Interactive Visualization of Large-Scale Oil and Gas Reservoir Simulation Models

Abstract

Accurate simulation of the processes in oil and gas reservoirs requires a detailed description of the geological and physical structure of the reservoir that leads to models containing billions of cells. Conventional approaches to the visualization of such large models are slow-going during the interactive processing of the data. This study is dedicated to the development of parallel algorithms suitable for interactive visualization of large-scale reservoir 3D models based on the preselection of data needed for constructing the model representation from the unstructured mesh. The key component of this workflow is the efficient slicing algorithm of the large mesh. We present the parallel implementation of a slicing algorithm for MPI CPU and multi-GPU computational systems in a form of a plugin program extension for an open-source software ParaView. We tested the performance of (1) our algorithm, (2) an existing commercial reservoir simulation software, and (3) the built-in tool for model slicing of ParaView. Our approach demonstrates an almost interactive visualization of the reservoir model with 1.9 billion cells of unstructured mesh with a slice viewing delay of 1.8 s using the GPU algorithm implementation.