Software tools that allow the user to automatically generate many models that capture and propagate uncertainty in the subsurface have existed for decades. Despite their existence, true adoption of these types of tools have been limited. Reasons for low acceptance/adoption can stem from:
Challenges in maintainability/transparency of the automated modeling workflows.
High investment cost in terms of time and resources required to set up and maintain the workflows.
Company processes for reservoir modeling not designed for an uncertainty centric paradigm
Expected deliverables from the modeling project not aligned with an uncertainty centric paradigm.
Hence, despite our best efforts to use state of the art tools, we often observe that modeling projects easily become victims of the Seven Wastes of Reservoir Modeling Projects: 1. Non-utilized human talent; 2. Defects, 3. Handoffs and delays; 4. Relearning and task switching; 5. Non-value adding features; 6. Overspecialization; and 7. Loss of information.
In this paper we evaluate learnings from the past 30 years of reservoir modeling projects in the oil and gas industry. Specifically looking at methods that aim at quantifying the uncertainty in the subsurface using multiple realizations – also known as ensemble-based methods. We will highlight what we perceive are the major challenges that companies currently face - the so called Seven Wastes of Reservoir Modeling Projects - and outline potential ways to overcome these challenges based on lessons learned from more than 100 modeling projects on fields world-wide over the last 10 years. Furthermore, we highlight the importance of not falling victim to the many cognitive bias effects that can limit us in improving our understanding and consequently the value of our reservoirs. We demonstrate how a combination of fit for purpose algorithms and process help put subsurface teams in a mindset of continuous learning that helps reduce the typical pitfalls seen in traditional reservoir modeling projects. The benefit of the proposed way of working is that subsurface teams greatly reduce the time it takes from incorporating new data or new hypotheses about the subsurface, until having a full ensemble of models that can be used to support the business decision at hand. As a result, we potentially reduce the cycle time from new data, or new ideas to conditioned models from months or years to hours or days.