Numerical investigation of the POD reduced-order model for fast predictions of two-phase flows in porous media

Abstract

Purpose This paper aims to present an efficient IMPES algorithm based on a global model order reduction method, proper orthogonal decomposition (POD), to achieve the fast solution and prediction of two-phase flows in porous media. Design/methodology/approach The key point of the proposed algorithm is to establish an accurate POD reduced-order model (ROM) for two-phase porous flows. To this end, two projection methods including projecting the original governing equations (Method I) and projecting the discrete form of original governing equations (Method II) are respectively applied to construct the POD-ROM, and their distinctions are compared and analyzed in detail. It is found the POD-ROM established by Method I is inapplicable to multiphase porous flows due to its failed introduction of fluid saturation and permeability that locate on the edge of grid cell, which would lead to unphysical results. Findings By using Method II, an efficient IMPES algorithm that can substantially speed up the simulation of two-phase porous flows is developed based on the POD-ROM. The computational efficiency and numerical accuracy of the proposed algorithm are validated through three numerical examples, and simulation results illustrate that the proposed algorithm displays satisfactory computational speed-up (one to two orders of magnitude) without sacrificing numerical accuracy obviously when comparing to the standard IMPES algorithm that without any acceleration technique. In addition, the determination of POD modes number, the relative errors of wetting phase pressure and saturation, and the influence of POD modes number on the overall performances of the proposed algorithm, are investigated. Originality/value 1. Two projection methods are applied to establish the POD-ROM for two-phase porous flows and their distinctions are analyzed. The reason why POD-ROM is difficult to be applied to multiphase porous flows is clarified firstly in this study. 2. A highly efficient IMPES algorithm based on the POD-ROM is proposed to accelerate the simulation of two-phase porous flows. 3. Satisfactory computational speed-up (one to two orders of magnitude) and prediction accuracy of the proposed algorithm are observed under different conditions.

Effect of salinity on oil production: review on low salinity waterflooding mechanisms and exploratory study on pipeline scaling

“…Numerical simulation is an effective approach to study mechanisms and phenomena especially when experimental studies are hard to perform [94][95][96][97][98][99]. Modeling the mechanisms behind low salinity waterflooding is the first step to mathematically understand and describe the observed phenomenon, as well as the key point in developing advanced enhanced oil recovery techniques.…”

A Galerkin‐free/equation‐free model reduction method for single‐phase flow in fractured porous media

“…According to the specific form of the original governing equations or discrete matrix equations, there are several ways to construct the corresponding POD reduced‐order model (POD‐ROM) . All of these procedures must establish accurate mathematical expressions of the POD‐ROM to obtain the spectral coefficients or POD time coefficients, which are needed for the reconstructions of approximation solutions.…”

“…The proper orthogonal decomposition reduced‐order model (POD‐ROM) can reduce the number of degrees of freedom for original high‐dimensional physical problems by combining the POD and the Galerkin projection method . It has been applied in many fields, such as turbulence flow, heat conduction and convective heat transfer, two‐phase flow, and gas flow . Recently, the POD method has been extended to simulate flow in porous media .…”

Three-dimensional simulation of acidizing process in carbonate rocks using the Darcy–Forchheimer framework

“…In order to avoid the instability issues, the model is solved implicitly in time, i.e. the transient term is discretized using the backward Euler scheme [46][47][48]. For spatial term, the central difference and upwind schemes are used to discretize the diffusion and convection terms, respectively.…”