Fast high-resolution prediction of multi-phase flow in fractured formations

•We describe and apply a new reduced order modeling technique that allows flow in fractured reservoirs to be accurately determined with significant reduction in computational cost.•The method does not require intrusive changes to codes, and readers can easily replicate the method; only low- and high-resolution models are needed.•The accuracy of the reduced order model is approximately determined using an error estimator.

The success of a thermal water flood for enhanced oil recovery (EOR) depends on a detailed representation of the geometrical and hydraulic properties of the fracture network, which induces discrete, channelized flow behavior. The resulting high-resolution model is typically computationally very demanding. Here, we use the Proper Orthogonal Decomposition Mapping Method to reconstruct high-resolution solutions based on efficient low-resolution solutions. The method requires training a reduced order model (ROM) using high- and low-resolution solutions determined for a relatively short simulation time. For a cyclic EOR operation, the oil production rate and the heterogeneous structure of the oil saturation are accurately reproduced even after 105 cycles, reducing the computational cost by at least 85%. The method described is general and can be potentially utilized with any multiphase flow model.