Recent Advances in Scale Prediction: Approach and Limitations

Abstract

Summary Numerous saturation indices and computer algorithms have been developed to determine whether, when, and where scale will form. However, scale prediction can still be challenging because the predictions from different models often differ significantly at extreme conditions. Furthermore, there is a great need to accurately interpret the partitioning of water (H2O), carbon dioxide (CO2), and hydrogen sulfide (H2S) between different phases, as well as the speciations of CO2 and H2S. This paper summarizes current developments in the equation-of-state (EOS) and Pitzer models to accurately model the partitioning of H2O, CO2, and H2S in hydrocarbon/aqueous phases and the aqueous ion activities at ultrahigh-temperature, ultrahigh-pressure, and mixed-electrolytes conditions. The equations derived from the Pitzer ion-interaction theory have been parameterized by regression of more than 10,000 experimental data from publications over the last 170-plus years using a genetic algorithm on the supercomputer DAVinCI at Rice University. With this new model, the 95% confidence intervals of the estimation errors for solution density are within 4×10–4 g/cm3. The solubility predictions of CO2 and H2S are accurate to within 4%. The saturation-index (SI) mean values for calcite (CaCO3), barite (BaSO4), gypsum (CaSO4·2H2O), anhydrite (CaSO4), and celestite (SrSO4) are accurate to within ±0.1—and for halite the values are within ±0.01—most of which are within experimental uncertainties. This model accurately defines the pH value of the production tubing at various temperature and pressure regimes and the risk of H2S exposure and corrosion. Furthermore, our model is able to predict the density of soluble chloride and sulfate (SO42−) salt solutions within ± 0.1% relative error. The ability to accurately predict the density of a given solution at temperature and pressure allows one to deduce when freshwater breakthrough will occur. In addition, accurate predictions can only be reliable with accurate data input. The need to improve the accuracy of scale prediction with quality data will also be discussed.