Machine Learning Methods to Speed up Compositional Reservoir Simulation (SPE 154505)

Compositional reservoir simulation is one of the most powerful techniques currently available to the reservoir engineer upon which most reservoir development decisions rely on. According to the number of components used to describe the fluids there is an increasing demand for computational power due to the complexity and the iterative nature of the solution process. Phase stability and phase split computations often consume more than 50% of the simulation total CPU time as both problems need to be solved repeatedly and iteratively for each discretization block at each iteration of the non-linear solver. Therefore, speeding up these calculations is a research challenge of great interest. In this work, machine learning methods are proposed for the solving of the phase behavior problem. It is shown that under proper transformations, the unknown closed-form solution of the Equation-of-State based phase behavior formulation can be emulated by proxy models. The phase stability problem is treated by classifiers which label the fluid state in each block as either stable or unstable. For the phase split problem, regression models provide the prevailing equilibrium coefficients values given the feed composition, pressure and temperature. The development of these models is rapidly performed offline in an automated way, by utilizing the fluid tuned-EoS model prior to running the reservoir simulator. During the simulation run, rather than solving iteratively the phase behavior problem, the proxy models are called to provide non-iteratively direct answers at a constant, very small CPU charge regardless of the proximity to critical conditions. The proposed approach is presented in two-phase equilibria formulation but it can be extended to multi-phase equilibria applications. Examples demonstrate the advantages of this approach, the accuracy obtained in the calculations and the very significant CPU time reduction achieved with respect to conventional methods.