A method for screening and designing doped carbon-based CO2 adsorbents based on multi-scale simulation-driven machine learning.

Abstract

This invention discloses a method for screening and designing doped carbon-based CO2 adsorbents based on multi-scale simulation-driven machine learning. The method is as follows: S1, model construction; S2, slit pore free volume descriptor V. f The process involves: S3, CO2 adsorption characteristic calculation; S4, eigenvalue extraction; S5, machine learning training set construction; S6, machine learning model training; S7, machine learning model screening; S8, configuration prediction of doped carbon-based adsorbents; and S9, guiding experimental synthesis. This invention is the first to introduce a free volume descriptor into a machine learning training model to reflect the size of the adsorption space accessible to CO2 under the coupling mode of micropores and doping sites. This invention uses density functional theory-grand canonical Monte Carlo-molecular dynamics multi-scale calculation data as a training set to drive machine learning, thereby comprehensively considering the complex coupling effects within carbon-based adsorbents and achieving accurate prediction and screening of functional structure grouping patterns for high-performance carbon-based adsorbents.

Description

Technical Field

[0001] This invention belongs to the field of flue gas CO2 adsorption and capture technology, and relates to a method for the directional design of carbon-based adsorbents, specifically a method for screening and designing doped carbon-based CO2 adsorbents based on multi-scale simulation-driven machine learning. Background Technology

[0002] CO2 capture, utilization, and storage (CCUS) is considered the only technological approach for the power industry to achieve near-zero CO2 emissions. Among these processes, CO2 capture is the most expensive and energy-intensive, making it a key factor determining system cost and energy efficiency. Physical adsorption capture of CO2 using porous adsorbents offers advantages such as easy CO2 recovery, harmless adsorbent recycling, and a CO2 adsorption / desorption temperature window that matches the flue gas environment, making it a crucial choice for next-generation low-cost, low-energy CO2 capture technology. Among various adsorbent materials, carbon materials possess advantages such as wide availability of raw materials, low cost, and good tolerance to flue gas environments, giving them significant application potential in the field of flue gas CO2 adsorption and capture.

[0003] The key to achieving efficient CO2 adsorption and capture lies in the precise directional design and preparation of high-performance adsorbents. For carbon-based adsorbents, the microporous structure and doping sites are crucial factors determining their CO2 adsorption performance. Currently, researchers mainly use experimental trial-and-error methods to adjust carbonization-activation process parameters, individually controlling the pore size or doping environment of carbon materials to attempt to prepare high-performance adsorbent materials. However, in practical applications, the micropores and doping sites within carbon-based adsorbents are interdependent, and their combination significantly impacts CO2 adsorption. Simultaneously, the presence of impurities such as N2 and H2O in flue gas, and the interaction between these impurities and the carbon-based adsorbent, also profoundly affects the CO2 adsorption process. This interdependence between the micropore space and doping sites in carbon-based adsorbents, coupled with the complexity of flue gas composition, generates a complex nonlinear coupling effect on CO2 adsorption that transcends the influence of individual pore size or doping. This means that it is difficult to achieve precise directional design of carbon-based CO2 adsorbents through conventional trial-and-error experiments, necessitating a new method for directional design of carbon-based adsorbents that considers the aforementioned coupling effects.

[0004] Recently, high-throughput computational screening methods based on grand canonical Monte Carlo simulations and machine learning have become a new paradigm guiding the development of high-performance adsorbents. Currently, researchers are using machine learning methods driven by grand canonical Monte Carlo simulation data to design ordered porous adsorbents such as MOFs (Nature 2019, 576, 253-256, CN118609683A, CN115169220A). However, for carbon-based adsorbents used for CO2 adsorption and capture in flue gas, the highly coupled structural characteristics of micropore space and doping sites, combined with the complex composition of flue gas, often simultaneously induce multi-scale effects such as surface polarization, occupancy of molecularly accessible space, and shift in optimal adsorption pore size. Machine learning models based solely on single-scale simulation data cannot accurately reflect these multi-scale effects, hindering the screening and targeted design of high-performance carbon-based adsorbents. Summary of the Invention

[0005] This invention provides a method for screening and designing doped carbon-based CO2 adsorbents based on multi-scale simulation-driven machine learning. This method uses a machine learning approach that couples density functional theory calculations, grand canonical Monte Carlo simulations, and molecular dynamics simulation data to screen and design doped carbon-based adsorbents.

