Method for preparing spinel type OER catalyst based on machine learning assistance

Abstract

The invention relates to a method for preparing a spinel type OER catalyst based on machine learning assistance, and relates to a method for preparing a catalyst. A series of machine learning algorithms are utilized, experimental data are taken as the basis, and the spinel type OER catalyst is designed in an aided manner. The unique advantages of machine learning in the aspect of recognizing the relation between overpotential and catalyst composition and synthesis and reaction parameters are fully utilized, and the incidence relation between the components, the structure, the process and the performance is established more quickly in a more labor-saving mode. Under the framework, input catalyst composition, synthesis and reaction parameters and test conditions are quantified, then relations between various conditions such as catalyst chemical composition (AxByCzO4 and a carrier), a synthesis method and roasting temperature and overpotential are fitted through various algorithms, and important factors influencing the OER activity are evaluated.

Description

Technical Field

[0001] The present invention relates to a method for preparing a catalyst, and in particular to a method for preparing a spinel OER catalyst assisted by machine learning. Background Art

[0002] With the depletion of fossil energy and the intensification of environmental pollution problems, the development of renewable clean energy is imperative. Hydrogen production by water electrolysis is one of the more promising technologies. The process is mainly divided into cathode hydrogen evolution reaction (HER) and anode oxygen evolution reaction (OER). Among them, OER involves a complex four-electron reaction process, and its slow kinetic characteristics seriously restrict the overall efficiency of hydrogen production by water electrolysis. Among them, precious metal-based catalysts such as iridium and ruthenium oxides have excellent OER catalytic activity, but their scarcity and high cost limit their industrial promotion. Based on this, it is crucial to design and develop low-cost, high-activity non-precious metal-based OER catalysts.

[0003] Transition metal-based catalysts such as Fe, Co, Ni oxides, hydroxides, phosphides, borides, etc. have good OER activity. Among them, spinel oxides are favored because of their low price, easy access, rich chemical states, and controllable structure. In practice, due to the complex reaction mechanism, the development process of spinel electrocatalysts is still highly dependent on the trial and error process, the experimental steps are cumbersome, and there is a lack of theoretical guidance. Long cycles and high costs limit the development of new materials. Machine learning, as a new data processing technology, has also played an important role in the research of OER electrocatalysts. Using machine learning algorithms, such as neural networks and support vector machines, to analyze large amounts of experimental data, establish structure-activity relationship models, and predict the catalytic performance of new materials can accelerate the material screening process and provide a good technical means for the prediction, simulation, and reaction conditions of heterogeneous catalytic reaction processes.

[0004] The traditional catalyst design process is lengthy and has the blindness of experimental “trial and error”. Summary of the invention

[0005] The purpose of the present invention is to provide a method for preparing spinel OER catalysts based on machine learning. The method is based on experimental data to assist in the design of spinel OER catalysts, making full use of the unique advantages of machine learning in identifying the relationship between overpotential and catalyst composition, synthesis and reaction parameters to establish the correlation between composition-structure-process-performance more quickly and economically. The method is low-cost, simple and efficient, does not require experiments, and is pollution-free.

[0006] The objective of the present invention is achieved through the following technical solutions: A method for preparing a spinel OER catalyst based on machine learning assistance, the method comprising the following steps: Step 1: Through literature research, various conditions such as the composition, support, synthesis method, calcination temperature, etc. of spinel OER catalysts are searched from the literature and included in the data set as input variables; Step 2: Include the overpotential as an output variable in the data set; Step 3: Before making model predictions, preprocess the data set and construct descriptors; Step 4: Randomly divide the data set samples obtained in step 3 into training set and test set; Step 5: Use multiple algorithms to build the model; Step 6: Evaluate the quality of the model by calculating the mean squared error (MSE) and the mean absolute error (MAE), and select the model parameters with the smallest MSE and MAE as the best model.

[0007] A method for preparing spinel OER catalysts assisted by machine learning, wherein all data processes of the method are analyzed based on Python.

[0008] Advantages and beneficial effects of the present invention: (1) The present invention uses a computer system to assist in the design of spinel OER catalysts through machine learning algorithms. The method is simple and fast. The atomic properties and structural parameters of all samples are quickly obtained using the spinel OER catalyst sample data collected from the literature; the converted data is used as the input variable and the overpotential is used as the output variable. The machine learning model is used to obtain the calculation results in just a few seconds; the calculation process is convenient and efficient; the overpotential of the spinel OER catalyst can be determined in advance through the prediction model, and samples that meet the requirements can be selected for experimental verification, which can improve the experimental efficiency and avoid blindness.

