A fractal physics-based data-driven model for water-flooding reservoir (FlowNet-fractal)

Abstract

This paper proposed a fractal physics-based data-driven framework for reservoir simulation (name as FlowNet-fractal) by integrating physics-based data-driven model with fractal theory. FlowNet-fractal enables fast history matching and production prediction of water flooding reservoirs by considering the fractal characteristics of reservoir permeability and porosity. Details of FlowNet-fractal calculation were given with an oil-water two-phase flow example. In the FlowNet-fractal, transmissibility, control pore volumes (PVs), fractal mass dimension and fractal index were separately defined in each one-dimensional connection element to map reservoir properties, which more specifically reflected reservoir heterogeneity than the traditional FlowNet method. In this paper, an example of simple heterogeneous reservoir model and two actual water-flooding reservoir cases with different scales were given. The calculation results showed that the FlowNet-fractal method outperformed the FlowNet method in both the convergence speed and the accuracy of history matching. Moreover, the heterogeneity of the reservoir model was also defined by the inversed fractal mass dimension and the fractal index. • An integrated fractal-physics-based data-driven model and fractal theory are proposed in this research as the FlowNet-fractal framework for reservoir simulation. • The FlowNet-fractal approach beats the FlowNet method in terms of computing correctness, convergence speed, and historical matching effect. • Using the fractal mass dimension and fractal index, FlowNet-fractal may not only highlight reservoir heterogeneity but also significantly boost calculation rate.