Mathematical and neural network prediction model of three-phase immiscible recovery process in porous media

Scaling analysis of fluids displacement in porous media is a reliable fast method to evaluate the displacement performance of different oil production processes under various conditions. This paper describes scaling of multiphase flow through permeable media with special attention to three-phase immiscible water alternating gas flooding processes. The procedure of inspectional analysis (IA) was used to determine relevant dimensionless groups. It was found that scaling immiscible water alternative gas (IWAG) displacement in a two-dimensional homogeneous anisotropic medium needs matching of seven dimensionless scaling groups. A series of numerical sensitivity studies were conducted to determine the magnitude of scaling groups and their interaction with recovery factors. None of the dimensionless groups individually could correlate the efficiency of the process in all the circumstances. Two predictor models, a mathematical model and an artificial neural network (ANN) model, were developed to map the effective combinations of the dimensionless scaling groups. For the mathematical model, a new combined dimensionless group which incorporates all the dominant scaling groups has been derived. Functional relationship between the combined group and fractional oil recovery is investigated. This relation has potential application in prediction of displacement efficiency. To prepare the ANN prediction model, scaling groups obtained from fine mesh simulation were used as input parameters. The agreement between the simulation methods and ANN model validate the applicability of the model to predict the recovery factors for the range of the groups which were not included in the simulations. The comparison study of the models showed the superior performance of ANN model compared to the mathematical model. The ANN model can be used to propose a proper development plan for a reservoir and/or to determine the optimal plan of operation. (C) 2014 Elsevier B.V. All rights reserved.