We report on an investigation into optimization techniques that could be useful for the automatic matching of production data. One key objective is to extract the maximum amount of information from the data. Our starting point is a series of grids of differing refinement starting at a coarse scale and an associated set of property parameterizations at increasing resolutions. Automatic optimization algorithms are investigated that utilize the differing scales. The aim is to develop optimization methods that can extract maximum information from the production data and are fast and robust enough for practical use.

The algorithms written for this study, which were initially developed for stand-alone single-phase flow models, have subsequently been implemented in a system using a standard commercial simulator. The results presented here are for synthetic models using a simulated truth case; the models are introduced in order of increasing levels of realism.

You can access this article if you purchase or spend a download.