Modeling of nitrogen solubility in normal alkanes using machine learning methods compared with cubic and PC-SAFT equations of state

Abstract

Accurate prediction of the solubility of gases in hydrocarbons is a crucial factor in designing enhanced oil recovery (EOR) operations by gas injection as well as separation, and chemical reaction processes in a petroleum refinery. In this work, nitrogen (N2) solubility in normal alkanes as the major constituents of crude oil was modeled using five representative machine learning (ML) models namely gradient boosting with categorical features support (CatBoost), random forest, light gradient boosting machine (LightGBM), k-nearest neighbors (k-NN), and extreme gradient boosting (XGBoost). A large solubility databank containing 1982 data points was utilized to establish the models for predicting N2 solubility in normal alkanes as a function of pressure, temperature, and molecular weight of normal alkanes over broad ranges of operating pressure (0.0212–69.12 MPa) and temperature (91–703 K). The molecular weight range of normal alkanes was from 16 to 507 g/mol. Also, five equations of state (EOSs) including Redlich–Kwong (RK), Soave–Redlich–Kwong (SRK), Zudkevitch–Joffe (ZJ), Peng–Robinson (PR), and perturbed-chain statistical associating fluid theory (PC-SAFT) were used comparatively with the ML models to estimate N2 solubility in normal alkanes. Results revealed that the CatBoost model is the most precise model in this work with a root mean square error of 0.0147 and coefficient of determination of 0.9943. ZJ EOS also provided the best estimates for the N2 solubility in normal alkanes among the EOSs. Lastly, the results of relevancy factor analysis indicated that pressure has the greatest influence on N2 solubility in normal alkanes and the N2 solubility increases with increasing the molecular weight of normal alkanes.

Vapor–Liquid Equilibria of Binary Systems Consisting of Nitrogen and n-Tridecane, n-Pentadecane, n-Hexadecane, and n-Heptadecane at Pressures of 1.5–9 MPa and Temperatures of 303–633 K─Determined with a Raman Spectroscopy-Coupled Microfluidic Device

“…Modeling using PC-SAFT for the binary system n -hexadecane/N 2 has been also carried out by Madani et al . using a machine-learning algorithm, as well as by Rowane et al and García-Sánchez et al However, Madani et al do not report any k ij PCS in tabular form, while the k ij PCS of Rowane et al cannot be directly compared to our fitted k ij PCS , since the pure compound parameters m , σ, and ε were obtained by a group contribution method. García-Sánchez et al considered only the VLE data of Lin et al for their fit and report with k ij PCS = 0.1816 and k ij PCS = 0.1860, two slightly different values depending on the objective function used.…”

“…The latter are furthermore visualized together with the PPR78-predicted binary interaction coefficients k ij, Td PR in the Appendix section. Modeling using PC-SAFT for the binary system nhexadecane/N 2 has been also carried out by Madani et al 37 using a machine-learning algorithm, as well as by Rowane et al 2 and Garci ́a-Sańchez et al 38 However, Madani et al 37 do not report any k ij PCS in tabular form, while the k ij PCS of Rowane et al 2 cannot be directly compared to our fitted k ij PCS , since the pure compound parameters m, σ, and ε were obtained by a group contribution method. Garci ́a-Sańchez et al 38 considered only the VLE data of Lin et al 9 for their fit and report with k ij PCS = 0.1816 and k ij PCS = 0.1860, two slightly different values depending on the objective function used.…”

Modeling crude oil pyrolysis process using advanced white-box and black-box machine learning techniques

“…XGBoost includes several parameters, making it a complicated model. In addition, hyperparameters are required to limit the danger of over-fitting and forecast variability 47 . The number of iterations (n estimators) and the learning rate are the two key hyper-parameters that avoid overfitting in XGBoost.…”

ML-SAFT: A machine learning framework for PCP-SAFT parameter prediction

“…To enable general and thermodynamically consistent predictions, one approach is to predict thermodynamic model parameters [10,11,12,13,14], which also enables simple use in existing process simulation packages. Given the predicted parameters, the thermodynamic model can in turn be used to predict thermodynamic properties.…”