Estimation of nitrogen-crude oil interfacial tension: Application of hybrid machine learning algorithms

Nitrogen gas injection has been identified as an effective Enhanced Oil Recovery (EOR) technique in the last years. Precise measurement of interfacial tension (IFT) under reservoir conditions is crucial for planning an effective gas-based EOR process. Conversely, measuring IFT through experimental means is expensive and difficult, requiring complex interpretation and costly devices and arduous procedures. This highlights the necessity of developing accurate and affordable models for estimating IFT. This article aims to propose a smart method using Adaptive Neuro-fuzzy Interference System (ANFIS) optimized by Genetic Algorithms (GA) and Particle Swarm Optimization (PSO) to precisely predict the N₂-crude oil IFT considering pressure, temperature, and density difference. The IFT system's sensitivity analysis revealed that the density difference of the phases is the most impactful parameter. Moreover, nearly all the collected experimental data is considered trustworthy for building the model. In conclusion, ANFIS-PSO was suggested to be more trustworthy than ANFIS-GA in relation to determination coefficient (ANFIS-PSO: 0.88 and ANFIS-GA: 0.78), and average absolute relative error (ANFIS-PSO: 11.08 and ANFIS-GA: 21.98) when validated against test data. The results obtained indicate that the ANFIS-PSO model developed can reliably be utilized to precisely estimate N₂-crude oil IFT for EOR studies and optimizations tasks.