[0006] The objective of this invention is achieved through the following technical solution:

[0007] A method for screening and designing doped carbon-based CO2 adsorbents based on multi-scale simulation-driven machine learning includes the following steps:

[0008] Step S1, Model Construction: Construct a slit-pore carbon model that couples typical doping sites with micropores of different sizes, and optimize the geometric configuration of the slit-pore carbon model using density functional theory.

[0009] Step S2, Slit Hole Free Volume Descriptor V f Extraction: Calculation of the free volume descriptor V of the slit aperture under the coupling between doped sites and micropore space. f ;

[0010] Step S3, CO2 adsorption characteristics calculation: The thermodynamic and kinetic characteristics of CO2 adsorption in the adsorbate are calculated using the giant canonical Monte Carlo method and the molecular dynamics method, respectively.

[0011] Step S4, Feature Extraction: Based on the multi-scale simulation calculation results of Step S3, extract the feature values ​​that affect CO2 adsorption performance: pore size, nitrogen atom content, oxygen atom content, electronegativity, dipole moment, charge, H2O volume fraction, and N2 volume fraction;

[0012] Step S5: Machine learning training set construction: Divide the density functional-grand canonical Monte Carlo-molecular dynamics multiscale computation data into training set and test set, use the training set to train the machine learning model, and use the test set to evaluate the machine learning performance of the model.

[0013] Step S6, Machine Learning Model Training: Using the slit aperture free volume descriptor V f Using the feature values ​​extracted in step S4 as input variables and CO2 adsorption as the target variable, the machine learning model is iteratively trained using the training set, and the input slit orifice free volume descriptor V is analyzed. f The feature values ​​extracted in step S4 are normalized to improve the generalization ability of the model.

[0014] Step S7, Machine Learning Model Selection: Based on prediction accuracy, identify the optimal machine learning model suitable for CO2 adsorption prediction, and based on feature interaction analysis, identify the feature values ​​that dominate CO2 adsorption.

[0015] Step S8: Prediction of the configuration of doped carbon-based adsorbents: Based on the characteristic values ​​of dominant CO2 adsorption, predict the configuration of doped carbon-based adsorbents with optimal CO2 adsorption performance.

[0016] Step S9: Guided experimental synthesis: Based on the predicted configuration of the doped carbon-based adsorbent with optimal CO2 adsorption performance, high-performance carbon-based adsorbent materials suitable for CO2 capture in flue gas are synthesized in a directional manner.

[0017] Compared with the prior art, the present invention has the following advantages:

[0018] (1) Compared with traditional experimental trial-and-error methods, the design method proposed in this invention can more accurately and efficiently guide the preparation of high-performance carbon-based adsorbents. Currently, the design and preparation of carbon-based adsorbents mainly rely on experimental trial-and-error methods to control the parameters of the carbonization-activation process to obtain optimal preparation conditions. Such methods are time-consuming and costly. More importantly, the multi-scale functional units within carbon-based adsorbents are interdependent, and it is difficult to decouple the complex coupling effects between different functional units through simple trial-and-error experiments, thus failing to achieve the precise and targeted preparation of high-performance adsorbents. This invention introduces machine learning methods to decouple the nonlinear superposition effects between multi-dimensional variables, effectively determining the key characteristics that determine CO2 adsorption performance, thereby realizing the targeted design and preparation of high-performance doped carbon-based adsorbents.

