Machine learning method and system for auxiliary construction of sludge carbon active sites
By integrating Extra Tree and XGBoosting models, the accuracy of the prediction of active site of sewage peat catalysts is solved, accurate prediction and efficient screening under different conditions are achieved, experimental costs are reduced, and bottlenecks in active site preparation are broken. New insights are provided.
Patent Information
- Application Number
- CN202411276017.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-12
- Publication Date
- 2024-12-06
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The prior art is difficult to accurately predict and regulate the active sites of sewage peat catalysts, especially under different pyrolysis parameters, which affects the precise construction and efficient screening of active sites.
Using integrated Extra Tree and XGBoosting machine learning models, combined with big data analysis, quantitatively evaluate the contribution of sludge raw material composition and pyrolysis parameters to active sites through pre-processing and model evaluation, and optimize the prediction model to accurately predict sludge catalysts active site.
It realizes accurate prediction of the active sites of pellet catalysts under different conditions, reduces experimental costs, breaks the bottleneck of the precise preparation of pyridine nitrogen active sites on the surface of Fe pellet, and provides new insights into the active sites of pellets. The construction of
Smart Images

Figure CN119092007A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of active site research of sewage sludge charcoal, and more particularly to a machine learning method and system for assisting in constructing active sites of sewage sludge charcoal. Background Art
[0002] Thermal treatment converts sewage sludge (SS) waste into functional carbon materials, providing an efficient and environmentally friendly method for high-value sludge recovery. The development of highly active catalysts is the key to improving the efficiency of Fenton-like catalytic degradation. In addition, the active sites of SS carbon-based functional materials are the key factors affecting their activity. Pyridinic N, as an important active site in sludge carbon catalysts, can not only improve the catalyst activity by enhancing the active center, but also affect the overall performance of the catalyst by adjusting the electronic structure of the catalyst surface. It is worth noting that Fe in SS raw materials affects the formation of pyridinic N active sites. During the pyrolysis process, the iron oxides in the raw materials first catalyze the conversion of protein nitrogen in the coke into nitrogen-containing functional groups (pyrrolic nitrogen, quaternary nitrogen, nitrogen oxides, etc.), fixing the nitrogen in the coke. Then, as the pyrolysis temperature increases, α-Fe is formed and reacts with pyridinic N in the semi-coke to further form Fe X N. As the pyrolysis temperature increases, Fe X N is decomposed and converted into N2. Secondly, Fe promotes the formation of structural defects on the surface of sludge carbon, which are excellent sites for the formation of active sites. Therefore, Fe plays an important role in the process of constructing catalyst active sites. In addition, under different pyrolysis parameters (heating rate, pyrolysis temperature and atmosphere), the SS carbonization process involves complex chain reactions between multiple components. Under existing scientific conditions, it is difficult to observe the changes in intermediate states and functional groups during the pyrolysis process, which has become a major obstacle to clarifying the formation mechanism of active sites, and thus affects the precise regulation of active sites.
[0003] Through artificial intelligence, with the help of big data analysis and machine learning algorithms, hidden information and complex nonlinear relationships can be discovered. However, there are some problems with traditional models. For example, random forests are not effective for high-dimensional sparse data, and XGBoost is very sensitive to outliers and noise. It is worth noting that Extra Tree is a variant based on random forests. It generates multiple decision trees by randomly selecting features and data subsets, and can handle high-dimensional data and noise well. Therefore, the present invention integrates the two models to achieve complementarity: the strong predictive ability of XGBoost is combined with the robustness of Extra Tree, which can process different types of data and improve the overall generalization ability of the model, thereby showing better stability and accuracy on different data sets. The prediction results of the two models are weighted averaged to obtain more accurate prediction results than a single model. Therefore, there is an urgent need for a machine learning method and system for assisting the construction of active sites of sludge charcoal, accurately predicting the key active sites on the surface of sludge charcoal catalysts and high-throughput screening of the construction conditions of the sites. Summary of the invention
[0004] In view of this, the present invention provides a machine learning method and system for assisting the construction of active sites in sludge charcoal. By integrating big data analysis and machine learning algorithms, the complex relationship between the raw material composition and preparation parameters of sludge-based catalysts and the composition of pyridine nitrogen sites is constructed.
