Unconventional/complex reservoirs

Machine Learning Optimizes Production in an Unconventional Reservoir

The paper investigates estimation of optimal design variables that maximize net present value for life-cycle production optimization during a single-well CO2 huff ‘n’ puff process in unconventional oil reservoirs.

Reservoir model and grid system
Fig. 1—Reservoir model and grid system. (a) 3D grid; cross-sectional view. (b) 2D grid; areal view.

The complete paper investigates the efficient estimation of optimal design variables that maximize net present value (NPV) for life-cycle production optimization during a single-well carbon dioxide (CO2) huff ‘n’ puff (HnP) process in unconventional oil reservoirs. The NPV is calculated by a machine-learning (ML) proxy model trained to approximate the NPV that would be calculated from a reservoir simulator run. The ML proxy model can be obtained with either least-squares support vector regression (LS-SVR) or Gaussian process regression (GPR).

Introduction

The design variables considered in this study include CO2 injection rate, production bottomhole pressure (BHP), duration of the injection, and duration of production time for each cycle. To the best of the authors’ knowledge, the complete paper represents the first study that considers the life-cycle optimization problem for a hydraulically fractured well in unconventional oil formations by using two different ML-based methods.

Modeling and Sensitivity Analysis

Reservoir and Fluid Model.

×
SPE_logo_CMYK_trans_sm.png
Continue Reading with SPE Membership
SPE Members: Please sign in at the top of the page for access to this member-exclusive content. If you are not a member and you find JPT content valuable, we encourage you to become a part of the SPE member community to gain full access.