Automated multi-scenarios well correlations

Stratigraphic correlation of well data is cornerstone both in regional syntheses and reservoir modelling workflows. However, this task demands high expertise and can be extremely tedious. Practically, geologists often lack time to explore all possible correlation scenarios and assess the corresponding uncertainties.

In this demo project we explore how well correlations can be partially automated and digitally assisted in order to efficiently consider multiple scenarios and evaluate the risk associated.

To do so we are working on an expert system based on a combination of Dynamic Time Warping algorithms and geological rules. The rules can be applied via the cost function within the dynamic time warping algorithm, as demonstrated by research from the University of Lorraine and the RING consortium  (of which IFPEN is a member).

With this approach, stratigraphic correlations can be generated and the uncertainty in the correlation can be assessed.

Future work includes testing the existing methods on large data sets, including knowledge infused cost functions. There is further potential in training convolution neural networks to describe the correlation for known geological contexts.