The paper discusses a novel workflow and tools for stochastic optimization under uncertainty, aimed at maximizing the value of exploration and early appraisal assets, using a public example of an East African offshore gas discovery.

Following an initial discovery well, the prospect enters an appraisal phase with tangible uncertainties in the resource base (P10/P90 ratio of 3.1), well productivity (P10/P90 ratio of 2.1), and cost estimates (30%/50 % tolerance). The choice of LNG train capacity and the development concept changeover point has a profound effect on the asset evaluation results used to attract new project partners.

These uncertainties raise a number of critical questions: How can a facility engineer elaborate on these key parameters given the threefold uncertainty in the resource base at this project stage? Will these choices change under a random walk in oil prices, or will they change given different attitudes towards risk by management? Does maximization of project after-tax value and of production entitlement imply identical values of control parameters in these circumstances?

The challenges are resolved by sampling the solution space with 54,000 full-cycle evaluation trials run in just over 13.5 hours using a combined power of an exploration decision support tool with fit-forpurpose and fit-to-data risks, resource, and value assessment that honors all risks and uncertainties; a workflow management tool for decision optimization based on a multiple-realization approach powered by various uncertainty and optimization methods; and a proprietary link between these two applications.

The original operator’s view on the facilities concept was found to reflect a P50 of reserves uncertainty and a P50 of worldwide LNG train capacity distribution. Optimizations were performed to address five progressively complicating problem formulations. The achieved outcomes align with conclusions that can be derived through a reasoned train of thought in an economy-of-scale context, indicating our solution is applicable for modeling situations without simple analytical solutions.