[0019] (2) The machine learning training set proposed in this invention is based on multi-scale simulation results. Compared with conventional machine learning models that rely solely on grand canonical Monte Carlo simulation data, it can more comprehensively reflect the multi-scale effects caused by the coupling between micropore space and doping sites within carbon-based adsorbents, achieving high-precision prediction and screening. Currently, machine learning prediction models for porous adsorbents mainly use single-scale grand canonical Monte Carlo simulation results as the training set. However, actual carbon-based adsorbents have amorphous structural characteristics, and the interdependence between micropore space and doping sites leads to complex multi-scale effects. Machine learning methods that rely solely on grand canonical Monte Carlo simulations cannot fully consider this multi-scale effect, making it difficult to achieve accurate prediction of adsorbents. This invention proposes using density functional theory-grand canonical Monte Carlo-molecular dynamics multi-scale calculation data as the training set to drive machine learning, thereby comprehensively considering the complex coupling effects within carbon-based adsorbents and achieving accurate prediction and screening of functional structure pairing patterns for high-performance carbon-based adsorbents.

[0020] (3) This invention is the first to propose a free volume descriptor to reflect the steric hindrance effect caused by the coupling between the micropore space and doping sites in doped carbon-based adsorbents. The machine learning model built based on this descriptor significantly improves prediction accuracy compared to traditional models that only consider pore size and surface chemical environment. To address the steric hindrance effect caused by the coupling between doping sites and micropore space in carbon-based adsorbents, this invention introduces the free volume descriptor into the machine learning training model for the first time to reflect the size of the adsorption space accessible to CO2 under the micropore-doping site coupling mode. The results of the examples show that after introducing the free volume feature value into the machine learning model, the predicted value and the actual value of R... 2 The value increased from 0.687 to 0.934, while the root mean square error decreased from 0.597 to 0.253. Based on the model prediction results, a nitrogen-oxygen co-doped carbon-based adsorbent with expanded pore size was prepared, exhibiting a CO2 adsorption capacity as high as 4 mmol g at room temperature and pressure. -1 Compared with carbon-based adsorbents prepared by conventional experimental trial and error methods, the adsorption capacity was increased by 130%. Attached Figure Description

[0021] Figure 1 A flowchart for screening and designing doped carbon-based CO2 adsorbents based on multi-scale simulation-driven machine learning;

[0022] Figure 2 To take into account the training results of machine learning models using free volume descriptors;

[0023] Figure 3 The CO2 adsorption isotherm at room temperature is obtained by preparing an N / O co-doped carbon-based adsorbent with expanded pore size, guided by a machine learning model that considers the free volume descriptor.

[0024] Figure 4The training results of the machine learning model without considering the free volume descriptor;

[0025] Figure 5 The CO2 adsorption isotherm at room temperature of the N / O co-doped carbon-based adsorbent prepared under the guidance of the machine learning model without considering the free volume descriptor. Specific implementation mode

[0026] The technical solution of the present invention will be further described below in conjunction with the drawings, but it is not limited thereto. Any modification or equivalent replacement of the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention shall be covered by the protection scope of the present invention.

[0027] The present invention provides a method for screening and designing doped carbon-based CO2 adsorbents driven by multi-scale simulation, as Figure 1 shown, the method includes the following steps:

[0028] Step S1, model construction: construct a slit pore carbon model coupling typical doping sites and microporous spaces of different sizes, and optimize the geometric configuration of the slit pore carbon model by density functional theory.

[0029] In this step, the size of the slit pore in the slit pore carbon model is 0.5 - 2.0 nm, the doped heteroatoms include individual and combined forms of N, O, B, and S, and 0 < heteroatom concentration ≤ 20 at%.

[0030] Step S2, extraction of the slit pore free volume descriptor V f : Calculate the slit pore free volume descriptor V f .

[0031] In this step, the slit pore free volume descriptor V f is the ratio of the accessible volume of gas molecules in the slit pore to the total volume.

