Performance of Data Segmentation ANN Model for Elemental Sulfur Solubility Prediction in Natural Gas Transportation Pipeline

Abstract

During the gas transportation, the pressure differential along the pipeline leads to the formation of unwanted impurities including elemental sulfur. The elemental sulfur formation is interestingly limited by the solubility of sulfur in the vapor phase. Our work applied the LM-ANN method (Levenberg-Marquardt Artificial Neural Network) to provide accurate and fast prediction of sulfur solubility in natural gas pipeline. Various sulfur solubility models were developed using the proposed method in order to predict sulfur solubility under two operating pressure ranges including; models for data with pressure lower than 25 MPa (LP model) and pressure higher than 25 MPa (HP model). The proposed models could capture the variation of gas composition and conditions (temperature and pressure) while the empirical model was required to be adjusted as the changing values of input variables. The proposed models provided a promising performance with R <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> and RMSE increased from the base model up to 0.9841 and 0.036 for LP model and 0.9913 and 0.0093 for HP model, respectively. The proposed models have advantages over the empirical models as they could cover the variation of gas composition and conditions while the traditional model require to be adjusted as the values of input variables change.