Machine Learning Optimizes Production in an Unconventional Reservoir

The full paper investigates the environment friendly estimation of optimum design variables that maximize internet current worth (NPV) for life-cycle manufacturing optimization throughout a single-well carbon dioxide (CO2) huff ‘n’ puff (HnP) course of in unconventional oil reservoirs. The NPV is calculated by a machine-learning (ML) proxy mannequin educated to approximate the NPV that will be calculated from a reservoir simulator run. The ML proxy mannequin might be obtained with both least-squares help vector regression (LS-SVR) or Gaussian course of regression (GPR).IntroductionThe design variables thought-about in this research embody CO2 injection charge, manufacturing bottomhole strain (BHP), period of the injection, and period of manufacturing time for every cycle. To one of the best of the authors’ data, the whole paper represents the primary research that considers the life-cycle optimization drawback for a hydraulically fractured nicely in unconventional oil formations through the use of two completely different ML-based strategies.Modeling and Sensitivity AnalysisReservoir and Fluid Model.

Recommended For You