Value of learning – practical benefits of an ensemble-based framework from fit-for-purpose models to value-driven decision

Typically, subsurface knowledge and operational aspects are key for the success of field-development strategy, well planning, and reservoir-management decisions. In general, reservoir models are used to: 1. quantify subsurface understanding, 2. obtain quality prediction, and 3. support decision making. In practice, it is challenging to build or calibrate reservoir models that preserve enough physical and operational representation to the degree at which quality decisions can be obtained. Therefore, it is necessary to identify the interlink between subsurface understanding and decision quality. The interlink is basically a collection of few subsurface and operational elements that have significant impact to decision - This interlink is referred to as the value of (subsurface and operational) learning. This paper demonstrates the use of an ensemble-based workflow to: 1. identify value of learning, 2. evaluate value of learning with a variety of decisions, and 3. access the consequences of value of learning. In this work, we used EVEREST, a technology for optimization under uncertainty co-owned by TNO and Equinor, to quantify the impact of both geological and operational uncertainties (e.g. drilling time, production time, rig arrival and departure availability) to decision quality. We derived value of learning from the computationally efficient and attractive approximate gradient (StoSAG method) used in EVEREST. The approximate gradient, which here serves as a sensitivity metric, is used to rank the influence of geological and operational parameters to the decision. The key subsurface and operational elements are captured and evaluated for a robust decision making. The proposed workflow is demonstrated with a synthetic but realistic REEK model. The paper addresses on how value of learning could be used in practice with the potential benefits for decision making. The results showed that the interlink between key parameters and key actions are crucial for robust decision making from which the consequences and optimization scenarios are evaluated systematically. In addition, the value of learning helps practitioner access relevant key information that connects key uncertainty to the key decision. The proposed method leads to decision maturation and support system that build the connection between subsurface knowledge, operational aspects, and decision making. (C) 2022 European Conference on the Mathematics of Geological Reservoirs 2022, ECMOR 2022. All rights reserved.