Computer Aided Molecular Design Of Additives In Naphtha Steam Cracking Process

Low-carbon olefins are important basic raw materials in the chemical industry,and petroleum cracking is currently the main process for producing low-carbon olefins such as ethylene and propylene in China.Under the background of " carbon peaking and carbon neutrality ",with the increase in refining capacity and the rapid development of new energy technologies such as lithium batteries and hydrogen energy,the domestic refining industry has overcapacity and urgently needs to accelerate structural adjustment and gradually transition to "refining and chemical integration".Therefore,developing new petroleum cracking technologies with higher yield of low-carbon olefins and better atom economy is of great significance.Currently,the petroleum cracking process has defects such as high reaction temperature,high energy consumption,and difficulty in controlling the selectivity of low-carbon olefins.Developing new cracking technologies has broad prospects.High-temperature thermal cracking of hydrocarbons follows a free radical mechanism,and the addition of cracking initiators can quickly generate free radicals,initiate chain reactions in advance,and achieve the purpose of reducing cracking temperature and regulating product distribution.However,the current experimental methods for exploring the cracking performance of initiators are costly and have a long development cycle.Moreover,the mechanism of action of initiators in initiating cracking and regulating product distribution in complex kinetic mechanisms is not fully understood and requires indepth analysis.In order to meet the needs of initiator design and development for cracking,a fast and efficient initiator intelligent screening method is urgently needed.Based on the mechanism modeling and reaction process simulation of petroleum hydrocarbon thermal cracking,this paper focuses on the structure design and screening of initiators,and develops a complex reaction kinetics analysis framework,constructs an automatic solution combining mechanism modeling with process simulation,and generates a structure-effect relationship database of initiators to explore the mechanism of action of initiators in regulating cracking product distribution.Using data-driven methods,a machine learning-based structure-property prediction model for initiators is established to achieve accurate prediction of the types of free radicals generated and high-throughput screening of initiator molecular structures.The main work contents are as follows:(1)In view of the large scale of cracking kinetics and the difficulty in obtaining a comprehensive understanding of complex reaction mechanisms,this paper develops a complex kinetics simplification and global reaction path automatic analysis algorithm framework.Based on the newly proposed cumulative flux analysis concept,the framework can restore the reaction mechanism without affecting the prediction accuracy,reduce the complexity and computational scale of the kinetics model.Its global path analysis visualization function can identify the main global reaction paths and automatically draw reaction network diagrams.Through the global reaction network diagram,users can easily extract the reaction path that has the greatest impact on the reaction throughout the entire reaction interval and quantitatively obtain the global conversion ratio relationship between various species,greatly simplifying the analysis of reaction kinetics.In the case of n-hexane cracking,the original 5390 reactions and 328 species of reaction kinetics were simplified to a skeleton mechanism containing 460 reactions and 94 species,and simulation calculations showed that the simplified kinetics maintained the same prediction accuracy.This framework can be widely applied to other fields containing complex kinetics analysis,such as combustion and atmospheric chemistry,and provides a new strategy for simplifying and analyzing large-scale chemical mechanisms.(2)Using Reaction Mechanism Generator,the reaction kinetics of n-hexane cocracking with initiators of different structures were generated.Combining the generated kinetic mechanism,the product distribution data of initiator-induced cracking were obtained through reactor simulation under the same reaction conditions.Due to the large number of molecules and complex calculations involved,this chapter established an automated database generation process and developed an automated solution combining mechanism modeling and process simulation,ultimately creating an initiator structure-performance relationship database containing 3711 sample points.For the anomalous values in the database,a density functional theory(DFT)-based quantum chemistry calculation method was employed,combined with the proposed kinetic simplification framework,to correct the kinetic parameters of the main reaction pathways.Furthermore,a functional group substructure search algorithm was developed to analyze the initiator structures and explore the relationship between different structural compositions and cracking performance.The analysis revealed that lower minimum bond dissociation energies of initiators corresponded to lower cracking initiation temperatures.Substructures such as carbon-nitrogen bonds(C-N)and hydroxyl groups(C-OH)promoted n-hexane cracking,while carbon-carbon double bonds(C=C)inhibited n-hexane cracking and promoted benzene production.Ketone groups(C-C(=O)C)promoted ethylene production,while carbon-nitrogen bonds(C-N)inhibited ethylene production but enhanced propylene yield.Carbon-carbon double bonds(C=C)promoted the production of butadiene and benzene.Overall,the addition of initiators not only lowered the starting temperature of hydrocarbon cracking but also altered the product distribution.(3)The evaluation of the performance of cracking initiators involves complex high-temperature cracking reactions,and predicting reaction performance based solely on their molecular structure may have certain limitations.Therefore,we hope to further learn from the study of initiator molecular structures to the study of cracking reactions.A machine learning-assisted high-throughput screening strategy was proposed to quickly identify high-performance cracking initiator molecular structures.Based on the single operating condition database mentioned above,combined with different reaction operating conditions,a multi-operating condition database containing 371,034 sample points was generated.The message passing neural network was constructed to directly predict the cracking product distribution from the molecular structure and reaction conditions.Through the trained model,new molecular structures were screened in a high-throughput manner.The results showed that the predicted values of ethylene and propylene yields of the screened initiators had a relative error of about 1%compared with the simulation values,and the relative error of butadiene yield was 7%.Compared with the blank cracking group,the recommended initiators can increase the yield of ethylene by.36%.propylene by 7.9%,and butadiene by 9%.The analysis found that the reaction temperature was the main factor affecting the cracking process product distribution.(4)The performance evaluation of cracking initiators involves complex hightemperature cracking reaction processes.Predicting reaction performance based solely on their molecular structure may have certain limitations.Therefore,we hope to further learn from the study of initiator molecular structures to the study of cracking reactions.A machine learning-assisted free radical reaction product prediction strategy was proposed to identify the types of free radicals generated by initiator molecules in the initial cracking reaction.A total of 15,781 free radical reactions were selected from the organic small molecule free radical reaction enthalpy database and the NIST reaction database as the learning data for the neural network model.The WLN graph convolutional neural network model was constructed,which showed an accuracy of 89%in predicting reactions on the validation set,achieving the prediction of the types of free radicals generated by the initiator in the initial cracking reaction.