Methodology to Predict the Conversion, Composition, Density, Viscosity, and Stability of Visbroken Heavy Oils from Feed Data and Reactor Conditions

Abstract Field visbreaking is one option to reduce diluent requirements in bitumen pipelines. To assess proposed visbreaking designs at the design stage, it is necessary to predict product properties based on limited data. This study proposes a methodology to predict visbreaking conversion, product properties, and product stability solely from typically available feed data and the reactor temperature and space time. The methodology adapts a previous approach based on feed data and conversion to eliminate the need for measured conversions. The methodology uses a previously developed correlation to predict the conversion from the equivalent residence time (ERT). The production composition (DSARA─distillates, saturates, aromatics resins, and asphaltenes) is predicted from previously developed correlations to conversion. Product density and viscosity are predicted from the conversion using an excess volume mixing rule and the expanded fluid viscosity model, respectively. The DSARA component properties are inputs to these models and are determined from previously developed correlations. The models are tuned to match the feed properties. A new conversion-based correlation is developed to predict the product stability versus asphaltene precipitation. In addition, a new approach is proposed to further tune the models to match product data when available. The methodology was tested on six different crude oils from different geographical sources. The oils were visbroken in a continuous induction heated reactor across a range of ERT. The feeds and visbroken products were characterized in terms of DSARA composition, density, viscosity, and stability (onset of asphaltene precipitation). The methodology predicted the conversions, compositions, product densities, and onsets with average deviations almost within the measurement uncertainty. The viscosity deviation of 97% was significantly higher than the measurement uncertainty but occurred over a viscosity reduction of 2–3 orders of magnitude. Tuning the model to match the product data reduced the average deviations to below the measurement uncertainty.