Performance Optimization Of Adsorption Carbon Capture And Preliminary Exploration Of Machine Learning

Adsorption carbon capture technology is one of the effective methods to reduce CO₂emission.However,the development of this technology is restricted by high energy consumption,and there is a large development space in energy consumption and efficiency.The optimization and improvement of adsorption carbon capture technology is a research hotspot in recent years.In this paper,the research method of numerical simulation is used to model the variable temperature adsorption carbon capture technology（TSA）,and the optimization strategy of adsorption carbon capture technology is studied based on the model.Firstly,this paper investigates the optimization method of adsorption carbon capture technology.The optimization method mainly has three directions:the optimal design of adsorption chamber,the optimal design of adsorption cycle process and the screening of adsorbent materials.Among them,cycle process optimization and adsorbent material screening are the research hotspots.The research shows that the main methods of cyclic process optimization are:genetic algorithm,particle swarm optimization algorithm and artificial neural network algorithm.In the selection of adsorbent materials,most scholars use to build prediction models to predict the performance of adsorbents in order to sort adsorbents.Then,this paper optimizes the cycle process of variable temperature adsorption carbon capture technology.The Toth model was selected as the adsorption isotherm model,and the zero dimensional equilibrium model of adsorption cycle was established.As the data source of cycle optimization,the model can simulate the evaluation indexes such as yield,specific energy consumption,purity,recovery and efficiency.Then,the purity,recovery and efficiency are selected as the optimization objectives,Using T_heat（heating fluid temperature）and T_cool（cooling fluid temperature）as decision variables,the multi-objective optimization of the adsorption process is carried out through genetic algorithm to find the best adsorption energy efficiency of an adsorbent under the optimal working condition,so as to achieve the purpose of cycle optimization.Mg-mof-74,zeolite 13X and activated carbon are selected as representative adsorbent materials for case analysis of adsorbent evaluation.Finally,a machine learning prediction model is established,which can predict the optimal cycle performance of adsorbent.The Toth model of 150 adsorbent materials is input into the zero dimensional equilibrium model,and then optimized by genetic algorithm.The results are used as the data set of machine learning.The prediction objectives set in the machine learning process are purity,recovery and efficiency,and the characteristic symbols are adsorption equilibrium isotherm parameters.Two machine learning prediction models are designed,namely the least square model and the random forest model.The results show that the model prediction effect is the best when the random forest algorithm is used for machine learning training,The determination coefficient（R²）of recovery prediction can reach 0.993,and the mean square error value（MSE）is 1.59E-05.In view of the good prediction effect of the random forest model,the characteristic importance analysis is carried out based on it.The results show that the△H value of N₂and CO₂has the greatest impact on the target prediction results.The accurate prediction results of the optimal cycle performance of the adsorbent can provide a basis for the subsequent selection of appropriate adsorbent materials.