[0032] Step S3, calculation of CO2 adsorption characteristics: Use the grand canonical Monte Carlo method and the molecular dynamics simulation method to calculate the adsorption thermodynamic characteristics and kinetic characteristics of CO2 in the adsorbate respectively.

[0033] In this step, the adsorbate gas components are a mixture of CO2, N2, and H2O, where 0 < CO2 concentration ≤ 100 vol%, 0 ≤ H2O concentration ≤ 10 vol%, the remaining gas is N2, and the environmental temperature is 298 K.

[0034] In this step, the giant canonical Monte Carlo method is specifically set as follows: the Metropolis algorithm is used to randomly select the particle's motion, and the motion distribution in each GCMC simulation is set to 20% exchange, 20% conformation, 40% rotation, and 20% translation. The maximum dimensions of the translation angle and rotation angle are respectively... The step size is 5°, the balance step size and the total step size are 1000000 and 10000000 respectively, and the simulation adopts a fully rigid model.

[0035] In this step, the molecular dynamics simulation method is specifically set as follows: the Dreiding force field based on the Lennard-Jones potential function is used, and the interaction with the potential field generated by the atoms includes the surface of the slit-pore carbon model and the gas; random initial atomic velocities are used, and the Nose-Hoover temperature control system and Maxwell-Boltzmann distribution corresponding to the simulation temperature are used to perform an EMD simulation of 1000 ps in the NVT ensemble with a time step of 1 fs.

[0036] Step S4, Feature value extraction: Based on the multi-scale simulation calculation results of step S3, feature values ​​affecting CO2 adsorption performance are extracted. Feature values ​​include: (1) pore size; (2) nitrogen atom content; (3) oxygen atom content; (4) electronegativity; (5) dipole moment; (6) charge; (7) H2O volume fraction; (8) N2 volume fraction.

[0037] In this step, the electronegativity of the model is calculated using the formula for group electronegativity:

[0038]

[0039] In the formula, N G q represents the total number of atoms in the group, q represents the total atomic charge of the group, and n is the total number of atoms in the model. i It is the number of atoms i in the model. It is the electronegativity of the i atom before bonding.

[0040] In this step, density functional theory is used to calculate the Mulliken charge of the geometrically optimized aromatic carbon clusters;

[0041] In this step, density functional theory is used to calculate the dipole moment of the geometrically optimized aromatic carbon clusters.

[0042] Step S5: Construction of machine learning training set: Based on the multi-scale simulation calculation results of step S3, the training set and test set are divided. The machine learning model is trained using the training set and evaluated using the test set.

[0043] In this step, the ratio of the training set to the test set is 9:1.

[0044] Step S6, Machine Learning Model Training: Using the slit aperture free volume descriptor V f Using the feature values ​​extracted in step S4 as input variables and CO2 adsorption as the target variable, the machine learning model is iteratively trained using the training set, and the input slit orifice free volume descriptor V is analyzed. f The feature values ​​extracted in step S4 are normalized to improve the model's generalization ability.

[0045] In this step, the machine learning model is trained by introducing gradient boosting regression (GBR), extreme gradient boosting (XGBoost), K-nearest neighbor (KNN) algorithm, or random forest (RF) algorithm to train CO2 adsorption capacity.

[0046] Step S7: Machine learning model selection: Based on prediction accuracy, find the optimal machine learning model applicable to CO2 adsorption prediction, and based on feature interaction analysis, find the feature values ​​that dominate CO2 adsorption.

[0047] In this step, the regression coefficient R is used 2 The root mean square error (RMSE) is used as an evaluation metric to identify the optimal machine learning model suitable for CO2 adsorption prediction, where:

[0048] Regression coefficient R 2 The calculation formula is:

[0049]

[0050] In the formula, N represents the number of data points.

