LAUSR

Abstract The accurate determination of hydrate formation temperature (HFT) is an extremely vital step in the context of designing processes containing hydrates. Due to the prohibitive time and the expensive cost of the experimental procedures, some empirical and theoretical approaches have been developed for estimating HFT. However, most of these prior approaches are associated with a lack of generalization, low accuracy, and the non-consideration of some paramount impacting factors. In this study, a reliable and simple-to-use correlation was implemented for predicting HFT of different natural gas types, including sour, acid, and sweet gases. The correlation was established using a powerful explicit-based machine learning technique, namely gene expression programming (GEP). A widespread database encompassing 279 experimental measurements was employed in the learning and testing phases of this method. Results showed that the outcomes of the correlation cohered with the real measurements of HFT. Besides, it was found that GEP-based correlation provided excellent prediction performance and it outperformed the best preexisting models. GEP-based correlation can predict HFT with an average absolute relative error (AARE) of 0.1397%. In addition, the generated hydrate curves by GEP-based correlation overlapped the real ones with a high degree of accuracy. Lastly, the statistical validity of GEP-based correlation was documented using outliers detection.