Data analytics in asset valuation under uncertainty: A case study of unexplored oilfields

Abstract

Valuing unexplored oilfields is challenging since it deals with multiple sources of uncertainty and covers a lengthy period. Previous research suggests that the major source of uncertainty in the valuation of unexplored oilfields is reservoir condition. It is usually represented by a bunch of parameters based on which reserve volume and production rates can be estimated. The values of these parameters are revealed only after exploration well is drilled; we refer them as post-discovery parameters. Prior to exploration drilling, data on some reservoir characteristics, called the pre-discovery parameters, are available. This research aims to develop a model to estimate the probability distribution of post-discovery parameters based on pre-discovery data. The model is data-driven, built using the database of proven reservoirs. We use the Bayesian network to develop the model and apply the k-fold cross validation to test the results. By adding a window parameter to the target variable, the model provides information regarding the trade-off between accuracy and confidence. In general, compared to the initial model that uses a priori clustering based on the reservoir’s lithology and depth, our Bayesian network model produces lower variances.