Enhanced prediction of total organic carbon in shale gas reservoirs using extended feature variables and random forests

Total organic carbon (TOC) is a critical parameter for evaluating shale gas reservoir quality and hydrocarbon potential. This study presents an innovative approach combining extended elastic impedance (EEI) and random forest (RF) algorithms to predict TOC distribution in shale gas reservoirs. We address two key challenges: limited feature variables and imbalanced sample sets. Our method automatically generates 222 feature variables through EEI transformation and employs synthetic minority oversampling technique (SMOTE) to balance sample distribution. Applied to a shale gas exploration area in Southwest China, the proposed method demonstrates superior accuracy compared to conventional approaches, particularly in identifying high-TOC sweet spots. The results are validated through horizontal drilling, confirming the method’s effectiveness in shale gas reservoir characterization.