Adaptive modeling of waterflooding process in oil reservoirs

Waterflooding is a common solution for enhancing oil recovery in oil reservoirs in which water is injected to the reservoir. For this purpose, injection rates should be determined to optimally control the desired oil production from each well to accomplish reservoir management strategies. This goal can be achieved by precisely modeling the waterflooding phenomena. In this paper, an applicable algorithm for modeling the waterflooding process in oil reservoirs has been developed. Using proper System Identification techniques will help to consider any change in the inherent reservoir specifications in addition to operating point variations due to injection alternation rate and also severe nonlinear nature of the reservoir and consequently to obtain adaptive model for the reservoir which is able to reflect all the mentioned phenomena. The reservoir has been considered as a Multi-Input/Multi-Output (MIMO) plant with several manipulated variables and outputs to mimic the variables dynamics of the reservoir during its operational life. A Recursive Least Square (RLS) approach has been utilized with proper forgetting factor and sliding window length selection to extract the reservoir dynamic model in the content of a MIMO model structure. The developed approach has been evaluated for SPE10 model as a well-known reservoir case study in this paper. The observed results demonstrate that the proposed method has acceptable capability, compared to the conventional off-line Least Square (LS) and Extended Least Square (ELS) approaches which may lead to Auto Regressive model with eXogenous input (ARX) and Auto Regressive-Moving-Average model with eXogenous inputs (ARMAX) structures, for modeling the dynamical behaviors of the reservoir to be employed as a powerful tool for control and optimization applications.