Attention U-Net as a surrogate model for groundwater prediction

Numerical simulations of groundwater flow are used to analyze and predict the response an aquifer system its change in state by approximating solution fundamental physical equations. The most classical methodologies, such as Finite Difference (FD) Element (FE) Methods, use iterative solvers which associated with high computational cost. This study proposes a physics-based convolutional encoder-decoder neural network surrogate model quickly calculate system. Holding strong promise cross-domain mappings, networks applicable for learning complex input-output mappings systems. manuscript presents Attention U-Net that attempts capture relations generates solutions hydraulic head whole domain given set parameters boundary conditions. accurately predicts steady highly heterogeneous locations piezometric up 3 wells input. learns pay attention only relevant parts generated field corresponds target samples great detail. Even relative coarse finite difference approximations proposed is shown be significantly faster than comparative state-of-the-art numerical solver, thus providing base further development presented models prediction.