This work presents a new workflow to obtain a better-constrained reservoir-scale model for an Alkaline-Surfactant-Polymer (ASP) injection pilot design. It is explained how the impact of uncertain parameters related to ASP flooding can be quantified, using calibrated core-scale simulation based on experimental results, and how the influential parameters range for future reservoir-scale simulation can be determined. Computational costs of core-scale model are therefore much lower, and the final reservoir model is better constrained.
ASP flooding feasibility implies core scale studies, where chemical formulations are validated in the laboratory under field conditions. In the objective of the pilot designing, a numerical model is constructed and calibrated to history-match the core flood sequences: Remaining Oil Saturation (ROS), surfactant-polymer (SP) and polymer-alkaline (PA) injection and eventually the chase water slug. In order to quantify the impact of ASP chemical parameters on the history match, the Global Sensitivity Analysis (GSA) was performed using Response Surface Modeling (RSM). To obtain the acceptable range of influential parameters for future reservoir-scale simulation, the Bayesian optimization is used.
Applying this methodology on a real reservoir core, the laboratory measurements are accurately reproduced. Nevertheless, once the core-scale model was matched, the transition to reservoir-scale model must be done. Due to a large number of parameters and their associated uncertainties, this transition is not straight-forward. Thus, an additional step in our workflow is included. A new methodology is applied to firstly quantify the impact of uncertain parameters related to ASP flooding (adsorption of surfactant on the rock, critical micellar concentration, water mobility reduction by polymer etc.). To do so, the RSM is used and influential parameters are identified. In this study, the surfactant adsorption coefficients are the most influential parameters while others related to SPA have a poor impact on experiment results matching. Secondly, the acceptable range of influential parameters for future reservoir-scale simulation and feasibility study is obtained during Bayesian optimization. Thus, instead of using a wide (prior) range of uncertain parameters values, refined (posterior) distribution laws can be used for future reservoir model.
While the classical approach consists in matching experimental results to obtain calibrated values of certain properties (that are then entered in the reservoir model) and finally determine the influential parameters at the reservoir scale, here the choice was made to determine influential parameters and characterize their impacts at the core scale. This step helps to better constrain the reservoir model. Ongoing work is using the results of this workflow for pilot design and risk analysis.