Sensitivity Analysis and Optimization of Surfactant-Polymer Flooding under Uncertainties

Abstract

Chemical flooding is currently one of the most promising solution to increase the recovery of mature reservoirs. In Surfactant-Polymer (SP) processes, several parameters should be taken into account to estimate the return on investments: concentrations of the injected chemical species, slug sizes, initiation times, residual oil saturation, adsorption rates of the chemical species on the rock, etc. Some parameters are design parameters whereas other ones are uncertain. For operators, defining the optimal values of the first ones while considering the uncertainties related to the second ones, is not an easy task in practice. This work proposes a methodology to help handle this problem. Starting from a synthetic reservoir test case where an SP process is set up, we select design and uncertain parameters which may impact the production. In the reservoir simulator, for the sake of flexibility, some of them are tabulated functions, which enables the user to input any data coming from any system. However, point-wise modifications of these curves would soar the number of parameters. Therefore, a particular parameterization is introduced. We then propose a methodology based on Response-Surface Modeling (RSM) to first approximate the oil production computed by a reservoir simulator for different values of our parameters and identify the most influential ones. This RSM is based on a Karhunen-Loève decomposition of the time response of the reservoir simulator and on an approximation of the components of this decomposition by a Gaussian process. This technique allows us to obtain substantial savings of computation times when building the response surfaces. Once a good predictability is achieved, the surfaces are used to optimize the design of the SP process, taking economic parameters and uncertainties on the data into account without additional reservoir simulations.