Probabilistic Forecasting in Dry-Gas Reservoirs and Monte Carlo Simulation and Decision Tree Comparison

by
Maria Estilita Portillo, M.S.E.

University of Texas at Austin, 2005
Supervisor: Larry W. Lake

Probabilistic methods have gained popularity and have become the preferred approach in many activities related to the petroleum industry, among them probabilistic forecasting. Two of the most common tools used in this context are Monte Carlo simulation (MCS) and decision tree (DT) analysis. However, few studies exist when production is sought from multiple reservoirs with uncertainty associated to each. We developed a reservoir simulator for dry-gas reservoirs that is able to handle multi-reservoirs and perform production forecasting in a probabilistic mode including a simple economic model. The simulator was used to perform both a deterministic and a probabilistic assessment of the performance of three Nigerian reservoirs under different development scenarios.

We found that for all the scenarios developed there is a compromise between an extended production plateau and the maximum net present value (NPV) of the project. We also used the tool to compare the results obtain from MCS vs. DT. We found that the DT approach can closely reproduce the results of MCS if enough “branches” of the tree or probable values for the uncertain variables are used. The reduction of the number of branches in the tree results in loss of the extreme values of the output distributions.

Back to theses index