Full text loading...
-
Reducing 3D uncertainty by an ensemble-based geosteering workflow: an example from the Goliat field
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, 3rd EAGE/SPE Geosteering Workshop, Nov 2021, Volume 2021, p.1 - 5
Abstract
A probabilistic decision support system for geosteering coupled with automated data assimilation techniques has been shown to outperform many humans at least in sandbox 2D experiments. Utilizing probabilistic decision support for optimal positioning of real wells requires a 3D probabilistic earth model compatible with standard tools, and rapid generation of synthetic logs of extra-deep measurements.
We introduce a method for representing a 3D earth model with depth uncertainty in stratigraphic surfaces. The surfaces are parameterized using a multi-scale technique, capturing correlations and uncertainties on both large and small scales. The probabilistic earth model is updated by an ensemble-based method that assimilates extra-deep EM measurements acquired while drilling. A recently developed Deep Neural Network model ensures rapid simulation of extra-deep EM measurements. However, the model assumes local layer-cake geology and thus can only interpret 1D sensitivity for each measurement position.
The proposed method is applied to a synthetic study based on geosteering in the Goliat field. The results demonstrate that data assimilation of the extra-deep EM measurements successfully reduces uncertainty in the 3D model and updates it towards the truth. Moreover, we demonstrate 3D visualization of the probabilistic model compatible with standard geo-modelling tools, useful for decision support.