[0051] The formula for calculating the root mean square error (RMSE) is:

[0052]

[0053] In the formula, Y i actual and Y i pred These represent the actual value and the predicted value, respectively. This represents the actual average value.

[0054] In this step, the feature interaction analysis method uses the SHapley Additive explanation (SHAP) value to deeply explain the contribution of each feature value. The SHAP value represents the average incremental contribution φj of each feature j, and the SHAP value is calculated using the following formula:

[0055]

[0056] In the formula, M represents the set containing all features; S is any subset of features other than feature j, |S| represents the number of features in set S, and |M| represents the total number of features; f(S) indicates that the model makes predictions only by observing features within set S; f(S∪{j}) represents the predicted value when feature j is included in set s, and |S|! (|M|-|S|-1)! serves as a weight to balance sets of different sizes. The machine learning model is trained using the training set and evaluated using the test set.

[0057] Step S8: Prediction of the configuration of doped carbon-based adsorbents: Based on the characteristic values ​​of dominant CO2 adsorption, predict the configuration of doped carbon-based adsorbents with optimal CO2 adsorption performance.

[0058] Step S9: Guided Experimental Synthesis: Based on the predicted configuration of the doped carbon-based adsorbent with optimal CO2 adsorption performance, high-performance carbon-based adsorbent materials suitable for CO2 capture in flue gas are synthesized in a directional manner, and experimental adsorption data are verified against theoretical prediction data.

[0059] In this step, the directional synthesis method is microwave-assisted activation, which obtains carbon-based adsorbents with the desired pore size and heteroatom distribution by changing the activation conditions.

[0060] Example 1:

[0061] Slit-pore carbon models with different pore sizes ranging from 0.5 to 2.0 nm were constructed, modified with typical N and O heteroatoms, where the heteroatom concentration was ≤20 at%. Density functional theory was used to geometrically optimize the constructed models, extracting the free volume and charge distribution. Grand canonical Monte Carlo simulations and molecular dynamics simulations were employed to calculate the adsorption behavior of CO2 in the constructed models at a partial pressure of 15 kPa and a temperature of 298 K. Based on the obtained adsorption capacity data, machine learning training and prediction were performed using pore size, free volume, nitrogen atom content, oxygen atom content, N2 vol%, H2O vol%, electronegativity, dipole moment, and charge as feature values. The machine learning training results are shown below. Figure 2 As shown, from Figure 2 The R-value can be seen from the predicted value and the actual value. 2 The mean square error is as high as 0.939, and the root mean square error is only 0.243, indicating that the machine learning model driven by multi-scale simulation data can accurately predict the CO2 adsorption performance of carbon-doped adsorbents.

[0062] Example 2:

[0063] Based on the prediction results of the machine learning model driven by multi-scale simulation data in Example 1, N / O co-doped microporous carbon materials with expanded pore sizes were prepared by microwave-assisted KOH activation. Specifically, a certain mass of coal powder, KOH, and different masses of melamine were mixed evenly, dried, and then heated at 200W microwave power for 10 minutes. Residual impurities were washed away with 1M hydrochloric acid to obtain the N / O co-doped microporous carbon material. Physicochemical structural analysis showed that it contained 3 at% N atoms and 5 at% oxygen atoms, with the pore size concentrated at 0.8 nm. The adsorption isotherm of the prepared pore-expanded N / O co-doped microporous carbon at 298 K is shown below. Figure 3 As shown, the adsorption capacity at 1 bar is 4 mmol g. -1 This further demonstrates that a machine learning model driven by multi-scale simulation data can enable the directional design and preparation of high-performance doped carbon-based adsorbents for CO2 adsorption and capture in flue gas.