[0009] (2) The present invention overcomes the shortcomings of traditional experiments and uses theory and calculation to assist in the design of the overpotential of spinel OER catalysts. It has low cost, simple method, easy implementation, and is suitable for popularization and application.

[0010] (3) The online prediction method of the present invention does not involve any experiments or use any chemicals in the entire process, does not generate any chemical pollution, and is in line with the concept of green environmental protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 The spinel catalyst (AB 2 O 4 ) Data visualization of eigenvalues; Figure 2 The current density is 10 mA / cm2 in Example 1 of the present invention. 2 The data distribution histogram of the target value overpotential when ; Figure 3 It is a Pearson correlation heat map between the eigenvalues ​​of Example 1 of the present invention; Figure 4 is the MAE of different models of Examples 1, 2, and 3 of the present invention; FIG5( a ) is the MSE of different models of Examples 1, 2, and 3 of the present invention; FIG5( b ) is a scatter plot of actual values ​​and predicted values ​​in the training set and the test set based on the SVM model in Example 4 of the present invention; Figure 6 This is the SHAP diagram of Example 4 of the present invention. DETAILED DESCRIPTION

[0012] The present invention will be further described below in conjunction with the embodiments shown in the accompanying drawings. Example 1

[0013] A method for designing a spinel OER catalyst based on machine learning, comprising the following steps: Step 1: Through literature research, various conditions such as the composition, support, synthesis method, calcination temperature, etc. of spinel OER catalysts are searched from the literature and included in the data set as input variables; Step 2: Include the overpotential as an output variable in the data set; Step 3: Before making model predictions, preprocess the data set and construct descriptors; Step 4: Divide the data set samples obtained in step 3 into training set and test set according to the ratio of 7:3; Step 5: Use the support vector machine (SVM) algorithm to build the model; Step 6: Calculate the mean squared error (MSE) and mean absolute error (MAE) of the model, which are 182.7 and 20.6 respectively. Example 2

[0014] This embodiment is basically the same as Embodiment 1, with the following special features: The data set samples obtained in step 3 of Example 1 were divided into a training set and a test set at a ratio of 8:2, and the model was constructed using the Random Forest algorithm. The mean squared error (MSE) and mean absolute error (MAE) of the model were calculated to be 190.3 and 21.0, respectively. Example 3

[0015] This embodiment is basically the same as Embodiment 1, with the following special features: The data set samples obtained in step 3 of Example 1 were divided into a training set and a test set at a ratio of 6:4, and a regression tree algorithm was used to build a model. The mean square error (MSE) and mean absolute error (MAE) of the model were calculated to be 193.8 and 22.1, respectively. Example 4

[0016] According to the MSE and MAE calculated in Examples 1, 2, and 3, SVM was selected as the optimal model, the optimal model was retrained using the optimized parameters on the training set, and a SHAP diagram between the characteristic values ​​and overpotential of the spinel OER catalyst was established.

[0017] In summary, the support vector machine (SVM) model of the present invention has the best fitting effect, with an average mean square error of 182.7 and an average absolute error of 20.6. This method can quickly design spinel AB 2 O 4 Catalysts can screen out the important features that affect the catalytic performance of spinel OER catalysts, greatly reducing the time and cost of screening catalysts.

[0018] The above describes the embodiments of the present invention in conjunction with the accompanying drawings, but the present invention is not limited to the above embodiments, and various changes can be made according to the purpose of the invention of the present invention. Any changes, modifications, substitutions, combinations or simplifications made according to the spirit and principle of the technical solution of the present invention should be equivalent replacement methods. As long as they meet the purpose of the invention of the present invention and do not deviate from the technical principles and inventive concepts of the method for assisted design of spinel OER catalysts based on machine learning, they belong to the protection scope of the present invention.

Claims

1. A method for preparing spinel OER catalysts based on machine learning, characterized in that: The specific steps of the method are as follows: (1) Through literature research, the composition, support, synthesis method, calcination temperature and other conditions of spinel OER catalysts were searched from the literature and included as input variables in the data set; (2) Include the overpotential as an output variable in the data set; (3) Before making model predictions, the dataset is preprocessed and descriptors are constructed; (4) Randomly divide the data set samples obtained in step (3) into training set and test set; (5) Use multiple algorithms to build models; (6) The reliability of the model is evaluated by calculating the mean squared error (MSE) and the mean absolute error (MAE), and the model parameters with the smallest MSE and MAE are selected as the best model.

2. The method for preparing spinel OER catalysts based on machine learning according to claim 1, characterized in that: The data flow is based on analysis using python.

Images