sets to train a machine learning model for processing regression problems, optimizes experimental conditions and identifies mechanisms based on the machine learning results of the two subtasks, and effectively optimizes performance and identifies mechanisms quickly and at low cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 Schematic diagram of the DE algorithm optimizing the process of removing sulfamethoxazole from water using peracetic acid activated with CoFe2O4 catalyst. DETAILED DESCRIPTION

[0021] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. Component models, material names, connection structures, control methods, algorithms and other features not clearly described in this technical solution are all considered to be common technical features disclosed in the prior art.

[0022] Example 1

[0023] This embodiment provides a method for decomposing heterogeneous catalytic peracetic acid tasks based on machine learning, comprising the following steps:

[0024] S1: Collect experimental data, and use data preprocessing methods to normalize the data and eliminate high correlation features; use the Sperman correlation coefficient method to eliminate high correlation features; the collected experimental data include: type of pollutants, concentration of pollutants, type of catalysts, concentration of catalysts, water matrix parameters, quenching experimental data, stirring speed, pH value, temperature.

[0025] In a specific embodiment, catalyst parameters can be obtained by electron microscopy and other related characterizations; the concentration of organic pollutants can be obtained by advanced liquid chromatograph or UV-visible spectrophotometer; inorganic anions in water quality parameters are determined by ion chromatography; the properties of pollutants are obtained by first-principles calculations; the remaining parameters can be directly recorded or obtained by calculation. The obtained data are standardized by the z-score method. Missing values ​​in the data are filled by the KNN method. Afterwards, the Sperman correlation coefficient method is used to determine the correlation between all features. According to the principle of retaining as many features with different meanings as possible, features with correlation coefficients greater than 0.9 are eliminated.

[0026] S2: The overall task is decomposed into two subtasks: performance optimization and mechanism identification, and corresponding data sets are constructed for the two tasks respectively; the data set of the performance optimization subtask includes: reaction rate constant, type of pollutant, concentration of pollutant, type of catalyst, concentration of catalyst, water matrix parameters, stirring speed, pH value, and temperature. The data set of the mechanism identification subtask includes: reaction rate constant, type of pollutant, concentration of pollutant, type of catalyst, concentration of catalyst, quenching experimental data, stirring speed, pH value, and temperature.

[0027] S3: Use the two preprocessed data sets to train a machine learning model for regression problems; the machine learning models for regression problems include: XGBoost, CatBoost, Bagging, LGBM, ANN, SVM, GP, KNN machine learning models. The two preprocessed data sets are imported into the machine learning model for regression problems respectively, and R 2 The corresponding machine learning model with the best performance is determined by RMSE, thereby obtaining the optimal models trained for the two problems respectively.

[0028] In this embodiment, for the above two subtasks, the best machine learning model is the CatBoost model.

[0029] S4: Optimize experimental conditions and identify mechanisms based on the machine learning results of the two subtasks.

[0030] The prediction results of the performance optimization subtask are used to optimize the reaction rate constant through an optimization algorithm; the mechanism identification subtask determines the possible reaction mechanism based on the predicted value of the quenching experiment. The optimization algorithm is a particle swarm optimization algorithm (PSO) or a differential evolution algorithm (DE) and other related algorithms.

[0031] In this example, the removal of sulfamethoxazole in water by peracetic acid activated by CoFe2O4 catalyst is taken as an example to optimize the concentration of peracetic acid, the concentration of pollutants, the concentration of catalysts, the water matrix parameters, the stirring speed, the pH value, and the temperature. The specific optimization process is shown in Figure 1 For this embodiment, the mechanism identification subtask uses the predicted k t / k0 value, to determine the possible reaction mechanism; where k t is the pseudo-first-order reaction rate constant after adding a certain amount of quencher, and k0 is the pseudo-first-order reaction rate constant without adding any quencher. According to the prediction results of tert-butyl alcohol and methanol, as shown in Table 1, it can be determined that the main mechanism of the catalyst to activate peracetic acid to remove sulfamethoxazole is through a non-radical process.

