A Review of Machine Learning Methods Applied to Gas Solubility Prediction

Gas solubility prediction is foremost in the petroleum industry, affecting reservoir management, production optimization, and safety considerations. This study presents a comprehensive review of machine learning methods used in gas solubility prediction. These methods are Artificial Neural Networks (ANNs), Adaptive Neuro-Fuzzy Inference System (ANFIS), Functional Network (FN), and Support Vector Machine (SVM). The study explores machine learning methods, highlights their capability to predict gas solubility, and discussed some case studies to illustrate their accuracy in gas solubility prediction. The ANNs exhibit adaptability in handling clustered data. However, they are constrained by dependencies on hidden layer configurations. This study also delves into ANFIS, a hybrid model combining ANN learning capacities with fuzzy logic, showcasing its robust predictions for gas solubility. Studies underscore ANFIS's accuracy, especially in uncertain and complex systems, and its ability to overcome fitting challenges. ANFIS models have an Average Absolute Percentage Error (AAPRE) of (4.631-4.719) % for gas solubility prediction. FN emerges as an alternative to ANNs, leveraging domain knowledge alongside data for enhanced insights. The studies indicate FN's satisfactory performance in gas solubility prediction, particularly in estimating gas solubility (Rs) with acceptable error rates. FN models have an AAPRE of 10.2012% for gas solubility prediction. This review also discusses SVM, a supervised learning model effective in regression analysis for gas solubility prediction. The SVM has a lower over-fitting risk and convergence with global optima, especially in the modified versions such as LSSVM for large-scale problems. The SVM models have an AAPRE of 9.0757-15.94 % for gas solubility prediction. This review on the machine learning models in gas solubility prediction underscores their importance in optimizing oil production, enhancing safety measures, and ensuring compliance with environmental regulations in the oil and gas industry.