Machine Learning and Kinetic Insights into Biomass and Biosurfactant Production for Enhanced Oil Recovery Applications

Enhanced oil recovery (EOR) using biosurfactants has gained significant attention due to its cost-effectiveness and environmental compatibility. In this study, we introduce a hybrid modeling approach that combines the Bayesian optimization algorithm (BOA) with support vector regression (SVR) to predict rhamnolipid production by Pseudomonas aeruginosa MYSAG. In addition, we assess the produced biosurfactant for its potential application in microbial EOR (MEOR) using experimental data for model development, and benchmark the predictive performance of the hybrid BOA-SVR model against the conventional response surface methodology (RSM) using multiple statistical indicators. Compared with the RSM model, the proposed hybrid model exhibits superior predictive accuracy, achieving more than a 95% improvement in mean absolute percentage error (MAPE), along with lower mean absolute error (MAE), relative error (RE), and root mean square error (RMSE), as well as a higher coefficient of determination (R²=0.9833). We use additional simulated data to validate the hybrid BOA-SVR model’s generalization ability, and analyze the kinetics of bacterial growth and biosurfactant production using the logistic and Luedeking-Piret (LP) models, revealing a growth-associated production pattern. We further evaluate the efficacy of the produced biosurfactants in MEOR through interfacial tension (IFT) and zeta potential measurements. At a concentration of 0.27 wt%, the IFT decreases significantly from 27 mN/m to 0.84 mN/m, supported by corresponding zeta potential data. Comparisons with commercial surfactants confirm the competitive performance of the laboratory-produced biosurfactants. Overall, integrating machine learning with kinetic modeling provides predictive and mechanistic insights into biosurfactant production, paving the way for more efficient and sustainable oil recovery strategies.