Deep Learning Model of Two-Phase Fluid Transport Through Fractured Media: A Real-World Case Study

Modelling of fluid flow in well’s vicinity in naturally fractured reservoirs is a commonly employed technique used for wells’ productivity enhancement, like acid stimulation. Unfortunately, a detailed model reflecting the complex geophysical structure of the porous media is a timely and computationally demanding task. In this paper, a deep learning model is proposed for solving Darcy equation coupled with the transport equation, based on physics-informed neural network (PINN) deep learning technology. Datasets obtained from the 3D numerical simulator are used to train and test our method. We test the sensitivity of our method to the type of optimizer and learning rate, time step size and the number of timesteps, DNN architecture, and spatial resolution. The results of computational experiments on a real-world problem prove a good numerical stability of the solution, its computational efficiency and high precision of the PINN model.