[0005] In order to achieve the above object, the present invention adopts the following technical solution:
[0006] In one aspect, the present invention provides a machine learning method for assisting in constructing active sites of sludge charcoal, comprising the following steps:
[0007] Collect sludge-based catalyst parameters and construct a sludge-based catalyst parameter data set for predicting active sites of sludge carbon-based catalysts;
[0008] Preprocessing the data in the sludge-based catalyst parameter data set, and dividing the preprocessed sludge-based catalyst parameter data set into a training set and a test set;
[0009] The training set is input into multiple machine learning regression algorithm models for training, and the prediction performance of multiple machine learning regression algorithm models is evaluated using the test set to select the best model for the prediction performance of each active site;
[0010] The best model for predicting the performance of each active site was analyzed, and the contribution of sludge raw material composition and pyrolysis parameters to the prediction of active sites was quantitatively evaluated. The optimal sludge raw material and preparation condition parameters of the best model were selected based on the contribution.
[0011] The sludge raw materials and preparation condition parameters are input into the integrated model to predict the active sites of the sludge charcoal, and the active site prediction results of the sludge charcoal are obtained.
[0012] Preferably, the sludge-based catalyst parameters are collected, including:
[0013] The input features of the dataset include: C%, N%, O%, Fe%, Pyrolysis temperature, Heating rate and Pyrolysis time and 1 output variable pyridinic N.
[0014] Preferably, data preprocessing selects standardization as a preprocessing method.
[0015] Preferably, the machine learning regression algorithm model includes:
[0016] SVM, KNeighbors, Bagging, Decision Tree, Extra Tree, XGBoosting, ARD Regression, Bayesian Ridge, and Random Forest.
[0017] Preferably, calculate R for the training and test datasets 2 and RMSE values to evaluate the prediction accuracy of the model; and screen decision tree, Extra Tree, XGBoosting, ARD regression, and Bayesian Ridge machine learning algorithms for further optimization.
[0018] Preferably, the Extra Tree and XGBoosting machine learning algorithms are integrated and optimized using Bayesian optimization.
[0019] Preferably, the integrated model of active site prediction performance is analyzed using feature importance and SHAP algorithm, and the contribution of sludge raw material composition and pyrolysis parameters to active site prediction is evaluated. The optimal pyrolysis time, pyrolysis temperature and heating rate conditions for pyridine N generation are obtained through partial dependence diagram analysis.
[0020] On the other hand, the present invention also provides a machine learning system for assisting the construction of active sites of sludge charcoal, which is used to implement any of the above-mentioned machine learning methods for assisting the construction of active sites of sludge charcoal, comprising a data acquisition unit, a preprocessing unit, a model evaluation unit, a contribution evaluation unit and an active site prediction unit connected in sequence:
[0021] A data acquisition unit, which collects sludge-based catalyst parameters and constructs a sludge-based catalyst parameter data set for predicting active sites of sludge carbon-based catalysts;
[0022] A preprocessing unit, used for preprocessing data in the sludge-based catalyst parameter data set, and dividing the preprocessed sludge-based catalyst parameter data set into a training set and a test set;
[0023] A model evaluation unit, used to input the training set into multiple machine learning regression algorithm models for training, and use the test set to evaluate the prediction performance of the multiple machine learning regression algorithm models, and select the best model for the prediction performance of each active site;
[0024] The contribution evaluation unit is used to analyze the best model for predicting the performance of each active site, and quantitatively evaluate the contribution of sludge raw material composition and pyrolysis parameters to the prediction of active sites, and select the optimal sludge raw material and preparation condition parameters of the best model according to the contribution;
[0025] The active site prediction unit is used to input sludge raw materials and preparation condition parameters into the integrated model to predict the active sites of sludge charcoal, and obtain the active site prediction results of the sludge charcoal.