[0032] Table 1 Quenching experiment predictions of the trained CatBoost model

[0033]

[0034] Components not described in detail in this embodiment are all existing components that can be purchased through public channels.

[0035] The above description of the embodiments is to facilitate the understanding and use of the invention by those skilled in the art. It is obvious that those skilled in the art can easily make various modifications to these embodiments and apply the general principles described herein to other embodiments without creative work. Therefore, the present invention is not limited to the above embodiments, and improvements and modifications made by those skilled in the art based on the disclosure of the present invention without departing from the scope of the present invention should be within the scope of protection of the present invention.

Claims

1. A method for decomposing heterogeneous catalytic peracetic acid tasks based on machine learning, characterized in that: The following steps are involved: S1: Collect experimental data and use data preprocessing methods to normalize the data and remove high correlation features; S2: Decompose the overall task into two subtasks: performance optimization and mechanism identification, and construct corresponding datasets for the two tasks respectively; S3: Use the two preprocessed datasets to train a machine learning model for the regression problem; S4: Optimize experimental conditions and identify mechanisms based on the machine learning results of the two subtasks.

2. The method for decomposing heterogeneous catalytic peracetic acid tasks based on machine learning according to claim 1, characterized in that: In S1, the collected experimental data include: type of pollutant, concentration of pollutant, type of catalyst, concentration of catalyst, water matrix parameters, quenching experimental data, stirring speed, pH value, and temperature.

3. The method for decomposing heterogeneous catalytic peracetic acid tasks based on machine learning according to claim 1, characterized in that: In S1, the Sperman correlation coefficient method is used to eliminate features with high correlation.

4. The method for decomposing heterogeneous catalytic peracetic acid tasks based on machine learning according to claim 1, characterized in that: In S2, the data sets for the performance optimization subtask include: reaction rate constant, type of pollutant, concentration of pollutant, type of catalyst, concentration of catalyst, water matrix parameters, stirring speed, pH value, and temperature.

5. The method for decomposing heterogeneous catalytic peracetic acid tasks based on machine learning according to claim 1, characterized in that: In S2, the data sets of the mechanism identification subtask include: reaction rate constant, type of pollutant, concentration of pollutant, type of catalyst, concentration of catalyst, quenching experiment data, stirring speed, pH value, and temperature.

6. The method for decomposing heterogeneous catalytic peracetic acid tasks based on machine learning according to claim 1, characterized in that: In S3, the machine learning models for processing regression problems include: XGBoost, CatBoost, Bagging, LGBM, ANN, SVM, GP, and KNN machine learning models.

7. The method for decomposing heterogeneous catalytic peracetic acid tasks based on machine learning according to claim 1, characterized in that: In S3, the two preprocessed data sets are imported into the machine learning model of the regression problem, and R 2 The corresponding machine learning model with the best performance is determined by RMSE, thereby obtaining the optimal models trained for the two problems respectively.

8. The method for decomposing heterogeneous catalytic peracetic acid tasks based on machine learning according to claim 1, characterized in that: In S4, the prediction results of the performance optimization subtask optimize the reaction rate constant through an optimization algorithm; the mechanism identification subtask determines the possible reaction mechanism based on the predicted value of the quenching experiment.

9. The method for decomposing heterogeneous catalytic peracetic acid tasks based on machine learning according to claim 8, characterized in that: The optimization algorithm is a particle swarm optimization algorithm (PSO) or a differential evolution algorithm (DE).

10. The method for decomposing heterogeneous catalytic peracetic acid tasks based on machine learning according to claim 1, characterized in that: In S4, for the performance optimization subtask, an optimization algorithm is used to optimize the reaction rate constant, thereby obtaining the optimal reaction rate constant, peracetic acid concentration, pollutant type, pollutant concentration, catalyst type, catalyst concentration, water matrix parameters, stirring speed, pH value, and temperature; for the mechanism identification subtask, the predicted k t / k0 value, to determine the possible reaction mechanism; where k t is the pseudo-first-order reaction rate constant after adding a certain amount of quencher, and k0 is the pseudo-first-order reaction rate constant without adding any quencher.

Images