The Solubility Prediction Of CO₂in Organic Solvents By Artificial Neural Network

In the global climate warming, carbon dioxide emissions increased situation, reduceenergy consumption and reduce carbon dioxide emissions have become a hot issue ofcommon concern in the world. After thirty years of reform and opening up, the economyin the China gained the success that attract worldwide attention at the same time, theconsumption of carbon dioxide emissions and energy in our country have beenrespectively ranked second in the world and the first. The current energy-saving emissionreduction, especially to reduce carbon dioxide emissions has become the whole world tomaintain sustainable economic and social development of consensus. Chinese as aresponsible big country in the world, to reduce energy consumption and reduce carbondioxide emissions has made a solemn promise to the world. Therefore, how to extract thereduction of CO2from atmospheric environment has become a very importantenvironmental problem.To reduce carbon dioxide emissions is a systems engineering, in addition to reducingthe consumption of fossil energy, strengthen the comprehensive utilization of energy, inaddition to the utilization of energy, through the ways of generating sequestration andtrapping industrial carbon dioxide in the process is important, the compression. Carbondioxide capture has become a hot topic of current research. Carbon dioxide capturemethods mainly have two kinds of physical sequestration and chemical absorption,chemical absorption mainly by organic solvent under high pressure dissolving absorbcarbon dioxide, so the carbon dioxide solubility in organic solvents is the key data of thetechnology promotion can application.Thus, in this paper work, the artificial neural network was applied to give theprediction of the solubility of carbon dioxide (CO2) in organic solvents. As a result, theabsorption of CO2can reach its theoretical level. Moreover, the solubility of carbondioxide (CO2) in organic solvents is very important and fundamental in chemical processdesign, especially for CO2capture. In this work, a feed-forward artificial neural network(ANN) model was applied to calculate/predict the solubility CO2in36common organic solvents. With critical properties of organic solvent and operating temperature andpressure as network input, the solubility of CO2in organic solvents were calculated withan optimized three-layer feed-forward ANN model. Application of the model of2769data points containing36solvents generated results with a squared correlation coefficientof0.96and an absolute average relative deviation (AARD) of12.78%from theexperimental values. The predicted results proposed by the ANN model coincide wellwith the existing experimental values.The main work of this study include:(1) In the first chapter, it mainly discuss about the reduction of CO2gases,whichincludes the methods of CO2absorption.(2) In the second chapter, the paper gives the details of the data and tables whichpresent the basic physical properties of organic solvents.(3) In the third chapter, the ways of creationg a fundamental model of artificial neuralnetwork is discussed, including the basic principles of the ANN.(4) In the fourth chapter, it mainly deals with the precise details of the solubilityprediction of artificial neural network, which includes three vital parts of setting up aartificial neural network, such as training, validation and test. In the final part of thepaper, it gives the prediction comparison between ANN and equation of states, whichshows that the solubility prediction of ANN is better than EOS.