[0026] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a machine learning method and system for assisting in the construction of active sites of sludge charcoal, which has the following beneficial effects:
[0027] (1) The method for predicting active sites of sludge charcoal of the present invention integrates Extra Tree and XGBoosting machine learning models, and only needs to input seven parameters: C%, N%, O%, Fe%, heating rate, pyrolysis temperature and pyrolysis time. The prediction model R 2 The value reaches 0.80, and the active sites can be predicted directly based on the raw material composition and preparation parameters. There is no need for large-scale mechanized experiments, which reduces the cost of manpower and material resources;
[0028] (2) The model successfully constructed by the present invention can accurately screen the optimal pyrolysis time, heating rate and pyrolysis temperature for the generation of pyridine N. The pyrolysis temperature of 500-600°C, the heating rate of 5-10°C / min, and the pyrolysis time of about 150 min are conducive to the generation of pyridine N sites;
[0029] (3) The method for predicting active sites of sewage sludge charcoal described in the present invention breaks the bottleneck of the precise preparation of pyridinic nitrogen active sites on the surface of Fe-containing sewage sludge charcoal. The developed method for predicting active sites of sewage sludge charcoal provides new insights into the application of machine learning in the field of active sites of sewage sludge charcoal. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.
[0031] Figure 1 A flow chart of a machine learning method for assisting in constructing active sites of sludge charcoal disclosed in the present invention;
[0032] Figure 2 A structural diagram of a machine learning system for assisting in constructing active sites of sludge charcoal disclosed in the present invention;
[0033] Figure 3 It is a Pearson correlation matrix analysis diagram disclosed in the present invention;
[0034] Figure 4 The model disclosed in the present invention is optimized by Bayesian 2 value;
[0035] Figure 5 It is the RMSE value after Bayesian optimization of the model disclosed in the present invention. DETAILED DESCRIPTION
[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0037] The embodiment of the present invention relates to a method for constructing active sites of sludge carbon with the aid of a machine learning model. By integrating big data analysis and machine learning algorithms, a complex relationship between the raw material composition and preparation parameters of sludge-based catalysts and the composition of pyridine nitrogen sites is constructed. The development of a new model can accelerate the screening of SS carbon-based catalysts. The proposed new model can be extended to the prediction and application of other active sites on the surface of Fe-containing SS carbon-based catalysts.
[0038] Reference Figure 1 As shown, the embodiment of the present invention discloses a machine learning method for assisting in constructing active sites of sludge charcoal, comprising the following steps:
[0039] Collect sludge-based catalyst parameters and construct a sludge-based catalyst parameter data set for predicting active sites of sludge carbon-based catalysts;
[0040] Preprocessing the data in the sludge-based catalyst parameter data set, and dividing the preprocessed sludge-based catalyst parameter data set into a training set and a test set;
[0041] The training set is input into multiple machine learning regression algorithm models for training, and the prediction performance of multiple machine learning regression algorithm models is evaluated using the test set to select the best model for the prediction performance of each active site;
[0042] The best model for predicting the performance of each active site was analyzed, and the contribution of sludge raw material composition and pyrolysis parameters to the prediction of active sites was quantitatively evaluated. The optimal sludge raw material and preparation condition parameters of the best model were selected based on the contribution.
[0043] The sludge raw materials and preparation condition parameters are input into the integrated model to predict the active sites of the sludge charcoal, and the active site prediction results of the sludge charcoal are obtained.
[0044] Furthermore, the sludge-based catalyst parameters were collected, including:
[0045] The input features of the dataset include: C%, N%, O%, Fe%, Pyrolysis temperature, Heating rate and Pyrolysis time and 1 output variable pyridinic N.
[0046] Specifically, data were collected from different articles using Google Scholar and Web of Science search engines, with the keywords “sludge-based catalyst”, “XPS”, “raw material”, “pyrolysis temperature”, “pyrolysis time and heating rate”. The retrieved literature was reviewed and the data that met the requirements were selected. A total of 878 peer-reviewed articles were reviewed, and 92 data points were retrieved and used to construct the data set.
[0047] Furthermore, data preprocessing selected standardization as the preprocessing method.
[0048] Specifically, refer to Figure 3 As shown in the figure, XGBoosting method is selected as a sample to compare the effects of standardization, PCA and no preprocessing. All other hyperparameters are set to the same value. Standardization preprocessing achieves the highest R 2 and the lowest RMSE. The RMSE value achieved by standardization preprocessing reached the lowest value, which was 0.1386. Therefore, standardization was selected as the preprocessing method.
[0049] Furthermore, the machine learning regression algorithm model includes:
[0050] SVM, KNeighbors, Bagging, Decision Tree, Extra Tree, XGBoosting, ARD Regression, Bayesian Ridge, and Random Forest.
