This article, written by Special Publications Editor Adam Wilson, contains highlights of paper SPE 182719, “Drill and Learn: A Decision-Making Work Flow To Quantify Value of Learning,” by R.G. Hanea, SPE, P. Casanova, and L. Hustoft, Statoil; R.B. Bratvold, SPE, University of Stavanger; and R. Nair, C. Hewson, SPE, O. Leeuwenburgh, and R.M. Fonseca, TNO, prepared for the 2017 SPE Reservoir Simulation Conference, Montgomery, Texas, USA, 20–22 February. The paper has not been peer reviewed.
Uncertainty assessment and reduction are often elements of high-quality decision making, although they are not, in themselves, value creating. Value can be created only through decisions, and any decision changes resulting from assisted history matching should be modeled explicitly. This paper presents a comparison of existing work flows and introduces a practically driven approach, referred to as “drill and learn,” using elements and concepts from existing work flows to quantify the value of learning (VOL).
The idea to apply numerical optimization methods to reservoir models in order to arrive at optimal field-development plans has been around for a long time. Early methods for optimization were quite limiting, however, in terms of the complexity of the problems that could be addressed. Recent developments in algorithms and computing power have made it possible to begin to address the full complexity of the field-development optimization problem, including a large number of decision variables of various types, a better characterization of geological uncertainty, handling of realistic platform constraints, and operating strategies of newly drilled wells.
Closed-loop reservoir management aims to incorporate new information into models and optimize field-development or reservoir-management strategies on a nearly continuous basis. The main assumption underlying the closed-loop-reservoir-management framework is that acquiring information can change decisions about how the field should be developed or operated such that certain performance objectives are improved. This assumption is identical to that underlying the concept of value-of-information (VOI) determination, which addresses the question of whether one should actually acquire specific data considering not only the expected effect on the system performance but also the cost of acquiring these data from which the information is to be extracted.
A first attempt to investigate the feasibility of applying a decision analytic VOI work flow to quantitative reservoir-model-based decision making tried to evaluate the increase in expected economic value with and without the use of specific data with associated measurement errors. The probabilistic aspect of a VOI evaluation was accounted for by the use of an ensemble of models that capture geological uncertainty, and the value was based on applying optimization methods to the model ensemble and evaluating the resulting strategies on synthetic truth models that were not part of the ensemble. Improvements to the computational complexity of the originally proposed work flow were suggested in subsequent work. Still, even with the suggested modifications, the computational cost of a formal VOI evaluation would remain prohibitive for many real-field models without further modification or simplification.