CORRELATIONS Prediction of Physical Properties of Hydrocarbons, Petroleum, and Coal Liquid Fractions Evagelos Retzekas, Epaminondas Voutsas,* Kostis Magoulas, and Dimitrios Tassios Thermodynamics and Transport Phenomena Laboratory, Department of Chemical Engineering, National Technical University of Athens, 9 Heroon Polytechniou Str., Zographou Campus, 157 80 Athens, Greece A simple method that uses the molecular structure and density as input parameters for the prediction of the normal boiling point (T b ), critical temperature (T c ), and critical pressure (P c ) of pure hydrocarbons is presented. For T b the average absolute error is 1.0% as compared to 3.3% for the Joback method and 2.9% for that of Stein and Brown. Its main advantage over the first method lies with large molecular weight compounds and that over the second with highly branched compounds. For the prediction of T c , the average absolute error is 1% similar to that of the Joback, Riazi, and Riazi-Daubert methods which, however, require knowledge of T b . Finally, for P c , the proposed method gives an average absolute error of 2.7% as compared to 3.9% for the Joback method and 4.2% and 4.8% for the T b -requiring methods of Riazi and Riazi- Daubert, respectively. The proposed method gives also better results for these three properties when compared to the recently proposed and more difficult to use group interaction contribution method of Marejon and Fontevila. Using data for pure hydrocarbons, correlations have been developed for the prediction of molecular weight (MW), T c , and P c of petroleum and coal liquid fractions. MW prediction gives an average absolute error of 4.1% as compared to 4.6% for the Riazi-Daubert method, and both methods provide better results for coal liquids than the Starling and “single-parameter” expressions. T c and P c predictions with errors of 1.2% and 5.5% are similar to those of the Riazi-Daubert method, but no conclusion can be reached about the reliability of these methods because of the small number of available data. 1. Introduction Knowledge of accurate physical properties is very important in the chemical, petrochemical, and petro- leum industries for the optimum design and evaluation of separation processes as well as for reservoir fluid modeling. For example, it is known that a small error in T c may lead to a very large error in the prediction of vapor pressure through equations of state. 1 The physical properties considered here are the normal boiling point and critical properties of pure hydrocarbons, which are considered in the first part of the paper, and the MW and critical properties of petroleum and coal liquid fractions considered in the second part. 2. Pure Hydrocarbons The most commonly used methods for T b prediction are the group contribution ones of Joback 2 and of Stein- Brown, 3 while for the prediction of critical properties, again that of Joback and those of Riazi 4 and Riazi- Daubert 5 are the most commonly used. The recently proposed group interaction contribution (GIC) method, 6 which claims improved performance over the classical group contribution methods, is also considered here. All methods are briefly described in the appendix. 2.1. Proposed Method. For the prediction of the normal boiling point and the critical properties of pure hydrocarbons, the following expression, which is a combination of the group contribution approach with an empirical term that includes density, is proposed: where Q stands for T b , T c , and P c ; F is the liquid density at 20 °C; MW is the molecular weight of the compound; a-d are constants that are the same for all hydrocar- bons but different for each property; N i is the number of times that group i appears in a compound; and G i is the value of the group. The group assignment used at this work is the same as the one proposed by Joback. 2 2.2. Results and Discussion. 2.2.1. Normal Boil- ing Point. Table 1 presents the database of 110 compounds used for the evaluation of the necessary parameters in eq 1, which are presented in Table 2, along with the prediction results for the database of 183 compounds used for validation. Table 1 also includes the results obtained with the methods of Joback and of Stein and Brown for comparison purposes. The overall per- formance, in the total of 293 compounds, of the three methods is also shown graphically in Figure 1, where the predicted T b values are plotted against the experi- mental ones. Finally, Table 3 compares the performance * Corresponding author. Tel.: +301 772 3137. Fax: +301 772 3155. E-mail: evoutsas@chemeng.ntua.gr. Q ) aF b MW c + d + ∑ i N i G i (1) 1695 Ind. Eng. Chem. Res. 2002, 41, 1695-1702 10.1021/ie010642a CCC: $22.00 © 2002 American Chemical Society Published on Web 02/23/2002