The capability of conventional oil characterization methods combined with a cubic equation of state to predict the phase behavior of heavy oil and solvent mixtures was evaluated using data for mixtures of bitumen, propane, and carbon dioxide. The bitumen was characterized based on simulated distillation (SimDist) data extrapolated using six different approaches. Each curve was divided into temperature-based pseudo-components with physical and critical properties determined from established correlations. Four sets of correlations were evaluated and modifications were proposed to extend the Twu critical property correlations to pseudo-components with normal boiling points higher than 730 degrees C. The phase behavior of the oil was modeled using the advanced Peng-Robinson equation of state and the interaction parameters were optimized to fit experimental saturation pressures of pseudo-binary mixtures of each solvent with the bitumen. The model was then used to predict saturation pressures and liquid-liquid phase boundaries of pseudo-ternary systems as well as asphaltene precipitation from n-heptane diluted bitumen. The best characterization of the bitumen was a Gaussian distribution of boiling points from the SimDist for the maltenes plus a Gamma distribution of the molecular weight of the asphaltene fraction. Characterizations with boiling points above 800 degrees C failed. Both the Lee-Kesler and Twu property correlations could be applied to heavy fractions but the Lee-Kesler correlations provided a better fit to the data. The model correctly predicted the saturation pressures and liquid-liquid boundaries for pseudo-ternary mixtures of propane-CO2-bitumen with an AARD of less than 6%. However, the model could not correctly predict asphaltene precipitation yields from bitumen at high dilution with n-heptane. (C) 2011 Elsevier B.V. All rights reserved.