[0064] Comparative Example 1:

[0065] The difference between this comparative example and Example 1 is that the free volume V is not considered in the machine learning training. f The descriptor was trained and used for machine learning training and prediction using only pore size, nitrogen atom content, oxygen atom content, N2 vol% , H2O vol% , electronegativity, dipole moment, and charge as features. The machine learning training results are as follows: Figure 4 As shown, from Figure 4 The R-squared value between the predicted and actual values ​​can be found in the data. 2 The mean square error is only 0.687, while the root mean square error is as high as 0.597, indicating that machine learning models that do not consider the free volume descriptor cannot accurately predict the CO2 adsorption performance of doped carbon-based adsorbents.

[0066] Comparative Example 2:

[0067] Based on the machine learning model prediction results of Comparative Example 1, N / O co-doped microporous carbon materials were prepared by ammonia physical activation. Specifically, deashed coal was placed in a horizontal tubular furnace, and ammonia was injected at a rate of 140 μL / min using a jet pump, with nitrogen gas at a rate of 80 mL / min as the gas carrier. After activation at 900℃ for 1 h, nitrogen-doped microporous carbon was obtained. Physicochemical structural analysis showed that its N and O heteroatom content and distribution were the same as in Example 2, while the pore size was concentrated at 0.7 nm. The adsorption isotherm of the prepared pore-expanded N / O co-doped microporous carbon at 298 K is shown below. Figure 5 As shown, the adsorption capacity at 1 bar is 1.74 mmol / g. -1 The pore size of the carbon-based adsorbent prepared in Example 2 is only 40% of that of the carbon-based adsorbent with expanded pore size, which further demonstrates that machine learning models that do not consider free volume cannot predict the screening of high-performance doped carbon-based CO2 adsorbents.

Claims

1. A method for screening and designing doped carbon-based CO2 adsorbents based on multi-scale simulation-driven machine learning, characterized in that... The method includes the following steps: Step S1, model construction: Construct a slit pore carbon model with the coupling of typical doping sites and micropore spaces of different sizes, and optimize the geometric configuration of the slit pore carbon model using density functional theory; Step S2, Slit Hole Free Volume Descriptor Extraction: Descriptor of free volume of slit pores under the coupling of doped sites and micropore space. Slit hole free volume descriptor This is the ratio of the accessible volume of gas molecules within the slit to the total volume. Step S3, CO2 adsorption property calculation: Calculate the adsorption thermodynamic properties and kinetic properties of CO2 in the adsorbate using the grand canonical Monte Carlo method and the molecular dynamics method respectively; Step S4, eigenvalue extraction: Extract the eigenvalues affecting the CO2 adsorption performance based on the multi-scale simulation calculation results of Step S3: pore size, nitrogen atom content, oxygen atom content, electronegativity, dipole moment, charge, H2O volume fraction, N2 volume fraction; Step S5, machine learning training set construction: Divide the density functional - grand canonical Monte Carlo - molecular dynamics multi-scale calculation data into a training set and a test set, use the training set to train a machine learning model, and use the test set to evaluate the performance of the machine learning model; Step S6, Machine Learning Model Training: Using the slit aperture free volume descriptor V f Using the feature values ​​extracted in step S4 as input variables and CO2 adsorption as the target variable, the machine learning model is iteratively trained using the training set, and the input slit orifice free volume descriptor V is analyzed. f The feature values ​​extracted in step S4 are normalized to improve the generalization ability of the model. Step S7, machine learning model screening: Find the optimal machine learning model applicable to CO2 adsorption prediction based on prediction accuracy, and find the eigenvalues dominating CO2 adsorption based on the feature interaction analysis method; Step S8, doped carbon-based adsorbent configuration prediction: Predict the configuration of the doped carbon-based adsorbent with the optimal CO2 adsorption performance based on the eigenvalues dominating CO2 adsorption; Step S9, guiding experimental synthesis: Based on the predicted configuration of the doped carbon-based adsorbent with the optimal CO2 adsorption performance, directionally synthesize a high-performance carbon-based adsorption material suitable for flue gas CO2 capture.

