Constructing Reservoir Flow Simulator Proxies Using Genetic Programming for History Matching and Production Forecast Uncertainty Analysis

Created by W.Langdon from gp-bibliography.bib Revision:1.3973

@Article{Yu:2008:JAEA,
  author =       "Tina Yu and Dave Wilkinson and Alexandre Castellini",
  title =        "Constructing Reservoir Flow Simulator Proxies Using
                 Genetic Programming for History Matching and Production
                 Forecast Uncertainty Analysis",
  journal =      "Journal of Artificial Evolution and Applications",
  year =         "2008",
  volume =       "2008",
  pages =        "Article ID 263108",
  keywords =     "genetic algorithms, genetic programming",
  URL =          "http://www.cs.mun.ca/~tinayu/Publications_files/JAEA.pdf",
  URL =          "http://downloads.hindawi.com/archive/2008/263108.pdf",
  DOI =          "doi:10.1155/2008/263108",
  size =         "13 pages",
  abstract =     "Reservoir modelling is a critical step in the planning
                 and development of oil fields. Before a reservoir model
                 can be accepted for forecasting future production, the
                 model has to be updated with historical production
                 data. This process is called history matching. History
                 matching requires computer flow simulation, which is
                 very time-consuming. As a result, only a small number
                 of simulation runs are conducted and the
                 history-matching results are normally unsatisfactory.
                 This is particularly evident when the reservoir has a
                 long production history and the quality of production
                 data is poor. The inadequacy of the history-matching
                 results frequently leads to high uncertainty of
                 production forecasting. To enhance the quality of the
                 history-matching results and improve the confidence of
                 production forecasts, we introduce a methodology using
                 genetic programming (GP) to construct proxies for
                 reservoir simulators. Acting as surrogates for the
                 computer simulators, the cheap GP proxies can evaluate
                 a large number (millions) of reservoir models within a
                 very short time frame. With such a large sampling size,
                 the reservoir history-matching results are more
                 informative and the production forecasts are more
                 reliable than those based on a small number of
                 simulation models. We have developed a workflow which
                 incorporates the two GP proxies into the history
                 matching and production forecast process. Additionally,
                 we conducted a case study to demonstrate the
                 effectiveness of this approach. The study has revealed
                 useful reservoir information and delivered more
                 reliable production forecasts. All of these were
                 accomplished without introducing new computer
                 simulation runs.",
  notes =        "Department of Computer Science, Memorial University of
                 Newfoundland. Chevron Energy Technology Company",
}

Genetic Programming entries for Tina Yu Dave Wilkinson Alexandre Castellini

Citations