[0051] Furthermore, we calculate the R of the training and test datasets. 2 and RMSE values to evaluate the prediction accuracy of the model; select decision tree, Extra Tree, XGBoosting, ARD regression and Bayesian Ridge machine learning algorithms for further optimization. 2 The RMSE values are all less than 0.1.
[0052] Furthermore, the Extra Tree and XGBoosting machine learning algorithms are integrated and optimized with Bayesian optimization. The Extra Tree and XGBoosting machine learning algorithms are integrated and optimized with Bayesian optimization. Figure 4 As shown, the optimized model R 2 The value is 0.80, refer to Figure 5 As shown in the figure, the RMSE value is 0.1285. When the Fe feature is deleted, the model R 2 The value dropped to 0.61, and the RMSE value increased to 0.1421. The optimization parameters are as follows:
[0053] (1) Set the xgb__n_estimator parameter to 309 and the etr__n_estimator parameter to 1092.
[0054] (2) Set the xgb__max_depth parameter to 21 and the etr__max_depth parameter to 12.
[0055] (3) Set the xgb__learning_rate parameter to 0.58.
[0056] (4) Set xgb__Subsample parameter: 0.81.
[0057] (5) Set xgb__colsample_bytree parameter: 0.98.
[0058] Specifically, (1) Adjust the parameters of xgb__n_estimators and etr__n_estimator: Increasing the number of trees usually improves model performance because the model will be more robust and reduce randomness. However, too many trees may increase computing time while having limited performance improvements. Set the parameters of xgb__n_estimators to 300-1000. Set the parameters of etr__n_estimator to 400-1100.
[0059] (2) Adjust the xgb__max_depth and etr__max_depth parameters: The maximum depth of the tree is one of the key determinants of model complexity. Increasing the maximum depth allows the model to learn more complex patterns, but may also lead to overfitting. Set the max_depth parameter: Set the max_depth parameter to 10-30. Set the etr__max_depth parameter to 10-30.
[0060] (3) Adjust the xgb__learning_rate parameter: A lower learning rate means that the model learns the data pattern more slowly, usually requiring more n_estimators parameters for training, which can improve the final performance of the model. Set the learning_rate parameter: 0.01-1.00.
[0061] (4) Adjust the xgb__Subsample parameter: the sample ratio for training each tree. This is a technique to reduce variance, which can be achieved by using only part of the data to train each tree. Set the Subsample parameter: 0.8-1.0.
[0062] (5) Adjust xgb__colsample_bytree parameter: the proportion of features that can be used by each tree. Each tree is trained by randomly selecting some features. colsample_bytree parameter: 0.8-1.0.
[0063] Furthermore, the feature importance and SHAP algorithms were used to analyze the integrated model of active site prediction performance, and the contribution of sludge raw material composition and pyrolysis parameters to active site prediction was evaluated. The optimal pyrolysis time, pyrolysis temperature, and heating rate conditions for pyridine N generation were obtained through partial dependence plot analysis.
[0064] Specifically, the best model for predicting the performance of each active site is analyzed, and the contribution of sludge raw material composition and pyrolysis parameters to the prediction of active sites is quantitatively evaluated. The optimal sludge raw material and preparation condition parameters of the best model are selected according to the contribution. The best model of pyridinic N active site is analyzed using feature importance and SHAP. For pyridinic N active site, the order of feature importance is: C%>pyrolysis temperature>N%>holding time>O%>Fe%>heating rate. In addition, pyrolysis temperature of 500-600℃, heating rate of 5-10℃ / min, and pyrolysis time of about 150min are conducive to the generation of pyridinic N site. Input sludge carbon's C% (24), N% (3.1), O% (35), Fe% (30), heating rate (℃ / min) (20), pyrolysis time (min) (60), pyrolysis temperature (℃) (700). Then the prediction result is displayed: the content of pyridinic N site (0.257%).
[0065] Specific implementation method 2: A method for predicting pyridine N active sites in sludge carbon in this implementation method is carried out by the following steps:
[0066] 1. Data collection: Search for sludge pyrolysis carbon-related content on web of science and Google Scholar, and enter the search keywords: including "sludge-based catalyst", "XPS", "raw material", "pyrolysis temperature", "pyrolysis time and heating rate", etc. Then review the retrieved literature to see if it meets the data we need. A total of 878 peer-reviewed literature were reviewed, and 92 data points were retrieved and used to construct the data set.
