Prediction Of The Solubility Of Polycyclic Aromatic Hydrocarbons In Supercritical Carbon Dioxide By Using Models

In petrochemical industry,supercritical fluid?SCF?is mainly used in supercritical fluid extraction?SFE?,for example,polycyclic aromatic hydrocarbons?PAHs?can be extracted from supercritical fluid.PAHs contain two or more fused benzene rings.But in the process of SCF,the key is to obtain the solubility of PAHs in supercritical carbon dioxide?SCCO₂?in order to choose parameters and operate equipment.Carbon dioxide is the most commonly used supercritical fluid,which is nontoxic,non-flammable,non-expensive and readily available at low cost.At present,there are four sorts of models used to correlate and predict the solute solubility in SCCO₂,which are experimental method,equations of state,semi-empirical and intelligent models.The experimental determination of solubility of compounds at various temperate and pressures is expensive and time-consuming.Equation of state need solute critical properties,but in many occasions,solute critical properties was unknown.To solve this problem,based on the theory that the molecules of a solute associate with the molecules of a gas with the formation of a solvato complex,we combined Support Vector Machine?SVM?with?Genetic Algorithms,GA?,and combined SVM with?Grey Wolf Optimization,GWO?to establish the new model.The main contents of this paper are organized as follows:?1?Based on the Chrastil and modified Chrastil,this paper proposed the five-parameter and six-parameter models to correlate PAHs solubility in SCCO₂,and make comparisons with 17 different models correlated from published literature.As the results showed that,the new model proposed the least AARD?AARD=7.53%,7.57%,8.01%?,and it can accurately correlate PAHs solubility in SCCO₂.?2?A SVM model combined with genetic algorithm?GA-SVM?was proposed to predict the solubility of PAHs in SCCO₂.Experimental data?467 data sets?for the PAHs accessible to various literature were used for training?329?and validation?141?.The results showed that the predictions of the proposed model are in excellent agreement with experimental data with the minimum AARD of 5.78%,the maximum AARD of 7.77%,and the overall AARD is 5.94%%.Furthermore,the new model was verified and evaluated,and the range of model's applicable temperature and pressure was given,which T=308.15?348.15K and p=7.98?35.59MPa.?3?On the base of SVM model and GWO,the new equation?GWO-SVM?was proposed.The new model successfully correlated solute solubility of 9 compounds?888 data points including 626 training data points and 262 testing data points?in SCCO₂.The results showed that the predictions of the proposed model are in excellent agreement with experimental data with the minimum AARD of 0.27%,the maximum AARD of 7.51%,and the overall AARD of 3.16%.Furthermore,the new model was verified and evaluated,and the range of model's applicable temperature and pressure was given,which T=308.05?348.15K and p p=7.83?35.59MPa.?4?At the same temperate and pressure conditions,semi-empirical with relatively high precision,GA-SVM and GWO-SVM model was evaluated.The results showed that GWO-SVM model with the AARD of 1.42%give better results than the other models.GWO-SVM model has stong generalization ability and high accuracy.At the same time,this equation can provide a practical method for model application in oilfield.