TX 75083-3836, U.S.A., fax 01-972-952-9435.
AbstractThis paper describes an investigation of abandonment decisions and shut-in policy as a function of uncertainty in oil price. We first review a fundamental error that is often made in predicting the outcome of, and hence making decisions about, systems that are subject to uncertainty: for many common models, the use of "best" estimates of the uncertain input parameters to the model does NOT result in the "best" estimate of the model's output ("best" is defined as average, or minimum error). The same argument applies to predicting output statistics, such as P10 or P90, from corresponding input statistics. This is part of the reasoning behind, for example, the use of geostatistical simulation models of the sub-surface, rather than smoothed, spatially-averaged models.In this work the focus is on decision errors caused by temporal averaging, specifically, the "smoothing out" of oil price fluctuations over time, and by restricting uncertainty investigations to the uncertainty in parameters of smoothed price models. We illustrate these points by application to determining optimal abandonment decision policies. We show that it is better to wait for a period after first going cash-flow negative, and how to estimate the length of that time. We also show that these conclusions are relatively insensitive to the oil-price model parameters. Further we show that, if maximizing NPV is the objective, then contrary to normal operating procedures, it is more economic to choke-back production in periods of low oil price.