Abstract
History matching with uncertainty quantification has been a topic of great interest over the last 10 years, with many algorithmic approaches developed, and many applications presented. Most of the applications focus on the uncertainty in the petrophysical properties of the reservoir, as that is by far the easiest to parameterise.
There are few papers that address the topic of structural uncertainty quantification, yet that is potentially the biggest component of subsurface uncertainty. The principal problem in accounting for structural uncertainty is the difficulty in parameterising subsurface structures, automatically gridding the reservoir model and ensuring that the grid will simulate in a reasonable time.
This paper presents a new approach to structural uncertainty quantification. We have adapted the immersed interface method from computational mechanics, used to model fluid flow in domains with moving boundaries, to reservoir simulation. The immersed interface method is used with an extended multipoint flux approximation to completely decouple the parameterisation of structural features from their representation on the grid.
Using this new technique, we are easily able to history match a 2D synthetic case where we simultaneously vary the top and bottom surfaces of a reservoir, the location and orientation of faults within the reservoir, and the location and size of pinchouts.