[0067] 2. Data preprocessing: Three different preprocessing methods were selected, namely standardization, PCA and no preprocessing. The correlation of input features was analyzed by Pearson correlation matrix (e.g. Figure 1 ). Here, the XGBoosting model is selected for processing to compare the impact of preprocessing differences. All other hyperparameters are set to the same value, and the RMSE value achieved by the standardized preprocessing is the lowest, which is 0.1486. Therefore, standardization is selected as the preprocessing method.
[0068] 3. Model screening:
[0069] Using the dataset as input, the performance of nine commonly used ML algorithms (SVM, KNeighbors, Bagging, Decision Tree, Extra Tree, XGBoosting, ARD regression, Bayesian Ridge, and Random Forest) in predicting active sites of sludge-based catalysts was evaluated and compared. 2The RMSE values are all less than 0.1.
[0070] 4. Model Optimization
[0071] Integrate Extra Tree and XGBoosting machine learning algorithms to further perform Bayesian optimization. The optimized model R 2 The value is 0.73, the RMSE value is 0.1314, and the optimization parameters are as follows:
[0072] (1) Set the xgb__n_estimator parameter to 360 and the etr__n_estimator parameter to 1050.
[0073] (2) Set the xgb__max_depth parameter to 25 and the etr__max_depth parameter to 12.
[0074] (3) Set the xgb__learning_rate parameter to 0.58.
[0075] (4) Set xgb__Subsample parameter: 0.81.
[0076] (5) Set xgb__colsample_bytree parameter: 0.98.
[0077] V. Model Analysis
[0078] The feature importance and SHAP were used to analyze the optimal model of the pyridine N active site. For the pyridine N active site, the order of feature importance was: C%>pyrolysis temperature>N%>holding time>O%>Fe%>heating rate. In addition, the pyrolysis temperature of 500-600℃, the heating rate of 5-10℃ / min, and the pyrolysis time of about 150min were beneficial to the formation of pyridine N sites.
[0079] 6. System Prediction
[0080] The active sites of sludge charcoal were predicted by inputting the elemental composition of sludge charcoal, C% (24), N% (3.1), O% (32), Fe% (0), heating rate (℃ / min) (15), pyrolysis time (min) (120), pyrolysis temperature (℃) (600). The predicted results were then displayed: the content of pyridinic N sites (0.224%).
[0081] Reference Figure 2As shown, the present invention also provides a machine learning system for assisting in the construction of active sites of sludge charcoal, the system is used to implement any of the above-mentioned machine learning methods for assisting in the construction of active sites of sludge charcoal, including a data acquisition unit, a preprocessing unit, a model evaluation unit, a contribution evaluation unit and an active site prediction unit connected in sequence:
[0082] A data acquisition unit, which collects sludge-based catalyst parameters and constructs a sludge-based catalyst parameter data set for predicting active sites of sludge carbon-based catalysts;
[0083] A preprocessing unit, used for preprocessing data in the sludge-based catalyst parameter data set, and dividing the preprocessed sludge-based catalyst parameter data set into a training set and a test set;
[0084] A model evaluation unit, used to input the training set into multiple machine learning regression algorithm models for training, and use the test set to evaluate the prediction performance of the multiple machine learning regression algorithm models, and select the best model for the prediction performance of each active site;
[0085] The contribution evaluation unit is used to analyze the best model for predicting the performance of each active site, and quantitatively evaluate the contribution of sludge raw material composition and pyrolysis parameters to the prediction of active sites, and select the optimal sludge raw material and preparation condition parameters of the best model according to the contribution;
[0086] The active site prediction unit is used to input sludge raw materials and preparation condition parameters into the integrated model to predict the active sites of sludge charcoal, and obtain the active site prediction results of the sludge charcoal.
[0087] The active site prediction module is used to obtain the data of the optimal input characteristics of the biochar, and input the data of the optimal input characteristics into the optimal model to predict the active sites of the biochar, so as to obtain the active site prediction results of the biochar.
[0088] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.
