Modeling and Simultaneous History-Matching of Multiple WAG Coreflood Experiments at Reservoir Conditions

Abstract

Abstract Good prediction of the performance of water-alternating-gas (WAG) processes relies on the proper estimation of three-phase relative permeabilities and their hysteresis with injection cycles. This is often a lengthy and expensive procedure requiring the numerical interpretation of several WAG corefloods, usually performed at reservoir conditions. In this paper, we propose an automated procedure based on multi-experiment optimization, leading to a consistent set of three-phase hysteresis parameters for all available experimental data. We apply it to two near-miscible WAG coreflood experiments that differ in the number and length of their injection cycles (long slugs versus short slugs). Both experiments were performed at reservoir conditions on a horizontal sandstone core, with light oil from a West African field and below minimum miscibility pressure. These experiments were carried out using extensive monitoring, including material balance at standard and reservoir conditions, full compositional analysis of liquids and gas produced at standard conditions, differential pressure measurements across the core and three-phase in situ saturations using a dual energy X-ray scanner. History-matching results were obtained by coupling our optimization tool with an in-house research reservoir simulator (IHRRS), combining an advanced EoS-based equilibrium relaxation model with a three-phase relative permeability hysteresis model. Optimization was performed with the Nelder-Mead algorithm, using a relatively large number (>10) of fitting parameters to model the relative permeability functions and their hysteresis. Multiple sets of parameters were easily obtained to match each experiment individually, suggesting that the history-matching of a single experiment is poorly constrained. As expected, muti-experiment optimization led to a better constrained but more challenging problem to solve; yet all available data could be matched reasonably well with a common set of parameters and several acceptable solutions were found. By providing a more robust estimation of three-phase hysteresis parameters, the proposed method increases the reliability of our simulation-based interpretations, necessary to evaluate the stakes of a WAG project.