2. The method for screening and designing doped carbon-based CO2 adsorbents based on multi-scale simulation-driven machine learning according to claim 1, characterized in that... In the said Step S1, the size of the slit pores in the slit pore carbon model is 0.5 - 2.0 nm, and the doped heteroatoms include individual and combined forms of N, O, B, and S, and 0 < the concentration of heteroatoms ≤ 20 at%.

3. The method for screening and designing doped carbon-based CO2 adsorbents based on multi-scale simulation-driven machine learning according to claim 1, characterized in that... In the said Step S3, the adsorbate gas components are a mixture of CO2, N2, and H2O, where 0 < the concentration of CO2 ≤ 100 vol%, 0 ≤ the concentration of H2O ≤ 10 vol%, the remaining gas is N2, and the ambient temperature is 298 K.

4. The method for screening and designing doped carbon-based CO2 adsorbents based on multi-scale simulation-driven machine learning according to claim 1, characterized in that... In the said Step S3, the specific settings of the grand canonical Monte Carlo method are as follows: Use the Metropolis algorithm to randomly select the movement of particles, set the movement distribution in each GCMC simulation to 20% exchange, 20% conformation, 40% rotation, and 20% translation, the maximum sizes of the translation angle and the rotation angle are 1 Å and 5° respectively, the equilibrium step length and the total step length are 1000000 and 10000000 respectively, and the simulation uses a fully rigid model; the specific settings of the molecular dynamics simulation method are as follows: Use the Dreiding force field based on the Lennard-Jones potential function, and the interaction with the potential field generated by atoms includes the surface of the slit pore carbon model and the gas; use random initial atomic velocities, use the Nose – Hoover temperature control system corresponding to the simulation temperature and the Maxwell - Boltzmann distribution, and perform 1000 ps of EMD simulation in the NVT ensemble with a time step of 1 fs.

5. The method for screening and designing doped carbon-based CO2 adsorbents based on multi-scale simulation-driven machine learning according to claim 1, characterized in that... In the said Step S4, calculate the electronegativity of the model using the formula for group electronegativity: In the formula, The total number of atoms in the representative group, The total atomic charge of the representative group. In the model The number of atoms yes Electronegativity of atoms before bonding; The Mulliken charge of the geometrically optimized aromatic carbon clusters was calculated using density functional theory. The dipole moment of the geometrically optimized aromatic carbon clusters was calculated using density functional theory.

6. The method for screening and designing doped carbon-based CO2 adsorbents based on multi-scale simulation-driven machine learning according to claim 1, characterized in that... In step S6, the machine learning model is trained by introducing gradient boosting regression, extreme gradient boosting, K-nearest neighbor algorithm or random forest algorithm to train CO2 adsorption amount.

7. The method for screening and designing doped carbon-based CO2 adsorbents based on multi-scale simulation-driven machine learning according to claim 1, characterized in that... In step S7, the regression coefficient R 2 The root mean square error (RMSE) is used as an evaluation metric to identify the optimal machine learning model suitable for CO2 adsorption prediction.

8. The method for screening and designing doped carbon-based CO2 adsorbents based on multi-scale simulation-driven machine learning according to claim 1, characterized in that... In step S7, the feature interaction analysis method uses SHAP values ​​to deeply interpret the contribution of each feature value, where the SHAP value represents the contribution of each feature. Average incremental contribution The formula for calculating the SHAP value is as follows: In the formula, This represents a set containing all features; Features Any feature subset other than Represents a set The number of features in the middle, Represents the total number of all features; The model represents the data obtained solely from the observation set. Predict based on features within; Indicates features Included in set The predicted value, As a weight to balance sets of different sizes.

9. The method for screening and designing doped carbon-based CO2 adsorbents based on multi-scale simulation-driven machine learning according to claim 1, characterized in that... In step S9, the directional synthesis method is a microwave-assisted activation method, which obtains a carbon-based adsorbent with the desired pore size and heteroatom distribution by changing the activation conditions.

Images