[0089] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A machine learning method for assisting the construction of active sites of sludge charcoal, characterized in that: The following steps are involved: Collect sludge-based catalyst parameters and construct a sludge-based catalyst parameter data set for predicting active sites of sludge carbon-based catalysts; Preprocessing the data in the sludge-based catalyst parameter data set, and dividing the preprocessed sludge-based catalyst parameter data set into a training set and a test set; The training set is input into multiple machine learning regression algorithm models for training, and the prediction performance of multiple machine learning regression algorithm models is evaluated using the test set to select the best model for the prediction performance of each active site; The best model for predicting the performance of each active site was analyzed, and the contribution of sludge raw material composition and pyrolysis parameters to the prediction of active sites was quantitatively evaluated. The optimal sludge raw material and preparation condition parameters of the best model were selected based on the contribution. The sludge raw materials and preparation condition parameters are input into the integrated model to predict the active sites of the sludge charcoal, and the active site prediction results of the sludge charcoal are obtained.
2. A machine learning method for assisting in the construction of active sites of sludge charcoal according to claim 1, characterized in that: Collect sludge-based catalyst parameters including: The input features of the dataset include: C%, N%, O%, Fe%, Pyrolysis temperature, Heating rate and Pyrolysis time and 1 output variable pyridinic N.
3. A machine learning method for assisting the construction of active sites of sludge charcoal according to claim 1, characterized in that: Data preprocessing Standardization was selected as the preprocessing method.
4. A machine learning method for assisting the construction of active sites of sludge charcoal according to claim 1, characterized in that: Machine learning regression algorithm models include: SVM, KNeighbors, Bagging, Decision Tree, Extra Tree, XGBoosting, ARD Regression, Bayesian Ridge, and Random Forest.
5. A machine learning method for assisting in the construction of active sites of sludge charcoal according to claim 4, characterized in that: Compute R for the training and test datasets 2 and RMSE values to evaluate the prediction accuracy of the model; and screen decision tree, Extra Tree, XGBoosting, ARD regression, and Bayesian Ridge machine learning algorithms for further optimization.
6. A machine learning method for assisting in the construction of active sites of sludge charcoal according to claim 5, characterized in that: Perform integrated optimization and Bayesian optimization on Extra Tree and XGBoosting machine learning algorithms.
7. A machine learning method for assisting the construction of active sites of sludge charcoal according to claim 1, characterized in that: The feature importance and SHAP algorithms were used to analyze the integrated model of active site prediction performance, and the contribution of sludge raw material composition and pyrolysis parameters to active site prediction was evaluated. The optimal pyrolysis time, pyrolysis temperature and heating rate conditions for pyridine N generation were obtained through partial dependence diagram analysis.
8. A machine learning system for assisting in the construction of active sites of sludge charcoal, characterized in that: A machine learning method for assisting in the construction of active sites of sludge charcoal as claimed in any one of claims 1 to 7, comprising a data acquisition unit, a preprocessing unit, a model evaluation unit, a contribution evaluation unit and an active site prediction unit connected in sequence: A data acquisition unit, which collects sludge-based catalyst parameters and constructs a sludge-based catalyst parameter data set for predicting active sites of sludge carbon-based catalysts; A preprocessing unit, used for preprocessing data in the sludge-based catalyst parameter data set, and dividing the preprocessed sludge-based catalyst parameter data set into a training set and a test set; A model evaluation unit, used to input the training set into multiple machine learning regression algorithm models for training, and use the test set to evaluate the prediction performance of the multiple machine learning regression algorithm models, and select the best model for the prediction performance of each active site; The contribution evaluation unit is used to analyze the best model for predicting the performance of each active site, and quantitatively evaluate the contribution of sludge raw material composition and pyrolysis parameters to the prediction of active sites, and select the optimal sludge raw material and preparation condition parameters of the best model according to the contribution; The active site prediction unit is used to input sludge raw materials and preparation condition parameters into the integrated model to predict the active sites of sludge charcoal, and obtain the active site prediction results of the sludge charcoal.
Citation Information
Patent Citations
Biochar active site prediction method and system
CN118016184A
Machine learning-based cement-based material performance optimization method and system
CN118277887A
Method for Predicting Benchmark Value of Unit Equipment Based on XGBoost Algorithm and System thereof
US20230213895A1
Cited By
Precise preparation method of functional group biochar based on machine learning
CN119626321A
