A method for simulating oil displacement by emulsion in a heterogeneous porous medium using a physically-informed neural network

Abstract

The utility model relates to the field of oil field development, enhanced oil recovery, hydrodynamic modeling of multiphase flows in porous media, computational modeling of filtration processes, as well as to the field of physically-informed machine learning. The objective of the utility model is to create a software-implemented method for simulating emulsion displacement in a porous medium, ensuring the prediction of the spatiotemporal distribution of pressure, phase saturation, concentration of mobile emulsion, amount of retained emulsion and changes in the filtration properties of the formation, taking into account the interconnected processes of multiphase filtration, emulsion transfer, emulsion retention and permeability reduction. The technical problem is solved in that in the method of modeling the displacement of an emulsion in a porous medium, which includes the input of the parameters of the porous medium, the parameters of the emulsion, the initial and boundary conditions, the use of physical equations of multiphase filtration, the equations of emulsion transport and the equations of emulsion retention, according to the utility model, the modeling of the said processes is carried out by means of a physically-informed neural network, which jointly approximates the spatiotemporal fields of pressure, phase saturation, mobile emulsion concentration, and retained emulsion amount. The technical result is the ability to spatiotemporally predict emulsion displacement parameters and changes in effective permeability without repeating the step-by-step solution of the full finite-volume problem for each forecast scenario.