The volumes of data generated by modern reservoir simulators can be huge. This leads to problems with CPU time and memory limitations when presenting the results for analysis. The issue is not just with handling the large arrays for massive 3D grids but also for the well and completion vectors, which can consume many gigabytes of disk space. The problem is compounded by the trend for many simulations of the same reservoir, whether for history matching or to explore multiple development scenarios.
A further issue is with data display. When the number of runs, wells, completions and timesteps gets large, it can be difficult for an engineer to assimilate the information — especially when comparing bulk simulation results with sporadic historical measurements.
This paper describes techniques to overcome these problems. We concentrate on well-related results such as rates and totals, and grid values in completed cells. These include recasting the results into a form suitable for direct access on disk, a load on demand architecture and lightweight in memory representations. In addition user interface ideas that concentrate information are presented.