Production Optimization of CO2 Huff-n-Puff Process in Unconventional Oil Reservoirs With Effects of Gas Relative Permeability Hysteresis, Geomechanics, and Capillary Pressure

Summary The present work introduces a production optimization framework that integrates key multiphysics phenomena like hysteresis in gas permeability and capillary pressure, geomechanics, and pore confinement in nanopores. Unlike the many previous works neglecting such phenomena, the framework enables one not only to perform a more realistic reservoir production optimization but also to understand how important it is to include such phenomena in life cycle production optimization of single-well CO2 huff-n-puff (HnP) processes in unconventional oil reservoirs. We use a commercial multiporosity compositional simulator, fully coupled with geomechanics, to simulate the CO2 HnP process. In the proposed framework, compositional fluid characterization incorporates capillary pressures within nanopores, while molecular diffusion remains consistently present throughout the simulations. The Killough’s hysteresis model is used for relative permeability calculations. The mathematical production optimization problem formulated in this study involves finding the optimal values of the design variables [which could be any combination of the controls like CO2 injection rate, production bottomhole pressure (BHP), injection, soaking, production periods, and overall length of each cycle] that maximize the net present value (NPV) subject to the bound and equality constraints of the optimization variables. We use and compare both the stochastic simplex approximate gradient (StoSAG) and the iterative least-squares support vector regression (LSSVR) proxy methods. Sensitivity studies based on our synthetic, but realistic, simulated examples show that nanopore confinement suppresses the saturation curve in pressure/volume/temperature (PVT) flash calculations, resulting in increased oil production during the primary production stage, whereas during HnP cycles, this mechanism has an adverse effect on production. In contrast, hysteresis generally has a positive effect on oil production over the cycles. CO2 injection around the minimum miscibility pressure point can enhance the recovery of oil, complemented by the positive effect of molecular diffusion. However, incorporating all these physical mechanisms increases system complexity, necessitating the development of a more realistic numerical optimization model. The same numerical model is used in the proxy-based optimization framework, and results show that the iterative LSSVR proxy method is highly efficient for performing production optimization in the complex CO2 HnP process, achieving computational times that are 3–10 times faster than those of the traditional optimization method based on the direct use of a compositional simulator, depending on the case.