ALBERT

Beschreibung: We propose a flexible framework for evaluating prospect dependencies in oil and gas exploration and for solving decision-making problems in this context. The model uses a Bayesian network (BN) for encoding the dependencies in a geologic system at source, reservoir, and trap levels. We discuss different evaluation criteria that allow us to formulate specific decision problems and solve these within the BN framework. The BN model offers a realistic graphic model for capturing the underlying causal geologic process and allows fast statistical computations of marginal and conditional probabilities. We illustrate the use of our BN model by considering two situations. In the first situation, we wish to gain information about an area where hydrocarbons have been discovered, and use the value of perfect information to determine which locations are the best to drill. In the second situation, we consider the problem of abandoning an area when only dry wells are drilled. For this latter, we use an abandoned revenue criterion to determine the drilling locations. The application is from the North Sea. Our main focus is the description, visualization, and interpretation of the results for relating the statistical modeling to the local understanding of the geology.

Beschreibung: We propose a flexible framework for evaluating prospect dependencies in oil and gas exploration and for solving decision-making problems in this context. The model uses a Bayesian network (BN) for encoding the dependencies in a geologic system at source, reservoir, and trap levels. We discuss different evaluation criteria that allow us to formulate specific decision problems and solve these within the BN framework. The BN model offers a realistic graphic model for capturing the underlying causal geologic process and allows fast statistical computations of marginal and conditional probabilities. We illustrate the use of our BN model by considering two situations. In the first situation, we wish to gain information about an area where hydrocarbons have been discovered, and use the value of perfect information to determine which locations are the best to drill. In the second situation, we consider the problem of abandoning an area when only dry wells are drilled. For this latter, we use an abandoned revenue criterion to determine the drilling locations. The application is from the North Sea. Our main focus is the description, visualization, and interpretation of the results for relating the statistical modeling to the local understanding of the geology. 2nd revised manuscript received November 15, 2010 Gabriele Martinelli holds a B.Sc. degree (2006) and an M.Sc. degree (2008) in mathematical engineering from Politecnico di Milano, Milan, Italy. He is currently a graduate student at the Norwegian University of Science and Technology. Jo Eidsvik got his M.Sc. degree (1997) from the University of Oslo and his Ph.D. (2003) from the Norwegian University of Science and Technology (NTNU), Norway. He worked for the Norwegian Defense Research Establishment and for Statoil. His current position is associate professor of statistics at NTNU. Ragnar Hauge has a master's degree in statistics from the Norwegian University of Science and Technology and a Ph.D. in statistics from the University of Oslo. He is currently an assistant research director at the Norwegian Computing Center, where he has been working since 1995. The main focus of his work has been stochastic modeling of facies and stochastic seismic inversion. Maren Drange Førland received an M.Sc. degree in statistics from the Norwegian University of Science and Technology in 2008. She is currently a research scientist at the Norwegian Computing Center.