Investigation on sulfur solubility in sour gas at elevated temperatures and pressures with an artificial neural network algorithm

Accurate prediction of the solubility of elemental sulfur in sour gas mixtures has been recognized as a key issue in the development of sour gas fields. Many experimental measurements and empirical models have shown the complicated relationships between sulfur solubility and sour gas properties. However, the accurate model that can be used to predict sulfur solubility in sour gas over a wide range of temperature and pressure is rare. Therefore, the objective of this work is to build an efficient model, namely T-S fuzzy neural network (T-S FNN), to investigate the sulfur solubility in sour gas. The model considers the reservoir pressure, temperature, the mole fraction of methane, hydrogen sulfide and carbon dioxide as input parameters and the sulfur solubility as target parameters. Subsequently, multiple experimental sulfur solubility data sets accessible to the literature are employed to train and test the model respectively. Finally, a series of studies are conducted to appraise the accuracy and generalization capability of the model. The result shows that the predicted solubility data had great agreement with experimental sulfur solubility with overall average absolute relative deviation of 5.35%, which proves the model in this work is feasible and effective. Additionally, it provides a new idea to the prediction of sulfur solubility in sour gas and helps operators to develop the sour gas reservoir better.